eCAADe 2014

Fusion
Volume 1

Volume 1 - eCAADe 32 | 1
 Editor
Emine Mine Thompson
Department of Architecture and Built Environment
Faculty of Engineering and Environment
Northumbria University

1st Edition, September 2014

Fusion - Proceedings of the 32nd International Conference on Education and research in Com-
puter aided Architectural Design in Europe, Newcastle upon Tyne, England, UK, 10-12 Septem-
ber 2014, Volume 1. Edited by Emine Mine Thompson. Brussels: Education and research in
Computer aided Architectural Design in Europe; Newcastle upon Tyne: Department of Archi-
tecture and Built Environment, Faculty of Engineering and Environment, Northumbria Univer-
sity.

ISBN: 9789491207075 (eCAADe)

Copyright © 2014

Publisher: eCAADe (Education and research in Computer aided Architectural Design in Eu-
rope) and Department of Architecture and Built Environment, Faculty of Engineering and En-
vironment, Northumbria University.

Cover Design: Northumbria Graphics

All rights reserved. Nothing from this publication may be produced, stored in computerised
system or published in any form or in any manner, including electronic, mechanical, repro-
graphic or photographic, without prior written permission from the publisher.

2 | eCAADe 32 - Volume 1
 eCAADe 2014

Fusion - Data Integration at its best

Volume 1
Proceedings of the 32nd International Conference on Education and research in
Computer aided Architectural Design in Europe

10-12 September 2014

Newcastle upon Tyne, England
Department of Architecture and Built Environment
Faculty of Engineering and Environment
Northumbria University

Edited by
Emine Mine Thompson

Volume 1 - eCAADe 32 | 3
4 | eCAADe 32 - Volume 1
Theme

Fusion- Data Integration at its best

This is the first volume of the conference proceedings of the 32nd eCAADe conference, held
from 10-12 September 2013 at the Department of Architecture and Built Environment,
Faculty of Engineering and Environment, Northumbria University in Newcastle upon Tyne,
England. Both volumes together contain 130 papers that were submitted and accepted to
this conference.

The theme of the 32nd eCAADe conference is Fusion- data integration at its best.

All quests for data integration in architecture and the construction industry lead ultimately
to FUSION, a synthesis of knowledge that transcends mere combination. FUSION is the
dream of a knowledge system that will enable multiple sets of data, in manifold formats, to
be presented in a unified view. This conference is exploring the possibilities for advanced
levels of data integration in the service of the representation and management of the natural
environment, and the design, visualisation and making of the built environment. These
proceedings are presenting the contributions which explore the elusive goal of FUSION in
architecture and related fields.

The first volume of the proceedings contains 65 papers grouped under seven sub-themes
(Towards Smarter Cities, Design Tool 1, CAAD Education, Fabrication, Shape-Form-Geometry,
Visualisation and Digital Heritage).

Emine Mine Thompson

Theme - Volume 1 - eCAADe 32 | 5

 Sponsors of the eCAADe 2014 Conference

Autodesk

Bentley

Northumbria University

6 | eCAADe 32 - Theme - Volume 1

Acknowledgements

Welcome to the 32nd proceedings of eCAADe in Newcastle upon Tyne, England - exactly fif-
teen years after the last eCAADe conference in England we are very happy to welcome you
back.

The idea of bringing the eCAADe conference to Newcastle dates back to October 2010. Since
then there have been many structural changes at Northumbria University and although many
unexpected things took place, the conference organisation was one of the constant threads
of life’s tapestry during this time.

Over the last four years many people helped to make this conference possible and it is my plea-
sure to acknowledge and thank them here. First of all, I would like to thank all Associate Deans
for Research and Innovation, initially Prof David Greenwood, and later on Prof Fary Ghassem-
looy, and then Prof John Woodward who supported the idea very enthusiastically. Also a huge
thanks goes to Margaret Horne (Director of Built Environment Visualisation Centre), Peter Bea-
cock (Director of Architecture), Ruth Dalton (Head of VRV Group) and Kevin Thomas (Head of
Architecture and Built Environment) for allowing time and space so that the organisation pro-
cess runs smoothly and for their trust and understanding.

Secondly, the previous conference organisers Henri Achten and Rudi Stouffs passed on their
experience and knowledge without hesitation - Thank you! Financial support was generously
provided by Autodesk and Bentley Systems and Northumbria University.

I would also like to thank to eCAADe council. The Council was very supportive from begin-
ning to the end and gave help with many aspects of the organisation. Both Presidents José
Duarte (until 2013) and Johan Verbeke (since 2013) encouraged and gave full support during
the organisation. Also, Bob Martens, as liaison with the conference host, and Nele De Meyere,

Acknowledgements - Volume 1 - eCAADe 32 | 7

 as administrative support, provided vital input.

This year, authors were challenged with the use of two different submission systems. Initially,
for the abstract submissions we utilised the OpenConf system which ran smoothly and we
thank Martin Winchester for his support on this. For the full paper submissions, Gabriel Wurzer
and his team allowed us to test-drive the ProceeDings tool in its Beta version. This is has been
a learning curve both for the ProceeDings team and for the organisation team. Although there
might still have some little issues to iron out, ProceeDings tool is a fantastic aid to the confer-
ence organiser. A huge thanks goes to Gabriel Wurzer, Wolfgang E. Lorenz and Gerda Hartl
who worked so hard to put together these proceedings; your efforts are highly appreciated.

The call for extended abstracts generated 240 submissions. With the help of the eighty-five
international reviewers, each submission was double-blind reviewed by three reviewers. 152
extended abstracts were accepted for full paper submission. Regrettably some authors with-
drew from submitting the final paper and eventually 130 papers made it to publication and
presentation. I thank and congratulate all the authors for their hard work and support on us-
ing the ProceeDings tool and finalizing their papers. I also owe a huge thanks to the reviewers
(see the reviewers list), who volunteered their time in order to assess the submissions.

For the third year in a row eCAADe is publishing two volumes of proceedings. We believe this
shows the growing interest for the eCAADe in general and also the continuing development
of computing and its reflection in architecture and design education and research. This year
another new approach with regards to the proceedings was evaluated: an option to provide
the proceedings as either electronic copy only or both hard and electronic copy. It is our belief
that this was a successful trail with the almost 50/50 divide on the request for the proceedings
(42% requesting electronic copy and 58% requesting both).

As the chair, I had support from the organisation committee, especially James Charlton, Ruth
Dalton, Andy Dunhill, Julie Clarke, Sara Walker, Bob Giddings, Minnie Fraser, David Green-

8 | eCAADe 32 - Acknowledgements - Volume 1

wood, Brian Agnew, and Oliver Jones. I further want to give my thanks to the Faculty Re-
sources Team, in particular Sue Ralph and the University Event’s team Shelley Brunsdon and
Marta Pyrek for their great assistance. Since 2010 the NewcastleGateshead Convention Bu-
reau, part of NewcastleGateshead Initiative provided excellent assistance for the conference
as well.

During the conference we had session chairs who led the twenty-five presentation sessions
and the student volunteers who assisted everybody throughout the conference - Thank you
all!

I am also very grateful to our Keynote Speakers; Christian Derix (Director of the Computational
Design Research group of Aedas Architects), Dr Andrew Hudson-Smith (Director of the Centre
for Advanced Spatial Analysis – CASA at The Bartlett, University College London) and Professor
Stephen Lockley (Chair of Building Modelling at Northumbria University and Research and Edu-
cation director of the BIM Academy).

A further special thanks also goes to Robert Aish (Bartlett School of Architecture and Visiting
Professor, University of Bath) and Professor Tom Maver (Research Professor, Mackintosh School
of Architecture and Emeritus Professor, University of Strathclyde) who arranged and co-chaired
the Rountable Session on “Design Computation in Practice” with the roundtable participants;
Shajay Bhooshan (Zaha Hadid Architects), Cobus Bothma (Kohn Pedersen Fox), Christian Derix
(Aedas Architect), and Francis Aish (Foster and Partners) who deserves my sincere thanks and
acknowledgements.

Finally I want to thank to my family for their understanding, patience and support over the last
four years while I was busy with the organisation of this conference.

eCAADe 2014 Conference Chair

Dr. Emine Mine Thompson

Acknowledgements - Volume 1 - eCAADe 32 | 9

 Organisation Committee

Andy Dunhill
Bob Giddings
Brian Agnew
David Greenwood
Dawn Whitton
James Charlton
Julie Clarke
Lawrence Hughes
Minnie Fraser
Oliver Jones
Ruth Dalton
Sara Walker

10 | eCAADe 32 - Acknowledgements - Volume 1

ProceeDings
A web-based word processor automating the production of conference
proceedings

Gabriel Wurzer1 , Bob Martens2 , Thomas Grasl3

1,2
Vienna University of Technology 3 SWAP
1,2
http://www.tuwien.ac.at 3 http://www.swap-zt.com
1,2
{gabriel.wurzer|b.martens}@tuwien.ac.at 3 tg@swap-zt.com

In this paper an online editing system for eCAADe papers is presented, which is
also the technology behind this volume. On the occasion of the eCAADe 1999
conference in Liverpool, a novel layout for the proceedings was developed. In the
course of forthcoming annual conferences, this became the distinctive "look and
feel" for eCAADe papers. Due to the complexity, professional typesetting was
required for and the authors were disconnected from the publication and layout
stage. This paper elaborates on the development and implementation of a
web-based tool, which takes care of the typesetting and delegates this activity to
the authors. Neither software installation is required, nor specific training must
be completed in advance. On top of this the degree of homogeneity can be raised
significantly, thus supporting the editors in charge to concentrate on the task of
harmonising the publication content.

Keywords: Word Processing, Proceedings Preparation, Cloud Computing

OVERVIEW now, the workflow suffered from a sharp cut between

To the annoyance of both authors and editors, word
processing packages are still in their infancy when
it comes to guarding a publication template against
modification and improper use. Modification, i.e. in-
serting or altering template styles, is a 'feature' that is
often seen when copying and pasting between doc-
uments or from the web. Improper use of a tem-
plate, on the other hand, is a failure to meet the se-
mantics of the entered content itself - think, for ex-
ample, of entering references incorrectly. On top of
this, the authorship is in charge of the final layout
and obliged to stay within the given page-limit. Until
the word-processing-stage and the loosely coupled
layout software stage.
During the past three years, we have been devel-
oping a web-based word processor that copes with
the mentioned problems: It enforces eCAADe's own
conference proceedings style and makes use of struc-
tured content with added semantics. The resulting
contributions can then be automatically compiled
into a proceedings book, or exported into LATEX for fur-
ther editing.
We will start with a brief outline of our solution,
but then immediately come to our main contribu-

Fusion - Data Integration at its best - Volume 1 - eCAADe 32 | 11

 tion: The development process behind such a mas-
sive web application. Decisions taken in that phase
are often far from obvious, which is why we would
like to share some of our insights with soul mates
who also want to bring their app to the web or into
the cloud. We will conclude with a brief statistical
overview of some of the data collected during the
submission process. This contribution targets hence
the wider eCAADe2014 conference theme of "Fusion"
in the sense of a collaboration tool aiming at data in-
tegration.
BACKGROUND AND RELATED EFFORTS
When speaking about the editing of conference pro-
ceedings, managing the time-line is the dominant
issue. In the given area of CAAD the (printed) pro-
ceedings ought to be finished by the beginning of
the conference; a post-conference publication is no
option, neither is preponing the deadline for sub-
mission. This means that within a timeframe of ap-
proximately 12 weeks (or so) roughly up to 150 pa-
pers with a significant amount of figures/tables/im-
ages need to be edited. Therefore, it makes sense
to delegate a justifiable amount of work to the au-
thors supporting the homogeneity of each individ-
ual contribution. An electronic publishing environ-
ment should be able to secure unintended violations
of implemented rules/restrictions. It should be men-
tioned that data entry via a structured web mask is
mandatory. The papers/contributions are to be cen-
trally stored and perpetual backup routines must be
implemented.
A first thought which instantly comes up is that
a gaggle of editors must have been tackling these
issues over and over. Possibly there exist different
views on the level of consistency and accuracy to be
achieved. It could, however, also be that some of
the work involved is regarded as unavoidable and di-
vinely ordained, be it alone the persistent use of up-
per and lower case in the title and headings. Indeed,
wide-ranging reflective publications are numerous
[such as Russey et al 1995; Fredriksson 2001] and
even a series of conferences has been dedicated to
12 | eCAADe 32 - Fusion - Data Integration at its best - Volume 1
this topical area [such as King 2000; www.elpub.net].
When searching for similar tools predominantly a
variety of tools can be retrieved, allowing to handle a
collection of already created PDF-files. ProceeDings
focusses on the contrary on the production of the
publication entry itself, carefully respecting the pre-
defined publication guidelines. As a matter of prin-
ciple, content management systems (for websites)
would allow to settle a similar task within a system of
distributed roles. However, this would in most cases
require a training, whereas ProceeDings claims to be
almost self-explanatory.
A solution like ProceeDings does not make the
guild of editors jobless as there is still a need for
taking care of the content of the publication as a
whole. Especially the setup of coherent sessions and
their composition in (parallel) sequences is rather de-
manding.
PROBLEM: FORMATTING
It is not the first time that the eCAADe has created a
web-based platform for its community: The Cumu-
lative Index on CAD (Martens and Turk 2000), for ex-
ample, was founded as paper archive for education
and research in architecture and urban planning. We
have targeted the same community with our web-
based word processor, however, it is more than likely
that we will find us serving a greater range of fields in
the future.
There were two main drivers that led to the de-
velopment of the tool:
• We were unsure what to do about Microsoft
Word's persistent tendency of breaking the
formatting of a document. Locking the tem-
plate had only mixed success, and likewise did
Springer's Manuscript template (which uses
Visual Basic for Application Macros for taming
Word to some extent, but: The user has to en-
able Macros and copy-and-pasting from the
web automatically allocates new paragraph
styles...).
 • We felt that a general-purpose word proces-
sor is ill-suited for the conference's needs, as
the user enters text but the program does
not enforce semantic and structural rules (see
next section) that lie at the heart of every pa-
per. LATEX does (or better: can do, if instructed)
all of this, however, it is not common in the
field and requires a certain degree of a learn-
ing effort. Having a larger audience switching
from Word to LATEX was, as can be imagined,
also not feasible.
SOLUTION: SYNTACTICALLY DUMB, SE-
MANTICALLY RICH
Microsoft Word's problems in keeping formatting in-
tact is the result of mixing text semantics (head-
ing, paragraph, bullet list,...) and styling (font fam-
ily, size,...). Copy-and-pasting brings out the worst
of both worlds, in the sense that the text is pasted
using the current paragraph format, which is auto-
matically updated to reflect the pasted text's styling.
Discarding styling information altogether and past-
ing only the text would be the obvious solution, how-
ever, there is no way of forcing this behaviour without
user intervention, e.g. by storing settings together
with the template.
Pasting "only text" does do away with superflu-

Figure 1
(a) Contributions as
set of paragraphs
which are
context-aware, they
know their place
within the text
body (example:
paragraphs cannot
be moved higher
than the
"keywords" section)
and also the editing
options according
to their type
(example: figures
offer an upload).
(b) Editing happens
via plain text entry
and markup, which
enables the tool to
discard superfluous
formatting as in the
given example.

Fusion - Data Integration at its best - Volume 1 - eCAADe 32 | 13

 ous styling, however, we completely loose text se-
mantics - the paragraph format, bold, italic, and so
on. In our web editor, we have therefore devised a
two-fold editing strategy that enables us both.
Semantically rich. We regard each contribution as
a set of paragraphs containing plain text (see Figure
1a). Paragraphs are context-aware: For example, a
heading cannot be moved further than the list of key-
words. One may not add an abstract block in the
body of the paper, or anything else but references in
the references section.
Syntactically dumb. Paragraphs are edited - one at
a time - simply by clicking them. In the simplest
case of a "paragraph" format, the application then
shows a text box where one can enter or paste plain
text (see Figure 1b). For paragraph types that ac-
quire structured information, such as the list of au-
thors or a reference entry, we present not one but
many plain text boxes in which the user types in the
needed information. Formatting beyond the para-
graph level is done using a wikitext-like syntax, for
example *bold* or /italic/. What wikitext markup is
available in each paragraph ultimately depends on its
type: For example, headings offer nothing but plain
text, while paragraphs furthermore enable bold, ital-
ics and equations done in this manner.
TECHNICAL OUTLINE
Our tool is composed of four technical layers:
• Editor Front-end: We chose to implement a
"Web 1.0" application which is based on clas-
sical request/response cycles, utilizing PHP
(served by nginx/php5-fpm) for generating
the HTML presentation (styled by CSS). Op-
tional functionalities such as the real-time pre-
view of formulae are handled via JavaScript
libraries (jQuery, MathJax) which degrade
gracefully in case of missing browser support.
As a matter of fact, the application displays
even on outdated browser versions (we tested
e.g. Internet Explorer 6 and the text browser
Lynx [1], which dates back to 1992).
14 | eCAADe 32 - Fusion - Data Integration at its best - Volume 1
• Editor Back-End: Instead of a classical
database, we store our documents as files.
The reason behind this is that load tests
showed database performance to drop when
memory is constrained, whereas files-based
access remains fairly stable even if the server
is under heavy load. Furthermore, we uti-
lize a memory database (memcached) for fast
access to back-end data and session manage-
ment.
• Printing: Documents are converted into LATEX
for printing. The actual generation of a
PDF is a very costly task, and therefore, we
have utilized a flexible "printer queue" (utiliz-
ing Beanstalk for PHP) where multiple other
servers could help in generating a print pre-
view of the currently edited document. On
the other hand, having too few resources to
serve all print requests will not overload the
server, since all clients are queued and only
some are served at a time.
• Book Compilation and Metadata creation:
The final compilation looks at a spreadsheet
stating (1.) in which volume is (2.) which
session, containing (3.) what contributions.
It then re-renders all PDFs in that order, giv-
ing the correct page numbers and also assem-
bling the list of contents in the process. The
resulting PDF are the proceedings, but not the
final product: We still have to do an extraction
of metadata (e.g. authors, title, references)
and creation of metadata for indexing, in for
example CUMINCAD, automatically.
All these parts work together so that authors can col-
laborate in producing a consistently formatted pro-
ceedings that saves the editor valuable time which
can now go into the actual editing process . Scientif-
ically though, these points are hardly of concern. In-
deed, what we are more interested in is how the com-
munity works with our tool, and when this work is de-
livered. An analysis of this is brought in the next few
sections.
LESSONS LEARNED
As eCAADe 2014 was the first time that a web-based
editing tool was used, we coined it ProceeDingsBeta .
Due to the constant correspondence with the au-
thors and monitoring the layout of the whole pro-
ceedings ourselves, we are able to give some insights
on what worked and what did not.
What worked
Inserting well-behaved text. Pasting paragraphs
from the clipboard as plain text and formatting that
with markup worked like a charm. The only excep-
tion were inline equations, which presented an own
problem (chars that are known to Microsoft Word but
not to Unicode, WingDings and the like).
Warnings. Most authors took a great deal of pride
in having their paper warning-free: For example, the
system would complain about too short or too long
abstracts, and people strongly responded to that.
That (in the end) the system was so liberal as to al-
low submission of every paper, regardless of warn-
ings, was not of concern. The people simply wanted
to help getting their work right, and this is one of the
most assuring things which we encountered during
review of all the final submissions. Some exceptions
(which we have to deal with, of course) were the ref-
erences, where the system would sometimes com-
plain wrongly about a URL being wrong, when in fact
they had only given an additional [accessed 2nd June
2014]. We are delighted by this ability to judge the
system by ones personal experience in scientific writ-
ing - some authors also left use NOTE paragraphs ex-
plaining what they want to appear in the paper and
why they could not accomplish - and thus this is one
of the points which we build on for the next version,
exposing e.g. more tools that we as editors had in
figure positioning and paragraph indentation ("dont-
indent" my paragraph, please).
Selecting the right kind of paragraph for the job.
For people not concerned with structured text, terms
such as "heading 1, 2, 3", "authored book" and "edited
book" might not be very descriptive. So it shook us
when we realized that people actually understand far
Fusion - Data Integration at its best - Volume 1 - eCAADe 32 | 15
more about paragraph formats than we had antici-
pated: After all, we had not given out a template in
LATEX instead of the web-based system because of fear
that this would be not understandable. The impres-
sion that we got is that that, even though LATEX might
be too much to stomach for the whole community,
the general concepts are clear for everyone - and this
includes structured markup such as the ones men-
tioned.
What did not work
A paper is not an image gallery. It would be a lie
to say that most authors added figures accompany-
ing the text. In fact, it was the other way round: The
text would accompany the figures! Given the very
limited abilities of LATEX in positioning the graphics,
this was as much of a pain for us as in layouting as has
been for the authors in editing. The future perspec-
tive on this is clear: (1.) Constrain the use of images
to cases where they are visualizing the text (every im-
age needs to be cross-referenced, as mentioned in
the User Guide), (2.) constrain the number of images
to a minimum, we think of 5-8 at the moment and (3.)
give more options for positioning the images in the
text, accompanied by a clear description over how
LATEX will attempt to position them.
Figure positioning. Figure positioning deserves
some more attention: LATEX will position a figure ei-
ther "right here" in the text or let it "float":
• "Right Here" means that images will be one
column wide. If there is not enough space ver-
tically in this column, LATEX will shift them to
the next column (or even the next page, 1st
column), leaving an ugly hole in the text. Au-
thors need to know that they need to close
these holes by inserting a "here" figure where
it has enough space, at best in the middle of a
column, surrounded by lots of text.
• "Floating" means that LATEX will put a figure ei-
ther at the top or at the bottom of the next op-
portunity. That again means: When a figure
is "beyond the top" of the page, it will insert
 it at the bottom of this page or at the top of
the next page. When the layouter has already
crossed the "the bottom" of the page, it will
insert it on "the top of the next page". In prac-
tise, this means that all floating pictures need
to be defined well before their insertion posi-
tion, so that the layout algorithm has them at
hand when going through the text. Arguably,
this is quite counter-intuitive. However, this is
nothing that we can work around, as LATEX is
built like that.
The mentioned points are even more enervating
when preparing for different kinds of output media:
For example, an eBook has the requirement that fig-
ures always appear exactly in the spot where they are
mentioned, i.e. "here". For a printed proceedings,
however, we may additionally use "floating" figures,
which may need to be defined before the spot where
they are actually referenced. Essentially this is the di-
vide between structure (as in eBooks) and layout (as
in printed proceedings). It will be our task to think
about ways in which we can bridge this gap, provid-
ing more options for positioning (e.g. figure "on an
own page", which was used during editing) and also
for vertical spacing before and after figures.
Formulae. Formulae were the main cause why a pa-
per did not print properly. This is no coincidence,
but the result of two diametrically opposed policies
on dealing with erroneous input: The web-based
formula viewer (MathJax) would simply ignore all
offending markup and display what it could make
sense of without complaining, while the printing al-
gorithm (LATEX) would immediately stop and report an
error.
Errors produced by LATEX are handled by ProceeD-
ings such that it shows an error page. It does, how-
ever, neither know what happened nor where the
error lies (i.e. no parsing of error text, yet). There-
fore, we display a generic error page that gives some
hints over what could have gone wrong, but is no use
when it comes to hunting down specific errors in a
formula - leading to a lot of despair and troubleshoot-
ing via mail. In further detail, our analysis shows that
16 | eCAADe 32 - Fusion - Data Integration at its best - Volume 1
there were three separate cases that led the printing
algorithm to fail:
• "Unicode" formula input into web editor:
What most people do is paste a formula from
Word into a paragraph. Technically, this is no
formula, but merely a set of characters that are
hopefully Unicode - and the system will ap-
propriately save them as normal text. In some
cases, however, characters that Word pastes
are simply not Unicode - they are symbols in
Windows encoding! So this process can fail
terribly, if we have no clue what this symbol
is (remark: this is likely, as we have one pro-
grammer (currently writing this) against the
rest of the Microsoft world). A better way
would be to input a formula markup, pre-
sented in the next bullet point.
• "Formula" input into web editor: With the
help of surrounding #, an author can insert an
inline formula (the other option would be to
make a standalone formula paragraph, which
does not need that). Some authors have taken
this hurdle, but kept pasting Unicode into
them. This can go well - the formula is lay-
outed in the formula font instead of the text
font - but this can also fail (when the charac-
ter is not defined in the formula font).
• Formula defined in ASCIIMath but LATEX is
too dumb to digest it: For those authors that
did embrace the formula syntax (technically:
ASCIIMath syntax, an easy way to enter even
the most complicated formulae), there were
two further hurdles: Either the formula was
not entered correctly or the the converter to
LATEX simply produced erroneous results; the
web editor would always produce a result, in
the sense of "I am happy with what you en-
ter; I will typeset what I understand and skip
the rest"; but LATEX would not work this way - it
would crash! So this went definitively wrong,
such things should never happen.
Summarizing, we should definitely parse the error
text given by LATEX and show the author what the er-
ror is, and where it lies in the paper. Another lesson
learned: Do not use the conversion from ASCIIMath
to LATEX, use the image produced by the web-based
formula viewer, at least for non-inline equations. In
that way, what you see would truly be what you get.
What did work, but most authors say it
didn't
References were a source of constant dispute be-
tween the editing team and the authors. As must
be said, we had lack of support in importing from
EndNote or BibTex, which is a shame fully taken. As
a result, the authors had to re-enter every reference
again, which was frustrating. However, the process
of having to review every reference again according
to eCAADe's needs has proven very effective in assur-
ing quality. In fact, it is clearly one of the conferences
where quality of references is of the highest stan-
dards, since we enforced semantic rules rather than
only taking "the data" that authors would provide
(for example, journal articles require a volume and a
page) which is far beyond what people would nor-
mally give us. Furthermore, it allowed authors having
different citation management systems to produce
one homogeneous reference section.
A BRIEF LOOK AT ABSTRACT SUBMISSION
ProceeDings was used for the editing of full papers,
while OpenConf has been employed for abstract up-
load (Word or PDF). Thus, strictly speaking, this part
should not appear at all in our paper, as it is some-
thing we are not concerned with. However, it might
nevertheless put some contrast on our later analysis
of the the final submission (see next section).
When looking at the number of submits to the
abstract submission site, we can note that these are
roughly four times of abstracts handed in. Further
analysing when the submits happened, we can see
that typical eCAADe submitter is:
• Well-behaved: The peak of the initial ab-
stract submissions is eight days before the ini-
Fusion - Data Integration at its best - Volume 1 - eCAADe 32 | 17
tial abstract submission deadline on 3rd Febru-
ary (there was an extended submission dead-
line on February 10th, which led people to
pause for two days before resuming a steady
stream of submissions).
• Occupied during the week: Most submis-
sions happen Mondays and Fridays between 12
und 16 hours UTC.
• Expecting an extended abstract submis-
sion: There are as much submission in the ex-
tended deadline period as before; this means
that people really assume that there will be an
extended abstract submission deadline, and
use this to correct their paper.
These three points are of course a very subjective in-
terpretation - the analysis of the final submission (see
next section) tries to put some more scrutiny in so as
to further narrow down what a conference organizer
can expect, on a statistical basis.
FULL PAPER SUBMISSION
The following is a statistical overview of the full pa-
pers submitted for eCAADe 2014 (sidenote: 164 pa-
per were initially accepted, 148 were present at the
time of the extended submission and 127 remained
in the final proceedings; either the authors failed to
complete their work or withdrew their paper after
submitting; the following data is a snapshot as per
18th June, two days after the extended deadline).
If ever there was a proof for a tendency to pro-
crastination in academia it is shown in figure 2. The
graph displays when authors started uploading their
papers (positive y-axis) and when they finally submit-
ted the paper (negative y-axis). The timeline along
the x-axis starts at the beginning of April, which is
when the first invitation to log-in to the system was
sent out, passes the initial deadline (D) on the 9th of
June and ends just after the extended deadline (ED)
on the 16th of June. Work on the papers did not start
to increase until about two weeks before the dead-
line, with a sizable portion not starting until after the
 Figure 2
Timeline showing
the dates the
authors started and
finished uploading
their data. The
initial deadline (D)
and the extended
deadline (ED) are
overlaid.

Figure 3
Number of revisions
and number of
items per paper.
Papers are sorted by
the order work on
them was started.
The moving
average of each
measure is overlaid.

18 | eCAADe 32 - Fusion - Data Integration at its best - Volume 1

 initial deadline had passed. It can also be seen that
the email sent out on the 5th of June announcing
the deadline extension caused a sudden drop in ac-
tivity. Procrastination can be seen even more clearly
with the paper submissions. Most waited until the ex-
tended deadline and some continued to work even
beyond the deadline.
However, the graph in figure 3 clearly shows that
the individual preferences in time planning have lit-
tle impact on the length of the paper. Here papers
are sorted by the date the authors started to upload
their data and displayed along the x-axis. The pos-
itive y-axis gives the number of revisions, the num-
ber of times the authors made a change like inserting,
changing or deleting data, on the negative y-axis the
number of items (paragraphs, headers, images, etc.)
is displayed as a measure of the overall length. Both
measures are overlaid by their moving average over
20 papers. While there is a clear tendency to fewer
revisions by papers started later, the overall number
of items remains fairly constant.
Figure 4
Distribution of
content items,
headers, graphical
items and
references (l.t.r.)
over all eCAADe
2014 papers.
Figure 4 shows the distribution of items over all pa-
pers. The items were grouped into content items
(paragraphs, lists, formulas and algorithms), headers,
graphical items (images and tables) and references.
Papers were then divided into 16 categories accord-
ing to their keywords and the item distribution was
calculated for each category (Figure 5).
CONCLUSION AND OUTLOOK
The implementation of ProceeDings in the frame-
work of eCAADe 2014 has delivered a treasure trove
of experience which will be used for further develop-
Fusion - Data Integration at its best - Volume 1 - eCAADe 32 | 19
ments. The tool was able to automate the production
of eCAADe proceedings, starting from an initial list of
accepted publications (coming from OpenConf ) and
ending with PDFs ready to be delivered to print pro-
duction.
During the whole process, the editors still have
an important role: As the technicalities of the paper
layout are dealt with, they can direct their attention
towards the content as such, i.e. focus on editing. In
that context, ProceeDings allows to monitor ongoing
developments within the community of submitting
authors at an early stage, i.e. before the (final) sub-
mission and eventually to deliver feedback. Till a rel-
atively late stage, authors can be involved in the pub-
lication process and the final outcome can be made
available at any time.
The number of opportunities to (unintention-
ally) overrule the template is shrinking and espe-
cially ongoing live experience will accelerate the
improvement of the interface. Most importantly,
the users can take care of the layout themselves:
You-Get-What-You-See (YGWYS) instead of You-Will-
Sometimes-See-What You-Will-Get (YWSSWYWG). In
this regard, the pre-publishing option (preprints)
might gain interest. The extraction of (coherent!)
metadata, required for indexing etc. has to be high-
lighted as well. It is unlikely that ProceeDings will
encompass the abstract/paper submission and re-
view stage. Here, a number of well-functioning (open
source) environments has been made available for a
longer period of time.
ACKNOWLEDGEMENTS
ProceeDings was programmed by Gabriel Wurzer
over the course three years. Bob Martens took the
tool into the eCAADe council, from which we got a
first "go" in getting it so far that it would be ready
for eCAADe 2015. Luckily, Emine Thompson gave us
the chance to test-drive the system even earlier, as
publication platform for the eCAADe 2014. We are
extremely greatful for this decision, since it involved
the risk that risk that some functionalities would still
be beta-quality. The paper importer proved to be

such a case, thanks goes to Marie Davidova and Mar-
tin Tamke for pointing out multiple issues. Likewise,
Thomas Grasl and Rudi Stouffs gave their input on
multiple features either missing or totally buggy, as
did a multitude of authors through our bug report-
ing system. Thanks goes as well to Gregor Hartweger,
Andrea Wölfer, Rudolf Scheuvens and Georg Penthor:
The first three made it possible that we got our own
server hosting the system (memory and CPU power:
sufficient), and the latter one made sure we could re-
locate our whole set-up into a new building when the
department was getting refurbished. From the ed-
itorial side, colleagues Wolfgang Lorenz and Gerda
Hartl helped tremendously in layouting the proceed-
ings. Speaking of layout, we also want to thank Henri
Achten for showing us how to do this with style, based
on his team's great work with the eCAADe 2012 pro-
ceedings. Our main volume of thanks goes to all au-
thors, who did most of the work: It is on their shoul-
ders that we all stand. Last but not least, here is to
20 | eCAADe 32 - Fusion - Data Integration at its best - Volume 1
Figure 5
Distribution of
content items,
headers, graphical
items and
references per
paper category.
you Martin Winchester, thank you for the OpenConf
support which you do every year, and for the chance
to pull an analysis of the abstract submission.
REFERENCES
Fredriksson, EH (eds) 2001, A Century of Science Publish-
ing: A Collection of Essays, IOS Press
King, P (eds) 2004, Digital Documents: Systems and Prin-
ciples: 8th International Conference on Digital Docu-
ments and Electronic Publishing, Springer, Berlin/Hei-
delberg
Martens, B and Turk, Z 1990, 'The Creation of a Cumula-
tive Index on CAD: "CUMINCAD"', ACADIA Quarterly,
19(3), p. 18–19
Russey, WE, Bliefert, C and Villain, C (eds) 1995, Text and
graphics in the electronic age : desktop publishing for
scientists, Wiley-VCH
[1] http://lynx.isc.org/
List of Reviewers
Sherif Abdelmohsen, Ain Shams University, Egypt
Henri Achten, Czech Technical University in
Prague, Czech Republic
Asterios Agkathidis University of Liverpool, United
Kingdom
Brian Agnew, Northumbria University, United
Kingdom
Seraphim Alvanides, Northumbria University,
United Kingdom
Aleksander Asanowicz, Bialystok University of
Technology, Poland
Claudio Benghi, Northumbria University, United
Kingdom
Andre Brown, University of Liverpool, United King-
dom
Gulen Cagdas, Istanbul Technical University Fac-
ulty of Architecture, Turkey
Gabriela Celani, University of Campinas, Brazil
Tomo Cerovsek, University of Ljubljana, Slovenia
James Charlton, Northumbria University, United
Kingdom
Julie Clarke, Northumbria University, United King-
dom
Birgul Colakoglu, Yildiz Technical University,
Turkey
Ruth Dalton, Northumbria University, United
Kingdom
Bauke de Vries, Eindhoven University of Technol-
ogy, Netherlands
Halime Demirkan, Bilkent University, Turkey
Dirk Donath, Bauhaus University Weimar, Ger-
many
Andy Dunhill, Northumbria University, United
Kingdom
Dietrich Elger, Koopx Architects Engineers, Ger-
many
Antonio Fioravanti, Sapienza University of Rome,
Italy
Thomas Fischer, Xi'an Jiaotong-Liverpool Univer-
sity, China
Pia Fricker, ETH Zurich, Switzerland
Harald Gatermann, HS Bochum University of Ap-
plied Sciences, Germany
Evelyn Gavrilou, University of Thessaly, Greece
Bob Giddings, Northumbria University, United
Kingdom
Thomas Grasl, SWAP Architects, Austria
Paul Greenhalgh, University of Northumbria,
United Kingdom
David Greenwood, Northumbria University,
United Kingdom
Christiane M. Herr, Xi'an Jiaotong-Liverpool Uni-
versity, China
Lawrence Hughes, Northumbria University,
United Kingdom
Mustafa Emre İlal, Izmir Institute of Technology,
Turkey
Umit Isikdag, Beykent University, Turkey
Angie Johnson, Northumbria University, United
Kingdom
Oliver Jones, Northumbria University, United
Kingdom
Anja Jutraž, University of Ljubljana, Faculty of Ar-
chitecture, Slovenia
Matevz Juvancic, University of Ljubljana, Faculty of
Architecture, Slovenia
Anetta Kepczynska-Walczak, Lodz University of
Technology Poland, Poland
Joachim Kieferle, RheinMain University, Germany
Erik Kjems, Aalborg University, Denmark
Michael Knight, University of Liverpool, United
Kingdom
Tuba Kocaturk, University of Liverpool, School of
Architecture, United Kingdom
Volker Koch, Karlsruhe Institute of Technology,
Germany
Branko Kolarevic, University of Calgary, Canada
Jose Kos, Federal University of Santa Catarina ,
Brazil
Alexander Koutamanis, Delft University of Tech-
nology, Netherlands
Andrew Li, Massachusetts Institute of Technology,
United States
List of Reviewers - Volume 1 - eCAADe 32 | 21
 Thorsten Loemker, ZAYED University Dubai, Johan Verbeke, KU Leuven, Belgium
United Arab Emirates Spela Verovsek, University Of Ljubljana, Slovenia
Alireza Maheri, Northumbria University, United Sara Walker, Northumbria University, United King-
Kingdom dom
Earl Mark, University of Virginia, United States Dawn Whitton, Northumbria University, United
Bob Martens, Vienna University of Technology, Kingdom
Austria Jerzy Wojtowicz, Warsaw University of Technol-
Jane Matthews, Curtin University, Australia ogy, Faculty of Architecture, Poland
Tom Maver, Glasgow Schol of Art, United Kingdom Stefan Wrona, Faculty of Architecture Warsaw Uni-
David Mcguinness, Northumbria University, versity of Technology, Poland
United Kingdom Tadeja Zupančič, University of Ljubljania, Faculty
AnnaLisa Meyboom, University of British of Architecture, Slovenia
Columbia, Canada
Volker Mueller, Bentley Systems, Incorporated,
United States
Michael Mullins, Aalborg University, Denmark
Marc Muylle, University of Antwerp, Belgium
Herman Neuckermans, KU Leuven –University,
Belgium
Rivka Oxman, Technion, Israel
Mine Ozkar, Istanbul Technical University, Turkey
Burak Pak, KU Leuven University, Belgium
Giuseppe Pellitteri, Dipartimento di Architettura -
Università di Palermo, Italy
Chengzhi Peng, University of Sheffield, United
Kingdom
Frank Petzold, TUM, Germany
Nuno Pinto, The University of Manchester, United
Kingdom
Ahmad Rafi, Multimedia University, Malaysia
Rabee M. Reffat, Assiut University, Egypt
Alan Richardson, Northumbria University, United
Kingdom
Peter Russell, RWTH Aachen University, Germany
Gerhard Schmitt, ETH Zürich, Switzerland
Davide Simeone, Sapienza University of Rome,
Italy
Martijn Stellingwerff, TU Delft, Netherlands
Rudi Stouffs, TU Delft, Netherlands
Kjeld Svidt, Aalborg University, Denmark
Emine Mine Thompson, Northumbria University,
United Kingdom
Bige Tuncer, Singapore University of Technology
and Design, Singapore
Jos van Leeuwen, The Hague University of Applied
Sciences, Netherlands

22 | eCAADe 32 - List of Reviewers - Volume 1

Keynote Speakers

Christian Derix
Christian Derix is director of the Computational Design Research group of Aedas Architects, which he
founded in 2004 in London, UK. CDR develops computational simulations for generative and analytical
design processes with an emphasis on spatial configurations and human occupation. Derix studied ar-
chitecture and computation in Italy and the UK and has researched and taught the subject at various
European universities since 2001. Currently he is associate professor at IE University Madrid and visiting
professor at the University of Sheffield. The work of CDR has recently won award commendations for
their Spatial Simulation framework at awards such as the 2010 Presidents Medal for Research in Practice
of the Royal Institute of British Architects (RIBA), the 2011 Italian Compasso d’Oro for algorithmic design
and user participation in industrial design for the VITA Shelving System or the Centre for Tall Buildings
and Urban Habitat’s (CTBUH) 2012 Innovation award for the computer-activated responsive façade of
the Al Bahar towers.

Andrew Hudson-Smith
Dr Andrew Hudson-Smith is Director of the Centre for Advanced Spatial Analysis (CASA) at The Bartlett,
University College London. Andy is a Reader in Digital Urban Systems and Editor-in-Chief of Future In-
ternet Journal, he is also an elected Fellow of the Royal Society of Arts, a member of the Greater London
Authority Smart London Board and Course Founder of the MRes in Advanced Spatial Analysis and Visu-
alisation at University College London . He is also course founder of the new MSc in Smart Cities and
Urban Analytics and the MRes in Smart Cities. His research focus on location based digital technologies
and he has been at the forefront of web 2.0 technologies for communication, outreach and developing
unique contributions. His research can be found at Digital Urban website. Finally, he is Editor of the
Open Journal of the Internet of Things.

Stephen Lockley
Professor Stephen Lockley is currently chair of Building Modelling at Northumbria University. From 1998-
2002 he was the director of the Construction Informatics Research Centre and Chair of Architectural In-
formatics at Newcastle University and has worked in industry as research and development director for
the RIBA (Royal Institute of British Architects) Enterprise division between 2002-2007. He has returned to
academia and his work now is in the area of the next generation of computer aided construction, Build-
ing Information Modelling (BIM), in particular researching and developing innovative software tools for
construction and design. He is a director of BuildingSmart UK and Ireland and sits on the British Stan-
dards Committee B555 for Construction Information. He is also research and education director of the
BIM Academy.

Keynote Speakers - Volume 1 - eCAADe 32 | 23

24 | eCAADe 32 - Keynote Speakers - Volume 1
Contents
5 Theme
7 Acknowledgements
11 ProceeDings
Gabriel Wurzer, Bob Martens, Thomas Grasl
21 List of Reviewers
23 Keynote Speakers

31 Towards Smarter Cities

33 CItyMaker Workshop
José Nuno Beirão, Letícia Mendes, Gabriela Celani
43 Decision support for inner-city development
Nils Seifert, Michael Mühlhaus, Gerhard Schubert, Dietrich Fink, Frank Petzold
53 Agents' movement_towards the reformation of public space
Maria Kerkidou, Anastasia Pechlivanidou-Liakata, Adam Doulgerakis, Alexandros
Sagias
63 A Mobile Sensing Kit for Urban Analysis
Nancy Diniz, Hai-Ning Liang
71 Urban Place and Networked Data
Trevor Hess, Eric Sauda
79 Multi-Perspective Urban Optioneering
Patrick Janssen, Rudi Stouffs
89 ICT-enabled Civic Empowerment and Participation: in Design, through
Design
Burak Pak, Johan Verbeke
99 The fusion of clever urban initiatives and digital applications
Matevz Juvancic, Tadeja Zupancic
109 Towards a dynamic evacuation system: developing methodologies to
simulate the evacuation capabilities of subway stations in response to a
terrorist attack with CBRNE weapons
James Charlton, Markus Brune
119 Simulation of Air flow, Smoke Dispersion and Evacuation of the Monument
Metro Station based on Subway Climatology
Zi Qian, Brian Agnew, Emine Mine Thompson
129 Building-up urban open spaces from shadow range analyses
Estefania Tapias, Shubham Soni

Contents - Volume 1 - eCAADe 32 | 25

 137 Aerodynamic strategy applied in an urban shelter design
Rafael Moya, Daniel Prohasky, Simon Watkins, Yan Ding, Jane Burry, Mark Burry
145 Computing the Urban Block
Elif Erdine, Evan Greenberg
153 Integrative Pedestrian Modelling Techniques based on Virtual Force Fields
Odysseas Kontovourkis, Despo Anagiotou
165 Wind sensing with real-time visualisations for Designers
Daniel Prohasky, Rafael Moya Castro, Simon Watkins, Jane Burry, Mark Burry
173 Shared Space Navigation
Charles Avis
181 Emergent articulation field in existing urban context
Peter Búš, Lukáš Kurilla, Henri Achten
191 Urban body network configurations
Eirini Androutsopoulou
201 A Digital Tool for Customized Mass Housing Design
Ahmet Emre Dincer, Gülen Çağdaş, Hakan Tong
213 Interrogating the Relation between E-Mobility Recharging Network Design
and Drivers' Charging Behaviour
Eiman Y. ElBanhawy, Ruth Dalton, Emine Mine Thompson

225 Design Tool 1

227 A Model for Sustainable Site Layout Design with Pareto Genetic Algorithm:
SSPM
Yazgı Aksoy, Gülen Cağdaş
239 The generation of possible space layouts
Jesper Thøger Christensen
247 Design ReExplorer: Interactive Design Narratives for Feedback, Analysis and
Exploration
Rodolfo F. Sánchez, Halil I. Erhan
257 On the Expressive Power of Programming Languages for Generative Design
António Leitão, Sara Proença
267 Thermal performance of patterned facades
Mehrnoush Latifi Khorasani, Jane Burry, Mahsa Salehi
277 Fuzzy computing for layout design in ill-defined, uncertain spaces
Asli Cekmis

26 | eCAADe 32 - Contents - Volume 1

287 A 3-Dimensional Architectural Layout Generation Procedure for Optimization
Applications : DC-RVD
Ioannis Chatzikonstantinou
297 Cellular Automata as a learning process in Architecture and Urban design
Mads Brath Jensen, Isak Worre Foged

303 CAAD Education

305 New Approaches in Architecture Education
Birgul Colakoglu
313 Digital Curation for CAAD Curricula
Ruggero Lancia, Ian Anderson
323 Integrating Sustainability in the Architectural Design Education Process
Ahmed Sarhan, Peter Rutherford
333 Digital Recipes
Miguel Paredes Maldonado
343 Augmented Reality in architectural studio learning
David Morton
357 Teaching and Designing for Augmented Reality
Andrzej Zarzycki
365 Framing Parametric and Generative Structures
Ivo Vrouwe, Burak Pak
373 Reprogramming Architecture
Elif Erdine, Alexandros Kallegias
381 Learning to be a Vault
David Stasiuk, Mette Ramsgaard Thomsen
391 Premise
Wendy W Fok

399 Fabrication
401 Design By Making
Serdar Aşut
411 Physical input-driven offline robotic simulation through a feedback loop
process
Odysseas Kontovourkis, George Tryfonos

Contents - Volume 1 - eCAADe 32 | 27

 423 Bio-inspired and fabrication-informed design strategies for modular fibrous
structures in architecture
Leyla Yunis, Ondřej Kyjánek, Moritz Dörstelmann, Marshall Prado, Tobias Schwinn,
Achim Menges
433 Isoprototyping
Pavlos Fereos, Marios Tsiliakos
445 Design to fabrication integration and material craftsmanship
Sina Mostafavi, Matthew Tanti
455 From rapid prototyping to automated manufacturing
Aant van der Zee, Bauke de Vries, Theo Salet
463 Material Swarm Articulations
Evangelos Pantazis, David Gerber
475 Digital Fabrication Technology in Concrete Architecture
Pedro Filipe Martins, José Pedro Sousa

485 Shape, Form and Geometry 1

487 Acoustically Efficient Origami Based Partitions for Open Plan Spaces
Sophia Vyzoviti, Nicolas Remy
495 Measurability of Loos' rejection of the ornament
Wolfgang E. Lorenz
505 Unsupervised Symmetric Polygon Mesh Mapping
Matthias Standfest
515 Constructing a shape grammar. The Ducal palace façade
Filipe Coutinho, José Pinto Duarte , Mario Kruger
527 Local-reconfigurable Freeform surface with plywood
Koki Akiyoshi, Hiroya Tanaka
537 Populating surfaces with holes using particle repulsion based on scalar fields
Timo Harboe Nielsen, Stephen Melville, Iain Sproat
547 Efficiency in Architectural Geometry Informed by Materials
Sevil Yazici

555 Visualisation
557 Digital Etching
Johan Voordouw
565 Data Integration In A Visual Mode
Anetta Kepczynska-Walczak

28 | eCAADe 32 - Contents - Volume 1

573 Linkography for evaluating ideas connectivity of Computer Aided
Design-based protocols
Huda Salman

583 Digital Heritage 1

585 Interactive Tabletops for Architectural Visualization
Bruno Figueiredo, Eduardo Castro e Costa, Bruno Araújo, Fernando Fonseca,
Joaquim A Jorge, José Pinto Duarte
593 Building Heritage Knowledge Framework using Context Ontologies
Silvia Gargaro, Elisabetta Leggieri, Antonio Fioravanti
603 Gamification of Shape Grammars
Serdar Aydin, Tian Tian Lo, Marc Aurel Schnabel
613 B(H)IM - Built Heritage Information Modelling
Davide Simeone, Stefano Cursi, Ilaria Toldo, Gianfranco Carrara
623 Digital fabrication as a tool for investigating traditional Chinese architecture
Di Li, Michael Knight, Andre Brown
633 Health and Safety Design by means of a Systemic Approach
Armando Trento, Antonio Fioravanti , Francesco Rossini
643 Digitally-Assisted Stone Carving on Canada's Parliament Hill
James Hayes, Stephen Fai, Phil White
653 Investigating a narrative architecture
Danilo Di Mascio, Tom Maver
665 Index of Authors

Contents - Volume 1 - eCAADe 32 | 29

30 | eCAADe 32 - Contents - Volume 1
Towards Smarter Cities
CItyMaker Workshop
An Urban Design Studio to experience the dynamic interaction between
design exploration and data flow on density-based indicators

José Nuno Beirão1 , Letícia Mendes2 , Gabriela Celani3

1
Faculty of Architecture, University of Lisbon 2,3 University of Campinas, Unicamp
1
jnb@fa.ulisboa.pt 2,3 {leticia|celani}@fec.unicamp.br

This paper shows the results of a workshop on parametric urban design. A

pattern based parametric urban design platform provides a common platform for
urban design and analysis by linking GIS with a parametric CAD environment.
Urban plans are developed by combining design patterns taken from a large set
of parametric urban design patterns acknowledged as a generic urban design
language. Urban plan instances are obtained through a specific composition of
patterns and a specific assignment of parameters to the patterns. The models
provide simultaneous analysis by confronting formal solutions with density
indicators that are automatically provided by a set of density calculation patterns.

Keywords: parametric urban design, CItyMaker, urban design workshop, design

methodology

INTRODUCTION must be taken into account, such as street, side-

The present paper is part of a research project that
uses parametric and algorithmic design methods for
the generation of public spaces in housing devel-
opments assuming that these methods contribute
to a wider and more in-depth exploration of solu-
tion spaces. The possibility of manipulating param-
eters for generating different alternatives on-the-fly
has had a great impact on design methods since
the early 2000's. Many discussions have been estab-
lished about the extent to which parametric design
can change the way we design, and about the rela-
tion between parametric design and other types of
computer-supported design methods, such as algo-
rithmic and generative design.
Parametric modelling is a powerful method for
generating urban plans, in which many variables
walk and block widths (Canuto and Amorim, 2012).
However, there is a limit to which extent parametric
modelling can generate diverse results without be-
ing combined with parametric design rules (Celani,
Duarte, Beirão and Vaz, 2011). The approach pre-
sented in this paper combines standard codes re-
currently used in the urban design practice or de-
sign patterns (Gamma et al, 1995), (Woodbury, 2010)
with parametric design to generate urban design so-
lutions. The combinatorial structure can be said to be
rule based due to the relational structure defined be-
tween the objects computed by patterns. In addition
to the urban design patterns a second set of patterns
provides calculations on density properties of the de-
signs.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 33

 In order to test the application of paramet-
ric modelling, a workshop was held at the Univer-
sity of Campinas in Brazil for the development of
a new neighbourhood located in the outskirts of a
metropolitan area, in Campinas. The set of tools
used in the workshop, here called CItyMaker (Beirão,
2012), consists of the use of geographic information
flow within a parametric design platform to support
decision-making in urban design.
OBJECTIVES
The goal of this paper is to present a methodology for
approaching the urban design process using para-
metric urban patterns to deliver a system of alterna-
tive solutions instead of a single, definitive solution.
This methodology was applied in a one month work-
shop in the University of Campinas. The experiment
consisted of having students develop an urban plan
for a new residential neighbourhood using a para-
metric design system and supporting decisions on
the calculation of density indicators. The area is lo-
cated in the outskirts of Campinas' metropolitan area,
in an empty space among existing low-income resi-
dential and industrial areas. Results were then anal-
ysed with the goal of assessing and refining the tools
and design methodology.
34 | eCAADe 32 - Towards Smarter Cities - Volume 1
DESIGN METHODOLOGY
A similar teaching methodology including paramet-
ric and algorithmic tools had been previously used in
two design studios (Duarte and Beirão, 2011) and two
workshops (Measurb - Faculty of Architecture, Tech-
nical University of Lisbon [1] and Tarlabaşi Datascope
- Faculty of Architecture, Istanbul Technical Univer-
sity [2]. In all events the fusion between rule-based
design and parametric design was addressed as a
methodological approach to the urban design prob-
lems given as design tasks to the participants. We ar-
gue that the use of parametric modelling alone is not
as effective as their combination with rules. In the
two workshops, a set of tools was given to the partic-
ipants to solve their design problems. The tools link
information provided by a geographical database
with a CAD parametric design environment. In de-
tail (see Figure 1), the proposed working environ-
ment integrates a geographic database (DB) with a
CAD software (in this case, Rhinoceros(C)) and a para-
metric Visual Programing Interface (VPI - Grasshop-
per(C)) including a set of predefined patterns previ-
ously programmed to generate typical urban design
actions (like grids, arterial streets and main squares)
as well as patterns designed to calculate density in-
dicators helping designers choosing their options
within the design space provided by the paramet-
ric tools. Such design system integrates analytical
and synthesis tools in the same working environment Figure 1
allowing designers to react on real time properties The structure of a
measured from the design models. For this reason, City Information
the design system has been called CItyMaker. The Modelling platform
name includes the acronym CIM which stands for City
Information Modelling, a term that congregates the
fusion between information flow and design explo-
ration. Information flow here means the calculation
of properties of the design in the context during the
design exploration process as well as real time as-
sessment of GIS data and generated data. CIM sys-
tems not only integrate analysis and synthesis, but
they also put together rule based design and para-
metric design. The rule based behaviour is accom-
plished by combining patterns or choosing between
available options within the patterns; the parametric
behaviour is intrinsic to the patterns defined in the
VPI where the input parameters are variables within
predefined ranges.
In the above mentioned workshops the students
were introduced to the several available design and
analysis tools. They had three days of thematic lec-
tures divided in two parts: (1) a theoretical lecture
introducing concepts to be applied and explored in
the practical lecture and (2) a practical lecture where
the tools were introduced and students solved small
applied problems using those tools. The two work-
shops had slight differences with respect to their the-
matic contents. The Lisbon workshop involved three
main themes: context analysis and program formu-
lation; design generation involving goals defined in
terms of density indicators; and evaluation. The Is-
tanbul workshop was more intensively focused on
analysis and included also three main themes: spatial
analysis; data mining; and integrated design genera-
tion. The two workshops substantiated the basis of
the Campinas workshop shown in this paper.
The workshop held at the University of Camp-
inas was attended by graduate and undergraduate
architecture students and involved an extended mix
of the previously mentioned workshops. The design
brief proposed the development of a new neighbour-
hood in the outskirts of a metropolitan area local-
ized in Campinas, Brazil. The students were asked to
examine the relationship between urban design ide-
als, urban design action, and the built environment
through readings, discussions and a final presenta-
tion of the proposed neighbourhood plan.
The main objective of the CItyMaker tool was to
integrate the flow of information related to design
configuration during the design process in order to
support decision-making in urban design. This tool
explores three aspects: (1) the integration of a plat-
form for parametric design with a geographic infor-
mation system; (2) a set of "design patterns" for ur-
ban design, generating small individual design ac-
tions, from which compound arrangements of those
"design patterns" generate urban plans; (3) interac-
tive calculation of properties and indicators associ-
ated with the urban design exploratory model.
The plans designed following the described
method produced interactive parametric design
models from which the calculation of precise urban
indicators is continuously updated at each change in
inputs, either numerical inputs or geometric config-
urational inputs.
The generative urban design system allows for
dynamic interaction between design exploration and
data flow on density-based indicators in order to sup-
port design decisions and henceforth improve de-
sign quality. The system also aims at responding to
input needs and provide dynamic information on the
measurements and properties of the solutions being
explored. The main idea is that the quality of design
decisions improves with the quality of the informa-
tion flow (Beirão, 2012).
The lectures in Campinas dealt with the follow-
ing themes: (1) shape grammars and rule based de-
sign; (2) parametric urban design; (3) 'spacematrix'
and density studies on the urban form; and (4) stud-
ies on spatial configuration by assessing the topolog-
ical relations of the elements of urban space includ-
ing an introductory lecture on 'space syntax' (Hillier
and Hanson, 1984). Additional support was given
in studio format following a thematic discussion on
finding the good city form (Lynch, 1984), (Duany and
Plater-Zyberk, 2005) with a specific emphasis on the
particular problems of the Brazilian city (Angélil and
Hehl, 2013).
CASE STUDY
Intervention area
The intervention area proposed for the development
of the urban design exercise in the Campinas work-
shop included irregular settlements, low and middle-
income housing developments and industries. The
area is located more or less halfway between the air-
port and the city centre. Close to the intervention
area there is a large sector of informal housing lo-
cated at the junction of two major state highways:
Santos Dumont and Anhanguera. The formation pro-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 35
 cess of this settlement began in the late 1990s, on the
grounds of a large empty area, referring to the history
of the development of low-income housing in Brazil-
ian cities (Ghilardi, 2012). Thus, the analysis and in-
tervention proposal in this workshop first proposes a
reflection on issues that relate to the theme of urban
form, the production of slums and the consequent
production of urban informality in Brazil.
The design brief included the development of an
urban design for this region considering the follow-
ing requirements: (1) housing for 5-10 thousand peo-
ple distributed by three predefined sets according to
their monthly income; (2) structures to facilitate and
encourage local trade; (3) facilities (schools, health
clinics, day care centre, parks, sports centre); (4) struc-
tures for production and service activities (possibly
related to airport activities, such as catering, laundry,
etc.); (5) infrastructure for transport connection to the
city and the airport.
Workshop
The workshop held at the School of Civil Engineering,
Architecture and Urban Design in the University of
Campinas was attended by sixteen graduate and un-
dergraduate architecture students during the month
of August 2013.
During the workshop the students had lectures
on concepts of urban design and analysed the di-
verse design ideals that influence cities and settle-
ments, and investigated how urban designers use
them to shape urban form. Students used a set of de-
sign tools defined as CItyMaker (Beirão, 2012). These
tools include a set of parametric design patterns pro-
grammed to design typical urban design operations
like streets, grids and public spaces, and a second
set of tools used to measure density properties of ur-
ban spaces following the conventions developed by
Berghauser-Pont and Haupt (2010). Therefore, the
tools comprise design and analytical tools that can be
used interactively within the same working platform.
In detail, the design patterns given to the stu-
dents included:
36 | eCAADe 32 - Towards Smarter Cities - Volume 1
Urban Design Patterns. Urban design patterns are
codes designed to replicate typical urban design ac-
tions. Different arrangements of these patterns pro-
duce different urban designs.
1. Street design pattern: This design pattern re-
ceives as input a set of polylines designed
in Rhinoceros or Grasshopper, a polygon as
intervention area and a value (variable) for
street width. It outputs the resulting grid (the
set of polygons resulting from the subtraction
of street surfaces from the intervention area);
the set of street surfaces; the set of street cen-
tre lines; and the total street length.
2. Rectangular grid pattern: The rectangular
grid pattern receives as input a polygon as in-
tervention area; a point (which sets the rota-
tion centre of the grid); and four input vari-
ables: grid rotation angle in degrees; street
width; block size in U direction; and block
size in V direction. Units are in meters expect
for the rotation angle. The pattern outputs
the grid (the set of blocks or islands found
within the intervention area); the street cen-
tre lines; and the total street length. The lat-
ter is calculated following the theory defined
in Spacematrix (Berghauser-Pont and Haupt,
2010), i.e., the length of the street defined as
boundary is considered only as half the length
because its length is shared with the neigh-
bouring areas.
3. Filter 1 (filtering small areas): Receives a grid (a
set of polygons representing islands in a grid)
and filters polygons with an area smaller than
an input value. Outputs the resulting grid and
its inverse - the filtered islands.
4. Filter 2 (polygon selection filter): This pattern
allows the selection of areas (polygons) within
a grid. The selection process separates the
polygons in a grid in two separate information
flows, one the set of selected polygons and
 the other its inverse, that is, the resulting grid
polygons. The inputs for selection are points
either defined automatically or manually. The
algorithm filters islands containing a point.
5. Maximum Building Height Distribution Pat-
tern: Maximum building height is commonly
set as a regulatory device for urban planning.
Typically this device is used by setting a max-
imum height or maximum number of floors
in an area. Usually such areas correspond to
reasonably large areas (district size) for which
uniform regulations are set. The present pat-
tern allows the designer to extend the regula-
tion diversity automatically to island level by
distributing a different constraint on the max-
imum number of floors per island. Planning
regulations can therefore be different per is-
land and also easily controlled parametrically.
The distribution of building height is defined
by setting locations as attractor points and
then manipulating the constraint distribution
according to a distribution function. This pat-
tern processes four inputs: the set of polygons
representing islands; and three variables: the
maximum number of floors; average floor
height; and average ground floor height.
6. Open Space / Coverage Distribution Pat-
tern: Following the Spacematrix theory Open
Space Ratio at island level (OSRi ) gives an in-
dicator on the amount of private open space
of the plan as it is giving information based
on the private open space found within an is-
land. OSR calculated at district level (OSRd )
is based on the total amount of open space
found in the district, private and public open
space. Therefore, the public open space ra-
tio (OSR) is given by the equation ∆OSR =
OSRd − OSRi (page 114). The open space
distribution pattern distributes private open
space differently per island. Similarly to the
previous pattern, this pattern considers a dis-
tribution function according to the position
of a set of attractor points. The pattern pro-
cesses three inputs: the set of polygons rep-
resenting islands; and two values varying be-
tween 0 and 1 used to set the minimum and
maximum accepted coverage (GSI) variation
range. The distribution pattern distributes
along the grid different coverage and pri-
vate open space within the GSI variation
range and according to the attraction effect
set through the distribution function.
Data Import Patterns. Data import patterns import
data from GIS databases. In the Campinas workshop
these patterns were not used because the GIS data
was incomplete and the needed data was already
available in the geometric model.
Analysis Patterns. The analysis patterns used in this
workshop were essentially density analysis patterns.
Other kinds of analysis were programmed directly in
Grasshopper by the students.
7. Spacematrix indicators patterns: Floor space
index (F SI) calculation; Ground space index
(GSI) calculation; Open space ratio (OSR)
calculation; Average number of floors (L) cal-
culation; and Network density (N ) calculation
patterns. The patterns are identical for every
level of aggregation, but this means that users
need to be aware that the input geometries
have to be consistent within the level of ag-
gregation being calculated. All the patterns
receive the base geometries as defined in the
spacematrix theory (pages 107-114).
8. Spacematrix indicator converter pattern: This
pattern was developed by Pirouz Nourian; it
receives 2 spacematrix indicators and outputs
the 2 indicators missing. The indicators are
Floor space index (F SI); Ground space index
(GSI); Open space ratio (OSR); and Average
number of floors (L).
The practical part of the workshop started with an
analysis of the existing urban tissues by measuring
Towards Smarter Cities - Volume 1 - eCAADe 32 | 37
 their density indicators. Students started by identi-
fying homogeneous areas and neighbourhoods for
which they calculated the above mentioned indica-
tors and produced maps for visualization of the sev-
eral topics of analysis (Figure 2).
From the several on-going analyses a discussion re-
garding the pros and cons of certain morphological
types was raised. Specifically, the issue of density
was discussed using parametric models to compare
the tower-based block with the traditional periph-
eral block breaking the myth that the tower-based
block would necessarily respond to higher density re-
quirements. Having acknowledged that such myth
was false the students engaged in a more critical atti-
tude towards the standard speculative construction
in Brazil and explored other design concepts in their
designs.
RESULTS
The results were evaluated in terms of the quality
of the designs in relation to the outcomes of simi-
lar projects developed by students with traditional
methods and the participants' impressions about the
workshop. The designs developed in this workshop
show a greater variety among the different teams
38 | eCAADe 32 - Towards Smarter Cities - Volume 1
than in regular workshops (Figure 3). They also
present a big variety of urban morphotypes within
a single project, i.e., students did not restrict the so-
lutions to a single grid for the entire area under de-
velopment but rather explored many variations avail-
able by the tools before freezing a solution. Regard-
Figure 2
ing students' opinions about the method, they have
Density analysis
deliberately acknowledged how the CItyMaker envi-
using the CItyMaker
ronment has helped them in generating a number of
analysis tools
different spatial configurations and testing different
parameters, thus generating a more diverse plan. A
local professor of urban design who was not familiar
with parametric design techniques was invited to as-
sess the results. She was surprised with the variety of
the results and the quality of the plans produced in
such a short period of time.
During the workshop students were allured to
use the tools to perform several urban analyses and
to measure density indicators as a means to support
design decision. A discussion questioning the mor-
phological solutions commonly developed by Brazil-
ian housing promoters was raised, in particular ques-
tioning the recurrent use of the isolated skyscraper
in gated communities. Several preconceptions re-
garding the formal expression of density were bro-
ken and reconsidered allowing the students to more
freely explore unconventional solution spaces. The
CItyMaker tools and spacematrix theory proved quite
useful to support this process especially because the
tools provide measurements with simultaneous vi-
sualization while still allowing for further formal ex-
ploration. At the final presentation, the set of den-
sity measures calculated from the model was manda-
tory. Students had to show the proposed indicators
that should rule their plans and also support the rea-
son for such indicators based on the developed anal-
yses. This task showed two contrasting behaviours.
On one hand students could easily produce measure-
ments taken from their designs with a readiness they
had never experienced before. Measuring proved to
be easy. On the other hand, measuring required a
lot of attention both methodologically while defining
the design model and interpretatively while consid-
 ering the meaning and impacts of the obtained mea-
sures as a way for supporting design decision. This
situation raised remarks regarding the necessary at-
tention to the parametric model development and
correct theoretical knowledge. Firstly, regarding the
development of the parametric model students were
alerted for the need of having a clear notion about
the concept of 'level of aggregation' in order to ob-
jectively define the indicators calculations at the re-
quired levels. Secondly, the meaning of density in-
dicators require detailed interpretation developed at
the light of the intervention context. The latter in-
volves a rather complex theoretical approach sup-
porting how density indicators relate, first to urban
form and moreover to urban qualities. Here, the ana-
lytical methods proved useful as they provided a bet-
ter insight on the behaviour of urban environments.
Although the workshop results were acknowledged
by all participants as positive and highly formative, it
is still possible to abstract from students comments
and presentations that better results could only have
been possible with more intensive lectures on urban

Figure 3
Two examples of
outputs produced
by the students.
Volumes indicate
constraints on
building height
defined at island
level. These
constraints were
generated
automatically using
attraction points
that distribute
height constraints
according to an
assigned weight.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 39

 design theory, especially on the identification and
meaning of quality indicators and in particular within
the context of Brazilian urban environments.
DISCUSSION
Since the Design Methods Movement - period of
discussions, conferences and publications on design
methods in architecture, engineering and industrial
design in the 1960s - two approaches have been
employed in order to optimize the design process
and provide greater flexibility for architectural de-
sign: parametric design and rule-based design. The
first strategy consists of defining topological rela-
tions between parts of a composition, while the defi-
nition of the precise measures is a subsequent phase
of the design process. In general these dimensions
can be selected from a desirable range, with mini-
mum and maximum values established beforehand,
usually taking into consideration values that are mul-
tiple of building components dimensions, resulting
in greater flexibility and variety (Monedero, 2000). A
strategy based on rules (rule-based design) (Broad-
bent, 1970) allows defining situations in which a par-
ticular element can be connected to another, and
how this may occur. This method allows a greater
variety of designs, since the application of the rules
in different orders can result in completely different
compositions.
In the exercises developed in this workshop this
could be easily noticed, for example, when some the
students subdivided the intervention area initially ac-
cording to a specific pattern, and then applied a dif-
ferent pattern within each generated cell. The sys-
tem allowed to easily make changes to each embed-
ded pattern in terms of type and/or parameters, al-
lowing for the evaluation of a virtually infinite num-
ber of possibilities.
Although the work platform is essentially para-
metric, the design patterns aggregation define the
topological structure of the model that is produced
in the design process, constituting a limited set of
rules that are interpreted by the corresponding para-
metric code. Therefore, the design contains an al-
40 | eCAADe 32 - Towards Smarter Cities - Volume 1
gorithmic and a parametric dimension that can be
represented more abstractly through a shape gram-
mar. In other words, the set of patterns available,
here called CItyMaker, is a generic grammar for ur-
ban design where each pattern encodes a typical ur-
ban design operation by means of a shape grammar
and is implemented in a parametric design platform.
Each instantiated urban plan can be seen as a spe-
cific grammar translated by the specific combination
of patterns with a specific assignment of parameters
(Beirão, 2014).
The workshop showed (1) how easily some stu-
dents unfamiliar with programming get enrolled in
the exploration of design solution spaces via para-
metric visual programming, (2) the possibility of con-
joining analytical models with design exploration
models which provide students with (3) the ability
of justifying their design solutions with supporting
analysis directly derived from the geometric models.
As stated in the beginning of this paper, the
workshop herein described was also part of a larger
research, proposing the use of parametric and rule-
based strategies for the development of public hous-
ing in Brazil. One of the plans developed in this
workshop was chosen as a basis for the next work-
shop, in which students used shape grammar to de-
sign a housing complex in which issues of popula-
tion growth, evolving plans and function mix were
addressed. We considered that it would not make
sense to develop this type of exercise on a traditional,
square-grid urban area.
Beyond the contribution to the field of rule-
based parametric design, this research was an oppor-
tunity to show how the fusion of different types of
knowledge can lead to new insights. In this case we
were able to fuse (1) research and teaching, (2) un-
dergraduate and graduate students, (3) Latin Amer-
ican and European cultures, (4) analysis and design,
(5) the use of parameters and rules combined in the
design process, (6) researchers specialized both in ar-
chitectural and urban design, and (7) the research de-
veloped by two PhD researchers.
ACKNOWLEDGEMENTS
We thank FAPESP (São Paulo Research Foundation)
for Leticia Mendes' PhD scholarship (process nº
2011/50139-0), and for supporting the Laboratory
for Automation and Prototyping for Architecture and
Construction (process nº 2012/10498-3), and José
Beirão's stay at Unicamp (process nº 2013/09362-2).
José Beirão would like to thank the collaboration of
Cristina Cavaco and João Cabral in the Measuring Ur-
banity event in Lisbon, as well as Ahu Sökmenoğlu
and Pirouz Nourian for the collaboration in the Tar-
labaşi Datascope Workshop. José Beirão also wants
to acknowledge the special collaboration that Pirouz
Nourian has shared during the past two years. Finally,
a special thanks to Gabriela Celani for her invitation
to the Campinas Workshop.
Ghilardi, FH 2012, O lugar dos pobres na cidade de
Campinas-SP: questões a partir da urbanização da
ocupação do Parque Oziel, Jardim Monte Cristo e
Gleba B, Ph.D. Thesis, University of São Paulo. Insti-
tute of Architecture and Urban Planning
Hillier, B and Hanson, J 1984, The Social Logic of Space,
Cambridge University Press, Cambridge
Lynch, K 1984, Good City Form, MIT press
Monedero, J 2000, 'Parametric design: a review and
some experiences', Automation in Construction, 9(4),
pp. 369-377
Woodbury, R 2010, Elements of Parametric Design, Rout-
ledge
[1] http://www.measurb.org/en/home.html
[2] http://tarlabasidatascope.wordpress.com/

REFERENCES
Angélil, M and Hehl, R 2013, Cidade De Deus – City
of God.Working with Informalized Mass Housing in
Brazil, Ruby Press
Beirão, JN 2012, CItyMaker: Designing Grammars for Ur-
ban Design, Ph.D. Thesis, Delft University of Technol-
ogy
Beirão, JN 2014, 'Gramáticas genéricas para o domínio da
cidade e urbanismo', Revista de Morfologia Urbana,
2(1), pp. 44-46
Berghauser-Pont, M and Haupt, P 2010, Spacematrix.
Space, Density and Urban Form, NAI
Broadbent, G 1970, Design in architecture: architecture
and the human sciences, John Willey & Sons, London
Canuto, R and Amorim, L 2012 'Establishing Parameters
for Urbanity', Proceedings: Eighth International Space
Syntax Symposium, Santiago de Chile
Celani, G, Beirão, JN and Duarte, J 2011 'Optimizing the
characteristic structure Combining shape grammars
and genetic algorithms to generate urban patterns',
Proceedings of eCAADe 2011, Ljubljana, pp. 491-500
Duany, A and Plater-Zyberk, E 2005, Smart Code (version
9.2), http://www.smartcodecentral.org/
Duarte, J and Beirão, jn 2011, 'Towards a methodology
for flexible urban design: designing with urban pat-
terns and shape grammars', Environment and Plan-
ning B: planning and design, 38(5), pp. 879-902
Gamma, E, Helm, R, Johnson, R and Vlissides, J 1995,
Design Patterns: Elements of Reusable Object-oriented
Software, Addison-wesley Reading, MA

Towards Smarter Cities - Volume 1 - eCAADe 32 | 41

42 | eCAADe 32 - Towards Smarter Cities - Volume 1
Decision support for inner-city development
An interactive customizable environment for decision-making processes in
urban planning.

Nils Seifert1 , Michael Mühlhaus2 , Gerhard Schubert3 , Dietrich Fink4 , Frank Petzold5
1,2,3,4,5
Technische Universität München
1
www.lib.ar.tum.de 2 www.ai.ar.tum.de
1,2
{nils.seifert|michael.muehlhaus}@tum.de

Re-densification of inner-city areas is a highly topical socio-political issue. In

order to meet the increasing demand for centrally located living space, the cities
have to amend building laws. At present, there are no analytical means of directly
comparing and assessing different strategies and measures aimed at identifying
the potential for internal development in specific urban quarters and learning
about the consequences for the citizens. In this publication, we describe an
interactive digital tool that monitors the key building codes and visualizes their
effects on the urban structure in real-time, so that this can serve as an informed
basis for debate and argumentation in the political decision-making and planning
process, consequently supporting the development of re-densification strategies
that are well-suited to their urban context.

Keywords: urban planning, redensification, decision support, visual

programming, versioning

INTRODUCTION trolled growth, city-wide planning strategies - sup-

Urbanization has increased dramatically since the
second half of the 20th century (Heilig, 2012). This
is due to several factors - mainly because of demo-
graphic change and higher energy prices, but also
because of tax procedures (e.g. "Gesetz zur Abschaf-
fung der Eigenheimzulage" [1]). With it the demand
for centrally located living space continues to rise
strongly, especially in larger cities (Adam and Sturm
2011). Many cities are already exploiting their current
building laws to the full and are now under pressure
to amend building legislation to meet the increasing
demand for affordable living space. To avoid uncon-
ported by a broad population - are essential (Fink et
al. 2011). But what effect will specific changes to leg-
islation have both on the target density and the ap-
pearance of the built environment? To answer this, it
is necessary to develop case-by-case strategies that
consider the existing building stock and its potential
for infill development.
In the political discussion that precedes any
amendments to building legislation, both the bene-
fits and consequences of different measures can be
difficult to assess. Public authorities have different
means to control the building development in cer-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 43

 tain areas. Examples are regulations of minimum
distances between buildings, clearance areas, limita-
tions of building heights and amount of levels, but
also in determining the zoning plan or development
plan. At this point the question arises, what effect
certain changes to the laws have, both on the target
density and the appearance of the built environment.
The redensification of inner city areas is a particu-
larly sensitive subject because such decisions have
a direct impact on the habitat of the citizens. For
a democratic decision-making process, it is essential
that the results and outcomes of alternative strate-
gies are communicated and made available for pub-
lic debate (Fink and Fischer 2012). At present, how-
ever, there are no analytical means for directly com-
paring and assessing different strategies and mea-
sures, identifying the potential for internal develop-
ment in specific urban quarters and learning about
the consequences for the citizens. This would pro-
vide the necessary basis for an informed debate.
REQUIREMENTS OF THE TOOL
In this publication, we describe an interactive digital
tool that monitors the key building codes and visual-
izes their effects on the urban structure in real-time
so that this can serve as an informed basis for debate
and argumentation in the political decision-making
and planning process and in turn support the de-
velopment of redensification strategies that are well-
suited to their urban context. As a planning sup-
port system (PSS), it enables planners to check vari-
ous strategies and their execution variants, to com-
pare several alternative approaches and to use their
newly acquired knowledge to further substantiate
their decisions (Klostermann 1997). In addition in-
dividual profiles and evaluation keys can be created
and then applied to various planning areas. The di-
rect visualization of results provides a quick overview
of the qualities of different scenarios to citizens and
authorities alike and provides information on the fea-
sibility of the scheduled targets.
On the one hand, the tool serves as a basis for
argumentation in the political decision-making and
44 | eCAADe 32 - Towards Smarter Cities - Volume 1
planning process, for example to verify the appropri-
ateness of proposed changes to legislation vis-à-vis
the existing building codes. In this context, the soft-
ware assists in determining the effects of changes to
building legislation. On the other hand, the software
is able to support decision-making in the develop-
ment of strategies for particular urban planning ar-
eas. An important aspect in this regard is to deter-
mine whether project design proposals are compati-
ble with the applicable building codes and which ad-
vantages and disadvantages they may have over one
another. The software makes it possible to consider
planning strategies by automating key computation
steps while continuously monitoring their fulfillment
of the relevant building codes.
To meet these demands, basic requirements of
the tool have been defined:
• Monitoring of building codes: the tool should
take into account the key building codes and
be able to visualize their effects on the urban
structure.
• Comparability: there should be a way to com-
pare different planning strategies and dif-
ferent states of work (Kwartler and Bernard
2001).
• Versioning: different approaches and the
correlations and relationships between them
should be plainly illustrated. There must be a
way to reasonably archive stored data.
• Automatic analysis results: the key parame-
ters and indicators - e.g. for urban density
- should be calculated automatically, making
it possible to directly compare different ap-
proaches (Gratzl et al. 2013). Results of plan-
ning measures should be visualized in real-
time.
• Usability: the way of human information pro-
cessing in general and particularly in plan-
ning and decision-making should be consid-
ered (Elgendy 2003). Therefore the opera-
 tional concept should be intuitive and clearly SETUP OF THE TOOL
structured. As a decision support system for inner-city planning,
a software environment was designed and imple-
mented as a prototype (project video [1]). The de-
• Extensibility: when new requirements arise or veloped software prototype is part of an interactive
needs are identified in the process of working expandable IT concept (see figure 1). It was imple-
with the tool, it must always be possible to in- mented in Java and evaluated by means of an ex-
tegrate the desired functionality into the sys- emplary inner-city planning scenario. The system is
tem at a later stage. This should be possible based on a software kernel that provides basic func-
without the need of any special programming tions regarding user interaction, data management
skills. and program interfaces. It can in turn be expanded
by plug-ins due to the modular software architec-
ture of the program. This furthermore allows to in-
Figure 1 tegrate existing software data bases into the system
the modular design (Degelmann and Miranda 1987). As a basic form of in-
of the system teraction, the kernel provides a visual programming
interface. It enables the user to develop and cus-
tomize desired functionality in real-time and thus im-
plement user defined analysis and calculation meth-
ods.
The application is therefore divided into two
modes: the planning mode and the programming
mode (see figure 2). The planning mode shows a
three-dimensional model of the selected building
block, its urban context and also the control elements
along with the results of the analyses. It helps users
to quickly examine and assess different hypotheses.
The programming mode, on the other hand, allows
users to develop their own analytical methods and
to adjust the functionality of the application using vi-
Figure 2 sual programming.
concept of the two
modes: planning Planning mode
mode (left) and The planning mode (see figure 3) can be described
programming as an interactive, graphically enhanced summary,
mode (right) where the results of planning measures can be visu-
alized in real time. The working area consists of three
main parts. In the input area on the left-hand side of
the screen, parameters can be changed in real time
with the help of controls such as sliders and buttons.
The central part of the screen shows an axonomet-
ric view of the interactive parametric city model with
the current state of the examined building block. The

Towards Smarter Cities - Volume 1 - eCAADe 32 | 45


user can see how building volumes have changed
compared to the initial situation and can also directly
manipulate the geometry of the buildings and check
whether they still fulfil the applicable building codes.
The output area on the right-hand side of the screen
shows the results of the analysis. With the help of dy-
namic graphs and diagrams, the user is provided with
direct feedback on whether targets can be achieved
and how different parameters have changed in com-
parison to the initial state of the building block (Few
2012). In this manner, important parameters can be
queried, allowing comparisons between different re-
visions and planning results.
Programming mode
The functionality of the tool is defined using pro-
gramming mode (see figure 4). The thus assembled
definitions allow the user to access the implemented
analysis and calculation methods in planning mode.
Elements of the input and output area, which are also
visible in planning mode can be integrated into the
46 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 3
the working area in
planning mode
(screenshot)
definitions for data input and output.
Visual programming has several advantages, es-
pecially for users with limited programming experi-
ence. Data flow and dependency graphs permit the
user to directly manipulate the represented object-
orientated structure (Hils, 1992). They can visualize
individual objects, their condition as well as the in-
terdependencies between them (Boshernitsan and
Downes 2004). Consequently, users with little or no
programming knowledge can work with the system
without having to know textual programming lan-
guages and their syntactic details.
Different stakeholders are involved in the plan-
ning process - each with different backgrounds and
perspectives (Achten 2002). Therefore it is necessary
to provide different ways interacting with the func-
tionality of the software. While on the one hand,
for decision-makers in the municipal administration,
the results of various planning considerations and
their visual significance are more likely important, it
must on the other hand, be possible for urban plan-
Figure 4
the working area in
programming
mode (screenshot)
ning professionals to quickly implement own calcu-
lation and analysis methods. To fulfill these require-
ments and with respect to extensibility, the applica-
tion workflow has three levels of complexity:
• Using the application in planning mode: rapid
results can be obtained by loading a pre-
prepared workspace with a predefined set of
functions. By using the control elements in
the input area on the left-hand side of the
screen, the user can change parameters and
see the effects of these changes on the digital
city model and the calculation results. In ad-
dition, planning variants can be exported as
data sheets and 3D models intended for fur-
ther planning and documentation.
• Development of functionality: individual cal-
culation and analysis methods can be de-
veloped using library components and the
graphical programming interface for use in
planning mode. The results of the calculations
can be visualized in the output area on the
right-hand side of the screen using different
kinds of customizable diagrams and display
elements. This way, the user can implement
own calculation methods based on the infor-
mation contained in the city model.
• Development of individual components: the
functionality of the system can be more fun-
damentally extended, for example to support
complex simulation methods, by program-
ming individual components and integrating
them into the workspace. The graphical pro-
gramming interface makes it possible to give
these modules customizable graphical repre-
sentations, and to link them to other compo-
nents.
As a consequence, when new requirements arise or
needs are identified in the process of working with
the tool, it is always possible to integrate the desired
functionality into the system at a later stage. We de-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 47
 termined the essential requirements for the Visual Figure 5
Programming Language (VPL) as follows: hierarchical
structure of the city
• Easy to use interface
model
• Exception handling

• Implementation of essential functionality like

copy and paste or saving and loading

• Basic components to handle Strings, Num-

bers, Lists and Maps.

• Support for loops

• Appealing graphics that can be customized

IMPLEMENTATION OF THE TOOL

Due to the modular architecture of the IT-concept,
different aspects of the program were developed as
independent libraries in JAVA. The main parts of the
software are described below.

3D city model
The data basis of the tool is a semantic three-
dimensional city model (Kolbe 2009). It is an infor-
mation model for the representation of 3D urban ob-
jects like blocks, land parcels, buildings and trees.
It defines the classes and relations regarding geo-
metric and semantic properties and the relations be-
tween them. Besides geometrical information such
as area sizes, volumes, distances and heights, owner-
ship structures are of crucial importance in the field
of building legislation. Therefore the data-model
has a hierarchical structure that consists of nested Figure 6
classes representing the logic structure of urban ob- the parametric
jects (building � plot � block � model, see figure 5). building model
The parametrical building class makes it possible to
change the geometry of buildings in accordance with
properties such as depth, height, number of floors or
roof shape (see figure 6). The objects degree of ab-
straction is adjusted to the information needed for
working with the building codes. While oriels and
small setbacks can be neglected, roof shapes and

48 | eCAADe 32 - Towards Smarter Cities - Volume 1

floor heights are vital in building legislation. The ge-
ometrical, topological and semantical data that are
required for the 3D city model of the tool can be ob-
tained using interfaces to different data sources such
as cityGML, ArcGIS or digital real estate cadasters.
Visual programming language (VPL)
The visual programing language is developed as a
standalone library for Java, containing all the basic
components to handle the common data types. New
functionality can be added by using the provided
protocol (Java: interface). The elementary compo-
nents of the VPL are nodes, connectors and connec-
tions. The functionality is represented by the nodes.
Each node can have one or more inputs and/or out-
puts. Inputs and outputs of the same data-type(s)
can be connected. These connections define the
data flow (Shu 1988).
One key aspect in the development of the VPL
was to provide the possibility to build patterns out
of individual functions, ie parts of the definition
(Alexander et al. 1977, Gamma et al. 1994). This pro-
vides a distinct advantage: once a certain function-
ality has been defined via graphical programming, it
can be saved and thus be reused. On this way partial
aspects of a project such as analyses or calculations
can be included into other projects.
To allow working with the visual programming
system, a range of components have been prototyp-
ically implemented in the application. They can be
selected from a component library which appears
in the upper left-hand corner of the workspace in
programming mode. The implemented components
range from general logic and mathematical func-
tions to more project-specific components for work-
ing with the urban model, along with graphic ele-
ments for data visualization and control elements.
The library is designed much like an extensible tool-
box so that components can be integrated into the
system at a later date.
Versioning
Planning is a complex process consisting of a se-
quence of decisions, in which falling back to earlier
versions or ideas or exploring alternative avenues is
an inherent part (Rittel and Weber, 1973). It is there-
fore not a linear process. The process of developing
ideas is better expressed as a tree structure. Branches
occur at key points in the design process, where the
idea is developed in different directions. At other
points, separate strands may also converge again to
merge different considerations into a new result. Be-
sides a simple, but yet effective visualization of the
interdependencies between different revisions as a
graph structure, the possibility to directly return to
any previous state of work and thus consider ideas
and elements of different revisions, brings in a signif-
icant advantage: correlations and relationships be-
tween different versions can be clearly illustrated.
Therefore the user can keep them in mind, but still
keep his focus on the desired aim. This in return, re-
duces his effort of documenting and storing data, es-
pecially in extensive projects.
To meet these requirements, a versioning system
was developed to manage all states of work. It can
be understood both as a data model and graphical
representation of the dependencies between differ-
ent work states and results. For data backup and
data exchange it is possible to save and load the
whole repository. In addition individual branches
can also be extracted, for example to integrate them
into other projects and to make single partial results
of different approaches comparable. The graphical
representation of the repository is placed at the bot-
tom of the working area. Here the user can see the
dependencies between different revisions to follow
the decent and the line of development of every
state of work. To directly compare different revisions
based on analysis and calculation results, they can be
placed in the output area on the right-hand side of
the working area via drag and drop.
Export and presentation of results
Planners and architects use digital data on the one
hand and haptic models and printouts on the other
hand. Regarding communication, discussion, archiv-
ing and submission to authorities, communicating
Towards Smarter Cities - Volume 1 - eCAADe 32 | 49
 data is profoundly crucial (Geertmann and Stillwell
2004). Various professionals and non-professionals
have diverse views on topics that need to be re-
spected. Large organisations such as administration
authorities have a high demand for information shar-
ing (Huber 1984). It is managed not only by digi-
tal data, but also via printed media. Therefore the
tool provides the possibility to create templates, de-
scribing the needed characteristics that are assem-
bled into a layout.
The results of the studies can be exported di-
rectly in a way that no further post-processing is nec-
essary. Data sheets showing selected properties (e.g.
floor area, number of residential units, floor area ra-
tio) are instantly available and can be used to com-
municate the results and act as a basis for further
discussion (see figure 7). To allow the results to be
reviewed using a real, haptic model, the tool has
50 | eCAADe 32 - Towards Smarter Cities - Volume 1
a three-dimensional geometric data exporter which
outputs data that can be used directly - i.e. without
the need for further preparatory steps - to produce
real models of the result using systems for digital fab-
rication (CNC-Milling, Rapid Prototyping, see figure
8).
EVALUATION
To evaluate the software prototype, we selected a
sample inner-city block consisting primarily of five
to six storey buildings plus roof level and existing Figure 7
garages, one and two storey workshops and residen- exported data
tial buildings in the courtyards within the block. A se- sheets, showing key
ries of studies were created showing various stages of parameters and
densification on the selected area (see figure 9). indicators of
different planning
CONCLUSION strategies
In this publication, we described an interactive cus-
tomizable digital tool that monitors the key building
codes and visualizes their effects on the urban struc-
ture in real-time, for supporting the development of
redensification strategies. After the requirements for
such a tool were defined, a software environment has Figure 8
been designed and implemented as a prototype. It 3d printed model,
was then evaluated by means of an exemplary inner- showing planning
city block. results
The current implementation of the tool is already
usable in a productive workflow. A range of basic
components have been implemented and success-
fully evaluated using a specific urban block as an ex-
ample. The facility to work directly in the parametric
urban model and the ability to extend the range of
functions at any time offer a suitable basis for a pro-
ductive workflow.
In future research, we intend to develop and
evaluate additional components in order to expand
the collection of analysis and simulation methods.
For example, sunlight simulation and shading analy-
ses are useful in dense inner-city situations, as are dis-
tance determination and wayfinding analyses. These
can provide further useful information to support the
decision-making process. In further development
of the tool new planning fields will be accessed by
Figure 9
series of studies,
showing various
stages of
densification of the
inner-city block
adding the needed functionality to the toolbox. Real
estate development for example is one of the ma-
jor drivers of city densification. Detailed and reli-
able data regarding building volumes, facade sur-
face area and other relevant factors should directly
be taken into account for calculations concerning
costs, finances and risk management. In this manner
the feasibility of different planning approaches could
also be monitored in real time.
REFERENCES
Achten, H 2002 'Requirements for Collaborative Design
in Architecture', DDSS 2002, pp. 1-13
Adam, B and Sturm, G 2011, Zurück in die Stadt: Gibt es
eine neue Attraktivität der Städte?, Bundesinstitut für
Bau-, Stadt- und Raumforschung (BBSR)
Alexander, C, Ishikawa, S and Silverstein, M 1977, A Pat-
tern Language: Towns, Buildings, Construction, Ox-
ford University Press, New York
Boshernitsan, M and Downes, MS 2004, Visual Program-
ming Languages: A Survey, University of California,
Berkeley
Dengelmann, LO and Miranda, V 1987 'Development of
Interfaces for CAD Processing in Architecture', ACA-
DIA Conference Proceedings 1987, North Carolina, pp.
95-104
Elgendy, H 2003, Development and Implementation of
Planning Information Systems in collaborative spatial
planning processes, Ph.D. Thesis, Universität Frideri-
ciana zu Karlsruhe
Few, S 2012, Show Me the Numbers: Designing Tables and
Graphs to Enlighten, Analytics Pr
Fink, D and Fischer, F 2012, 'Neue Werkzeuge für die
dichte Stadt', Bauwelt, 195, pp. 46-49
Fink, D, Fischer, F, Spengler, J, Gebhard, A, Schnetzer,
B, Schinabeck, J, Wagner, T and Häusler, R 2011,
Langfristige Siedlungsentwicklung München: Quali-
fizierte Verdichtung, TUM
Gamma, E, Helm, R, Johnson, RE and Vlissides, J 1994, De-
sign Patterns. Elements of Reusable Object-Oriented
Software, Prentice Hall
Geertman, S and Stillwell, J 2004, 'Planning support sys-
tems. an inventory of current practice', Computer,
Environments and Urban Systems, 28, pp. 291-310
Gratzl, S, Lex, A, Gehlenborg, N, Pfister, HP and Streit,
M 2013, 'LineUp: Visual Analysis of Multi-Attribute
Rankings', IEEE Transactions on Visualization and
Computer Graphics, 19, pp. 2277-2286
Heilig, GH 2012, World Urbanization Prospects: The 2011
Revision, United Nations publications, New York
Hils, DD 1992, 'Visual Languages and Computing Survey:
Data Flow Visual Programming Languages', Journal
of Visual Languages and Computing, 3, pp. 69-101
Huber, GP 1984, 'Issues in the Design of Group Decision
Support Systems', MIS Quarterly, 8, pp. 195-204
Klostermann, RE 1997, 'Planning Support Systems:
A New Perspective on Computer-Aided Planning',
Journal of Planning Education and Research, 17 (1), p.
45–54
Kolbe, TH 2009, 'Representing and Exchanging 3D City
Models with CityGML', in Zlatanova, L (eds) 2009, 3D
geo-information sciences, Springer, Berlin, p. 15–31
Kwartler, M and Bernard, RN 2001, 'CommunityViz: An
Integrated Planning Support System', in Brail, RK
and Klostermann, RE (eds) 2001, Planning support
systems. Integrating geographic information systems,
models, and visualization tools, ESRI Press, Redlands,
CA, pp. 285-308
Towards Smarter Cities - Volume 1 - eCAADe 32 | 51
 Rittel, H and Webber, M 1973, 'Dilemmas in a general
theory of planning', Policy Sciences, 4(2), pp. 155-169
Shu, NC 1988, Visual Programming, Van Nostrand Rein-
hold Company Inc
[1] https://vimeo.com/79028980

52 | eCAADe 32 - Towards Smarter Cities - Volume 1

Agents' movement_towards the reformation of public
space
Step 1: select | implement | observe crowd rules

Maria Kerkidou1 , Anastasia Pechlivanidou-Liakata2 , Adam Doulgerakis3 ,

Alexandros Sagias4
1,2
National Technical University of Athens, Greece
3
Demokritos National Center of Scientific Research
4
University of Piraeus
1
mkerkid@gmail.com 2 deste@central.ntua.gr
3
adoulgerakis@iit.demokritos.gr 4 sagialeko@gmail.com

In order to enable designers to envision the behaviour of pedestrians with

reference to specific environments, computational models of crowds and their
movement become indispensable tools of evaluation as well as tools of creativity.
In this paper, the model under development constitutes a generic model which
incorporates ideas about agent-based systems. The simulation program
comprises a support system for the designer to place virtual users in a context
that bears analogous environmental traits of the area under study. The design
problem which is addressed by the implementation deals with public squares for
which the programmatic demands involve a broad spectrum of users of diverse
idiosyncrasies. Our study attempts to elucidate how the variation in preferences
of pedestrian movement which depend on various personal, situational and
environmental factors, may influence the current use of a selected public space
and underpin qualitative alterations compared to its initial design. The intent of
the methodology is not to create a predictive tool of naturalistic human movement
but to explore how spatial configuration can be assessed and developed through a
simulation model of pedestrian behaviour.

Keywords: Crowd simulation, Spatial behaviour, Pedestrian movement, Public

space

INTRODUCTION of the fact that their principal aims are diverse, their
Crowd simulation has attracted the interest of sev- specific scopes bring into focus pedestrian move-
eral disciplines which, among others, are involved in ment as a way of grasping and measuring traits of
organization and formation of public space. In spite human spatial behaviour. The design of spaces that

Towards Smarter Cities - Volume 1 - eCAADe 32 | 53

 address the plethora in needs of an individual as well
as the diversity of a crowd's consistency (as it occurs
in public spaces) may benefit from an envisioning of
how it could perform given particular assumptions
on the usage conditions and the behaviours of the
autonomous entities composing the crowd which
would populate it. Outlined here is an investiga-
tion of pedestrian movement behaviour in the pub-
lic space of a town square in the city of Athens, based
on agent interaction and rule-based behaviour struc-
ture. The current project has been developed as a
conceptual paradigm as well as a testing benchmark,
aiming to produce a simulation of movement and be-
haviour of people in a public place, particularly the
Syntagma Square, taking into account:i) simulated
preferences and needs for each simulated agent sep-
arately, ii) the structure of space on a hierarchical and
functional level, iii) the diverse qualities and condi-
tions that the spatial clusters host and iv) the impact
that the agents' aggregation and movement has over
the other agents decision making and movement.
PEDESTRIAN MOVEMENT: MODELS
Simulation models
Research on pedestrian behaviour and more specif-
ically movement has illustrated that there can be
different simulation models which are in terms of
scale, broadly categorised into three levels: macro,
meso, and micro (Turner and Penn, 2002). Pedestrian
studies have initially focused on a macroscopic level
suited to transportation modeling, largely describ-
ing the characteristics of the flow rather than those
of the individual pedestrians (Al-Gadhi and Mahmas-
sani, 1991; Blue and Adler, 1998). The system's basic
scope is space allocation for pedestrians in the pedes-
trian facilities without much interest in the direct in-
teraction between pedestrians (Teknomo, 2002). On
the other hand, as Turner and Penn (2002) describe,
the mesoscopic level studies present a good example
of urban-level simulation, referring though mainly to
traffic modeling than pedestrian flow (Nagel et al.,
1996; Daganzo, 1994). Microscopic pedestrian stud-
ies, in contrast to the aforementioned, conceive ev-
54 | eCAADe 32 - Towards Smarter Cities - Volume 1
ery pedestrian as an individual and measure the be-
haviour of pedestrian interaction (Helbing and Mol-
nár, 1997; Hoogendoorn and Daamen, 2007; Pauls et
al., 2007). Currently, literature's interest has shifted
towards the models that operate at the level of in-
dividual pedestrians, trying to provide an overview
of the impact that infrastructural change may have
on pedestrian activity and behaviour (Kerridge et al.,
2001). By incorporating the developed project in the
design process of public space, one gains insight on
the composite nature of crowd behaviour and move-
ment and the complex outcomes that these pro-
cesses have in the actual use of public space. Since
knowledge of the possible outcomes of such com-
plex systems cannot be acquired beforehand, that
is before the actual use of the space by the users,
the development of simulations and their incorpo-
ration in the design workflow may have a tremen-
dous impact in the production and the evaluation
of public space. During the project development,
apart from pertinent research concerning human be-
haviour, knowledge and techniques developed and
used by the game industry have been of significant
impact, since currently, realistic simulation of crowd
behaviour and movement has been proven a key fea-
ture in the acceptability and the engaging factor of
games. To that end, algorithms, methods and solu-
tions have been implemented, expanded and cus-
tomised, within Unity 3D game engine, in order to re-
spond to the necessities and needs of this particular
research.
Implementation models
As simulation techniques have been ameliorating
over the last decade with advances in computa-
tional technologies, various pedestrian simulation
models have been studied through sophisticated be-
havioural algorithms. In terms of implementation
characteristics, different approaches have been fol-
lowed which can be classified mainly as:
Cellular Automata-based models (Blue and
Adler, 1998; Hoogendoorn et al., 2001; Burstedde
et al., 2001) whose dynamic behaviour is founded on
the formula describing the state of a cell for the next
time step, contingent upon the state of its neigh-
bouring cells. Broadly, in CA-based models as one
cell vacates, another cell is being occupied, thus ren-
dering movement.
Force-based models (Helbing and Molnár, 1995;
Okazaki and Matsushita, 1993) in which a physics ap-
proach is used in order to model pedestrians. Crowd
movement in this case, is subject to behavioural
forces resembling to gases, fluids and granular me-
dia.
Agent-based models (Schelhorn et al., 1999; Ker-
ridge et al., 2001; Kukla et al., 2003) which exploit the
analogies of interactions taking place among an as-
sembly of autonomous agents with those deriving
from the dynamic system of pedestrian behaviour.
DESIGN PROCESS: MOVEMENT
So far, modeling and simulation of pedestrian move-
ment scenarios rely on the differentiated assump-
tions that stem from each researcher's goal, concep-
tual framework and tool of implementation. Most
models of pedestrian movement behaviour address
the prediction and assessment of pedestrian perfor-
mance in the existing built environment. Methods in
which movement behaviour influences and defines
spatial design and organization, supporting the de-
sign process on its various stages, have not been ad-
equately studied and tested. Nonetheless, the appli-
cation of models, that can be scripted based on hu-
man movement behaviour rules and data, can sig-
nificantly contribute to tackling design tasks whose
complexity (as met in urban environments) due to
asymmetrical and conflicting gravitational factors
can often hinder the investigation of design possibili-
ties. Therefore, the goal of this research is to develop
a generic model that will allow different designs to
emerge, be compared and selected according to the
way that users negotiate space through movement.
Thus, the structure of such a model should follow
closely the human movement principles that con-
dition the pedestrian behaviour in urban space, in-
dicating the need for detail and suitability of a mi-
crosimulation approach (e.g. obstacle avoidance)
that could probably also extend to a meso-scale one
(e.g. individuals planning multi-stop trips). In such
a model every pedestrian is treated as an individual
unit whose movement characteristics are valorised.
Respectively, the aggregate level of the pedestrian
crowd is formed by taking into account the values of
movement parameters and by complying with pre-
determined behaviour rules of acting and interact-
ing. Insights to these elements can be acquired from
pertinent literature that deals with empirical data.
Of these, the most significant include walking speed
and spatial use. Walking speed is affected by sev-
eral factors which are contingent on personal charac-
teristics of pedestrians (e.g. age group, gender), fea-
tures of the urban infrastructure (e.g. attractiveness,
protection) environmental conditions (e.g. weather,
ambience), as well as pedestrian density (e.g. free
flow, traffic hindrance) (Daamen and Hoogendoorn,
2003; Hoogendoorn and Daamen, 2007). Spatial use,
on the other hand, comprises both lateral and lon-
gitudinal measurements of pedestrian walkway and
can vary according to the number and type of enti-
ties (obstacles, other pedestrians etc.) that can alter
the initial walking trajectory and speed (Willis et al.,
2004).
DESIGN PROCESS: AGENT MODEL
Specifically, this approach aspires to delineate the
behavioural elements concerning the pedestrian ex-
perience of space which are transparently embed-
ded within the environment and have not neces-
sarily been integrated to the initial design process.
In order to incorporate diverse pedestrian charac-
teristics and embody the range of factors influenc-
ing pedestrian behaviour an agent-based approach
is implemented, offering the ability to simulate au-
tonomous heterogeneous entities inside an environ-
ment, which enables their perception, interaction
and action (Bandini et al., 2005). Agent models have
been used to simulate various implications of pedes-
trian movement and are distinguished among dif-
ferent approaches for their quality of being akin to
Towards Smarter Cities - Volume 1 - eCAADe 32 | 55
 reality and therefore inherently suited to simulating
people and objects in very realistic ways (Castle and
Crooks, 2006). As Bhatta explains (2010), these mod-
els, while operating in a bottom-up way, have the
ability to simulate the simultaneous operations of
more than one agent, in an attempt to re-create and
predict the actions of complex phenomena. For ex-
ample, agents have been employed in order to: ex-
amine how the elaborate organisation of urban net-
works affects pedestrian movement (Resnick, 1997),
to establish the correlation of pedestrian behaviour
with route choice (Schelhorn et al., 1999); to study
the dynamics of pedestrian behaviour in streets (Ker-
ridge et al., 2001).
IMPLEMENTATION: DESIGN TASK
The design problem which is addressed by the im-
plementation deals with public space and in particu-
lar, public squares. The example of Syntagma square
(figure1) in the city of Athens offers variety in types of
users, levels of actions and interactions, environmen-
tal features and conditions etc. The selected town
square is located in the city centre and constitutes a
landmark and point of interest with frequent use dur-
ing day and night. Its size in spite of its significant
role in socio-political and commercial activities is rel-
atively moderate compared to other European town
squares. The eastern side of the square is higher than
the western. Both, the northern and the southern
side of the square include two green areas with shade
trees, while in the centre of the square a large wa-
ter fountain is located. Moreover, Syntagma Square
comprises a hub for many forms of public transporta-
56 | eCAADe 32 - Towards Smarter Cities - Volume 1
tion and is located within walking distance from sev-
eral historical sites, government buildings, shops and
offices. It is within the scope of this research to em-
ploy micro-scale principles with regard to the de-
tailed movement coordination of agents (e.g. colli-
sion avoidance, speed variations, interactions rules)
in combination with features that can also appear
in meso-scale trip organisation and wayfinding (e.g.
decision making, arrangement of connectivity graph
among goal locations) following simple rules of inter-
actions. The model's aspiration is to enable the visu-
alisation of human movement into virtual spatial con-
figurations, and thus, provide a testbed for animation
experiments that can contribute to the designer's un-
derstanding and creative thinking.
Figure 1
The implementation consists of a tentative
Syntagma Square
framework for a pedestrian movement model which
_Town square in the
incorporates agent-based modeling techniques, as-
city of Athens,
signing values of various parameters to individual
Greece Original
agents representing pedestrians, and defining rules
Photo: A.Savin
that act upon them.The model is informed by ob-
28/6/2013 Licensed
servational studies of human movement behaviour
under the terms of
in urban spaces providing measurements regarding
CC-BY-SA 3.0/FAL
preferred speed and distance that people like to
maintain around themselves. Agent systems emerge
as a conceptual paradigm to simulate the interaction
of autonomous agents in an environment; specifi-
cally, they facilitate the investigation of the individ-
ual's behaviour in micro-level associated to the pat-
terns emerging through the interplay of numerous
individuals in macro-level (Popov, 2009). A multi-
agent system is composed of a number of possibly
heterogeneous agents that act and interact within
and possibly with their environment. Furthermore,
aiming at dynamic environmental conditions and
physical interactions, the current research utilises a
cross-platform game engine that facilitates the simu-
lation, Unity 3D. Nowadays, video game culture and
artefacts are highly accessible and provide new me-
dia for designers to extend the physical spaces of
built architecture into meaningful virtual domains
and gather useful experience with yet to be con-
structed projects (Burrow and More, 2005). The
 model developed here synthesises algorithms con-
cerning the navigation and collision-free pathfind-
ing process of the pedestrian agents, as well as their
steering behaviour in cases of static and dynamic
threats such as other agents in complex environ-
ments. By doing so, the implementation aims to
combine the advantages of the constituent algo-
rithms (presented in the following section) in order
to identify the selected town square's current use in
terms of pedestrian movement and delineate possi-
ble behavioural elements that have not been inte-
grated to the initial design process.
Figure 2
Layers of the model
Model structure: constituent tools
The model can be briefly described as an agent-
based model that simulates movement behaviour
within a 3D virtual environment that represents the
area under study (figure 2). The structure of the
model is outlined in the diagram presenting the key-
features that compose the implementation script
(figure 3). The pedestrian agents act according to
predefined rules, navigational preferences and activ-
ity goals that the program user has set. Variation to
the aforementioned is embedded according to differ-
entiated: agent types (based on personal character-
istics of pedestrians: age, gender, level of mobility,
size etc.), trip specifications (walking purpose, flexi-
bility of activity agenda), and environmental features
(surroundings' characteristics, function of the pedes-
trian area and weather conditions). The agents may
choose to modify their navigational plans in response
to their surroundings, the behaviour of other agents
or the activities that take place nearby presuming
they fit in their agenda. The decision making pro-
cess relies on the agent's activity goals that are taken
into account by a multiplicative factor (based on the
agent's activity hierarchy and the proximity of the
challenging activity) returning a weighted outcome
that defines the agent's current state. The selection
of action alternatives depending on weight can lead
to always picking the one with the highest weight.
This feature leads to a deterministic model. As such,
the model may result to having several actions dom-
inating whereas others are being ignored.
Furthermore, as is often the case with actions
that have very similar weight totals, an unfair disre-
gard for the runner up occurs. As Kukla et al (2003) in-
dicate, a solution to this shortcoming is to make a ran-
dom choice, such that actions with a higher weight
have simply greater chance of being selected. Fol-
lowing the outline of its state, the agent runs through
the specifics of its forthcoming actions taking into ac-
count the distance between its current location and
the following target's location. This information is
stored and translated within the abstract data struc-
ture of a navigation mesh (figure 4). Navigation mesh
is a technique to describe the traversable surface of
the 3D environment using a set of convex polygons.
Due to its simplicity and high efficiency in 3D rep-
resentations, navigation mesh has become a main-
stream choice for 3D games. According to the num-
ber of sides of polygons, navigation mesh can be cat-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 57

egorized into triangulation and polygonization. It is
a highly intuitive floor plan, which agents can use for
navigation and pathfinding in virtual worlds (Cui and
Shi, 2011). Thereupon, the A* algorithm is applied in
order to configure the optimum path (Hart and Nils-
son, 1968). A* constitutes a robust heuristic search
method for many problems, pathfinding just being
one of them. For pathfinding in particular, A* algo-
rithm repeatedly examines the most promising un-
explored location it has traced. When a location is ex-
plored, the algorithm has completed its task if that lo-
cation is the goal; otherwise, it makes note of all that
location's neighbours for further exploration (Cui and
Shi, 2011).
The square's moderate size and geometrical charac-
teristics facilitate the endeavour in terms of agent-
perception, allowing the assumption that actual
pedestrian vision can in this specific case, be es-
tablished through a territorially pre-informed agent.
Based on the above, A* can be applied avoiding en-
cumbering computational cost. In different spatial
configurations where familiarity of the surroundings
58 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 3
Diagram of model
structure
cannot be included in the predefined knowledge of
the agent, adjustments of the program should take
place. For the refinement of the resulting path, the
implementation of funnel algorithm computes the
shortest Euclidean path between two points in a tri-
angulated polygon in linear time (Hershberger and
Snoeyink, 1991). Hence, in order to reach the activ-
ity target, the path is analysed in a series of "local"
targets. During their attempt to reach a particular
goal location, the agents navigate through the vir-
tual environment and interact with features of the
environment as well as with other agents. Apart
from carrying information regarding their heading
and route formation, the agents also convey instruc-
tions on how to deal with obstacles that remain sta-
Figure 4
tionary or have dynamic behaviour such as its own.
Syntagma Square
Rule-based approaches use various conditions in or-
_Navigation Mesh
der to identify the situation of an agent and once
that is defined, then the rules compute what steer-
ing decision the agent should make (Kapadia and
Badler, 2013). The development process has been
based on a range of computerised tools and libraries,
which being modified according to the requirements
of the specific case study, have been combined in
order to address the needs of the simulation. The
assembly of algorithms include primarily the Recast
& Detour (open source toolset developed in C++,
implementing navigation mesh, and A* pathfind-
ing) by Mononen[1]; the respective Unity adapta-
tion, CAINav, implemented by Pratt[2]; as well as, the
OpenSteer (C++ library helping to construct steering
behaviours for autonomous characters in games and
animation) by Reynolds[3], and its transcription to
UnitySteer (builds upon OpenSteer and OpenSteer-
DotNet, ported to Unity) by Méndez[4].
 Case study: Syntagma Square or immobile entities. Particularly, each agent begins
The experimental model is based on empirical data its itinerary at a given walking speed and navigates
that are retrieved from related literature (Willis et al., through space towards a final location target accord-
2004; Daamen and Hoogendoorn, 2003; Ishaque and ing to the respective type of user that the agent rep-
Noland, 2008) focusing on features that compose the resents. The pedestrian type is also intertwined with
individual pedestrian's characteristics and situational the complementary activity agenda that an agent is
factors that affect walking speed and spacing be- presupposed to undertake should the opportunity is
haviour: i) age, ii) gender, iii) level of mobility, iv) trip given during its journey. At the same time, spacing
preferences and purposes, which are assigned to dif- preferences (distances maintained from other pedes-
ferent pedestrian types that assemble the pedestrian trians or infrastructure objects), environmental at-
crowd of the square: i) children/adults/elder, ii) wom- tractiveness (shaded zones during summer or shel-
en/men, iii) carrying weight/walking with company, tered spaces in rain, lighted paths during night-time),
iv) travelling towards work/going for shopping/see- steering restrictions or affordances are also consid-
ing the sites. The program is implemented by us- ered and calculated providing a range of walking
ing mean values of the aforementioned parameters speeds and avoidance distances. Syntagma Square
from observational data in order to provide a generic is being studied as a confined area with specific en-
testbed for further calibration in the future. How- try points from which agents appear. These pedes-
ever, it is within the intentions of this study to coor- trian gateways are points of pedestrian street cross-
dinate the agents according to measurements deriv- ings (with traffic lights regulating the crowd's move-
ing from the town square and determine the over- ment), stairs that connect the upper with the lower
all composition of the pedestrian population associ- part of the square, metro station entrances near the
ated with the area per se in terms of different types of centre of the square. During a model run, agents are
people and their activities, so that the designer can despatched from these entry points and start their
have an imminent estimation of the spatial use un- navigation to the modelled area (figure 5).
der study. The changing pedestrian circulation ac-
Figure 5
cording to day and night hours or diverse weather
Agents _Navigation
conditions can also be of interest based on attrac-
tiveness and comfort. Broadly, each pedestrian rep-
resented by an autonomous agent moves in virtual
space (whose general plan is presumed as familiar
or easily discernible due to the square's geometrical
characteristics) towards a predefined target location.
Initially, the emerging movement patterns are under-
pinned by the assumption that agents follow a path
that comprises relatively straight lines and select the
more short and suitable route that meets their re-
quirements and purposes. However the actual path
followed, affected by local interactions, is not merely
Each agent has an intended destination and within
a linear path abiding by the fact that actual people do
the flexible or rigid limits of its preconfigured activ-
not move in strictly straight trajectories and don't al-
ity schedule, its movement evolves as a sequence of
ways follow the shortest path (Conroy Dalton, 2001;
closer targets, like a sequence of nodes inside the
Golledge, 1995). The agents' navigation includes ful-
navigation mesh. Should the surroundings within
filling predefined needs and interacting with mobile
a certain range induce the initiation of an activity,

Towards Smarter Cities - Volume 1 - eCAADe 32 | 59

 then the agent pursues its updated goal, refines its
schedule and redefines its path respectively based on
weighted factors that satisfy the agent's needs and
preferences through the shortest possible route.
Rescheduling and adjusting represent tasks
which are computed on the fly, including the vari-
ations in speed, direction and spatial use which con-
tribute to the detailed behaviour of agents. Varia-
tions of this sort, allow different and unpredictable
behaviours to occur in the dynamic setting (Haklay
et al., 2001). The 3D model of the square offers an
elaborate virtual environment in which the agent
population negotiates space, and thus, the program-
user has the opportunity to observe the movement
decisions as well as the crowd's overall pedestrian
behaviour while analysing or organising spatial con-
figurations of the modelled area.
CONCLUSIONS
The amount of information embedded in built envi-
ronment questions the self-sufficiency of traditional
design representations that can be significantly en-
riched by incorporating alternative methods such as
animation techniques and virtual environments, in
the design process. Enhancing the simulation model
with diverse types of agents (representing pedestrian
users) to act autonomously, and interact, enables the
exchange of information between agents and the en-
vironment to take place in virtual terms and be im-
printed onto the architectural creation procedure. It
is within the scope of this paper and from an archi-
tectural point of view to examine the attempt to de-
velop a pedestrian model in three-dimensional envi-
ronment that follows principles of human movement
behaviour. Within the framework of a case study that
deals with public space, the program has been found
useful to observe and evaluate movement behaviour.
By observing the situated virtual users while navigat-
ing on their own simulated volition through the vir-
tual model of the environment under study, the de-
signer has the opportunity to discover under differ-
ent terms the reciprocal relation between space and
its users. The model's geometry is representative of
60 | eCAADe 32 - Towards Smarter Cities - Volume 1
built-form offering direct visible results of the agents'
behaviour inside the 3D environment and therefore,
achieves to promote potential exposure of any hid-
den parameters facilitating the role of the designer.
Being based on the established familiarity with the
3D visualisation tools which contemporary architects
use, the model takes precedence over more abstract
approaches. Based on the above, the simulation
model can be used in different levels of design inves-
tigation, like micro and meso (see Simulation models
section), providing the rules of movement behaviour
that the scale of the model necessitates.This can be
very helpful when imperative decisions on the selec-
tion of design solutions are made as it provides the
basis for scientifically justified solutions where other-
wise decisions could only stem from theoretical as-
sumptions or basic and biased experience. The inves-
tigation, through simulation models, of how the vari-
ous personal, situational, and environmental factors,
shape pedestrian behaviour, reflects the designer's
need for an interface through which a putative ur-
ban space can be constructed and explored; thus, en-
abling a more comprehensive appreciation of the use
and formation of space. Combining empirical data
with several techniques of movement simulation ap-
plied in the film and game industry, the experimen-
tal model can be used as a complementary investiga-
tive design tool for pedestrian behaviour in relation
to spatial organisation, requiring further expression
and elaboration by traditional manual acquirements.
FUTURE WORK
As research objectives for further investigation, prin-
cipal goal will be to identify potential inadequacies of
the present design and articulation of space in ques-
tion, and stress the need for improvements (these
may refer to mere addition of signage systems or to
intricate suggestions of re-routing pedestrian paths).
Additional aim will be to obtain empirical data de-
riving from video recordings of the pedestrian be-
haviour exhibited at the public square in question
and compare the simulation's results to the findings
inferred by the collected observation data analysis.
Furthermore, as a following step in the development
of the project, the interaction of the crowd with the
structure and the features of space will be imple-
mented not only as a one way relationship "space
affecting crowd", but also as a dialectical relation-
ship "space and crowd reciprocally influencing each
other". Based on the aforementioned, crowd be-
haviour and movement will be studied as a factor
that shapes and transforms the crowd's surrounding
space in the early stages of the design process rather
than after construction.
ACKNOWLEDGMENTS
This research has been co-financed by the European
Union (European Social Fund - ESF) and Greek na-
tional funds through the Operational Program "Edu-
cation and Lifelong Learning" of the National Strate-
gic Reference Framework (NSRF) - Research Funding
Program: Heracleitus II. Investing in knowledge soci-
ety through the European Social Fund.
REFERENCES
Al-Gadhi, SA and Mahmassani, HS 1991, 'Simulation
of crowd behavior and movement: fundamental
relations and application', Transportation Research
Record, 1320, pp. 260-268
Bandini, S, Manzoni, S and Vizzari, G 2005, 'Crowd Mod-
eling and Simulation', in Van Leeuwen, JP and Tim-
mermans, HJP (eds) 2005, Recent Advances in Design
and Decision Support Systems in Architecture and Ur-
ban Planning, Springer Netherlands, pp. 161-175
Bhatta, B 2010, Analysis of Urban Growth and Sprawl from
Remote Sensing Data, Springer
Blue, VJ and Adler, JL 1998, 'Emergent fundamental
pedestrian flows from cellular automata microsimu-
lation', Transportation Research Record, 1644, pp. 29-
36
Burrow, A and More, G 2005 'Architectural designers and
the interactive audience', Proceedings of the second
Australasian conference on Interactive entertainment,
Sydney
Burstedde, C, Klauck, K, Schadschneider, A and Zittartz,
J 2001, 'Simulation of pedestrian dynamics using a
two-dimensional cellular automaton', Physica A: Sta-
tistical Mechanics and its Applications, 295(3-4), pp.
507-525
Castle, CJE and Crooks, AT 2006 'Principles and concepts
of agent-based modeling for developing geospatial
simulations', Working Paper 110, Centre for Advanced
Spatial Analysis, University College London, London
Cui, X and Shi, H 2011, 'A*-based Pathfinding in Modern
Computer Games', International Journal of Computer
Science and Network Security, 11(1), pp. 125-130
Daamen, W and Hoogendoorn, SP 2003, 'Controlled ex-
periments to derive walking behaviour', European
Journal of Transport and Infrastructure Research, 3(1),
pp. 39-59
Daganzo, CF 1994, 'The cell transmission model: A dy-
namic representation of highway traffic consistent
with the hydrodynamic theory', Transportation Re-
search Part B: Methodological, 28, pp. 269-287
Conroy Dalton, RA 2001 'The secret is to follow your
nose', Proceedings of the 3rd International Symposium
on Space Syntax, Atlanta, pp. 47.1-47.14
Golledge, RG 1995 'Path selection and route preference
in human navigation: a progress report', Proceedings
of Spatial InformationTheory-Lecture Notes in Com-
puter Science, pp. 207-222
Haklay, M, O'Sullivan, D and Thurstain-Goodwin, M 2001,
'"So go downtown": simulating pedestrian move-
ment in town centres', Environment and Planning B:
Planning and Design, 28(3), pp. 343-359
Hart, PE, Nilsson, N and Raphael, B 1968, 'A Formal Ba-
sis for the Heuristic Determination of Minimum Cost
Paths', IEEE Transactions on Systems Science and Cy-
bernetics, 4, pp. 100-107
Helbing, D and Molnár, P 1995, 'Social force model for
pedestrian dynamics', Physical Review E, 51(5), pp.
4282-4286
Helbing, D and Molnár, P 1997, 'Self-organization phe-
nomena in pedestrian crowds', in Schweitzer, F (eds)
1997, Organization of Complex Structures: From Indi-
vidual to Collective Dynamics, CRC Press, pp. 569-577
Hershberger, J and Snoeyink, J 1991 'Computing mini-
mum length paths of a given homotopy class', Pro-
ceedings of 2nd Workshop, WADS, Ottawa, pp. 331-
342
Hoogendoorn, SP, Bovy, PHL and Daamen, W 2001,
'Microscopic pedestrian wayfinding and dynam-
ics modelling', in Schreckenberg, M and Sharma,
SD (eds) 2001, Pedestrian and evacuation dynamics,
Springer-Verlag, Berlin, pp. 123-154
Hoogendoorn, SP and Daamen, W 2007 'Microscopic
Calibration and Validation of Pedestrian Models:
Cross-Comparison of Models Using Experimental
Data', Proceedings of Traffic and Granular Flow ’05, pp.
329-340
Towards Smarter Cities - Volume 1 - eCAADe 32 | 61
 Ishaque, MM and Noland, RB 2008, 'Behavioural Issues into human pedestrian behaviour in the built envi-
in Pedestrian Speed Choice and Street Crossing Be- ronment', Environment and Planning B Planning and
haviour: A Review', Transport Reviews: A Transna- Design, 29(4), pp. 473-490
tional Transdisciplinary Journal, 28(1), pp. 61-85 Willis, A, Gjersoe, N, Havard, C, Kerridge, R and Kukla,
Kapadia, M and Badler, NI 2013, 'Navigation and steer- R 2004, 'Human movement behaviour in urban
ing for autonomous virtual humans', WIREs Cogn Sci, spaces: implications for the design and modelling
4, p. 263–272 of effective pedestrian environments', Environment
Kerridge, J, Hine, J and Wigan, M 2001, 'Agent-based and Planning B, 31(6), pp. 805-828
modelling of pedestrian movements: the questions [1] http://arges-systems.com/blog/2009/07/08/unitysteer-
that need to be asked and answered', Environment steering-components-for-unity/
and Planning B: Planning and Design, 28(3), pp. 327- [2] http://opensteer.sourceforge.net/
341 [3] http://www.critterai.org/projects/cainav/
Kukla, R, Willis, A and Kerridge, J 2003 'Application of [4] https://github.com/memononen/recastnavigation
context-mediated behavior to a multi-agent pedes-
trian flow model (PEDFLOW)', 82th Annual Meeting of
the Transportation Research Board (TRB), Washington
DC
Nagel, K, Barrett, CL and Rickert, M 1996 'Parallel traf-
fic microsimulation by cellular automata and appli-
cation for large-scale transportation modeling', Los
Alamos National Laboratory unclassified report LA-
UR-96-50, New Mexico
Okazaki, S and Matsushita, S 1993 'A study of simula-
tion model for pedestrian movement with evacu-
ation and queuing', Proceeding of the International
Conference on Engineering for Crowd Safety, London,
pp. 271--280
Pauls, JL, Fruin, JJ and Zupan, JM 2007, 'Minimum
Stair Width for Evacuation, Overtaking Movement
and Counterflow-Technical Bases and Suggestions
for the Past, Present and Future', in Waldau, N,
Gattermann, P, Knoflacher, H and Schreckenberg,
M (eds) 2007, Pedestrian and Evacuation Dynamics
2005, Springer, pp. 57-69
Popov, N 2009 'Utilising Agent Based Models for Simu-
lating Landscape Dynamics', Proceedings of Cumulus
38ºSouth: Hemispheric shifts across learning, teaching
and research, Melbourne
Resnick, M 1997, Turtles, Termites, and Traffic Jams: Ex-
plorations in Massively Parallel Microworlds (Complex
Adaptive Systems), MIT press, Cambridge, MA
Schelhorn, T, O'Sullivan, D, Haklay, M and Thurstain-
Goodwin, M 1999 'STREETS: an agent based pedes-
trian model', Proceedings of Computers in Urban Plan-
ning and Urban Management
Teknomo, K 2002, Microscopic Pedestrian Flow Character-
istics: Development of an image processing data col-
lection and simulation model, Ph.D. Thesis, Tohoku
University
Turner, A and Penn, A 2002, 'Encoding natural move-
ment as an agent-based system: an investigation

62 | eCAADe 32 - Towards Smarter Cities - Volume 1

A Mobile Sensing Kit for Urban Analysis
For more legible, quantifiable intangible and temporary data

Nancy Diniz1 , Hai-Ning Liang2

1
Xi'an Jiaotong-Liverpool University 2 Xi'an Jiatong-Liverpool University
1,2
{nancy.diniz|haining.liang}@xjtlu.edu.cn

This paper describes the design and testing of a mobile sensing kit for real-time
urban site analysis. It aims to contribute to a discussion on the importance of
sensing-based mapping of urban phenomena following our assertion that current
delivery of urban analysis methodologies, with their emphasis on visuo-spatial
analysis only, fail to create maps that reflect the nuanced and layered
interrelationships between the people and the physical environmental space they
live in. In essence, the paper will bring to light the set up components and
deployment of a mobile sensing kit that allows for usually mapped static invisible
data (air quality, temperature, humidity) to be mapped as visible data. In other
words, we want to explore how real-time geo-referenced data collection can
enhance traditional data mapping and visualization methods enabling designers
to understand better the urban space.

Keywords: Environmental data collection, mobile sensing, intangible data

collection and visualization, city modeling, site surveying with open hardware,
DIY electronics

INTRODUCTION AND BACKGROUND
This paper highlights the marginalisation of intangi-
ble and temporary data in conventional urban anal-
ysis. Even though extensive and extremely valuable
research has been put into sensory analysis in the
urban space (Degen, Monica Montserrat and Rose,
Gillian 2012) these analogue mapping approaches
lack in visual and dynamic visualization of their find-
ings. This paper also addresses the limited GIS tech-
niques in spatial design that tend to be weighted
heavily toward visual, surface-based data (land use,
pedestrian flows, slope analysis, exposure, viewshed,
etc.). Within this domain, our goal is to transform as-
pects of the intangible world into a form that is legi-
ble and quantifiable. This paper aims at setting out a
sensorial base analysis framework that allows for the
use of electronics and sensory nodes (temperature,
light, air quality, bio-feedback data) that can allow for
an informed design that is able to provide design re-
sponses that have relationship to sensor parameters.
A long-term aim is an increased critical understand-
ing of local spatial characteristics and design impli-
cations for students of urban planning and architec-
ture.
We propose that a design framework based on
sensory information can enhance traditional walk-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 63

 Figure 1
Examples of
traditional site
analysis
visualizations on
flow (of people,
buses, cars,
bicycles).

Figure 2
The Snapshots of
animated
visualizations
display
time-dependent
information with
emphasis on
sensorial and
proxemics
experiences.

64 | eCAADe 32 - Towards Smarter Cities - Volume 1

along and surveys methods (Degen at al 2012)
and GIS data vizualization methods and increase
our visual understanding of urban interrelationships.
Hence, in our research we place primarily a deliber-
ate emphasis on the invention of novel methods for
understanding the site through sensing-based data
collection. This emphasis on data collection, inter-
pretation, and manipulation is a direct reaction to
the perceived risk posed by a "limited" understand-
ing of a place or a context, in which designers re-
act predominantly to existing physical features and
culturally based design goals. The common issue in
spatial design potentially fails to acknowledge spe-
cific site conditions that would remain undetected
by our senses alone and therefore new supporting
tools are needed to capture and represent the un-
detected data. The conceptual depth of the con-
ventional 'site visit' in architecture and urban design
is called into question. Alternative methods for site
recognition form the context of this exercise, includ-
ing "time-based" mapping paradigms using anima-
tion techniques to visualize data (Diniz et al 2012) and
"emotional mapping" by measuring the sweat on the
palm hands of users exploring the space resulting in a
series of bio-maps (Nold 2009) demonstrate how dif-
ferent ways of mapping space by leveraging percep-
tual and other sensorial understanding, leads to the
formation of variables (or parameters) which at the
same time turn themselves as catalysts for other vari-
ables. The potential for a constantly evolving rein-
terpretation of the perceptual experience and for as-
sociated paradigm to shift suggest a multiplicity of
design possibilities for urban areas that also need to
adapt to the new requirements of contemporary liv-
ing.
PREVIOUS EXPERIMENTS USING TIME-
BASED URBAN MAPPING VISUALIZATION
This paper build on previous research (Diniz et al
2012) supporting the idea of mapping urban space
experience in real-time incorporating people's fluxes,
activities, smells and densities in space. We did an ex-
ercise using animation techniques in an attempt to
provide practical tools and techniques to communi-
cate temporal sensorial information to assist in the
analysis and design of urban sites. From our study,
we used three simple techniques: animation, tradi-
tional site analysis drawings and parallel small multi-
ples or layered visualizations. Animation was used as
an easy-to-implement and intuitive way to communi-
cation time-dependent data. In more traditional vi-
sualizations, they rely heavily on the use of arrows,
colour contrasts, and simple shapes as visual high-
lights (Fig. 1). On the right, the maps make use of lay-
ering of information to highlight relationships. With
this kind of mapping is difficult to incorporate anima-
tion into them, especially as a post design feature. In
contrast, Fig. 2 was conceptualized with animation
being an integral part. Simple polygons describing
the urban morphology were used in the background
to deemphasize the information (so as to highlight
the other, foreground information). Finally, parallel
small multiples differed from layered visualizations,
with each one having varied benefits. Given access
to both during the analytical process can bring forth
their complementary benefits.
EXPERIMENTAL SET UP FOR A MOBILE UR-
BAN SENSING PROTOTYPE
Drawing on a previous sensor kit named "Ambient
Sensor Kit (ASKit)" by (Melsom, J.; Fraguada, L.; Fack-
lam, F.; Pecegueiro Curado, F. (2011) and 'Citizen Sen-
sor'; [1] as precedents to our project, the project de-
velops a sensor pack and data contextualization sys-
tem that allows users to collect readings of real time
data from their immediate surroundings using sen-
sors by recording the environmental conditions (air
quality, temperature, light, and heart-rate). The main
components of the hardware prototype implemen-
tation are an Arduino-based controller board, sen-
sors (air quality sensors, light, noise, temperature and
heart-rate) and a smart phone as an interface (Fig 3).
The project has also developed an online interface
that maps people's routes within the environs they
live; recording the data in real-time (Fig 4).
We use an Arduino mega board, as this version
Towards Smarter Cities - Volume 1 - eCAADe 32 | 65

has enough analogue and digital ports for the many
sensors to work together. Analogue signal from the
sensors is received via the IO port , and our program
runs on the Arduino control board to translate and
sample them into digital information. The power
management and other additional services are pro-
vided by our program.
Next we will detail the smart phone application
and online mapping platforms in all its components
as illustrated in (Figs 5 and 6)
A Service is an application component represent-
ing either an application's objective to perform a
longer-running operation while not interacting with
the user or to supply functionality for other applica-
66 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 3
The mobile sensit
kit components on
the left, the smart
phone interface
and screenshots of
the prototypes
being developed
and tested by users.
Figure 4
Screen shots of the
geo-referenced
online visualization
platform.
tions to use. An activity is a single, focused thing that
the user can do. Almost all activities interact with the
user, so the Activity class takes care of creating a win-
dow for you. A Fragment is a piece of an application's
user interface or behavior that can be placed in an Ac-
tivity, as we see in the various menus and interactive
steps in the smartphone interface application (Fig 3).
We have 3 services in our application: Bluetooth, GPS
and Cloud services. The Bluetooth Service, it receives
data from Bluetooth and unpacks it, and trims it, so
we get data correspond to the different sensors. In
order to spreads the data to all other components.
The Bluetooth service packs the data into a broad-
cast message. All other components have its Broad-
Figure 5
Diagram of the
sensing kit
prototype
components.

Figure 6
Smartphone API
and online
mapping platform
architecture

Towards Smarter Cities - Volume 1 - eCAADe 32 | 67

 cast Receiver, it is asynchronous and event-based, it
is triggered as the same time as the data is being
received. The GPS Service provides the location in-
formation to the phone, it called 'Google Play Ser-
vice API' to control the inner GPS chip to get the GPS
data. These data consists of Latitude, Longtitude, Ac-
curacy, Altitude, Speed and Local Time. When the
location is updated, this service writes the data to
the context for the application for global access. The
Cloud Service is connected with Yeelink.net server.
Yeelink is a distributed realtime computation plat-
form. It can process unbounded streams of data, do-
ing real time data processing. We just use it to store
our data. After the Cloud Service receives the broad-
cast, it sends the sensor data and the GPS informa-
tion with a timestamp via http protocol. Because sen-
sor data is a kind of stream, it is associated with time,
episodic smoothly continuous. The Yeelink merges
all the streams at the entrance point in their system
and splits it into various streams by stream id. The ad-
vantage of stream management system as opposed
as a normal relationship Database is that we can pro-
cess the data like a stream even it has been put in
the warehouse. Finally, we explain the online map-
ping feature. We use google map as the base We-
bGIS system. And we use jquery to build a friendly
user interface browser. Then a significant amount of
JavaScript visualization libraries are written and im-
ported to show the data in the form of heat maps.
The application gets the stream via the Yeelink Ap-
plication Programming Interface (API), and presents
the latest location and data or historical data for the
sensor kit. The stream use json as the format and http
protocol. The application sets a timer for five seconds
as an updating period. Then it requests the Yeelink
every five seconds for new updated data. Therefore
the stream data is based on a timeline range previ-
ously programmed, in this case every five seconds.
DATA COLLECTION TESTING AND FUTURE
WORK
The data was captured in different measurement ses-
sions; the kit incorporates a GPS chip to determine
68 | eCAADe 32 - Towards Smarter Cities - Volume 1
the position on earth and the orientation of the sen-
sors in each data collection session. Each data cap-
ture session lasted for about 15 minutes, they hap-
pened in different cities in China (Suzhou and Shen-
zen) and in two consecutive days in the morning
and afternoon in the same position and orientation.
This paper takes the potential of real time mapping
and visualization as a point of departure to discuss
and rethink traditional urban site analysis and de-
sign. The aim and scope of this paper is to provide
new visualization tools which could be implemented
with widely available open source software and low
cost DYI electronics and hardware. The assumption
is that real time mapping can allow for different anal-
ysis and evaluation of existing urban conditions and
therefore support different interpretations and simu-
lations of urban spaces to enhance urban spatial ex-
perience. To test our assumption, we described the
design, the hardware and software methodologies
involved in the prototypes implementation and the
results of data collection with the sensing kit. In the
future we want to conduct systemic testing and eval-
uation of the prototype by setting up an exercise of
dynamic mapping in a task for urban site analysis by
architecture students and explore how this method-
ology may provide opportunities for different design
process approaches, as opposed to traditional, static
and compartmentalized urban design site analysis.
The online visualization application (Fig. 8) needs fur-
ther development to allow a translation of the quan-
tifiable data at the moment only visualized on the
Yeelink platform (Fig 7). We are in the early stages
of understanding the potentials of dynamic mapping
and the many aspects of their creation, and encour-
age others to conduct further research in this chal-
lenging application domain. A recurring question for
future development of our research remains: will we
be able do we design differently by capturing, and vi-
sualizing unquantifiable and intangible data?
Figure 7
Screenshots of the
Yeelink application
where we can
visualize all
historical data
captured by the
sensing kit..

Figure 8
Screenshots of user
tests on the online
visualization
platform.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 69

 ACKNOWLEDGEMENTS
We thank the following Xi'an Jiaotong University
B(Eng) Architecture students: Gu Mengxue, Tao Yang
and Siyao Wang, for working on the design and tests
of the prototypes. Also to Jun Li and Guogen Chen
for the hardware and software implementation. Fi-
nally we thank XJTLU for sponsoring this project with
a SURF research grant.

REFERENCES
Degen, MM and Rose, G 2012, 'The sensory experiencing
of urban design: the role of walking and perceptual
memory', Urban Studies, 49(15), p. 3269–3285
Diniz, N, Anderson, B, Liang, H-N and Laing, R 2012 'Map-
ping the Experience of Space', Proceedings of SIGraDi
2012, Brasil - Fortaleza, pp. 550-553
Facklam, F and Pecegueiro Curado, F 2011, 'Data Driven
Parametric Design', PARC Journal, 7, pp. 26ce:1-
26ce:26
Nold, C (eds) 2009, Emotional Cartography – Technologies
of the Self, www.emotionalcartography.net
[1] http://thesis.jmsaavedra.com

70 | eCAADe 32 - Towards Smarter Cities - Volume 1

Urban Place and Networked Data
Space, Content and Time

Trevor Hess1 , Eric Sauda2

1,2
University of North Carolina at Charlotte
1,2
{ahess5|ericsauda}@uncc.edu

This paper explores the relationship between physical space, and virtual networks
in the city. Topic modeling is utilized to provide robust descriptions of virtual
conversations occurring over city-wide social networks, and data driven events
are used to describe virtual projections of physical events. By comparing the
prevalence and pervasiveness of topics and events in the city, the paper seeks a
comprehensive understanding of the city as an increasingly interconnected entity.

Keywords: place, network, urban, topic modeling, events

INTRODUCTION tion produces a unique set of conversational topics.

Mobile information technology has transformed our
understanding of time and space. Cities have histor-
ically been spatial centers affording easy communi-
cation and exchange, but information networks have
altered this dynamic, problematizing the spatial role
of cities. While some theorists claim that information
technology intensifies urban activity, these debates
highlight a central concern of contemporary urban-
ism: as information technology changes our under-
standing of time and space, how will urban space be
transformed?
To study the transformation of urban space and
communication, researchers utilize geolocated Twit-
ter data to investigate the spread of ideas through
the city. This examination was conducted through
two studies, one comparing topics discussed in
unique urban places, and another comparing topics
discussed at different proximities to a sporting event
in the city.
The first "differential location" study, allows for
an analysis of topics that are linked to physical loca-
tions, and puts forth the hypothesis that each loca-
By analyzing correlations in the volume and content
of the data, we can identify global topics that occur
across all locations, local topics that are connected to
a physical place.
The second "big game" study examines the
spread of ideas from an event that occurs at a spe-
cific place and time. For this study we identified a
American football game that would garner signifi-
cant social media interaction citywide. These games
have widespread viewership both at the stadium, in
restaurants throughout the city and in the surround-
ing suburbs allowing for a complete understanding
of the way urban and rural conditions affect digi-
tal/virtual participation. By dividing the space sur-
rounding the stadium into proximal regions we could
study the correlations between distance from the
event, and the topics and micro events that city in-
habitants discussed.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 71

 PREVIOUS WORK
Urbanism and Technology
As early as 1964, the dynamics of communication
and urban spatial arrangements were reconceptual-
ized by Melvin Webber based on the emergence of
public communication (telephones, radio, television)
and interactions that transcends place Accessibility,
rather than propinquity, would become the genera-
tor of place. Communities based on interest are more
important than those based on physical adjacency,
in effect anticipating the emergence of virtual, web-
based communities.
Castells' theory of the 'space of flows' keeps
space central to urban form. But the flows of capital,
information, technology, organizational interaction,
and images create and define places uniquely suited
to accept and situate these flows. Thus, 'places do not
disappear, but their logic and their meaning become
absorbed in the network'.
Varnelis' recent studies describe how digital net-
works are reconfiguring our relation to place by en-
abling simultaneous presence in both physical and
networked place. These studies of urban places
such as cafés note both the use of the location and
the coincident use of laptops, smart phones and
tablets that connect people with extended, distant
networks, combining localized time and space with
the flow of information.
Social Media and Space
SENSEable City Lab [1] has developed real-time urban
monitoring system using cell phone data for the eval-
uation of spatial and temporal dynamics, displayed
by superimposing three-dimensional graphics over
urban maps. While this approach is not capable of
capturing content, it is one of the earliest examples
of simultaneously representing space and data.
An analysis of the spatial distribution of Four
Square data in three cities found that each city has
a unique distribution (Bawa-Cavis 2011). Each city
exhibited polycentric disbursement of activity hubs.
The spatial concept of unique arrangements of dis-
persed and clustered hubs is an effective visualiza-
72 | eCAADe 32 - Towards Smarter Cities - Volume 1
tion, but there is no use of either temporal or topical
information in this analysis.
Researchers have examined geo-located tweets
collected from peers attending a conference in San
Diego [2] . The researchers performed content and
spatial analyses on the data, mapping how individu-
als move from place to place in a city, and how to bet-
ter predict individuals' movements based upon their
location. In addition, the tweets were analyzed indi-
vidually for their content, in an effort to determine
the perceived "feel and taste" of different urban areas
around the city.
METHODOLOGY
Data Collection
To conduct a holistic study of urban data networks,
two approaches to study were selected. Data was col-
lected using the Twitter API, which allows for param-
eterized searches of past tweets.
The "differential location" study focused on place
and its significance to topics and events. Three
unique, similarly sized locations were identified
throughout the city. The places chosen: Uptown (the
city center), the Charlotte, North Carolina airport, and
Southpark Mall (a large shopping center) were hy-
pothesized to create unique conversations that could
be related to events occurring at the location.
The "big game" study identified a major Ameri-
can football event, a Carolina Panther's playoff game,
that was hypothesized to garner significant citywide
viewership and participation through social media.
The surrounding area was divided into proximal
regions that corresponded to the immediate area
around the stadium, and to the urban occupation
patterns present throughout the city (urban, subur-
ban, and rural). The boundaries for these circular re-
gions were as follows: one quarter, one half, one,
two and a half, five, ten and twenty-five miles. To
ensure that the data collected was relevant to the
event, the following hashtags: "BofAStadium", "Cam-
Newton", "GoPanthers", "KeepPounding", "NFLPlay-
offs", "PantherNation", "Panthers" and "SFvsCAR". All
hashtags were related to the slogans, the stadium
name, team names or players names, and were used
as search parameters.
Trend Analysis
A trend analysis was conducted on the data to pro-
vide a baseline comparison between locations in the
"differential locations" study and proximal regions in
the "big game" study. To study the amount of partic-
ipation the total volume of tweets was qualified for
each location and region. To judge the originality or
uniqueness of participation the number of retweets
were compared across locations and regions. Addi-
tionally, tweets containing an "@" sign (representing
tweets that were directed at other users) were also
counted as an approximation of social interaction
where users were communicating with each other.
By monitoring and comparing these three dimen-
sions of virtual participation, researchers were able to
analyze the change in methods of participation.
Topic Analysis
A topic analysis was carried out on both "differen-
tial location" and "big game" datasets using the Stan-
ford Topic Modeling Toolbox. This type of analysis
produces a set of topics (each a collection of words)
that appear together throughout the tweet docu-
ment collection. There are two phases to a topic anal-
ysis. First a model is trained to produce a set of top-
ics. This is done to a combined set of all documents
in the dataset. In this phase we combined the tweets
from all three locations, and all proximal areas and
trained a model for each dataset. The model was
trained for thirty topics to capture the full range of
topics that occurred in each dataset and was filtered
using a dynamic stop word list. It is important to
note that the number of topics is not calibrated to a
feature in the data, rather it is a number selected by
the researchers. For each location or region, promi-
nent topics that greatly contributed to conversations
were identified and compared. By comparing topics
that were present in different subsets of the data, re-
searchers were able to analyze and quantify overall
relationships.
Event Analysis
The event analysis began by identifying outliers in
the data, where an inordinate number of tweets were
produced over a finite amount of time (daily for "dif-
ferential location" and 15-minute intervals for the
"big game" study). These peaks were identified by
calculating the median standard deviation of each se-
ries, and then filtering time segments by the num-
ber of deviations away from the median. To en-
sure that events were significant outliers, only vol-
ume peaks greater than two deviations away from
median were identified as events. The median stan-
dard deviation was selected over the standard devi-
ation as a method for outlier identification for its ro-
bustness with data containing extreme outliers and
non-normal distributions. By comparing the topics
that contributed to this event peak, we are able to de-
termine whether events are related, even if they oc-
cur simultaneously.
FINDINGS
Differential Location
Trend Analysis. A preliminary analysis of trends at
each location revealed a strong correlation with pop-
ulation density. Of the total 31,300 tweets collected,
the Uptown location produced 67.9% of tweets, ap-
proximately three times the amount produced by
the Charlotte airport, and seven times the volume of
Southpark Mall. The difference can be explained by
the large concentration of skyscrapers and high-rise
buildings in the city center that serve as a place of
work and residence throughout the day and night.
This is contrary to the airport and shopping center,
which both have regular operating hours, and do not
have permanent around the clock residents leading
to a lower number of overall tweets.
An analysis of tweet composition revealed that
the percentage of retweets was highest at the air-
port and shopping center. In these two locations,
retweets were responsible for 31.2% and 32.5% of
tweets respectively. Uptown, retweets were only re-
sponsible for 17.1% of tweets. The most social in-
teraction (tweets containing the "@" sign) were also
Towards Smarter Cities - Volume 1 - eCAADe 32 | 73
 found at Southpark Mall and Charlotte airport rep-
resenting 64.6% and 68.2% of tweets respectively.
Uptown the percentage of tweets directed at other
users was significantly lower at 48.6%. The highest
concentration of tweets containing hashtags was Up-
town at 29.4%. At the airport and shopping cen-
ter, the number of tweets containing a hashtag was
20.9% and 21.4% respectively.
A comparison of these percentages (shown in
figure 1) reveal a correlation between the airport and
the mall in the way that people tweet. At these loca-
tions they are 20% more likely to direct their tweet at
another user and 8% more likely to include a subject,
or hashtag in the text. Alone, this information does
not provide insight into the content of each tweet,
but allows us to gain preliminary insight into relation-
ships that may occur in the data.
Topic Analysis. The topic model that was trained on
data from all locations identified discussions of Hal-
loween, a television show, and a concert taking place
in the city see table 1. While topics do not contain
explicit descriptions of their content, they contain a
set of words that are associated with a conversational
topic.
At the airport, three prominent topics were iden-
tified that represented between 6.3% and 10% of all
tweets from this location. Topic 5 refers to Johnson-
Wales University (a school in the city) and the Bob-
cats (a professional women's soccer team in the city).
74 | eCAADe 32 - Towards Smarter Cities - Volume 1
Topic 11 refers to the premier of a television show
called "girlcode" and includes terms that correctly
identify the time of the premier. Topic 26 refers to
the Brewz Music Festival, a popular event in the city.
Topics 5 and 11, two of the dominant topics at the
airport, were prominent at the shopping center. To-
gether they represented 14% and 19.5% of tweets, re-
spectively. In Uptown, the most prominent topic was
26 at 5.1% of all tweets, and was shared with the air-
Figure 1
port.
Tweet volume over
In this topic data we again see a correlation be-
time at the
tween the airport and shopping center, first identi-
Charlotte Airport
fied during the trend analysis. At these two locations
prominent topics are responsible for a much greater
share of total tweets than they are at the city center.
Uptown, were tweets evenly distributed among top-
ics, leading to a low number of topics that were re-
sponsible for a large number of tweets. This could
be a reflection of the greater diversity of users tweet-
ing from this location or a higher number of physical
events occurring at the city center.
Event Analysis. Events were identified at each loca-
tion by visually, and statistically analyzing the tweet
volume each day see figures 1-3. Each location pro-
duced a different set of events that occurred over
the course of the study period, and each event was
broken down into topics that contributed to the in-
crease in volume. At the airport three events were
identified on the following dates: 10/14/2014, 10/30,
and 10/31. The topics involved in events at the three
locations are detailed in table 2. Of this list of top-
ics, three were previously identified as prominent
topics at this location, and the 10/30 event corre- Table 1
sponds with the Halloween holiday identified in 9. Prominent
At the shopping center three events were identified "differential
on the dates 10/20, 11/3 and 11/4. Three of the top- location" topics and
ics that contributed to these events were previously their associated
identified as prominent overall topics at this location. terms
Since none of these events occur simultaneously with
events at the airport, the possibility that events at
one location influence the events at the other loca-
tion is ruled out. Uptown, three events were iden-
tified on the dates 10/17, 10/20 and 10/31. Of the
 three events, the event on 10/20 occurs simultane-
ously with an event at the mall and the event on
10/31 with an event at the airport. While there is no
correlation between the topic discussed on the 10/20
event, the 10/31 event contains topic 9, which dis-
cusses Halloween and is the only significant event in
the dataset that occurs simultaneously at another lo-
cation and contains at least one similar topic.
Figure 2
Tweet volume over
time at Southpark
Mall
Figure 3
Tweet volume over
time at the Uptown
location
Table 2
Prominent topics in
each
location-based
event
Figure 4
Distribution of
tweet volume
across proximal
regions Two minor events were identified Uptown, that
occurred on the same days as Southpark Mall. The
events contain similar topics 7 and 26 which discuss
a concert venue and the Brewz Music Festival, but
these events were not statistically significant, and
thus, cannot be considered proof of similarity.
From this analysis we identified major events oc-
curring at each location. Two events occurred simul-
taneously with an event at another location, but only
one was identified as being caused by a similar topic
or physical event. While this cross-location event
is evidence of a citywide global event, most events
identified at each location were local events and did
not influence tweet volume at other locations.
Big Game
Trend Analysis. In total 71,650 tweets were collected
during the week surrounding the game, 60,000 of
which were produced during the game. The distri-
bution of tweet volume in each proximal region is
shown in figure 4.
Of the data collected, 90% of the tweets origi-
nated from the region two and a half to five miles
from the stadium. The second most active region
was ten to twenty-five miles from the stadium. When
the tweet volume was scaled according to the area
of each proximal region, the two and a half to five
mile range still represented an inordinate amount of
the tweets, but areas closer to the event are shown
to be more significant as their tweet production per
square mile was closer than the overall volume ini-
tially suggested. The lowest percentage of retweets
are present in the .25 to .5 mile radius surrounding
the stadium and in the 10 to 25 mile area with 22.3%
and 28.1% of the tweet being unoriginal. Between,
.5 and 10 miles from the stadium the percentage of
retweets is between 47.5% and 49.9%. The highest
percentage of retweets is highest at the stadium at
73.8%.
This suggests that people who are physically in-
volved in the event do not fully participate in the vir-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 75

tual, data-driven event by writing comments about
events during the game, and remote users play a
much more significant role in the virtual event net-
work. The role of remote network participation can
also be seen in the percentage of tweets directed at
other users. At the physical location of the event,
65.2% of tweets are directed at other users. This fig-
ure is opposed to the percentage of directed tweets
in the one-half to one mile region and two and a half
to five mile region surrounding the stadium. In these
regions 80.9% and 91.9% of tweets are directed at
other users.
Topic Analysis. For each proximal area, we identified
topics that inordinately contributed to the region's
overall conversations see table 3.
Then to gain insight into high-level relationships
between regions, topics that were prominent in mul-
tiple areas were identified. In this identification pro-
cess and region comparison, topic 11 was present in
six out of the seven areas, topic 3 was present in five
areas, and topics 1 and 21 were present in 4 out of
the seven areas. Each area contained four to eight
76 | eCAADe 32 - Towards Smarter Cities - Volume 1
Table 3
Prominent "Big
Game" topics and
their associated
terms
prominent topics except for the area within a quar-
ter and half a mile of the stadium. Here, topics were
evenly distributed across all topics and a prominent
topic could not be identified. The greatest diversity
of prominent topics was seen furthest from the sta-
dium at the range of 10 to 25 miles from the sta-
dium. This range contained eight prominent topics,
and contained three topics that were not prominent
in any other area (13, 25, 28). Across all areas, promi-
nent topics were responsible for between 6.5% and
19.9% of all tweets.
For areas within a 2.5 mile radius of the stadium
there was a strong correlation between the topics
discussed. Here, topics 3, 7, 11, and 21 were present
in all areas as prominent topics (excluding the 0.25-
0.5) mile range. On average these prominent topics
are responsible for 13.9% of all tweets, within a range
of 7.8% to 18.7%. In areas 2.5 to 25 miles from the sta-
dium only topic 11 is prominent in all three regions,
and only three topics, 1, 3, and 4 are prominent in two
areas. In total, there were 12 prominent topics across
all three regions (between 2.5 and 25 miles). This sug-
Table 4
Prominent topics in
each micro event

gests that as the distance from an event is greater, the ing the event overall. Local events on the other hand
diversity of the conversation increases, while conver- were either unique to a particular area, or appeared
sations close to the event more focused. infrequently in different proximities, and did not ap-
Event Analysis. On the day of the sporting event, pear in simultaneously with events in other areas.
Figure 5 daily tweet volume increased from an average of
Tweet volume over 1,000 tweets to 60,000 tweets. This extreme outlier
time 0-0.25 mile in the data confirms the existence of an event, and
from the event allows for a finer grained analysis occurring on small
time scales. To study the micro events, that together
adds up into the large daily event, the data was orga-
nized into time segments of fifteen minutes. In this
Figure 6 fine grained analysis, a total of 110 statistically signif-
Tweet volume over icant events were identified across all seven proximal
time 0.25-0.5 mile regions in 32 different time segments. A sample of
from the event the events an their contributing topics can be seen
in table 4. The distribution of tweet volume through-
out the day in each region is shown in figures 5-11.
Within a range of a half mile of the stadium only 5
Figure 7
events were identified, and the number of events in-
Tweet volume over
creased with the distance from the stadium. In the
time 0.5-1 mile from
32 time segments that produced events, 23 were re-
the event
lated to another proximal area by the major topics
that contributed to the event. This shows a strong
correlation between events in different proximal re-
Figure 8 gions over the course of a single event.
Tweet volume over Of the topics that appeared across different prox-
time 1-2.5 miles imal areas in coordinated events, only topic 10 was
from the event not found in the list see table 3. This observation
brings into view two types of topics involved in
events. Global topics, identified as topics that are
present across multiple proximal areas were promi-
nent throughout the dataset and important describ-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 77


CONCLUSION
In our analysis of location-based and event-centric
tweet collections we discovered two spatial scales
that characterized tweets in each study, the local/-
tangible scale and the global virtual scale. These two
distinct scales describe events as either: spatial and
persistent, at a particular location, or as virtual con-
structs that occur on a city-wide or national scale dur-
ing an event.
In the study of urban place, a unique set of top-
ics dominated the discussion at each location. While
there was some similarity between the dominate top-
ics at each location, these topics were not simulta-
neously involved in events, and thus little to no cor-
relation was found. In rare cases, a statistically in-
significant event was identified that related in both
its time of occurrence and contributing topics. The
insignificant correlations may provide some insight
into events occurring between places, but without
being a reoccurring or significant feature of the data
it cannot be claimed as conclusive evidence.
In the study of urban events, a set of topics dom-
78 | eCAADe 32 - Towards Smarter Cities - Volume 1
inated across all study areas. These prominent topics Figure 9
were present throughout the study period and ap- Tweet volume over
peared as significant contributors to all proximal ar- time 2.5-5 miles
eas. While these global/persistent topics dominated from the event
social interaction throughout the city, distance to a
physical event had some impact on conversation top-
ics that were unique to a proximal area. Distance to
the event also had a profound impact on the volume Figure 10
of tweets, and the originality of participation. Here, Tweet volume over
there is conclusive evidence that proximity to a phys- time 5-10 miles
ical event has an impact on the method of virtual in- from the event
teraction. Close to the event, people are physically
engaged in the physical event and are less likely to
contribute original ideas. As distance increases and
event participants are removed from physical inter- Figure 11
action, they increasingly rely on virtual networks to Tweet volume over
remotely participate in the event. time 10-25 miles
Through these two studies, we believe that top- from the event
ics and events are uniquely useful conceptual struc-
tures for understanding a networked city. Together
they allow for a combination of media, participant,
and spatial relationships to be studied in a way that
avoids categorical and false equivalence errors. Us-
ing concepts from Latour's actor network theory, we
believe it is possible to describe topics and events in
the city in a more fulsome manner.
REFERENCES
Bawa-Cavia, A 2011 'Sensing the Urban: Using Location-
Based Social Network Data in Urban Analysis', Work-
shop on Pervasive and Urban Applications, San Fran-
cisco, CA
[1] http://senseable.mit.edu/papers/pdf/ CalabreseR-
atti2007SCLWorkingPaper.pdf
[2] http://www.designeresearch.com/urban-digitalism.pdf
Multi-Perspective Urban Optioneering
Patrick Janssen1 , Rudi Stouffs2
1,2
National University of Singapore
1
patrick@janssen.name 2 stouffs@nus.edu.sg

This paper investigates the state-of-the-art with respect to simulation-based

planning support systems in order to draw a set of requirements and best
practices for an urban planning and design framework that enables multiple
stakeholders with differing perspectives to systematically explore design options,
leveraging the latest analysis and simulation techniques. From these
requirements and best practices, the foundations and structure of such an urban
planning and design framework are developed. A number of technological and
methodological challenges are identified for future investigation.

Keywords: Urban planning and design, optioneering, simulation-based planning

support systems

INTRODUCTION tasks (Batty 1995; Klosterman 1997). They usually

With the increasingly complex nature of future mas-
ter plans, there is a need for comprehensive urban
planning and design frameworks that leverage the
latest analysis and simulation techniques in order to
enable planning and design options to be systemat-
ically developed, evaluated, and analysed. In order
to be effective in supporting collaborative decision-
making, such frameworks must concurrently address
the varying spatial and temporal scales of urban plan-
ning and the varying requirements and interests of
different stakeholders. Currently, no such frame-
works exists and, as a result, collaborative decision
makers are unable to effectively reconcile their dif-
ferent viewpoints and objectives and only a very lim-
ited number of suboptimal planning options are ex-
plored using mainly ad-hoc manual approaches that
are time-consuming and error-prone.
Such frameworks fall under what are referred to
as Planning Support Systems, which are systems fully
dedicated to support and improve the performance
of those involved in undertaking specific planning
consist of a combination of planning-related theory,
data, information, knowledge, methods and instru-
ments that take the form of an integrated framework
with a shared graphical user interface. They are typ-
ically used to provide projections forward to some
point in the future or may involve some estimation
of the impacts that result from some form of devel-
opment (Geertman and Stillwell 2003; 2009).
This research focuses on planning support sys-
tems that enable stakeholders to apply complex
types of simulation at varying spatial and temporal
scales. Below, existing simulation-based planning
support systems are reviewed and an approach re-
ferred to as 'Urban Optioneering' is described. A
framework for simulation-based planning support
systems is then proposed, consisting of a method
and a platform. Finally, the discussion section high-
lights a number of technological and methodologi-
cal challenges that have to be addressed in order to
achieve and implement the proposed framework.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 79

 SIMULATION-BASED PLANNING SUPPORT
SYSTEMS
Current urban planning and design practices typ-
ically use off-the-shelf impact-analysis software,
mainly based on point scoring systems that use
checklists and other methods that do not provide
adequate feedback for planning sustainable envi-
ronments. For example, AECOM's Sustainable Sys-
tems Integration Model (SSIM; Brown and Kellen-
berg 2009) is a method and tool for analysing urban
plans using a set of key sustainability indicators. The
method requires a small number of alternative plans
to be developed, and then allows users to make se-
lections for a variety of predefined systems and op-
tions, and to interactively see a cost-benefit analysis
based on these indicators. However, the indicators
rely on relatively simplistic calculations performed in
MS Excel. For non-spatial indicators (such as energy
use, water use, carbon emissions, and development
costs), the calculations make extrapolations based
on representative data defined by experts. For spa-
tial indicators (such as connectivity, access to local
services, and access to transit), the analysis consists
mainly of various types of network analysis that do
not take into account the complex dynamics of urban
environments.
Even when adopting planning support systems
that have the ability to run more advanced types
of simulations, the feedback they provide focuses
on isolated impacts of land-use planning; the inter-
dependencies between different aspects as well as
holistic overall considerations have to be done 'by
hand' by the planning teams. For example, one of the
more accessible systems is the CommunityVIZ plan-
ning and analysis platform (Walker and Daniels 2011),
available as an extension to ArcGIS. CommunityVIZ
has a comprehensive modelling framework that in-
cludes a set of built-in models for the dynamic sim-
ulation of complex urban phenomena, and also sup-
ports the ability to plug-in custom models defined by
end users. However, the core of the modelling frame-
work is proprietary and as a result the interactions
between the various models cannot be controlled by
80 | eCAADe 32 - Towards Smarter Cities - Volume 1
end-users. Modelling complex phenomena by creat-
ing networks of linked models is therefore not possi-
ble.
The need for more comprehensive methods that
integrate simulation-based planning support sys-
tems is increasingly recognised and accepted by
practitioners. The complexity of planning urban de-
velopment arises by the interaction of its compo-
nents, and the fact that these interactions can lead to
unexpected, counterintuitive results. Unfortunately,
most of the existing solutions oversimplify such inter-
action.
This insight has led to numerous recent efforts in
the development of systems that integrate a number
of domain specific models (e.g., land-use, transporta-
tion and energy-supply). However, these mono-
lithic systems hard-code these domain specific mod-
els in ways that are not easily modified by end-
users, thereby essentially limiting the use of these
systems to a very narrow range of planning ques-
tions. SynCity system (Keirstead et al. 2009) imposes
an energy perspective; CitySim system (Robinson et
al. 2009) imposes a resource-flow perspective; and
UrbanSim system (Waddell 2002) imposes a trans-
port and land-use perspective.
In addition, these types of systems can typically
only be applied within a narrow range of scales,
thereby hindering a truly multi-scale approach to ur-
ban design. Such monolithic models reach the lim-
its of feasibility and practical usability due to huge
data demand, limited life span and costs of mainte-
nance (Conway and McClain 2003; Davis and Ander-
son 2004). Most of the academic approaches lack the
continuity and support to achieve a level of maturity
and industrialization that can make it usable for prac-
titioners.
Multi-Perspective Optioneering
Optioneering is a collaborative decision making
methodology that systematically explores a wide
spectrum of options early on in the design process
by iteratively developing, evaluating, and analysing
alternatives (Holzer and Downing 2010, Gerber at al.
 2012). Urban optioneering applies this methodology tions of single-perspective models (Lane 2006; Morin
at an urban scale, allowing urban planning and de- 2005; Mikulecky 2001). Single-perspective models
sign options to be developed in a more rigorous man- are reducing the complexity to certain aspects and in
ner. this way are losing relations outside the partitioned
In the design of complex engineering systems, frame, which cannot be brought back to life, even
optioneering has become an important tool to guide when different models are applied in parallel to each
the interaction between different experts, to test the other. According to Seck and Honig (2012) a major
consistency of proposed technical solutions and to reason for these limitations is caused by the fact that
evaluate their impacts. Variants of the optioneer- simplified hierarchical structures of technical systems
ing approach, such as virtual prototyping, have been are imposed on non-technical systems and thus in-
applied in many fields of design and engineering, herit "a strong reductionist world view".
from the design of airplanes and cars to the design To overcome the principal limitations of single-
of complex building constructions (El Khaldi et al. perspective models, the development of multi-
2010). Leading architectural offices and develop- perspective models is widely discussed (Seck and
ment companies are using optioneering in their col- Honig 2012; Kingston 2007; Frank 2002). Although
laborative design processes, enabling highly special- the discussion until now did not result in concrete
ized experts from all over the world to contribute to applications, various proposals have been made at a
the design solutions. In this context optioneering has conceptual level.
been proven as a model that supports complex deci-
Figure 1
sion making.
Modelling complex
For the application of optioneering at the urban
phenomena
level, methods and systems have to be developed
through multiple
that are capable of engaging with the high levels
perspectives (After
of complexity inherent in urban planning and de-
Seck and Honig,
sign. Urban planning and design processes have to
2012).
integrate economic, environmental and social issues,
have to cover different scales in space and time, and
have to address multi-actor environments.
These processes therefore require various per-
spectives (e.g., financial, environmental, social, oper-
ational) to serve the different actors involved. Fur-
thermore, due to the fact that urban plans are typi-
cally related to long-term decisions, they also have to
face changing conditions and demands, with unfore- Seck and Honig (2012) propose a complex sys-
seen and unexpected phenomena being the norm. tem (the natural system) as an expansible collection
From that point of view, current modelling and sim- of perspectives, where each perspective, although
ulation approaches are lacking in open system think- associated with its own formal aspect system, is re-
ing and "the capability of covering multiple system lated to "multiple non-isomorphic decompositions
perspectives at once and in different levels of details that may influence each other" (see Figure 1). In this
to provide better understanding of the systems" (Tek- way, the fixed hierarchy is replaced by flexible modu-
inay et al. 2010). larity; different formal aspect systems are connected
In the scientific debate about modelling of com- by bridges, involving different steps of decoding and
plex systems many authors emphasize the limita- encoding. Such multi-perspective modelling is in-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 81

 tended to "capture the tangledness of the systems Figure 2
that result when we observe the world from different The proposed
perspectives" (Seck and Honig 2012). Urban
The challenge is to elaborate the theoretical con- Optioneering
cept of multi-perspective urban optioneering into a Framework
practical framework for urban planning and design consisting of an
that reflects the multi-actor environment of the ur- optioneering
ban system. method and an
optioneering
PROPOSED FRAMEWORK platform.
One of the fundamental characteristic of multi-
perspective optioneering is that it is impossible to
predict in advance the perspectives that will be most
relevant to the stakeholders involved. This unpre-
dictability results from the fact that urban plan-
ning and design is fundamentally an unstructured
or 'wicked' process characterised by (1) multiple ac-
tors with differing, legitimate values and opinions; (2)
high uncertainty; (3) aspects of irreversibility; (4) no
clear solutions; (5) being fraught with contradictions;
(6) being persistent and unsolvable (Rutledge et al.
2008). Any optioneering framework that is a closed
system, based on the assumed relevance of a prede-
fined set of perspectives, is therefore bound to fail.
In order to support multi-perspective optioneer-
ing, a radically different type of framework is pro-
posed that is inherently open (Axelos 2006). The
openness of the framework hinges on enabling
stakeholders to define their own customized mod-
els that reflect the perspectives that are most rele-
vant to them, and to the scenario and context being
Urban Optioneering Method
explored. These customized models are defined as
An Urban Optioneering Method is proposed that en-
networks of loosely coupled components, including
ables questions relating to specific urban planning
a variety of domain specific simulation engines and
and design scenarios to be explored using multi-
data sets (Altintas 2011; Deelman at al. 2008; Curcin
perspective models. The perspectives can be related
and Ghanem 2008).
to the tasks of different planning agencies, to specific
The proposed framework consists of two parts:
interests of different stakeholders or to different po-
an Urban Optioneering Method and an Urban Op-
litical, social, economic or environmental priorities.
tioneering Platform. The method and the platform
For each question to be explored, customized
are shown in Figure 2, and will be described in more
models reflecting the differing perspectives need to
detail in due course.
be created, and the outcomes of those models need
to be explored. The method consists of an adaptive-
iterative process consisting of two nested loops: the

82 | eCAADe 32 - Towards Smarter Cities - Volume 1

adaptive loop and the iterative loop (see Figure 2).
The adaptive loop comprises three key activities:
defining questions, creating models, and exploring
outcomes.
• Urban questions are defined, possibly by dif-
ferent stakeholders in the planning process,
with differing and possibly conflicting con-
cerns.
• Urban models are created as executable
models, consisting of coupled legacy single-
domain urban models and data sets.
• Urban outcomes are explored by iteratively
executing models and analysing results, in
particular, including cross-perspective analy-
sis of data and result inconsistencies.
The iterative loop is part of the process of exploring
model outcomes. For any given multi-perspective
model, many variants can be explored based on
differing assumptions and using differing data sets.
These outcomes may involve the estimation of im-
pacts on the present and projected into the future
based on differing assumptions and data sets. Fur-
thermore, the results from running model variants
can be investigated using a variety of data analytics
techniques. This process of iterative exploration will
result in the accumulation of evidence and will culmi-
nate in actionable feedback, which may then trigger
new or modified questions to be posed.
The proposed method enables stakeholders to
ask complex questions. For example, the question
may be posed: to what extent would the electri-
fication of road transport reduce air conditioning
use in residential flats? Since electric cars are qui-
eter and cleaner, they will produce less noise and
air pollution, thereby resulting in more people opt-
ing for natural ventilation over air conditioning. The
model would have to include a complex set of do-
main specific tools and data sets. First, the behaviour
of residents within flats may be predicted using a be-
haviour model, with input requirements that include
localised environmental conditions (including tem-
perature, humidity, and ventilation) and localised
pollutant levels (including noise and air pollution). In
order to predict the environmental conditions, en-
vironmental simulations would be required, using
weather data and detailed urban models. These sim-
ulations may also include micro-climate simulations,
such as heat island simulations. In order to predict
the pollution levels, noise mapping and pollutant
transport simulations would be required. Both the
environmental simulations and the pollutant simula-
tions may in turn require data from other simulations,
such as traffic simulations and wind simulations. This
original question therefore results in a complex cas-
cading network of simulation engines and data sets.
Urban Optioneering Platform
An Urban Optioneering Platform is proposed that
will enable stakeholders to fluidly build and explore
computable multi-perspective models, consisting of
loosely coupled legacy simulation engines and data
sets. This process of exploration includes the abil-
ity to analyze and compare various partial models at
varying temporal and spatial scales, thereby allowing
conflicts and inconsistencies to be discovered.
Models are divided into three layers: compo-
nents, networks, and dashboards.
• Model components are the basic elements
from which models are built and may include
legacy simulation engines, data sets, and data
mappers. Such components are central to the
modularity feature of the platform, as they al-
low additional components to be added to
the platform as they become available and rel-
evant.
• Model networks are executable networks that
reflect sets of values and beliefs specific to the
perspectives being applied. Such models are
created by coupling together selected sets of
model components, including legacy simula-
tion engines.
• Model dashboards are customised graphical
Towards Smarter Cities - Volume 1 - eCAADe 32 | 83
 user interfaces and data-mashups associated
with one or more models, where each model
may reflect a different perspective. Such
dashboards will allow input and output data
from multiple perspectives to be analysed in
an integrated environment. At its simplest,
this may consist of some sliders for defining
input data and some graphs and charts for dis-
playing output data. However, it may also in-
clude complex spatiotemporal data manipu-
lation and data analytics.
The Urban Optioneering Platform consists of two ap-
plications: one application for building models and
another application for exploring models. They are
defined as distinct applications since they represent
fundamentally different modes of working, requiring
different skill sets.
The Model Builder application provides tools
building all three layers of a model. For building
model components, a set of tools is required to help
users to wrap existing computational objects such as
legacy simulation engines, data sets, and data map-
pers. The resulting components will be archived as li-
braries of components to be embedded within larger
models. For building model networks, a set of tools is
required to help users to create multi-domain model
networks from selected sets of components. These
tools will allow models to be visually constructed as
a network of components interconnected by wires.
Features such as advanced type checking and de-
bugging will need to be provided to help users build
valid models. For building model dashboards, a set
of tools is required to help users build customized
dashboards from predefined user interface building
blocks using visual drag-and-drop techniques.
Once a multi-perspective model has been built,
the outcomes of the model then need to be ex-
plored. The Model Explorer application is conceived
as a cloud-based application for deploying, execut-
ing, and managing multi-domain, multi-perspective
models. On the front end, the application provides
a graphical user interface for deploying models. On
the back end, the application provides automated
84 | eCAADe 32 - Towards Smarter Cities - Volume 1
scheduling and data management procedures for ro-
bust fault-tolerant parallel execution of models and
data analytics tasks.
The Urban Optioneering Platform aims to radi-
cally improve the way with which stakeholders are
able to leverage the latest computational tools and
techniques to explore critical questions that impact
decisions in urban planning and design. The system
enables diverse stakeholders to gather the evidence
required to take positions, which can then be used as
a basis for further discussions and negotiations.
DISCUSSION
In order to implement the proposed Urban Option-
nering Framework, a number of technological and
methodological challenges need to be tackled. Be-
low, three fundamental challenges are discussed:
building models; linking components; and making
decisions.
Challenge 1: Building models
The challenge of model building focuses on develop-
ing a system that enables stakeholders to build com-
plex models from a set of re-usable modular compo-
nents
An approach needs to be developed that does
not require advanced technical skills, as stakehold-
ers, such as urban planners and designers, cannot
be expected to have such skills. Previously, a dis-
tinction was made between two modes of working:
building a model and exploring a model. As a result,
it might be suggested that the stakeholders should
refrain from getting involved in the model building
process, which might be better left to people with
more advanced technical skills. However, the pro-
cess of building models cannot be subcontracted, for
two reasons: the people building models need to un-
derstand the urban planning and design issues; sec-
ond, the people exploring models need to under-
stand the technological issues, such as the limits and
constraints of the models being explored. It is there-
fore important to try and minimize the gap between
building models and exploring them.
 One approach to achieving this is to create a
set of tools that allow stakeholders to be directly in-
volved in the building of models from predefined
components. With this approach, a distinction is
made between building model components and
building model networks. Model components, in-
cluding various legacy domain-specific analysis and
simulation programs and data sets, are built and
tested by researchers who have the technical skills
and are specialists in their corresponding field of sci-
ence. Model networks are then built by stakeholders
by assembling predefined components using graph-
ical interfaces that do not require any programming
or other advanced technical skills.
With regards to building model networks, two
approaches can be identified, referred to as tightly
coupled component models versus loosely coupled
component models. Tightly coupled component
models are developed in a standard programming
language using a set of modular programming li-
braries for different single-domain models and the
resulting program is then compiled into a single ex-
ecutable. (For example: Leavesley et al. 1996; Dah-
mann 1997; Watson et al. 1998; Rizzoli et al. 1998;
Reed et al. 1999; Krahl 2000; David et al. 2002;
Voinov et al. 2004; Rahman et al. 2004; Kolbe at
al. 2005; Ahuja et al. 2005; Müller 2009). Loosely
coupled component models are developed by cou-
pling various analysis and simulation programs at the
data level. These components are typically legacy
programs developed as stand-alone executables that
are wrapped in order to enable data exchange via
input and output files (For example: Sydelko et al.
2001; Babendreier and Castleton 2005; Bernholdt et
al. 2006; Fortube et al. 2008; Tan et al. 2012). The
problem with the tightly coupled approach is that it
still requires a significant amount of software engi-
neering and programming knowledge. The loosely
coupled approach is therefore seen as being prefer-
able.
Challenge 2: Linking components
The challenge of linking components focuses on how
to link together domain-specific analysis and simula-
tion programs.
An approach needs to be developed that allows
components to be interactively linked in complex
ways. One promising approach is scientific workflow
systems (Deelman et al. 2008, Altintas 2011; Toth et
al. 2012). Such systems exhibit a common reference
architecture that consists of a graphical user interface
(GUI) for authoring workflows, along with a work-
flow engine that handles invocation of the applica-
tions required to run the solution (Yu and Buyya 2005,
Curcin and Ghanem 2008). Nearly all workflow sys-
tems are visual programming tools in that they allow
processes to be described graphically as networks of
nodes and wires that can be configured and recon-
figured by users as required (McPhillips 2009). Nodes
may represent analysis or simulation programs, while
wires represent the flow of data, linking an output of
one node to an input of another node.
With regards to the dataflow between such
nodes, a more difficult problem is the interoper-
ability issues that invariably exist between various
analysis and simulation programs (Janssen et al. In
press). Such programs typically require independent
domain-specific data models that are efficient within
the domain, but are difficult to share across domains.
Existing interoperability efforts such as IFC, CityGML,
and gbXML attempt to bridge domains, and there are
also efforts to extend or amalgamate such schemas
to improve generality. However the challenge is non-
trivial given the large set of domains that may be rel-
evant.
In the domain of urban planning and design,
such interoperability issues include in particular the
ability to step up and down between different spa-
tial and temporal scales of the data (also known as
multi-resolution modelling). In general, it is feasi-
ble to overcome such incompatibilities using a range
of data aggregation and compensation techniques
(Reynolds et al. 1997). For example, data aggre-
gation may involve combining sets of data at the
Towards Smarter Cities - Volume 1 - eCAADe 32 | 85
 precinct level in order to characterize the neighbour-
hood, while data disaggregation may do the reverse
by taking data at the neighbourhood level and ap-
portioning it to precincts based on various heuris-
tics. Data compensation comes into play when cer-
tain data sets are missing or deficient. For exam-
ple, when aggregating data from the precinct level to
the neighbourhood level, missing data may need to
be synthetically generated based on typical patterns
and distributions.
Finally, an important element of the proposed
framework is the development of bridges that con-
nect distinct model networks, for example by using
the outputs of two or more perspectives as inputs for
other perspectives, in this way replacing the fixed hi-
erarchical order of a single perspective model by flex-
ible modularity.
Challenge 3: Making decisions
The challenge of making decisions focuses on how
models can be used to support decision making in
urban planning and design.
An approach needs to be developed that cre-
ates models that have the potential to generate ac-
tionable feedback. At a building scale, this can
be achieved by analyzing complete design options.
Due to the fact that design constraints and per-
formance targets are often relatively well defined,
actionable feedback can be generated by various
techniques for optimizing and ranking alternative
options, such as Evolutionary Multi-objective Opti-
mization (EMO), Multiple-Criteria Decision Analysis
(MCDA), and Multi-disciplinary Design Optimization
(MDO). However, at an urban scale this may not be
feasible due to the fact the complexity of urban prob-
lems means that reductive techniques for optimizing
and ranking are not applicable, which in turn leads to
feedback that is often ambiguous and contradictory.
In order to be able to manage the complexity,
urban optioneering methods may need to focus on
questions that are more narrowly defined, but that
have a direct impact on how design options are de-
veloped and that are still relevant to a range of dif-
86 | eCAADe 32 - Towards Smarter Cities - Volume 1
fering perspectives. It is proposed that through a
more complex process of modelling and counter-
modelling (Greenberger at al. 1976) using a variety of
different types of models reflecting distinct perspec-
tives, stakeholders will gradually gather the evidence
required to take positions on questions that directly
impact decision making in urban planning and de-
sign. An important part of this evidence is an analy-
sis of the inevitable conflicts and inconsistencies that
will arise between the different models.
CONCLUSION
The state-of-the-art in planning support systems was
presented in order to draw requirements and best
practices for the design and development of a frame-
work that supports the complexity of the urban plan-
ning and design process. The foundations and struc-
ture of such an urban planning and design frame-
work were drawn. Finally, the major challenges that
have to be addressed in order to achieve and imple-
ment this framework were identified as an agenda for
future research.
ACKNOWLEDGMENTS
The approach proposed in this research emerged out
of discussions within the Urban Prototyping Group
[1] at the National University of Singapore. The group
was set up in June 2011 as a cross-faculty collabo-
ration between the School of Design and Environ-
ment (SDE) and the School of Computing (SoC). The
group aims to develop collaborative methods and
systems for urban information and simulation mod-
elling, leveraging growing computing power and in-
creasing availability of urban metabolic data.
REFERENCES
Ahuja, LR, Ascough II, JC and David, O 2005, 'Developing
natural resource models using the object modeling
system: feasibility and challenges', Advances in Geo-
sciences, 4, p. 29–36
Altintas, I 2011, Collaborative Provenance for Workflow-
driven Science and Engineering, Ph.D. Thesis, Univer-
sity of Amsterdam
Axelos, K 2006, 'The world: being becoming totality', En-
vironment and Planning D: Society and Space, 24, p.
643–651
Babendreier, JE and Castleton, KJ 2005, 'Investigat-
ing uncertainty and sensitivity in integrated, mul-
timedia environmental models: tools for FRAMES-
3MRA', Environmental Modelling & Software, 20, p.
1043–1055
Batty, M 1995, 'Planning support systems and the
new logic of comp', Regional Development Dialogue,
16(1), pp. 1-17
Bernholdt, DE, Allen, BA and Armstrong, R 2006, 'A com-
ponent architecture for high-performance scientific
computing', Int. J. of High Performance Computing
Applications, 20, p. 163–202
Brown, I and Kellenberg, S 2009, 'Ecologically engineer-
ing cities through integrated sustainable systems
planning', Journal of Green Building, 4(1), p. 58–75
Conway, RW and McClain, JO 2003, 'The conduct of an
effective simulation study', INFORMS Transactions on
Education, 3(3), p. 13–22
Curcin, V and Ghanem, M 2008 'Scientific workflow sys-
tems - can one size fit all?', Proceedings of CIBEC 2008,
Cairo, p. 1–9
Dahmann, JS 1997 'High level architecture for simula-
tion', Proceedings of First International Workshop on
Distributed Interactive Simulation and Real Time Ap-
plications, p. 9–14
Walker, D and Daniels, T 2011, The Planners Guide to Com-
munityViz: The Essential Tool for a New Generation of
Planning, APA Planners Press
Davis, PK and Anderson, RH 2004, 'Improving the com-
posability of DoD models and simulations', Journal
of Defense Modeling and Simulation, 1, p. 5–17
Deelman, E, Dannon, G, Shields, M and Taylor, I 2008,
'Workflows and e-science: an overview of workflow
system features and capabilities', Future Generation
Computer Systems, 25, p. 528–540
Fortube, D, Gijsbers, P, Gregersen, J and Moore, R 2008,
'OpenMI: real progress towards integrated mod-
elling', in Abrahart, RJ, See, LM and Solomatine,
DP (eds) 2008, Practical Hydroinformatics, Springer,
Berlin, p. 449–464
Frank, U 2002 'Multi-Perspective Enterprise Modeling
(MEMO): conceptual framework and modeling lan-
guages', Proceedings of HICSS35, p. 1258–1267
Geertman, S and Stillwell, J 2003, Planning Support Sys-
tems in Practice, Springer, Heidelberg
Geertman, S and Stillwell, J 2009, 'Planning Support Sys-
tems: Content, Issues and Trends', in Geertman, S
and Stillwell, J (eds) 2009, Planning Support Systems
Best Practice and New Methods, Springer, Dordrecht,
pp. 1-26
Gerber, D, Lin, SH, Pan, B and Solmaz, AS 2012 'Op-
tioneering: A New Basis for Engagement Between
Architects and Their Collaborators', Proceedings of
SimAUD 2012, p. 11:1–11:8
Greenberger, M, Crenson, MA and Crissey, BL 1976, Mod-
els in the Policy Process: Public Decision Making in the
Computer Era, Russell Sage Foundation, New York
Holzer, D and Downing, S 2010, 'Optioneering: A New
Basis for Engagement Between Architects and Their
Collaborators', Architectural Design, 80(4), p. 60–63
Janssen, P, Stouffs, R, Chaszar, A, Boeykens, S and Toth,
B In press 'Custom Digital Workflows with User-
defined Data Transformations via Property Graphs',
Proceedings of DCC'14, London
Keirstead, J, Samsatli, N and Shah, N 2009 'SynCity: an
integrated tool kit for urban energy systems mod-
elling', Proceedings of Fifth Urban Research Sympo-
sium, Marseille
El Khaldi, F, Ni, R, Culiere, P, Ullrich, P and Aboitiz, CT 2010
'Recent Integration Achievements in Virtual Proto-
typing for the Automobile Industry', Proceedings of
FISITA 2010
Kingston, J 2007, Multi-Perspective Modeling for Knowl-
edge Management and Knowledge Engineering, Ph.D.
Thesis, University of Edinburgh
Klosterman, R 1997, 'Planning support systems: a new
perspective on computer-aided planning', Journal of
Planning Education and Research, 17(1), pp. 45-54
Kolbe, TH, Gröger, G and Plümer, L 2005, 'CityGML: in-
teroperable access to 3D city models', in van Oost-
erom, PDP, Zlatanova, DS and Fendel, EM (eds) 2005,
Geo-information for Disaster Management, Springer,
Berlin, p. 883–899
Krahl, D 2000 'Extend: the Extend simulation environ-
ment', Proceedings of WSC '00, p. 280–289
Lane, D 2006, 'Hierarchy, Complexity, Society', in Pumain,
D (eds) 2006, Hierarchy in Natural and Social Sciences,
Springer, p. 81– 119
Leavesley, GH, Restrepo, PJ, Markstrom, SL, Dixon, M
and Stannard, LG 1996, The Modular Modeling Sys-
tem (MMS): User‘s Manual, United States Geological
Survey
Mikulecky, DC 2001, 'The emergence of complexity: sci-
ence coming of age or science growing old?', Com-
puters & Chemistry, 25, p. 341–348
Morin, E 2005, Introduction à la pensée complexe, Seuil,
Paris
Müller, JP 2009, 'Towards a formal semantics of event-
based multi-agent simulations', in David, N and Sich-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 87
 man, JS (eds) 2009, Multi-Agent-Based Simulation IX, elling & Software, 9, p. 285–304
Springer, Berlin, p. 110–126 Waddell, P 2002, 'UrbanSim: modeling urban develop-
Rahman, JM, Seaton, SP and Cuddy, SM 2004, 'Mak- ment for land use, transportation and environmen-
ing frameworks more useable: using model intro- tal planning', Journal of the American Planning Asso-
spection and metadata to develop model process- ciation, 68(3), p. 297–314
ing tools', Environmental Modelling & Software, 19, p. Watson, RA, Hornby, G and Pollack, JB 1998 'Model-
275–284 ing Building-Block Interdependency', Proceedings of
Reed, M, Cuddy, SM and Rizzoli, AE 1999, 'A frame- PPSN V, p. 97–108
work for modelling multiple resource management [1] http://sde.nus.edu.sg/rsh/UrbanPrototyping/SDE_-
issues: an open modelling approach', Environmental UP_index.html
Modelling & Software, 14, p. 503–509
Reynolds, P, Srinivasan, S and Natrajan, A 1997, 'Con-
sistency Maintenance in Multi-Resolution Simula-
tions', ACM Trans on Modeling and Computer Simula-
tion, 7(3), pp. 368-392
Rizzoli, AE, Davis, JR and Abel, DJ 1998, 'Model and data
integration and re-use in environmental decision
support systems', Decision Support Systems, 24, p.
127– 144
Robinson, D, Haldi, F, Kämpf, JH, Leroux, P, Perez, D,
Rasheed, A and Wilke, U 2009 'CitySim: comprehen-
sive micro-simulation of resource flows for sustain-
able urban planning', Proceedings of 11th Int. IBPSA
Conf., Glasgow, p. 1083–1090
Rutledge, DT, Cameron, M and Elliott, S 2008, 'Choos-
ing Regional Futures: Challenges and choices in
building integrated models to support long-term re-
gional planning in New Zealand', Regional Science
Policy & Practice, 1, p. 85–108
Seck, MD and Honig, HJ 2012, 'Multi-perspective mod-
elling of complex phenomena', Computational and
Mathematical Organization Theory, 18, p. 128–144
Sydelko, PJ, Hlohowskyj, I, Majerus, K, Christiansen, J and
Dolph, J 2001, 'An object-oriented framework for
dynamic ecosystem modeling: application for inte-
grated risk assessment', Science of the Total Environ-
ment, 274, p. 271–281
Tan, WC, Haas, PJ, Mak, RL, Kieliszewski, CA, Selinger, PG,
Maglio, PP, Glissman, S, Cefkin, M and Li, Y 2012
'Splash: a platform for analysis and simulation of
health', Proceedings of IHI 2012
Tekinay, C, Seck, M, Fumarola, M and Verbraeck, A 2010
'A context-based multi-perspective modeling and
simulation framework', Proceedings of WSC 2010, p.
479–489
Toth, B, Janssen, P, Stouffs, R, Chaszar, A and Boeykens, S
2012, 'Custom Digital Workflows: A New Framework
for Design Analysis Integration', International Jour-
nal of Architectural Computing, 10(4), pp. 481-500
Voinov, A, Fitz, C, Boumans, R and Costanza, R 2004,
'Modular ecosystem modeling', Environmental Mod-

88 | eCAADe 32 - Towards Smarter Cities - Volume 1

ICT-enabled Civic Empowerment and Participation: in
Design, through Design
Burak Pak1 , Johan Verbeke2
1,2
KU Leuven Faculty of Architecture
1,2
http://arch.kuleuven.be
1,2
{burak.pak|johan.verbeke}@kuleuven.be

This paper aims to discuss the potentials of novel modes of participatory design
in relation to the latest developments in Information and Communication
Technologies (ICT). The first part of the study involves the extraction of the basic
principles from the extraordinary cases of the Medical Faculty Housing by
Lucien Kroll (1976) and Cedric Price's Fun Place (1965) in which various forms
of ICT-enabled participation were conceived. In the second part, we reframe the
existing ICT tools and strategies and elaborate their potentials to support the
modes of participation performed in these two cases. As a result, by distilling the
created knowledge, we introduce a model of ICT-enabled design participation
which exploits a set of collective action tools to support sustainable ways of
self-organization and bottom-up design.

Keywords: Participatory architectural design, crowdsourcing, crowdfunding,

self-organization

INTRODUCTION planning practices in the USA as well as Team 10's re-

The concept of bottom-up participation was brought
onto the agenda of architectural design prominently
after the Second World War and gained traction with
the movements of 1968 (Jencks 2011).
Habraken (1961) was one of the first to intro-
duce the idea of spatial self-determination as a cit-
izen right. Habraken's concept of a supporting/en-
abling megastructure was not novel at that time;
it can be traced back to Nieuwenhuys's New Baby-
lon experiments between 1956 and 1978, Friedman's
Mobile Architecture in 1958 and Fuller's structural
designs (Lobsinger 2000). His ideas also resonated
with André Lurçat's ground-breaking participatory
design practices for the reconstruction of Maubeuge
in 1945 (Kroll 2014), Jane Jacob's (1961) critique of
volt against CIAM in 1959 and the Situationists (Math-
ews 2006).
In parallel with these contributions, in 1962, Wal-
ter Segal developed a low-cost housing solution suit-
able for self-build while trying to address his own
problem of providing a temporary home for his fam-
ily. This practice evolved into the development of
"The Segal Method" and several participatory design
projects in which 27 families worked with architects
to design and build their own homes (Broome 1995).
Following these developments, Bernard Rudof-
sky (1964) organized an influential exhibition in the
Museum of Modern Arts (MoMA) with the title "Ar-
chitecture without Architects". His work was a clear
criticism of the discipline and the role of the archi-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 89

 tect as an authoritative figure; a recurring theme in
Habraken's, Segal's and Team 10's contributions.
In this intellectual climate, director Joan Little-
wood commissioned Cedric Price to design an in-
formal and dynamic entertainment center: the Fun
Palace. This center has never been realized, but it is
still known as one of the most prominent participa-
tory design cases to inspire numerous architects; in-
cluding Richard Rogers' and Renzo Piano's Pompidou
Center (1976).
Around the same time period (1969-1978), Lu-
cien Kroll orchestrated the design of the Medical
Faculty of the Catholic University of Louvain (UCL)
with the student organization "La Maison Medicale-
La MéMé". The result was an ambitious piece of "anar-
chitecture", challenging every possible aspect of the
architectural practices of the time. It became an "icon
of democratic architecture" (Poletti 2010) for Kroll's
alteration of the usual hierarchical relationship be-
tween the architect and the user during the process
-and most importantly- the development of novel
design interventions to enable bottom-up participa-
tion.
In both of the projects referenced above (Fun
Palace and La MéMé), the designers attempted to
use various technologies and computational meth-
ods for augmenting participatory design processes.
However, the capacity and potentials of these were
limited at the time and Internet was non-existent.
Acknowledging this gap, this paper aims at the dis-
cussion of potentials of modes of participatory de-
sign in Price's and Kroll's projects in relation to the
latest developments in Information and Communica-
tion Technologies (ICT). In accordance with this aim,
the research questions to be explored in this paper
are:
1. What are the participatory design principles
of the Fun Place and La MéMé?
2. Which forms of ICT use were conceived in
these?
3. Can these serve a basis for a novel integrated
90 | eCAADe 32 - Towards Smarter Cities - Volume 1
participation model that incorporates state-
of-the-art ICT?
In this context the first part of our study involves the
extraction of the basic principles from the two refer-
enced cases. In the second part, building on the find-
ings of our case study, we reframe the existing ICT
tools and strategies in terms of space and time and
elaborate their potentials for supporting the revealed
modes of participation. In conclusion, by distilling
our findings, we introduce a model of ICT-enabled
design participation which exploits a set of digital
and non-digital collective action tools to support sus-
tainable ways of self-organization and bottom-up de-
sign.
A CASE STUDY OF KROLL'S LA MÉMÉ AND
PRICE'S FUN PALACE
In this section, informed by the introduced back-
ground, we will discuss the influential projects of
Kroll and Price with a focus on two main aspects: a)
the employed participatory design principles and b)
the ICT tools and strategies used to enable bottom-
up participation. The order of presentation repre-
sents the increasing complexity and relevance to ICT-
enabled participation.
Case 1: La MéMé
The project was initiated when the Catholic Univer-
sity of Louvain (UCL) decided to move its Medical Fac-
ulty to Brussels, Saint-Lambrechts-Woluwe. The uni-
versity authorities made an exceptional decision and
presented the preliminary design of the Medical Fac-
ulty Housing to the student committee. The students
rejected the project and contacted Kroll for his ser-
vices (Kroll 1987).
As a close follower of Pierre Bourdieu, Kroll ques-
tioned every aspect of the institutionalized practices
with the contributions of the spirited students of UCL.
He intended to create an open design process, "an ac-
tion open to new necessities and to decisions that are
always provisional and incomplete" (Kroll 1987). He
aimed at establishing an intellectual climate through
which a kind of friendly organization would emerge
 to result in a homeopathic kind of architecture (Kroll
2005).
Kroll organized meetings with the committees
and discussion groups. In these meetings he re-
ceived conflicting ideas. Instead of flattening out all
the differences of approach and attitudes he tried
to incorporate them into the design process (Kroll
1997). This was a creative refutation of the idea of
"consensus".
Throughout the project, the students were em-
powered to participate in two forms: getting in-
volved in the design process and through the partic-
ipation opportunities provided by the architectural
design per se (Figure 1).
Figure 1
Medical Faculty
Student Housing by
Lucien Kroll. The
users participated
in the decision
process and the
architectural
product enabled
them to shape and
reshape their Kroll developed a flexible structure system which he
surroundings. called "wandering columns" based on a loosely de-
fined grid. He collaborated with a professor of com-
puter engineering to manipulate the grid to support
the irregular and heterogeneous shell of the build-
ing. He designed the artificial ground around the
project to provide raw space for further development
(the aspects of wandering columns and his long-term
vision for expansion are more evident in the Alma
Metro Station which was built as an extension to the
project).
The "infill" -inherited from Habraken- is hypo-
thetically removable: demountable window frames,
moveable partitions and prefabricated sanitary units.
The architect used his own interpretation of the
Habraken's SAR module but refuted the idea of func-
tional zones (Kroll 1987). According to the princi-
ples of co-habitation, the infill can be torn down by
the users, which encourages them to take initiative
in planning and re-planning their environments. The
plan would always be incomplete.
In La MéMé, Kroll did not see aesthetics as the
central point of design. Through this project, he
strongly criticized what he called the "easel archi-
tecture": aesthetically pleasing but isolated from the
people, culture and community.
ICT tools and strategies for participation in
Case 1
In his book "Architecture of Complexity", Kroll (1987)
reserved a whole chapter to the computers. Instead
of computer aided design (CAD), he suggested com-
puter use in design (CUD) as a more appropriate term
for describing his vision. He stressed the importance
of open-endedness and heavily criticized the inflexi-
ble artificial intelligence practices of the time which
led to self-contained, closed and repetitive results.
In contrast, Kroll envisioned the drafting soft-
ware as a potential tool that allows open-endedness
through which the architectural product and the so-
cial relationships can be involved in the design and
manufacturing process. However, the communica-
tion technologies were not developed enough to
fully realize the social part of the potential.
The computer-based social interactions he fore-
saw were limited to three dimensional drawings
which he found useful for the communication of early
ideas to the inhabitants. He suggested that infinite
interactions were required to deal with the infinite di-
versity of the real world.
In close contact with the users, he employed vari-
ous algorithms to create diversity and differentiation
and presented a library of components that can be
combined according to user needs (Kroll 1987). He
tested a computational method -anthromorphism-
to allow a type of architecture with variant building
programs and devised the role of the architect as a
developer of "types" which can be varied by the in-
habitant. The diversity of the outcomes was unman-
ageable due to the technological limits of the time.
As a result he worried that the process would lead to
the Taylorist practices that he criticized (Kroll 2013).
Towards Smarter Cities - Volume 1 - eCAADe 32 | 91
 In the same chapter relating to the ICT use, Kroll
(1987) described another possible role for the com-
puter: evaluation and modification. During the de-
sign process, a custom program provided compar-
isons between the choices of components designed
by the architect and enabled rapid updates of these
specific components throughout the process. Fur-
thermore, by creating associated representations of
the components, Kroll used a computer to generate
façade drawings to be revised and detailed by the de-
signer. But he stressed that this process can never be
reliant on automation, which Kroll found "an absurd
and unhealthy claim".
Case 2: Fun Palace
In the Fun Palace project, Cedric Price, Joan Little-
wood and numerous intellectuals from different dis-
ciplines collaborated in the design of a flexible the-
ater in London's slums with a dynamic program. The
project aimed fusing information and communica-
tions technologies and industrial building principles
"to produce a machine capable of adapting to the
needs of users" (Price, 1965).
In contrast with Kroll's refutation of Le Corbus-
ier's metaphor of architecture as a "machine-to-live-
in", Cedric Price adopted and developed it further.
The project was an attempt at exploring improvisa-
tional architecture with the means of cybernetics and
information technologies (Mathews 2005).
Fun Palace did not have a fixed floor plan and in-
tended to "encourage random movement and vari-
able activities" (Lobsinger 2000). Mobile compo-
nents such as flying escalators, walkways, and ac-
tivity enclosures were carried by a megastructure

Figure 2
The Fun Palace
project by Cedric
Price and Joan
Littlewood
proposed a
dynamic program
that joins ICT and
industrial building
principles to
produce
architecture
capable of adapting
to the needs of
user. Illustrations
are from the CCA
Library Database
[1].

92 | eCAADe 32 - Towards Smarter Cities - Volume 1

and transported by a crane when necessary [1]. The
suggested time and place specific facilities covered
jam sessions, dance and science playgrounds, teach-
ing film, drama therapy, modeling and making areas
and music stations with instruments on loan (Landau
1984).
Similar to La MéMé, Fun Palace was not primar-
ily an aesthetical exploration. The building was con-
ceived to be super-functional and adapt to the peo-
ple's needs in a sustainable manner (Figure 2).
ICT tools and strategies for participation in
Case 2
As introduced in the previous section, the interdis-
ciplinary design team which Price collaborated with
included an English cybernetician and psychologist,
Gordon Pask. During the project, several practices
were proposed by Pask for the cybernetic regulation
of day-to-day activities (Mathews 2006). In this sense,
Fun Palace would be an ongoing conversation be-
tween the building and its users - "an assemblage of
interactive systems of interaction" [2].
Pask (1969) defined a number of domains of in-
terest for cybernetic interventions. Among those
were the Fun Palace and environment, visiting pat-
terns, mechanical and architectural considerations,
provision of specific participant activities, interactive
activities, individual participant situations (teaching
machines), controlled group activities, conditioning
systems and cybernetic art forms (Mathews 2005).
As a proof of concept, Pask created an apparatus
to collect feedback from the users after the realiza-
tion of the project (Figure 3). The proposed tool was a
"physical communication system" which he planned
to be used informally in one of the theaters to "ac-
commodate an invited audience" (Pask 1969). The
audience would be responding to a variety of activ-
ities using this tool and would be able to transform
the theater based on their preferences.
Through this exercise, Pask questioned to role of
the users and explored novel ways of participation in
an open-ended and performative manner.
CONCLUSIONS: DEVELOPING A MODEL OF
PARTICIPATION {IN} AND {THROUGH} AR-
CHITECTURAL DESIGN
Kroll's and Price's works can be considered as pro-
totypes of participatory, bottom-up architectural de-
sign. But it is necessary to differentiate between
these two projects. First of all, Price and his team
failed to realize the Fun Palace. Although it was de-
signed to be built, it can be seen as a proof of concept
for a utopic project.
On the other hand, La MéMé was partially re-
alized and served as a semi-functional prototype
through which many inspiring ideas were experi-
mented. It still stands in Brussels as a heterotopia be-
tween the ideal and the real, frozen in time.
In this context, in both projects, two intercon-
nected modes of participation were evident:
1. Participation {in} the design process
2. Participation {through} the design product
These two modes were significant in the ways they
enable the users and architects to co-produce archi-
tectural designs in a sustainable and participatory
manner. To start with, Participation {in} the design
process is similar to today's widely recognized inter-
pretation. It involves practices that "allow various ac-
tors to contribute to the overlapping phases of the
planning and decision-making" (Horelli and Wallin
2010).
In the case of La MéMé, Kroll has arranged nu-
merous meetings with committees and discussion
groups to empower the student groups (although
the level of participation and openness were chal-
lenged in the following years).
Price, on the other hand, did not believe that the
user needs can be precisely forecasted. The user par-
ticipation model he conceived would take place post-
occupancy. However, he shared his authority with
several intellectuals such as Littlewood who acted
as an essential part of the design team. Instead of
pursuing traditional consultation meetings, he asked
for the participation of an interdisciplinary commit-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 93
 tee to collaboratively design an enabling type of ar-
chitecture that facilitates participation to the highest
known extent.
The second and the most interesting mode
observed in the presented cases is participation
{through} the design product. This kind of empow-
erment takes place when various spatial qualities
of the architecture enable the inhabitants to shape
and reshape their own living environments. In both
projects, Kroll and Price aimed at the participatory
creation of infinitely flexible interactive spaces which
94 | eCAADe 32 - Towards Smarter Cities - Volume 1
represent the diversity of the needs of the inhabi-
tants. The forms of their designs were intended to be
altered to accommodate the changing needs of the
users.
In the La MéMé case, the dynamic elements were
the "infill": demountable window frames, moveable
partitions and prefabricated sanitary units. The Fun
Palace project envisioned mobile components such
as flying stairs, walkways and modular activity enclo-
sures.
Furthermore, in both of the cases a structure inde-
Figure 3
Fun Palace; diagram
for a cybernetics
theatre by Gordon
Pask (1965).
pendent from the infill (Kroll's wandering columns
and Price's megastructure) was used for facilitating
the dynamism of the architectural program.
Besides the participation modes discussed
above, the case study enabled the extraction of sev-
eral common principles conceived in both of the
projects. Among these, the most recurrent ones
were:
• Orchestrated self-organization
• Intense focus on the impact which architec-
ture can make on the users
• Incorporation of user variety and differences
into the architectural design process as well as
the product
• Incomplete, dynamic program as an enabler
for the continuous representation of the user
needs
• Embracing spontaneity and improvisation in
the design process
• Development of design rules or systems that
"regulate" the building in an open-ended way
Building on the principles and the participation
modes introduced above, it is clear that a more in-
novative approach to participation is possible. In this
context, we reframe participation as a reflective self-
organization practice which includes interactions {in}
the design process as well as {through} the design
product.
In this approach, ICT tools fuse two cycles of co-
operation in which the output of one process is trans-
formed by a second process and transferred to the
other one as input (Figure 4, on the next page). It is
important to note that the model does not suggest
a completely digitalized mode of operation: rather
it involves, augmentation and enhancement of tradi-
tional participatory practices through the use of ICT
tools and strategies.
The first cycle (Figure 4, on the left) involves a type of
social knowledge construction, building social capi-
tal. This capital is transferred to the second cycle (Fig-
ure 4, on the right) through which the users gather re-
sources, take action, accumulate experience and give
feedback to the first cycle.
This open-ended process involves several steps
which can be supported by different types of ICT-
enabled participation accessible today:
• Crowdsourcing: collection and sharing of
lived experiences
• ICT-enabled public meetings: information,
ideation and integration
• Crowdfunding: fund-raising, resource collec-
tion
• Event-based participation (blended activi-
ties): intervention
Operation Principles
First of all, from the perspective of cybernetics, the
role of the architect in our model cannot be neutral
or observant. It should be one of active participant in
the system. According to the second order cybernet-
ics, observers of a system cannot see how it works by
standing outside. They are engaged with the system
being observed in a cybernetic manner. Thus when
observers observe a system, they affect and are af-
fected by it (Foerster 1974).
Second, sustainable and informed participation
{through} the design product requires the develop-
ment of open-ended cybernetic systems which en-
able user organization, feedback and intervention in
a self-regulatory, indeterminate manner and without
a limited end-state.
Third, according to the law of requisite vari-
ety, for the appropriate regulation of participation,
"the variety in the regulator must be equal to or
greater than the variety in the system being regu-
lated." (Ashby 1958). Therefore the proposed model
should not be closed but open to external input of
different actors at various stages.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 95

Forth, the model should embrace reflexivity in
other words, circular relationships between cause
and effect.
A Hypothetical Use Case in Practice
Let's we take a large-scale housing project as a hypo-
thetical case for demonstrating the potential of our
participation model.
Following the cooperation cycles introduced in
Figure 4, the participatory design process starts with
crowdsourcing through which the needs and re-
quirements are collected in a structured manner.
Then, these are converted into several alternative de-
sign ideas and integrated into the context by the ar-
chitect, with the continuous ICT-enabled feedback of
the users.
Afterwards the users are asked to fund the
project through crowdfunding. If the process suc-
ceeds, the project gets transferred to the second
cycle and constructed with the contribution of the
users.
If the funding process fails, the design process
cycle is repeated: through a new crowdsourcing
96 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 4
ICT-enabled civic
participation in and
through design:
fusing two cycles of
cooperation
practice, the user feedback is collected to identify the
problems of the design, and to develop new alterna-
tive projects, which will then be asked to be crowd-
funded by the users.
The contribution of our model becomes more
evident after the construction of the architectural
project. Following the experiences of the resi-
dents, post-occupancy feedback is collected through
crowdsourcing.
When necessary, novel ideations on how to im-
prove the architectural design are created with the
continuous ICT-enabled feedback of the users and in-
tegrated into the existing context. Examples of these
ideas can be making interventions regarding the
communal or personal spaces, removing/adding par-
titions or reconfiguring the rules for co-habitation.
In the following stage, the users are again asked
to fund the changes through crowdfunding. If there
is enough support, the process moves to the second
cycle and suggested interventions are made. Follow-
ing the intervention, the user feedback is collected
and the participation continues to take place when
necessary. In this context, it becomes possible to de-
velop habitats which can adapt to the users' needs in
a sustainable manner.
Future Potentials of the Proposed Model
Reflecting on the latest developments in ICT, it is pos-
sible to claim that sensor networks and smart struc-
tures can play an important role on the gathering of
feedback as well as the support of user interventions.
For the operation of the model, it is necessary
to create sustainable multidimensional virtual repre-
sentations. In this sense, developing participatory
and more user-focused building information model-
ing (BIM) methods is crucial. These can also be used
as a basis for advanced visualizations to be displayed
during the ICT-enabled public meetings.
Furthermore, integrated with the information
model, web-based and mobile social platforms can
play a vital role for the collection and sharing of user
experiences through crowdsourcing as well as raising
funds through crowdfunding.
In the near future, low-cost robotic manufac-
turing methods have a potential to unlock self-
production practices, which can also be integrated
into the proposed model after the crowdfunding
step.
As suggested by Kroll (2014) in his response to
this paper, one the biggest challenges facing the in-
troduced practices will be "helping to organize social
groups and supporting the consistency of their dis-
cussions in a democratic manner, against the author-
itarian business-oriented approaches."
ACKNOWLEDGEMENTS
We would like to thank Lucien Kroll and Ranulph
Glanville for their valuable contributions to this pa-
per.
REFERENCES
Ashby, WR 1958, 'Requisite Variety and its implications
for the control of complex systems', Cybernetica (Na-
mur), 1(2), pp. 83-99
Architectural Association, ML 2000, 'Cybernetic Theory
and the Architecture of the Performance: Cedric
Price’s Fun Palace', in Goldhagan, J and Legault,
R (eds) 2000, Anxious Modernism, MIT Press, Cam-
bridge, MA
Banham, R, Barker, P and Price, C 1969, 'Non-Plan: an ex-
periment in freedom', New Society, 13(338), pp. 2-21
Broome, J 1995, 'Segal Method Revisited', Architects Jour-
nal, 202(20), pp. 53-55
Charles, J 2011, The Story of Post-Modernism: Five
Decades of the Ironic, Iconic and Critical in Architec-
ture, Wiley, London
Foerster, H 1974, Cybernetics of Cybernetics, University of
Illinois, Urbana Illinois
Habraken, JN 1961, De dragers en de mensen - Het einde
van de massawoningbouw, Scheltema & Holkema,
Haarlem
Kroll, L 1987, The Architecture of Complexity, MIT Press,
Cambridge, MA
Kroll, L 2013, 'Friendly Architecture', OASE Journal , Nai
Publishing, 90, pp. 28-31
Kroll, L 2014 'Private feedback of architect Lucien Kroll
the authors (Pak, B, Verbeke J.) in response to the
draft version of this conference paper', Document
sent via email in 15/06/2013 (in French), p. 2
Kroll, L 1997, 'Anarchitektuur', in Besch, JD (eds) 1997,
Componenten 2, Omtrent de Modernisering van de
architektuur, Publikatiebureau Bouwkunde Delft.,
Delft
Kroll, L 2005, 'Animal Town Planning and Homeopathic
Architecture', in Blundell, J (eds) 2005, Architecture
and Participation, Taylor and Francis, Kindle edition
Landau, R 1984, 'A Philosophy of Enabling: Cedric Price,
Works II. Cedric Price', AA (Architectural Association)
Files, 8, pp. 3-7
Mathews, S 2006, 'The Fun Palace as Virtual Architec-
ture Cedric Price and the Practices of Indeterminacy',
Journal of Architectural Education, 59(3), pp. 39-48
Pask, G 1969, 'The architectural relevance of cybernetics',
Architectural Design, 9, pp. 494-496
Poletti, R 2010, 'Lucien Kroll:Utopia interrupted', Domus,
937, p. .
Price, C 1965, 'The Fun Palace Project', Architectural Re-
view, 74, pp. 15-27
Rudofsky, B 1964, Architecture without Architects, Mu-
seum of Modern Art, New York
[1] http://cel.cca.qc.ca
[2] http://people.bath.ac.uk/jeh33/Publications/Harding
Towards Smarter Cities - Volume 1 - eCAADe 32 | 97
98 | eCAADe 32 - Towards Smarter Cities - Volume 1
The fusion of clever urban initiatives and digital
applications
Teaching architects and urban designers how to make apps for the public
involvement

Matevz Juvancic1 , Tadeja Zupancic2

1,2
University of Ljubljana, Faculty of Architecture
1,2
www.fa.uni-lj.si
1,2
{matevz.juvancic|tadeja.zupancic}@fa.uni-lj.si

Smart cities demand digitally smart designers. Teaching architects and urban
designers how to combine clever initiatives and tailored digital applications to
their advantage when trying to involve members of the public is one of the
important aspects of contemporary designer education. The paper demonstrates
an example of such an effort, highlights the reasoning behind teaching and
methods used, follows the learning path of students (and teachers) and analyses
the process, illustrating it with student work, offering architectural, pedagogical
and reflective implications for the purposes of similar introductions of topics in
architectural and urban design curricula.

Keywords: digital applications, apps, urban initiatives, non-experts, public

participation

INTRODUCTION: SMART CITIES DEMAND
DIGITALLY SMART DESIGNERS
Architects and urban designers are (arguably) well
skilled in the use of digital applications, especially
those used for professional purposes. Even if not,
mobile devices and digital applications permeate our
social lives and exchanges to the extent that regard-
less of the profession, we are all living in the digi-
tal domain. But when it comes to the fusion of digi-
tal applications and public involvement, the majority
of designers tend to follow social media trends and
the use of the existing digital applications. The tools
and services they provide are not always completely
tailored to the designers' needs and are often lack-
ing specific functions and visual upgrades that would
help in expert - non-expert communication.
On top of the conformity, the students of archi-
tecture and urban design often perceive the field of
digital applications and mobile devices to be outside
their scope of skills and job descriptions. However,
smart cities (Batty, 2013) will demand 'digitally smart'
designers. With their insight into the field, into the
needs and expert - non-expert communication, they
are precisely that needed driving force behind the in-
spiration and development of fresh, tailored and spe-
cialized digital applications to be used in design pro-
cesses when non-experts and lay-public are involved.
While we can interpret the urban users' needs

Towards Smarter Cities - Volume 1 - eCAADe 32 | 99

 Figure 1
”My Favourite
Places Drawn” is a
prototype of a
crowd sourcing
web based
application, which
would enable the
users to upload and
present their
favourite urban
places in a
sketch-like manner.
The researches may
use it to extract
detailed and
meaningful data
from the
photographs of the
urban places and at
the same time
graphically equalize
the visual material
for comparison and
research purposes
(designed by M.
Juvančič and S.
Verovšek).
and wishes indirectly from the hints the users leave a process, merging different tools, etc. They are not
in space with their adaptations and interventions meant to replace existing social media or other ex-
(Sanches and Frankel, 2011) or with the more tradi- isting applications but rather satisfy a specific need.
tional approaches, such as town hall meetings, work- There are several successful public participation plat-
shops and questionnaires, we can reinforce the ef- forms and application examples (Rucker and Whalen,
forts with the use of specialized apps, with the added 2012) and the best ones bring added value to the
benefit of actively engaging the user into the dia- user - who can e.g. express his/her opinion, make
logue, asking for the feedback and through his re- a suggestion, reinvent or reorganize the space, etc.
sponses, actions and initiatives inviting him to be- - or to the designer/researcher - who e.g. benefits
come a co-creator of urban places. from the crowd-sourced collection of data collected
The use of such applications is only meaningful or suggested by the users, gets the user perspective,
when they provide unique added value to the design is able to communicate with the usually anonymous
process as no other tool can, either by being widely and elusive lay-public, etc.
accessible, mobile, real-time processing, automating

100 | eCAADe 32 - Towards Smarter Cities - Volume 1

 MATERIALS AND METHODS 1. The students first needed to define physi-
Course settings: know the box but think cal place of their interventions out-of-the-box
outside of it thinking is only possible if you know what the
To activate and support the development in the box is and as we will later describe, the space
above mentioned direction, we have involved the itself greatly influences and inspires the pro-
topics of development of digital applications for fos- posed solutions / applications.
tering public participation and data collection into
an elective, one semester course titled Space & Me- 2. The involvement of public is a process with
dia. Based on our previous research into the public in- procedural, logistical and technical as well as
volvement in urban design and architecture, presen- design aspects: start with the definitions of
tations for non-experts, collaboration with the gen- the problems in space, propose a complex
eral public (Mullins, M., Zupancic, T. and Juvancic, M.: solution and the involvement of public on
2002), and previous course experience, the initial set- all levels, come up with fresh and innovative
tings, goals and outcomes were clearly defined: actions and only after that start finding out
where and how a digital application would
Figure 2 help support the process.
The on-site
(illustrated) 3. The digital application had to be clearly pre-
questionnaires to sented, all its functionalities implied and
gather the data and demonstrated to the extent that a digital de-
demands of the signer or a programmer would be able to un-
present users have derstand the ideas and this would represent
been used, the foundation for the potential software ap-
simulating some of plication development.
the roles their apps
would later take The course was introduced by showcasing differ-
over, and to ent initiatives involving public participation (done by
respond, adjust and teachers and students) across different countries (the
steer their student attendance was international). The cases
proposals. ranged from installations in public spaces that in-
volved or animated the public (e.g. Dominoes at
Mladi levi festival in 2011, see Note 1) to more elab-
orated initiatives (e.g. Beyond the Construction Site
running for the third consecutive year, see Note 2)
using all kinds of communication channels. The
examples ranged from single events or short-term
actions (week long) to long-term processes (cou-
ple of months, years). Teachers of the course also
presented ways of conceptualizing and demonstrat-
ing digital applications and functionalities with the
help of mock-ups, prototypes and animations (as de-
scribed in Buxton, 2007). Following the above men-
tioned steps and the discussions on each level. Stu-
dents also used illustrated on-site questionnaires to

Towards Smarter Cities - Volume 1 - eCAADe 32 | 101


research users' opinions and needs and then set to
design a public involvement design process, concept
for the application (app) that would support this pro-
cess and final urban design process proposal combin-
ing the lay public, the initiatives and the proposed
apps.
The crowd sourcing application as a show
case for app prototyping
My Favourite Places Drawn. The crowd sourcing ap-
plication prototype My Favourite Places Drawn (Fig.
1), developed within our research laboratory (Note 3)
has been used as a demo showcase of how to formu-
late, conceptualize and prototype such an applica-
tion, incorporating the initial idea, definition of goals
and specific aims the app should fulfil, definition of
targeted audience, benefits for the researchers/pro-
ducers and users, intuitive design, definition of func-
tionalities, developing the illustrations and a work-
ing mock-up prototype that can be used either in the
discussions with the experts who will eventually de-
velop the application for distribution or for the fine-
tuning and evaluation purposes.
My Favourite Places Drawn is a prototype of
a web based application, which would enable the
users to upload and present their favourite urban
102 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 3
Students used
different ways of
conceptualizing
and demonstrating
digital applications
and functionalities
with the help of
digital and
analogue
mock-ups,
prototypes and
animations.
places in a sketch-like manner. In this sense it is also
a tool for easily turning photographs into drawings
that can be saved, exported, coloured or printed. The
researches may use it to extract detailed and mean-
ingful data from the photographs of the urban places
and at the same time graphically equalize the visual
material for comparison and research purposes.
The functionalities. The conversion starts with the
import of the photograph (drag&drop), turning the
photograph into line drawing (edge finding filter),
adding the drawn elements from the libraries of peo-
ple, urban furniture, greenery, transportation, etc.
The elements can be stretched, flipped, put into fore-
ground, background, deleted, locked, etc. It is also
possible for the user to trace the edges of the back-
ground image 'by-hand' thus completely converting
the uploaded photograph into the drawing.
It will be possibe to vote for the best urban place
or drawing and the list of most viewed, most liked
and most commented drawings will be visible to
other non-registered visitors. Users who will want to
add their photographs and use the application will
have to register (for free) and give consent to their
creations being used for the research purposes.
Figure 4
“Pimp your city”
and “Campaigniser”
were the two
initiatives proposed
by the group of
students and they
each included sub-
actions such as
“graff customizer”,
dealing with graffiti
design, “lighting
customizer”, “adopt
a part of the city”,
“animals back in
town”, “fresh air Benefits for the user. The user is able to present his searched, tested and surveyed; walkability, pleasant-
factory”, “furniture favourite urban place to a wider audience and what is ness and similar indexes established and fine-tuned;
customizer”, etc. more appealing even convert it to a drawing. Further a wide range of micro-urban environments can be
The initiatives are uses can also be envisioned: the photographs quickly compared and dissected.
extensively become a colouring book, the drawings are usually
supported by the clearer and have less clutter for further work on them.
cross-platform
RESULTS
They can be changed and modified. The urban scene The process yielded two proposals from two groups
digital apps can be recreated, modified or reinvented, etc. By se- of students, both dealing with different places (in
facilitating lecting different themes and having the known ele- character, centrality and users) in the city of Ljubl-
competitions, ments already in place in the drawing, the user can jana. One has the square-like characteristics, it repre-
voting for have fun with completely changing the character of sents the crossroad of different paths, users and traf-
favourites, the drawing (e.g. child-like drawing, robot city, etc.) fic modalities, and the other is a less defined place
scheduling,
Benefits for the researchers. The researchers bene- without the distinctive shape, primarily used for tran-
proposal, on-line
fit from the crowd-sourced collection of urban places sition. Each of them has its own set of advantages
merchandise
created by the users. The tool offers a way of ex- and hindrances that students tried to identify and ad-
shopping,
tracting detailed and meaningful data from the pho- dress with their proposals. The spaces were chosen
donations, etc. The
tographs of the urban places and at the same time to keep the task tangible and less hypothetical, to
students also dealt
also graphically equalizing the visual material for force designers to think about actual problems and
with the graphic
comparison and research purposes (e.g. eliminating respond to them accordingly and to rethink the pro-
design. (students:
the photographers' skills, use of effects, lighting and posed applications in the terms of their specifics but
M. Bourget-Mauger,
image quality from the equation). By placing the ele- also their generic nature - applications which could
A. Chifflet)
ments with certain attributes and characteristics into be used in similar cases or even cover broader spectre
the drawing, the urban places will not only get recre- of city places. Both groups used the on-site analogue
ated in a uniform and comparable visual manner but questionnaires to gather the data and demands of
will also be digitized in a way to enable the applica- the present users, in a sense simulating some of the
tion of statistical and qualitative analysis. Cause and roles their apps would later take over, and to respond,
effect relationships in urban places can thus be re- adjust and steer their proposals (Fig. 2) towards the

Towards Smarter Cities - Volume 1 - eCAADe 32 | 103

 feedback they have received. The instructions for niture ("furniture customizer"), urban surfaces ("graff
the presentation of their concepts have been less de- customizer", dealing with graffiti design) and urban
fined. ambients ("lighting customizer"), which are directly
They were told to use any means presented ini- influenced/co-designed by the members of the pub-
tially that would demonstrate their app concept and lic. "City Campaigniser" fosters direct involvement
its functionalities (Fig. 3). The visualizations and of the urban users by making them adopt a physi-
graphics of the proposals can be best described as cal part of the place or otherwise tying a person to
draft sketches and are not compatible with the rich a particular object, thus creating a stronger bond be-
visual language of the interfaces we use in our daily tween the place and the user, while also raising the
life. They were used solely to clearly and effec- awareness of environmental, societal or cultural is-
tively present the idea to the teachers and IT experts. sues. The initiatives include the adaptation of a part
For their final presentations both groups used Pretzi of the city for gardening or other activities ("adopt a
(Note 4), which itself allows for more dynamic, visual part of the city"), sponsoring the implementation of
and interactive presentation. greenery in cityscapes ("fresh air factory"), reminding
users of other cultural or natural heritage and values
Pimp your City and City Campaigniser ("animals back in town"), etc.
The first group, dealing with the square-like place,
Figure 5
has upon analysis established that while the place
Each of sub- actions
being particular it also has characteristics of simi-
involves the
lar town squares across European old parts of cities.
members of the
Their proposal for the public participative design pro-
public and is
cess and supporting apps is appropriately generic in
extensively
nature and can be utilised in many similar situations
supported by the
whether in the same or other cities. While in need
cross-platform
of some minor urban design changes, which the pro-
digital apps, which
posal does not address, it rather focuses on the ini-
usually have a three
tiatives and apps that would mobilize change in the
part structure. The
appropriate direction.
process is in the
"Pimp your city" (Note 5) and "City Campaigniser"
forefront, the form
(Fig. 4) are two sets of loosely connected initiatives
and the materiality
proposed by the same group of students and they
as well as the
each includes sub-actions. Each of such actions in-
graphic design of
volves the members of the public and is extensively
the app follow in its
supported by the cross-platform digital apps, which
footsteps.
usually have a three part structure: a competition for
the best proposal coming from the users or a vot-
ing by the users choosing the best design proposed
by experts, a gallery of successful cases or an e-shop
with the merchandise and an event schedule that in-
forms about the happenings connected to this initia-
tive (Fig. 5).
"Pimp your city" deals with the physical change
of urban spaces through the redesign of urban fur-

104 | eCAADe 32 - Towards Smarter Cities - Volume 1

Urban CO.design.ER
The second group took a transitional, less articu-
lated and loosely defined space with undefined iden-
tity as their starting point of thinking about possi-
ble procedural, event and app supported initiatives
that would help improve the experience for the pub-
lic user. The complexity of the situation has resulted
in a complex procedural public participation pro-
cess proposal, where the co-design and co-decision-
making stand in the limelight (Fig. 6). The name - "ur-
ban.CO.designer.ER" - suggests the cooperative na-
ture and the emergency intervention that the place
is in need of. The process would aim to take into an
account all the age categories of the space users, giv-
ing them the possibility to express their ideas about
the place. According to the age groups, the means
of communication would range from traditional me-
dia and 'messengers' (e.g. letters per post) to the
new means of communication like game based apps
(e.g. for children and youth), specialized mobile apps,
desktop applications and social networks. The ex-
perts would then group and combine ideas into sev-
eral design/action proposals and the users would be
again asked to vote for their favourite. The process al-
lows for the members of the public to both co-design
or at least influence the design and participate in the
decisions.
The students then focused on one particular app
(running on mobile devices, desktops and on-site
interactive kiosks) that would enable the users to
suggest ideas (through intuitive, touch screen based
interface), choose among pre-determined changes
(e.g. change of colour) and observe them in real-time
presentations, pick urban furniture, propose events,
and share all this with all other interested members
of the public and with designers (Fig. 7).
For the rough prototype they have used a mock-
up paper model, where the user 'clicks' were simu-
lated by changing the hand drawn elements of the
user interface and through which the functionalities
of the proposed app could be initially tested.
Not being able to go through the motions of
their proposed co-design and co-decision making
process and fully involving the public into their
project, they nonetheless came up with the interven-
tion such an app could yield. The proposal involves a
mobile food stall and moving urban furniture, plant
pots and lights, which can be rearranged per pub-
lic taste and frequently changed by the users. While
the food stall brings the additional and much needed
programme to the place, the moving parts introduce
the dynamic nature and liveliness to the sub passage.
Their proposal not only ensures the involvement of
the public during the planning but also encourages
it during the implementation and use.
DISCUSSION
The course has been teaching - and not research-
oriented. The following findings are not based on
the precise and laboratory like data collection and
have not been quantitatively tested, but can be use-
ful when similar topics and teaching are introduced.
Particularities of a given space inspire- and
are reflected in- the apps concepts
The place of reference has proven to help in the pro-
cess of conceptualizing the digital applications. The
problematics in the space has contextualized and
brought perspective as well as specialization direc-
tion of apps needed. The two spaces chosen by
the students were diametrically opposite, both with
certain problems - one particular, mono-functional,
with clearly defined needs and the other, more com-
plex, articulated and multi-functional, with less dis-
cernible needs and obvious solutions, which are very
much reflected in the proposed initiatives and apps .
While the mono-functional space gets a tailored app
and urban design response to stir up the space-use
in new directions, the more articulate but complex
space benefits from more generic but nonetheless
fresh fusions of initiatives and app variations.
The choice determined the approach and further
development of concepts which led to specialized
solutions and apps on one hand and to more generic,
more widely applicable proposals for different cities
and neighbourhoods on the other.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 105
 Digital app concepts arise to genuine needs
and they alone are able to fill the void
Both proposals used the apps to address and intro-
duce much needed but specific functionalities that
cannot be served otherwise. While one is used
to gather broad perspective of existing and future
users' suggestions for the space development and re-
design, the other supports several initiatives to en-
liven and define the given space and other spaces in
the city. Both are designed to work across several

Figure 6
The complexity of
the situation has
resulted in a
complex procedural
public participation
process proposal,
where the public
co-design and
co-decision-making
stand in the
limelight.

106 | eCAADe 32 - Towards Smarter Cities - Volume 1

Figure 7
“urban.CO.designer.ER”
is designed to
involve the public in
“ER” interventions
in Fabijanjev bridge
surroundings. The
goal is to improve
the grey and boring
area around the
factory and the
bridge with the
help of users’ and
passers’-by
suggestions, where
digital app comes platforms and on different devices, including mo- the application has not been the aim of the course - it
handy for data biles. They encourage widespread non-expert par- is still important and possible to equip them with the
collection and ticipation, foster voting for proposals, scheduling of rudimentary lingo, methods and knowledge, for col-
co-decision making events and mobilize the once anonymous user - all on laboration among different experts in the field and
in order to make it the levels previously impossible through traditional make the software development seem less intimidat-
into a lively and means (or possible but resulting in costly actions). ing and out of reach. At the same time the designers
joyful space. can bring a fresh perspective, both in a sense of needs
(students: I. The initiatives provide the drive, the apps for-, use of- and new ideas into the sometimes rigid
Krznarić, M. facilitate them IT world of apps and interfaces.
Mikócziová, K. It is the fresh and new ideas about the initiatives that
Ignaciuk) promise change in the given places. The process is CONCLUSION: FUSION ACHIEVED ON SEV-
in the forefront, the form and the materiality as well
ERAL LEVELS
as the graphic design of the app follow in its foot-
The education of an architect or an urban designer
steps. The initiatives, the place and the intuitive and
has to reflect the current trends in media and digital
easy-to-use app that facilitates public involvement,
application use by involving these topics into the ex-
feedback, multidirectional communication between
isting curricula and developing particular skills. Only
experts and users, decision making, act as a system
in this manner will the future designers be able to
and create added value (in some cases they could not
take an active part in the design of smart cities and
come into existence without one of them). They rein-
apps, which will bring the public involvement to new
force and support each other, while each as a stand-
levels.
alone could not produce the same synergetic effect.
In line with the conference theme, the paper
traces several fusions that rise above mere amalga-
Insight into the field of digital apps - not a
mation: a) fusion of skills - skills of an architect but
change of a profession
also of digital app designer, b) fusion of digital tools
While it is usually neither possible nor viable to teach
and fresh initiatives - the fresh initiatives provide the
the students of architecture and urban designers the
drive, the apps with tools facilitate them, c) fusion of
higher level skills of programming, needed for the re-
expert and non-expert participation and collabora-
alization of a finished digital product - the making of
tion - always elusive mix and balance are given more

Towards Smarter Cities - Volume 1 - eCAADe 32 | 107

 credibility and do-ability through the use of digital
apps.
While the mainstream currents always produce
much needed critical reflections (Note 6), the e-world
is a given fact facing the urban designers and ar-
chitects of tomorrow which should be reflected in
the curricular development in architectural and ur-
ban design education.
ACKNOWLEDGEMENTS
The ideas presented here and the successful running
of the course Space and Media would not have been
possible without the help of our colleagues and co-
tutors Anja Jutraz and Spela Verovsek. We also have
to give credit to students Marie Bourget-Mauger, Axel
Chifflet, Krzystof Ignaciuk, Iva Krznaric and Martina
Mikocziova, who participated in the course and ded-
icated their full attention and energy beyond our ex-
pectations.
ENDNOTES
1. The dominoes in the form of breeze building
blocks were set all over the city of Ljubljana
leading to the venue of the festival, triggered
and when the last one fell the festival started.
[1]
2. Beyond a Construction Site is a community-
based garden intervention on an empty and
long abandoned construction site in the cen-
tre of Ljubljana. Neighbouring residents and
other interested can tend for the urban gar-
dens and socialize. [2]
3. My Favourite Places Drawn app prototype
has been developed by M. Juvancic and S.
Verovsek and is stemming from their research
of visual urban vocabularies. [3]
4. Prezi® by Prezi Inc. is a cloud-based presenta-
tion software and storytelling tool for present-
ing ideas on a virtual canvas that can run also
in an offline mode.
108 | eCAADe 32 - Towards Smarter Cities - Volume 1
5. The apps have been named by the students
as branding was seen as an important part of
the process and it denotes (or should at least
reflect) the app's intended function. Not all
the names for apps are completely relevant or
spelled correctly but nevertheless give them
their own identity.
6. Adam Greenfield's book 'Against the smart
city: the city is here for you to use' is such
an example, which points to where the cur-
rent vision of smart cities could lead us with
dire and unwanted consequences (Green-
filed, 2013). It is also a reminder that we
should not overcomplicate the use of urban
environments and not to over e-design them.
REFERENCES
Batty, M 2013, The New Science of Cities, The MIT Press,
Cambridge, London
Buxton, W 2007, Sketching User Experiences: Getting the
Design Right and the Right Design, Morgan Kauf-
mann, San Francisco
Greenfield, A 2013, Against the smart city (The city is here
for you to use), Do projects
Mullins, M, Zupancic, T and Juvancic, M 2002 'Experts
and Users Efficiency and Accuracy in the Presen-
tation of Design Ideas Across Networks', Connect-
ing the Real and the Virtual - design e-ducation, 20th
eCAADe Conference Proceedings, Warsaw, pp. 430-
439
Rucker, DG and Whalen, PF 2012 'Online Public Partic-
ipation Platforms and Applications.', Wise Economy
Workshop & New World Public Engagement
Sanches, MG and Frankel, L 2011 'Visual anthropology
and public design: can the association between
these fields generate valuable insights into the di-
verse patterns of urban behaviour?', N.F.M. Roozen-
burg, L.L. Chen and P.J. Stappers (eds), Diversity and
unity: Proceedings of IASDR2011, the 4th World Con-
ference on Design Research, 31 October – 4 Novembe,
Delft, the Netherlands, pp. 1-11
[1] http://vimeo.com/29250411
[2] http://www.obrat.org/onkraj-gradbisca-beyond-
construction-site
[3] http://predmet.fa.uni-lj.si//mmprostor/myfaveplacesdrawn/
myfaveplacesdrawn_prototype.htm
Towards a dynamic evacuation system: developing
methodologies to simulate the evacuation capabilities
of subway stations in response to a terrorist attack with
CBRNE weapons
James Charlton1 , Markus Brune2
1
University of Northumbria, Newcastle upon Tyne, United Kingdom 2 Department
of Geography, Ruhr-University of Bochum
1
j.charlton@northumbria.ac.uk 2 markus.bruene@rub.de

Events in recent times have highlighted the vulnerability of underground public

transportation to possible terrorist attacks. A key question therefore is how an
evacuation can be accomplished from underground stations safely. The strategy
"go up and take the nearest exit to the surface" might not be the best response.
Evidence from the Daegu subway station fire in 2003, investigated by Tsujimoto
(2003) and Jeon and Hong (2009) establish that smoke or toxic airborne
substances from a terrorist attack tend to use the same direct routes used by the
fleeing passengers and as result significant injuries or fatalities can occur. This
study proposes the concept of a dynamic evacuation system which would guide
subway users along safe routes. To test how this system may operate, the study
discussed combines measurements from tracer gas experiments with climate
measurements to establish how toxic agents spread in subway stations under
certain conditions and combines these results with those from pedestrian
simulations applied to calculate evacuation times for possible escape routes. By
integrating the resulting dataset from these methods, an evidence base of how a
dynamic evacuation system may work can start to form.

Keywords: pedestrian simulation, subway climatology, CBRNE, subway

evacuation, tracer gas experiments

INTRODUCTION the individual strands of research (subway climatol-

This paper introduces an initial strategy to aid in
the evaluation of different evacuation procedures
for subway stations in response to a terrorist attack
with chemical, biological, radiological, nuclear and
explosive (CBRNE) weapons. The paper introduces
ogy, tracer gas experiments and pedestrian simula-
tions), to highlight the development and application
of the proposed methodologies to develop an un-
derstanding of how effective a dynamic evacuation
system may be. The study discussed, demonstrates

Towards Smarter Cities - Volume 1 - eCAADe 32 | 109

 that by integrating appropriate datasets a greater un-
derstanding of the impact of different evacuations
routes on an evacuees health can be understood. It
is shown that by controlling the routes pedestrians
use to exit a subway station, the number of fatalities
can be reduced. It is highlighted that a dynamic evac-
uation guiding system based on subway climatology
would take into account the location and cause of the
evacuation, the resulting dispersal of gas, smoke, etc.
and the subway climatology at the time. In doing so,
the system could identify the most endangered areas
and guide passengers via an adaptive escape route
using audio and visual techniques. Information on
the evolution of the emergency situation could also
simultaneously be relayed back to the rescue forces
to help plan the rescue and evacuation procedures,
by being able to answer questions as to where to con-
centrate the rescue teams. It is envisaged that such
a system would help to plan the rescue and evacua-
tion procedures by identifying areas to concentrate
the rescue teams, identifying the most endangered
areas and providing an adaptive escape route system
for the passengers. However, before such a system
could be implemented, this paper concludes by high-
lighting the need for further study and how the latest
advances in simulation software can further aid in our
understanding and prediction of similar situations.
CASE STUDY: ALEXANDERPLATZ, BERLIN.
To aid in development of the proposed methodology
this paper highlights the application and results from
the assessment of the subway station, Alexander-
platz - Berlin, Germany. The subway station Alexan-
derplatz (figure 1) is located in the center of Berlin
and is the most used interchange station in the city.
Built from 1910 - 1913 according to the plans by Al-
fred Grenander, the underground station originally
only consisted of one line. From 1926 to 1930 two ad-
ditional lines were adjoined, also built according to
Grenander's conception. Today the subway station
still consists of these three subway lines, the North
and South running U2 and U8 and the East and West
running U5. Alexanderplatz subway station is also di-
110 | eCAADe 32 - Towards Smarter Cities - Volume 1
rectly connected to a local over ground train station.
Due to the importance and complexity of Alexander-
platz, the station has nine exits to the outside, one
exit to the local train station and another to an office
building. As a result it is an ideal subway station to
aid in the development and application of the initial
proposed methodology.
STRAND 1. SUBWAY CLIMATE
The relatively young research field of "Subway Cli-
matology" was developed in the late 1990s (Pflitsch
and Küsel 2003). The workgroup of Cave & Subway
Climatology (Ruhr-University Bochum) investigated
the special climate conditions in subways since 2000
(Pflitsch et al. 2010). The BMBF (German Federal
Ministry of Education and Research) funded the re-
search projects OrGaMIR and OrGaMIRplus (Cross-
organizational hazard prevention to protect human
life and critical infrastructures by optimized preven-
tion and reaction) to examine the special climate in
subway systems (Blennemann, F, 2005). From this re-
search it was established that the energy budgets of
subway stations differ from classical climatology: ra-
diational heating and cooling processes are absent,
the soil heat flux is significant higher due to higher
surface volume ratio and higher ground temperature
in urban agglomerations, anthropogenic energy re-
lease from trains, technical equipment and passen-
gers also has to be taken into account. On the other
hand energy loss due to latent and sensible heat
transfer is highly reduced. These climatic key fac-
tors cause a general over-warming of subway tunnels
compared to the outer atmosphere from late sum-
mer to winter. This effect can become even stronger
if a fire breaks out in a station by adding additional
heat to the climate. The effect of penetrating cold
air and rising warm air at the subway systems open-
ings to the outer atmosphere drives natural air flow
in the tunnels. This natural air flow is not stable, dif-
ferent aboveground weather situations leads into al-
ternating air flow direction inside the subway tunnels
(Pflitsch and Küsel 2003).Therefore, in the event of a
terrorist attack with CBRNE substances inside a sub-
Figure 1
Overview of
subway Station
Alexanderplatz and
positions of
measurement
points,top: cross
section, bottom:
plan views. More
exits can be found
on line U2 and U8.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 111

 way station, such climatology would have a major ef-
fect on the dispersion of the airborne substances re-
leased (Brüne et al., 2012).
Climatology results: Alexanderplatz
In order to gain an understanding of the climatol-
ogy of Alexanderpaltz, continuous long-term air flow
measurements (A, B, figure 1) were set up in the adja-
cent tunnels of the U5 subway platform as part of the
OrGaMIRplus Project. The data gathered, established
that when there was no train movement, there was
little air movement in the tunnel, allowing thermal
effects to dominate; air heated up inside the station
and moved due to buoyancy effects to upper levels
and to the outside. In addition with the very low air-
flow velocity at the adjacent tunnels (a and b) it leads
to the assumption that released tracer gas would not
penetrate the tunnels, but instead rise up to the up-
per level and find its way through station entrances
to the outside.
For the period of the study discussed in this pa-
per, six additional ultrasonic anemometers (I-IV, fig-
ure 1) were placed at the mezzanines on level -3
to measure the air flow inside the subway station.
Again, the mobile measurements mostly showed up
streaming air masses from the U5 platforms (level -4)
to the mezzanine west (III) and the mezzanine east
(V, VI). The exception to this was at location IV on
the mezzanine west. At this location cold air pen-
etrates from outside, therefore downstream airflow
was recorded.
STRAND 2. TRACE GAS METHODOLOGY
To aid in the simulation of a terrorist attack with
CBRNE weapons, sulphur hexafluoride (SF6) was
used as a tracer gas to exanimate the behaviour
of airborne substances. SF6 is nonvenomous, non-
smelling, non-visible and easily detectable due to its
very low occurrence in the atmosphere of 0.005 ppb.
Although sulphur hexafluoride is initially six times
heavier than air, on release it mixes very quickly with
the ambient air and behaves like it (Pflitsch et. al.,
2010). Sulphur hexafluoride can therefore be consid-
112 | eCAADe 32 - Towards Smarter Cities - Volume 1
ered as an appropriate safe gas to simulate a terrorist
attack with CBRNE weapons.
In the development of the proposed strategy the
release of the tracer gas took place outside of opera-
tional hours in order to replicate best practice, where
ventilation systems are switched off in the event of a
fire or terrorist attack. Sanchez et. al. (2000) mod-
elled the number of victims associated with a ter-
ror attack using chemical or biological weapons and
concluded that the best strategy is to shut down the
ventilation and stop the trains. Similarly many old
subway stations do not have ventilations system at
all. So the operational break in which the tracer gas
simulations were carried out, reflects the response of
subway operators to the hazardous circumstances re-
sulting from a CBRNE attack. Furthermore, turning
the fans off allows subway climatology to play the
leading role in the movement of the gas through-
out the station. Therefore the empirical dispersion of
tracer gases obtained from this study can be used to
accurately analyse the evacuation simulations.
During the study an amount of 4kg SF6 was re-
leased for two minutes in the middle of the U5 plat-
form. In different locations students (1-22, figure 1)
took samples every minute with 60 ml syringes. The
syringes were closed with a rubber plug and the air
was examined by gas chromatography afterwards.
Results from Tracer gas dispersion
Following the release the gas the first high values
were observed only 4 minutes at measurement point
8, located in the mezzanine west, with a reading of
1090 ppb (see figure 2). At the same time a greater
contamination reading was recorded in the above ly-
ing shopping street level at measurement point 16,
with a reading of 6630 ppb. In the remaining areas of
the subway station, including the lowest level (level -
4) where the gas was released, readings lower than
1000 ppb were recorded for a period up to 6 min-
utes following the start of the release. After this pe-
riod of time, significantly high reading were recorded
for measurement points 5 and 14 located on level -4
and the east mezzanine level -3, respectively. These
 values stayed high for the remaining experimental
time. In contrast, over a similar time frame, record-
ings at measurement points 8 and 16 began to de-
crease. Recordings at measurement point 16 gradu-
ally decrease from 5 minutes after the release, while
recordings at point 8 remained high before more dra-
matically falling after 11 minutes. The initial lack of
build-up of tracer gas on level -4 is believed to be re-
lated to the dimensions of the space. The ceiling of
platforms U5 is more than 5m high and at its widest
spot measures 36m.
Figure 2
Timeline of SF6
Values of selected
Measurement
points.
Summarizing the tracer gas readings, during the
first 5 minutes gas rises from the lower level through
the western mezzanine into the shopping street
level. After this period the trajectories moved to the
eastern part of the station and the concentration re-
mained high. Both pathways lead to the shopping
street where high concentration could also be found.
STRAND 3. PEDESTRIAN SIMULATIONS
To examine the evacuation capabilities of Alexan-
derplatz, the study adopted the use of agent based
pedestrian simulation software capable of repre-
senting the attributes and behaviour of individual
Figure 3 agents within given environments. The selected soft-
Percentage ware, Legion SpaceWorks, allows entities to be pro-
distribution of grammed to have; varying degrees of prior knowl-
agents within the edge regarding a buildings layout (e.g. commuters
pedestrian vs. tourists), different mobility (e.g. children, adults,
simulation model impaired, runner, etc.), or different global origins (e.g.
(Level -4) UK, Asia, North American, etc.). All of these varying
options and interactions define the speed and route
in which agents will navigate through a space. Pri-
marily used by architects and civil engineers, unlike
its predecessor Legion Studio, Legion SpaceWorks
does not require the user to define the paths they
take, but instead to only define the origin and exit
points of agents. Route modifiers are then added to
the model to trigger a change in the movement, ac-
tivity or destination of the pedestrian, allowing for
a more organic and realistic pedestrian flow to be
achieved. Using these available options and 2D CAD
plans of each platform and concourse area within the
station Legion SpaceWorks was used to create evac-
uation simulations of Alexanderplatz.
For the study discussed, it was assumed that
2050 agents were at their starting position at the
lowest level on line U5, this represent the highest
demand according to the German Railway Author-
ity, the other subway lines where neglected. Of the
2050 agents placed throughout level -4, 236 agents
were located on each platform, distributed as shown
in Figure 3, and 789 agents were distributed evenly
throughout each of the trains. The speed profiles
for each of the agents were configured to repre-
sent Northern European commuters. The agent's re-
action time to an alarm sounding were defined in
accordance with British Standards for a transporta-
tion hub (British Standards Institution, 2004) which
refers to six key categories of fire evacuation analy-
sis; alertness, familiarity, building type, staff training
and alarm type. In accordance with these guidelines,
in each of the scenarios simulated, no agents reacted
to the alarm sounding in the first 90 seconds, the ma-
jority of the agents reacted around 165 seconds and
all agents had reacted within 240 seconds.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 113
 Results from pedestrian simulations Figure 4
Two different multi agent pedestrian simulations Simulation 1 -
were set up: Cumulative Mean
Density map
• Simulation 1: The agents have free choice to
find the shortest route

• Simulation 2: Agents were forced to evacu-

ate through the west route only. This decision
was based in tracer gas results and should be
the most uncontaminated evacuation route.

In simulation 1, where the agents had free choice

to find the shortest route out of Alexanderplatz, the
simulations showed that the majority of the agents
Figure 5
chose to exit using the central stairs at either end of
Simulation 2 -
the platform, via level -3 and exiting on level -2 (see
Cumulative Mean
figure 4). The agents that chose to exit via the west
Density map
central stairways were quickly subjected to conges-
tion on level -4, which was sustained for several min-
utes and spread to the west central stairway located
on level -3 as more and more agents tried to evacuate
via this route. Following the evacuation from level -4,
the agents were able to remain at a consistent speed
while exiting the model.
In the 2nd simulation agents were guided along
a pre-defined route away from the dispersal of gas.
Agents were guided to the relevant stairs at the far
west side of level -4. Agents continued up these sets
of stairs until they reached level -1 where they were The egress times for each of the simulations are
navigated to use only exit Ap III/3 located on the summarized in table 1. It is shown that the agents
south side of level -1 west (see figure 5). Although were able to exit Alexanderplatz the quickest when
this route was chosen to prevent agents exiting via they had free choice to decide to shortest route (sim-
routes highly contaminated with lethal gas, the lim- ulation 1). Agents were slowest to exit the model
ited number of exits resulted in high levels of con- when they were guided on a pre-defined path, as in
gestion, most significantly for those agents located simulation 2. In the 2nd simulation it took 17:05 min-
on platform 2. Agents exiting via the defined route utes for all the agents to exit the model, even though
from platform 2 became obstructed by those agents the first agents began to exit the model in a compara-
exiting from platform 1. The obstruction resulted ble time to the agents in simulation 1. This significant
in a buildup of agents from platform 2 on the se- increase in time is a result of the congestion caused
lected stairs ways, resulting in no significant move- by agents exiting platform 1 and blocking agents ex-
ment from agents exiting from platform 2 for a period iting from platform 2, due to the agents only be able
of 3 minutes.

114 | eCAADe 32 - Towards Smarter Cities - Volume 1

 to exit via one exit (Ap III/3). The resulting bottleneck
caused agents exiting via the pre-defined route from
platform 2 to make no significant movement for a pe-
riod of 3 minutes.
Table 1
Time of egress for
different levels
subdivided by the
two simulations
COUPLING TRACER GAS AND PEDESTRIAN
SIMULATION OUTPUTS
From the SF6 tracer gas readings a contamination
for different parts of the station could be derived
for every minute of the study. Similarly, the pedes-
trian simulations provide the location of agents ev-
ery 0.6 seconds. By knowing both the position of
each agent and the changing concentration levels of
SF6 throughout the station, it is possible to interpo-
late the amount of inhaled gas by each agent as they
move through the model and get a definitive under-
standing of how each evacuation route would affect
the agent's health. In order to get a more realistic
assessment of human safety during the evacuation
time at this stage of the study it was assumed that
phosgene (CCIO2) was used instead of SF6. 2.7 kg of
phosgene is needed to have the same concentration
levels like 4 kg of SF6 (see Table 2).
Table 2
Chemical properties
of SF6 and
Phosgene and the
mass equivalents
causing the same
concentration
levels
Results of coupling outputs
Considering the Acute Exposure Guideline Levels
(AEGL) of the US Environmental Protection Agency
[10], two values can be used to examine the influence
on human health:
• AEGL-2 (600 ppb for a 10 minutes expo-
sure period): is the airborne concentration of
phosgene which it is predicted that the gen-
eral population, including susceptible individ-
uals, could experience irreversible or other se-
rious, long-lasting adverse health effects or an
impaired ability to escape.
• AEGL-3 (3600 ppb for a 10 minutes expo-
sure period): is the airborne concentration of
phosgene which it is predicted that the gen-
eral population, including susceptible individ-
uals, could experience life-threatening health
effects or death.
From simply overlaying the outputs from the pedes-
trian simulation with the contamination maps pro-
duced from the tracer gas experiments it is possible
to carry out an initial examination into those agents
which would have been exposed to high levels of gas.
Figure 6 shows the position of the agents from sim-
ulation 1 and 2 after 5 minutes, against the corre-
sponding contamination map. It is clear that those
agents which have chosen the quickest route (simu-
lation 1) have also moved through the areas of high
contamination located on the west mezzanine on
level -3 and the shopping street located on level -2.
Based on the results shown, a safety assessment
of different evacuations can be made from the to-
tal numbers of agents reaching the different levels.
Table 3 summarizes these results and indicates the
number of agents that exceed the two levels. As
expected from the initial examination, in simulation
1 the agents using the west staircases between the
platforms U5 and the shopping street begin to ex-
ceed the AEGL-2 level when they are on the west
mezzanine level. Those agents that are slow to re-
spond and as a result are caught in the congestion
that builds on these staircases are also the first agents
to reach the deathly AEGL-3 level in this location. As
the evacuation continues more deaths are expected
on the shopping street level. From the completed
Towards Smarter Cities - Volume 1 - eCAADe 32 | 115
 Figure 6
Concentration and
position of bots in
simulation 1 and 2,
5 min after release
and exists used for
pedestrian
simulations 1,2,3

study, the results show that in simulation 1, 362
agents will die during their evacuation of Alexander-
platz, whilst 1060 will have long-lasting health issues.
This study could not take into account that agents
who exceed AEGL-2 level of contamination may evac-
uate slower, increasing the congestions build up and
also that dead agents may become obstacles for the
other agents. In the 2nd simulation, even with the
low contamination, due to the long evacuation time,
it was established that 212 agents will exceed the
AEGL-2 level before they are able to exit the subway
station, but no agents will exceed the AEGL-3 level.
The combined tracer gas results and pedestrian sim-
ulation shows that all the 212 agents that exceed the
AEGL-2 level will do so in the congestion build up on
level -1 as they try to exit via a single exit. On this
level, there are two more exits available, so the num-
ber of harmed passengers would significantly lower.
Table 3
Summary of agents
exceeding AEG
levels
In comparing the two simulations it is clear that
agents exiting via a controlled, pre-defined route are
able to do so in the safest manner, regardless of the
increased evacuation time. The results show that the

116 | eCAADe 32 - Towards Smarter Cities - Volume 1

majority of the agents in simulation 2 remained un-
der 600 ppb and no agents exceeded the AGEL-3
level and therefore no agents would have died exit-
ing via this route.
CONCLUSION: TOWARDS A GUIDED EVAC-
UATION SYSTEM
This study aimed to develop methodologies to high-
light the advantages a dynamic evacuation system
could have on the safety of subway user evacuating
subways in response to a terrorist attack with chem-
ical warfare. By combing tracer gas results based
on subway climatology with outputs from pedestrian
simulations the study established how different evac-
uation routes have different risk levels. Generally it
was showed that staying as long as possible in lower
levels or taking guided routes away from the disper-
sal of gas could be a good strategy to leave the area
of risk even if it takes longer. It was shown that the
subway climate conditions, especially the airflow in
tunnels outlines the basis one element for a dynamic
evacuation guiding system, which can be proposed
from the results of this study. Such systems would
take into account the location and cause of the evac-
uation, the resulting dispersal of gas, smoke, etc. and
the subway climatology at the time. In doing so,
the system could identify the most endangered areas
and guide passengers via an adaptive escape route
using audio and visual techniques. In catastrophic
circumstances people want to go out as quickly as
possible to the over ground, so intelligent evacua-
tion systems may guide people against their natural
flight behavior. Information on the evolution of the
emergency situation could also simultaneously be re-
layed back to the rescue forces to help plan the res-
cue and evacuation procedures, by being able to an-
swer questions as to where to concentrate the rescue
teams.
Although this study made a significant start in
developing methodologies and an evidence base for
the concept of a dynamic evacuation system, the ap-
plied methodologies are still not fully formed to allow
for a more extensive study. The tracer gas data pre-
sented in this paper is based on a single tracer gas
experiment and is only valid for this specific climatic
situation. Although this limitation does not distract
from the aim of this paper to show how combining
results from tracer gas experiments with pedestrian
simulations can be used to assess evacuation proce-
dures for underground stations during an emergency
situation, it does not allow for other simulations to be
executed or for future evacuations simulations to be
carried out without first collecting first hand subway
climatology results. As such, in order to develop a
more sound methodology for a more dynamic study,
it can be proposed that a relationship between urban
weather trends and subway climatology needs to be
established. It is certainly within the scope of acces-
sible simulation software to help establish this trend.
In doing so, the climatology of a subway station dur-
ing a given state of weather could be predicted, al-
lowing for further, dynamic studies, to be carried out.
REFERENCES
Blennemann, F 2005, Brandschutz in Fahrzeugen und Tun-
neln des ÖPNV. Fire protection in vehicles and tunnels
for public transport, Alba-Fachverl
Brune, M, Pflitsch, A, Agnew, B and Speigel, J 2012 'Dy-
namics of natural air flow inside subway tunnels',
Proceedings from the fifth International Symposium
on Tunnel Safety and Security, New York, pp. 329 -
337
Jeon, G and Hong, W 2009, 'Characteristic Features of
the Behavior and Perception of Evacuees from the
Daegu Subway Fire and Safety Measures in an Un-
derground Fire', Journal of Asian Architecture and
Building Engineering, 8, pp. 415 - 422
Pflitsch, A, Brune, M, Ringeis, J and Killing-Heinze, M 2010
'ORGAMIR: Development of a safety system for re-
action of an event with emission of hazardous air-
borne substances - like a terror attack or fire - based
on subway climatology', Proceedings from the fourth
International Symposium on Tunnel Safety and Secu-
rity, Frankfurt au Main
Pflitsch, A and Kusel, H 2003, 'Subway-Climatology -
New Research Field for the Management of Possible
Catastrophes in Subway Systems', in Pyka, J. L., Du-
bicka, M, Szczepankiewicz-Szmyrka, A and Sobik, M
(eds) 2003, Man and Climate in the 20th Century, Uni-
versity of Wroclaw, pp. 384-394
Towards Smarter Cities - Volume 1 - eCAADe 32 | 117
 Sanchez, J. G., Coke, L, Wasmar, F and Policastro, A. J.
2000 'Comparison of Strategies in Managing C/B Re-
leases in the Subway Environment', Proceedings of
Tunnel Management: 3rd International Conference,
Sydney
Tsujimoto, M 2003 'Issues raised by the recent subway
fire in South Korea', ICUS INCEDE newsletter: Interna-
tional Center for Urban Safety Engineering, Tokyo

118 | eCAADe 32 - Towards Smarter Cities - Volume 1

Simulation of Air flow, Smoke Dispersion and Evacuation
of the Monument Metro Station based on Subway
Climatology
Zi Qian1 , Brian Agnew2 , Emine Mine Thompson3
1,2,3
Faculty of Engineering and Environment, Northumbria University
1,2,3
{zi.qian|brian.agnew|emine.thompson}@northumbria.ac.uk

This research is creating a working laboratory in Newcastle Monument metro

station to understand the details of how the over ground climate influences the
internal airflow and the impact this has on evacuation strategies. It is intended to
link weather data with the background air flow in the station and identify the
main driving forces for the dispersion of smoke or toxic agents throughout the
station. The subway air flow will be evaluated and then interfaced with a VR
simulation of the station and adjoining tunnels to produce a visual diagnostic and
predictive tool. This data, with existing pedestrian movement modelling, will
assist in identifying the correct evacuation procedures to minimise the exposure
time of individuals to a smoke filled environment. The final outcome of this work
will be a decision making process with regard to the initiation of effective
measures to minimize the impact of fire within a subway that can be used at the
initial design period to evaluate proposed evacuation routes.

Keywords: Subway Climatology, Smoke and Fire CFD Simulation, Tunnel

Safety, Emergency Ventilation

INTRODUCTION ground station passages (Yang et al., 2006) making

The first urban railway system and the world's first
underground line (Metropolitan) opened on 10th
January, 1883, in London to cope with population
growth and increasing traffic congestion. Under-
ground railway systems have become one of the im-
portant facilities of urban transportation however
there is tremendous potential danger once a fire oc-
curs in a subway system or a tunnel. These types
of fires have some distinctive characteristics such
as complicated burning process, rapid spread and
high temperatures that disperse smoke very quickly.
This creates problems in narrow and enclosed under-
evacuation difficult. Research shows that over 60% of
all deaths in subway fires are attributable to people
being either wholly or partially overcome by gas or
smoke (Levy, 2010). The exposure time of individuals
in a smoke environment depends upon their speed
of movement and the direction they choose to travel.
National statistics imply that the provision and cor-
rect location of emergency exit signs is of paramount
importance (Chen, Chien, et al., 2003) as is provid-
ing effective smoke control systems and providing
correctly located emergency exits and escape routes.
The related research of verification of the effective-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 119

 ness of smoke control system have been done by Tan
(Tan et al., 2012). The research did at a typical subway
station of Taipei propose an innovative smoke con-
trol scheme with a platform edge door (PED), which
turns out to be much more efficient in evacuating
smoke than that currently used (Chen, Guo, et al.,
2003). This however requires an understanding of
the air flow that is driving the motion and dispersion
of the smoke. The aim of this research is to develop
an understanding of the air flow in the Monument
Metro Station in Newcastle upon Tyne and to link this
with the movement of smoke or other agents. Pro-
ducing a simulation of the dispersion of smoke and
evacuation of passengers in a virtual reality (VR) en-
vironment will inform the decision making process in
the event of a fire or emergency to assist in evacua-
tion strategy. The VR data obtained from this study
will extend the existing Virtual NewcastleGateshead
data base (Morton et al., 2012).
This program of work includes laser scanning of
the Monument Metro Station to capture an accurate
model of the station from point cloud data which is
then used as a background to display the air move-
ment determined by using CFD models. The air flow
will be evaluated and then interfaced with the VR
simulation to produce a visual diagnostic and pre-
dictive tool that will display the background air flow
within the station. This information can then be com-
bined with pedestrian movement simulation created
in the Legion software to determine the likelihood of
pedestrians being overcome by smoke and to deter-
mine the most effective escape and evacuation strat-
egy for a particular incident.
This research is timely because several existing
subway systems in Europe have reached the age
which need to be refurbished. There is also a need
in some countries for the construction of new sub-
ways and tunnels. This research will strengthen deci-
sion making processes; improve the effectiveness of
simulation in producing a safer environment in un-
derground stations. This research methodology will
also be applicable to many different types of large
building and structures such as shopping centres, air-
120 | eCAADe 32 - Towards Smarter Cities - Volume 1
ports even large ships as the air movement is inti-
mately linked to climatic conditions, ventilation sys-
tems, and local energy release from machinery such
as electric motors
Justification of Monument Station for Re-
search
The Monument Metro Station is the principal station
on the underground section of the Tyne and Wear
Metro system and is used by almost 6 million pas-
sengers per year (NEXUS, 2013). The total length of
the underground section is 4 km encompassing 6 sta-
tions as shown in figure 1. The underground section
of the north-south line goes from Jesmond station
to Central Station and the east-west line has open-
ings at Manors and St James. The two lines cross at
the Monument station which is situated fully under-
ground with the highest level being the ticket lobby
area. It has four platforms on two tracks that cross at
different levels. The platforms are accessed by escala-
tors or lifts from the ticket hall and there are connect-
ing stair wells between the different platform levels.
The pedestrian entrances/exits to the station are in
the Eldon Square shopping centre and at Grey Street
and Blackett Street. This station is widely used by
commuters travelling to the city centre.
Figure 1
Underground
Stations of the Tyne
and Wear Metro
System (NEXUS,
2013)
SUBWAY CLIMATOLOGY
From the early 70s, subway evacuation and tunnel
safety in emergency situations in case of fire has
aroused great concern (Apte et al., 1991; Policastro et
al., 1998). The knowledge about airflow interdepen-
dencies with the outer atmosphere and investigated
the effects of varying ventilation velocities of under-
ground transport system is urgently needed. Sub-
way Climatology as a new research field for the man-
agement of possible catastrophes in subway-systems
which was proposed by Pflitsch and Küsel based on
their work on cave systems. (Pflitsch et al., 2000;
Pflitsch, 2001; Pflitsch et al., 2003). The aim of this
present Subway Climatology research is to gain a bet-
ter understanding of how air mixes between the sub-
terranean levels and street-level (regional airflow) in
a subway station and to identify the main driving
forces for the dispersion of smoke or toxic agents
throughout the station. It has been shown (Pflitsch
et al., 2003) that a natural background air flow ex-
ists in a subway system that is active during the nor-
mal operation hours and influences the spreading of
gases. This air flow is mainly driven by stack effect
(Chen, Chien, et al., 2003) and independent of the
train movements and re-establishes quickly after the
passing of a train. The natural background air flow
is driven by the over ground climate so it exhibits a
seasonal variation similar to that found in the airflow
of barometric caves. Understanding the influence of
the background air flow on the ventilation of the sub-
way system is important because it has a strong in-
fluence on the direction and strength of the internal
air flow and air exchange within the station. This air
exchange is strongly influenced by chimney effects
within the underground buildings (for instance in es-
calator wells). A recent study (Pflitsch et al., 2013)
has shown that, in the event of a disaster, airborne
toxins in subway systems are dispersed mainly by air
movements caused by natural ventilation and chim-
ney effects that are produced by the pressure differ-
ential between station levels and street levels. Warm
air within a station naturally flows upwards towards
the over ground entrances by way of stairs and esca-
lators. The addition of energy to the air from the esca-
lator motors enhances the buoyancy effect accelerat-
ing and dispersion the air. However flow reversals or
recirculation has been noticed between different lev-
els in a station due to the interaction of the natural
air flow with the station ventilation system (Pflitsch
et al., 2012). This can lead to a blocking of normal
exit routes making them unsafe and in the worst case
making them death traps. As a consequence the in-
fluence of natural ventilation should be included in
the disaster planning in order to develop effective
strategies for the reaction to fires and the dispersal
of smoke. Thus knowledge about the internal air-
flow and the interdependencies with the outer atmo-
sphere is urgently needed.
Metro and subway system construction is on the
increase world-wide with longer tunnels being built,
often as parts of ever more complex transport sys-
tems. Every metro system is unique in terms of
method of construction and layout of the stations
(Beard, 2009). It is clear that the behaviour of a
subway system should be considered together with
the over ground climatology. The background air
flow is very complex but recent advances in comput-
ing power and developments in programming have
made it possible to approach an understanding of
these flows using computational methods. Unfortu-
nately the presentation of the air flow data is very dif-
ficult for non-experts (such as the metro operators)
to understand so it is essential that a better means of
displaying this information is also developed.
RESEARCH METHODOLOGY
The objectives of the project will be met by provid-
ing a platform that will promote shared understand-
ing and transfer of knowledge and expertise and by
furthering the development of Subway Climatology
provide a safer environment for subway passengers.
A working laboratory has been created at the Monu-
ment station by a consortium of Northumbria Univer-
sity, Ruhr University Bochum, Newcastle University
and Nexus. Several parameters such as air flow ve-
locity and direction, temperature and humidity data
are being recorded and processed to reveal details of
the sensitivity of the airflow momentum and energy
transfer capacity on external factors such as the wind
direction at the tunnel portals, the local over ground
weather and internal factors. The work includes laser
Towards Smarter Cities - Volume 1 - eCAADe 32 | 121
 scanning of the station to produce a VR representa-
tion that will be used as a background for the display Figure 2
of air movement determined by using CFD but the Point Cloud model
VR files will also be used to generate a CFD model of of Monument
the station. This will be combined with other models Metro Station (1)
such as those to predict flow in stair wells (A. Zohra-
bian et al., 1989; A. S. Zohrabian et al., 1989). This will
also be accompanied by tracer gas tests to validate
the CFD results. The air flow data will then be used
to determine the efficacy of evacuate routes, deter-
mine egress time and identify possible bottleneck lo-
cations in the station by combining the airflow re-
sults with an evacuation simulation using the propri- Figure 3
ety software LEGION. The eventual outcome will be Point Cloud model
an assessment of the evacuation scenario of the sta- of Monument
tion that will inform designers and architects of the Metro Station (2)
impact of building design features on the smoke dis-
persion and thus enhance security in the event of a
fire in the subway.

Laser scanning and 3D modelling

A full laser scan of the Monument station, which con-
tains 68 scans, has been produced. This includes the
four platforms and tunnels, the concourse area, esca- Figure 4
lators and stairs and part of the surround building at Model of two
the three exits as show in Figure 2 and 3. Figure 7 is a tunnels and
perspective view of Monument Metro station on the platforms created in
city centre map which shows the position of the un- 3ds Max
derground structure.
The point cloud data has been converted man-
ually to a 3D polygon mesh using the software 3ds Figure 5
Max. A point cloud plugin provides a simple solu- Location of the
tion to import point cloud data into 3ds Max called measuring points
Clouds2Max. This is a very effective method that around Monument
minimises the noise of the point cloud to produce Metro Station
smooth polygon models which can satisfy the accu-
racy required of CFD modelling. A simple model in
3ds Max of two tunnels and platforms located at the
same level is shown in figure 4.

122 | eCAADe 32 - Towards Smarter Cities - Volume 1

Figure 6 AIR FLOW MEASUREMENT AND WEATHER
Ultrasonic DATA ANALYSE
anemometers The measurement of temperature, humidity, wind
installed in the speed and direction have been done at 20 points (as
tunnel show in Figure 5) around the Monument Station in-
cluding 3 points outside the exit at street level. These
measurements have been taken 2 to 3 days every
month from October 2013 and will be repeated un-
til September 2014. This data will be used to validate
the CFD modelling of air flow inside the station.
Figure 7 Ultrasonic anemometers have been used in rele-
Location of the vant research (A Pflitsch et al., 2000) and installed in-
three measuring side the tunnel at Monument Station, Central station
points at street level and Haymarket Station since 2009 as show in Figure
6. The measure data will used to define the boundary
conditions in CFD modelling.
The three measurement points (as show in Fig-
ure 7) at the street level outside the three exits are
Figure 8 intended to measure the micro climate of Newcastle
October City centre where the Monument station is located.
Temperature Trend Figures 8 and 9 show the results of comparing the
line (Logarithmic) temperature, minimum and maximum wind speed
of the measured data with that recorded in a local
weather station located at Northumbria University.

TRACER GAS EXPERIMENT

Different types of tracer gas experiments were com-
plete in February 2014 by staff and students of
Northumbria University and the Ruhr University
Bochum. The tracer gas tests were carried out using
Figure 9 Sulphur Hexafluoride (SF6), which is commonly used
October Maximum in dispersal analysis (Andreas Pflitsch et al., 2012).
Wind Speed Trend The special sensors for detecting the concentration
line (Logarithmic) of SF6 were designed and developed by the part-
ners at Bochum, figure 10. A series of tests were per-
formed including releasing the gas in a moving train
and several tests were performed by releasing the gas
at different locations in the station.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 123


In experiment 4, the SF6 gas was released at the
lower platform during an operational break. From
previous research the natural background air flow
had been identified as flowing from south to north
in the then current weather condition. So it was
expected that the SF6 would migrate northwards
through the tunnel. A number of airflow sensors and
SF6 sensors were placed throughout the station as
shown in figure 11. Also shown is the SF6 release
point indicated by the shape representing a gas bot-
tle.
124 | eCAADe 32 - Towards Smarter Cities - Volume 1
The concentration of the SF6 as it moved Figure 10
through the station is shown in figures 12, 13 and Release of SF6 at
14. The gas was released at 02:29:00, and within 2 platform, SF6
minutes the sensors number 11, 13 and 14 detected sensors and
very high concentrations of SF6 as shown in figure 12. ultrasonic
The other sensors on this platform did not detect SF6 anemometers
confirming the direction of the natural background
air flow. Then a large amount of gas went to the up-
per platforms 3 and 4 through the connecting stair
well as indicated by the red line (sensor 19) in figure
13. The tracer gas dispersed over the whole level of
platforms 3 and 4 within 4 minutes and proceeded
to flow down the tunnels and also up towards the
concourse area through the escalator link as show in
figure 13 (red line of sensor 1). The tracer gas sen-
sors at the station exits, 5, 16 and 18 in the figure 14
indicated that most of the gas dispersed into Eldon
square shopping centre (pale blue line sensor 16).
Other tracer gas experiment results also showed
Figure 11
The location of
anemometers, SF6
sensors and release
point for test 4.
 the dispersion of the gas throughout the whole PEDESTRIAN SIMULATIONS
stations under different conditions. The path of The information obtained from the tracer gas exper-
the tracer gas was most unexpected and this indi- iments has been examined alongside the result of
cates the complexity of the flow regime that exists the evacuation simulation produced by the Legion
in stations making it difficult to predict the disper- software. As the research did at Alexanderplatz in
sion route of smoke and devise effective evacuation Berlin which combining climate data from the tun-
routes. nels, readings from tracer gas experiments and the
Figure 12 results of the pedestrian evacuation, conclude into
Concentration line a better prognosis of safe and unsafe evacuation
diagram of SF6 routes in certain situations (Brüne et al., 2014). The
sensors located at pedestrian simulation of Monument metro station
platform 1 and 2 considered the evacuation from 2 trains containing
250 people in each train at two different platforms at
different levels. It was also considered that 50 peo-
ple would be waiting on each platform and 50 people
would be at the concourse area. The numbers of peo-
ple evacuation with time is shown in Figure 15 and
the evacuation route indicating the concentration of
people is shown in figure 16.
It can be seen that the pedestrians are evacu-
Figure 13 ate from 2 minutes to 7 minutes after the beginning
Concentration line of the evacuation. Overlaying the tracer gas results
diagram of SF6 on this data provides an indication of the number of
sensors located at causalities that may be expected from an emergency
platform 3 and 4 incident. It can be seen that passengers on platform 1
south of the gas release point can safely exit the plat-
form by way of the escalator to the concourse area.
The tracer gas dispersed on the lower platforms (1
and 2) migrates through the stair well to platforms
3 and 4 (sensor 19) in around 4 minutes and then
travels to the concourse area via the escalator in 7
minutes (sensor 1). This means that the passengers
on platforms 3 and 4 and in the concourse area near
Figure 14
these platforms are at greatest risk. The evacuation
Concentration line
route from platforms 3 and 4 is less hazardous. It can
diagram of SF6
be seen that there is some congestion on the stairs
sensors located at
at the main exits to the concourse area but of great
concourse area
concern is the migration of the tracer gas dispersion
to Eldon square shopping centre through exit 2 (sen-
sor 16).
So for this situation the scenario of pedestrian
evacuation will be: First, the passengers on platforms
3 and 4 should be evacuated as quickly as possi-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 125

 ble to the concourse area. The passengers on the
lower platforms should leave the station by walking
down the track southwards against the background
air flow. This will take them away from the gas release
site and reduce the congestion at the main exits to
the station allowing speedier evacuation of the pas-
sengers on platforms 3 and 4.
CFD SIMULATION APPLICATION IN TUN-
NEL AND SUBWAY STATIONS STUDIES
The tracer gas experiment combined with the pedes-
trian simulation gives a highly accurate results of
evacuation plans but the tracer gas experiment only
represents one weather condition so does not give a
complete picture. In order to fully understand the air
flow dynamics in the station it will be necessary ex-
126 | eCAADe 32 - Towards Smarter Cities - Volume 1
tend this work to account for the seasonal variation of
the weather. Previous research by Wilby (Wilby, 2005)
shows that the local microclimate in cities is strongly
influenced by the so called "heat island effect" in
which the temperature and airflow are strongly influ-
enced by the presence of buildings. This has been
observed in Newcastle and reproduced by measure-
ments and CFD simulation of the area around the
Figure 15
Monument for a limited number of cases. The near-
Numbers of people
est weather station recording data on an hourly ba-
evacuation with
sis to the Monument station is on the university site
time
approximately one half kilometres to the east and on
the top of a 5 story building. The tests performed and Figure 16
shown in figures 8 and 9 show there is a loose cor- Density of people at
relation between the local weather data at the sta- evacuation route
tion and that recorded by the weather station. This is
not accurate enough to be a reliable estimate of the
situation at the monument but the weather station
data can be used as an input to a CFD model of the
area around the station. A whole year weather data
can be simulating through this model as methodol-
ogy to predict accurate climate outside station base Figure 17
on weather station data. Then use this micro climate CFD simulation of
data as an input of boundary condition in metro sta- Newcastle city
tion CFD modelling. This has been done and is shown centre micro
in figure 17 which shows the CFD model developed climate
from the Virtual Newcastle Gateshead City Model.
In the fullness of time it is hoped to produce a
comprehensive CFD model of the entire Monument
station taking into account the variable local climate
conditions, different operation situations and differ-
ent fire locations. Relative researches (Simcox, et al.,
1992; Chen, et al., 2003; Wen et al., 2007; Xiaojun,
2008; Gousseau, et al., 2011) of analyses and sim-
ulation ventilation, temperature distribution, smoke
dispersion and smoke control carried out using CFD
simulation in a tunnel or subway station. Numeri-
cal (CFD) three-dimensional simulations are provid-
ing the advantages such as accuracy, convenience,
and cost-saving and provides the results in a format
that can easily be post-processed to create anima-
tion.
CONCLUSION
The program of work outlined in this paper is in-
tended to develop a methodology and a tool capa-
ble of analysing the airflow in subway systems that
would include a CFD study of fire simulation and
smoke movement and dispersion in a subway sys-
tem. This research brings together complementary
groups working in the fields of Subway Climatology,
Virtual Reality, Pedestrian Simulation, Human Factors
and Emergency Preparedness for gaining a better in-
sight of the impact of a fire in a subway system. This
research integrates domains of knowledge that hith-
erto have been used in isolation and in so doing
brings a holistic approach that strengthens the de-
cision making processes and improves the effective-
ness of simulation and design exercises and will pro-
duce a safer environment for the travelling public.
The final outcome of this research can be assist-
ing in operational planning for a possible incident
for the Emergency Services, can be used as a train-
ing tool to familiarize maintenance workers with de-
tails of the tunnels for the subway operator and can
also be used to simulate emergency incidents. It will
be possible to assess escape and evacuation strategy
for a particular incident. For the architects, design-
ers and planners it will be used as an early stage de-
sign tool to assess the functionality of proposed new
tunnels and station designs and examine the result
of proposed changes in the layout of existing sta-
tions. This research methodology can be extended
to many other types of buildings or structures such
as airports, and shopping centres that are suscepti-
ble to the same risks as subway systems.
ACKNOWLEDGEMENTS
The author would like to express thanks for the sup-
port given to this project by Northumbria Univer-
sity (especially Dr. James Charlton and late Graham
Kimpton), the Ruhr University Bochum and Nexus
(Tyne and Wear Passenger Transport Executive).
REFERENCES
Apte, V., Green, A. and Kent, J. 1991 'Pool fire plume flow
in a large-scale wind tunnel', Fire Safety Science 3rd
Int. Symposium
Beard, A. N. 2009, 'Fire safety in tunnels', Fire Safety Jour-
nal, 44(2), pp. 276-278
Brüne, M., Charlton, J, Pflitsch, A and Agnew, B 2014
'Coupling tracer gas experiments with evacuation
simulation An empirical approach to assess the ef-
fectiveness of evacuation routes in subway stations',
ISTSS
Chen, F., Chien, S. W., Jang, H. M. and Chang, W. J. 2003,
'Stack effects on smoke propagation in subway sta-
tions', Continuum Mechanics and Thermodynamics,
15(5), pp. 425-440
Chen, F., Guo, S. C., Chuay, H. Y. and Chien, S. W. 2003,
'Smoke control of fires in subway stations', Theo-
retical and Computational Fluid Dynamics, 16(5), pp.
349-368
Gousseau, P., Blocken, B., Stathopoulos, T. and Heijst,
G. J. F. 2011, 'CFD simulation of near-field pollutant
dispersion on a high-resolution grid: A case study
by LES and RANS for a building group in downtown
Montreal', Atmospheric Environment, 45(2), pp. 428-
438
Levy, C. J 2010, 'Moscow attack a test for Putin and his
record against terror', New York Times, March 30th, p.
A1
Morton, P. J, Horne, M, Dalton, R and Thompson, E.
M 2012 'Virtual City Models: avoidance of obso-
lescence.', Digital Physicality-Proceedings of the 30th
eCAADe Confer-Comparative evaluation of the two
methodologies.
Pflitsch, A. 2001, 'Investigations on air currents in un-
derground public transportation systems', Meteorol-
ogische Zeitschrift, 10(4), pp. 239-246
Pflitsch, A, Arnold, U and Combs, P 2000 'An initial study
of linkages between subterranean and street-level
climate in New York City', Paper presented at the
Biometeorology and Urban Climatology at the Turn of
the Millennium. ICB-ICUC.
Pflitsch, A, Bruene, M, Steiling, B, Killing-Heinze, M, Ag-
new, B, Irving, M and Lockhart, J 2012, 'Air flow mea-
surements in the underground section of a UK light
rail system', Applied Thermal Engineering, 32, pp. 22-
30
Pflitsch, A, Brüne, M, Killing-Heinze, M, Ringeis, J, Agnew,
B and Steiling, B 2013, 'Natural Ventilation as a Factor
Controlling the Dispersal of Airborne Toxins in Sub-
way Systems in a Disaster Situation', Journal of Trans-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 127
 portation Safety & Security, 5(1), pp. 78-92
Pflitsch, A and Flick, B 2000, 'Proof and characterization
of slowest air currents using sonic anemometers in
urban and topographical climatology research', The
Journal of Applied Sensing Technology, 17, pp. 75-82
Pflitsch, A and Küsel, H 2003, 'Subway-Climatology–New
research Field for the Management of possible
Catastrophes in subway systems', Man and Climate
in the 20th Century, Studia Geograficzne, 75, pp. 384-
394
Policastro, A and Coke, L 1998, 'Protecting US subway
systems from chemical/biological terrorism', Transit
Policing, 8(1), pp. 5-20
Simcox, S, Wilkes, N and Jones, I 1992, 'Computer simu-
lation of the flows of hot gases from the fire at King's
Cross underground station', Fire Safety Journal, 18(1),
pp. 49-73
Tan, J, Xie, Y, Wang, P, Wang, T and Hu, B 2012, 'Experi-
ments for Verification of the Effectiveness of Smoke
Control System in a Typical Subway Station', Przegląd
Elektrotechniczny, 88, pp. 78-81
Wen, J. X., Kang, K, Donchev, T and Karwatzki, J 2007, 'Val-
idation of FDS for the prediction of medium-scale
pool fires', Fire Safety Journal, 42(2), pp. 127-138
Wilby, R 2005 'Urban heat island and air quality of Lon-
don', http://www.asp.ucar.edu/colloquium/2004/CH/
presentations/AirQualityTutorialBackground.pdf, UK
Xiaojun, C 2008, 'Simulation of temperature and smoke
distribution of a tunnel fire based on modifications
of multi-layer zone model', Tunnelling and Under-
ground Space Technology, 23(1), pp. 75-79
Yang, G.S, An, Y. L, Peng, L. M and Zhang, J. H 2006, 'Simu-
lation of smoke flow and longitudinal ventilation in
tunnel fire', Transactions of Nonferrous Metals Society
of China, 16(3), pp. 741-746
Zohrabian, A, Mokhtarzadeh, M and Reynolds, A 1989
'Buoyancy-driven air flow in a closed half scale stair-
well model: velocity and temperature measure-
ments', The 10th AIVC Conference, Dipoli, Finland
Zohrabian, A. S, Mokhtarzadeh-Dehghan, M. R,
Reynolds, A. J and Marriott, B. S. T 1989, 'An experi-
mental study of buoyancy-driven flow in a half-scale
stairwell model', Building and Environment, 24(2), pp.
141-148
[1] http://www.nexus.org.uk/metro

128 | eCAADe 32 - Towards Smarter Cities - Volume 1

Building-up urban open spaces from shadow range
analyses
Estefania Tapias1 , Shubham Soni2
1
ETH Zurich, Chair of Information Architecture 2 Indian Institute of Technology
Bhubaneswar
1
https://www.ia.arch.ethz.ch/ 2 http://www.iitbbs.ac.in/
1
tapias@arch.ethz.ch 2 ss19@iitbbs.ac.in

This paper explores an alternative approach for the creation of new built forms
based on solar access analysis. Consolidated on urban areas under development,
the denominated 'inverted' approach is focused on the generation of recreational
open spaces based on shadow conditions caused by existing built forms, and as a
starting point for the construction of new urban envelopes as possible
development areas. Unlike the existing method of the 'solar envelope', the
'inverted' approach shows an alternative procedure for the construction of built
forms, based on pedestrian comfort caused by solar access in urban spaces
rather than on indoor performance affected by the penetration of sunlight into
buildings. As a method for the creation of urban envelopes, this approach
attempts to enhance pedestrian comfort according to the study of solar access in
urban areas. The 'inverted' approach is based on sun path data and is developed
as a generative procedure, where the results of shadow range analyses and the
different urban objectives work as input parameters for the generation of urban
envelopes. Based on this methodology, two Grasshopper® custom components
are developed.

Keywords: urban open spaces, solar access, shadow range simulation,

generative modelling

INTRODUCTION nificantly more heat than rural areas (Howard, 1818).

Climate is the pattern of variation in precipitation,
temperature, humidity, sunshine, wind velocity, phe-
nomena such as fog, frost, and hail storms, and other
measures of the weather that occur in a given region
over long periods of time [1]. Urban climate refers to
climatic conditions in an urban area that differs from
the climate of its rural surrounding and is attributed
to urban developments. Cities absorb and retain sig-
This warmth of cities in contrast to their surrounding
is known as 'Urban Heat Island' or UHI (Oke, 1976). Ur-
ban climate conditions affect how cities will develop
in the future, not only because of the impact on the
environment or on the energy consumption of build-
ings, but also on the human comfort in urban spaces.
While there is a growing field on bio-climate archi-
tecture where knowledge on how to design climate-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 129

 sensitive indoor environments is increasing, there is
still a lot to explore on the construction of the urban
open spaces according to microclimatic conditions
(Calthorpe, 2011).
With the understanding of urban microclimate
and its effects on architecture and urban design, the
application of different strategies to control solar ac-
cess have been fundamental when it comes to de-
signing buildings and urban areas (Erell; Pearlmutter;
Williamson, 2011). Therefore, it is important to con-
sider the study of solar exposure where the degree
of exposure to solar radiation is one of the main con-
trols on microclimate conditions. These strategies
can be divided into two purposes; the solar access
for buildings and the solar access for pedestrians
(open public spaces) (Erell; Pearlmutter; Williamson,
2011). The first one is the basis for the 'solar enve-
lope' developed by Ralph L. Knowles. The latter one
is related to the outdoor thermal comfort, and is the
basis for the 'inverted' approached introduced in this
paper.
Outdoor spaces are important for cities as these
provide daily pedestrian traffic and different outdoor
activities contributing to urban livability and vital-
ity (Chen; Ng, 2011). Promoting the use of streets
and outdoor spaces by pedestrians will benefit cities
from physical, environmental, economical, and so-
cial aspects (Hakim et al., 1998). In this way, ensur-
ing that people are comfortable in outdoor spaces
is essential to high-quality urban living. Over the
past few decades, making outdoor spaces attractive
to people, and ultimately used by them, has been
increasingly recognized as a goal in urban planning
and design (Chen & Ng, 2011; Gehl & Gemzøe, 2004;
Carr, et al., 199) Among many factors that determine
the quality of outdoor spaces, the urban microcli-
mate is an important one. Pedestrians are directly
exposed to their immediate environment in terms of
variations of air temperature, relative humidity, wind
speed, and solar radiation. Therefore, people's sensa-
tion of thermal comfort is greatly affected by the local
microclimate (Chen & Ng, 2011).
The outdoor thermal environment is impacted
130 | eCAADe 32 - Towards Smarter Cities - Volume 1
by the built environment in terms of; anthropogenic
heat (Ichinose et al., 1999) ground surface covering
(Lin et al., 2007), evaporation and evapotranspira-
tion of plants (Robitu et al., 2006) and shading by
trees and man-made objects. The outdoor thermal
comfort is generally studied on the urban micro-scale
level, which is affected by the Canopy Layer Heat Is-
land (CLHI) (Voogt, 2004; Oke, 1995; Oke, 1976). As
shade can block incident solar radiation, some stud-
ies have discussed shading effect on thermal envi-
ronments. For example, street orientation and the
height/width (H/W) ratio have been measured to as-
sess the shading levels in some studies (Emmanuel
et al., 2007; Johansson, 2006). In the context of ur-
ban planning, how outdoor thermal environment in-
fluence thermal sensations of people and their be-
haviour (use of outdoor spaces) is of great interest for
designing urban spaces.
This paper investigates a method for the creation
of urban open spaces based on shadow conditions
caused by existing built forms, and as a starting point
for the construction of new urban envelopes as pos-
sible development areas.
'SOLAR ENVELOPE' APPROACH: SOLAR
ACCESS FOR BUILDINGS
The referred concept of 'solar rights' is the guaran-
tee of exposure of a building to direct sunlight in a
predetermined period in order to meet the require-
ment for indoor passive solar heating. The building
surfaces that must receive the benefits of sunlight are
usually vertical and equator-facing. This information,
within other requirements of solar access for build-
ings, leads to establishing limits on height or volume
of buildings. Such limits were referred as 'solar en-
velopes' by Ralph L. Knowles in 1974 as a framework
for architecture and urban design to provide solar ac-
cess needs according to building geometry, and sup-
port solar energy for future developments (Knowles;
Berry, 1980).
Nowadays there are simple ways and tools to
measure sun path movements. The most know is the
Stereographic sun-path diagram, which represents
 annual changes in the path of the Sun through the
sky in a single 2D diagram. These diagrams provide a
summary of solar position that the designer can re-
fer to when considering shading requirements and
design options. Based on this diagram, environmen-
tal analysis tools can generate visualizations on how
the sunlight affects the built environment. One way
to measure this effect is to analyse how shadows are
form according to building obstacles. The 'solar en-
velope' builds-up from this knowledge in order to
create a method to avoid shadowing to neighbour-
ing buildings in desirable angles. Recently, the en-
vironmental analysis Diva for Rhinoceros® Grasshop-
per® developed a 'solar envelope component' based
on Knowles theory and which "creates the largest
build-able volume that will not shade its neighbours
for a specified period of time throughout the year"
[2].
THE 'INVERTED' APPROACH: SOLAR AC-
CESS FOR PEDESTRIANS
This paper proposes a method for controlling so-
lar access, like the 'solar envelope', but for avoiding
shading in potential recreational open spaces, rather
than avoiding shading for neighboring buildings.
The method called 'inverted' approach is developed
by means of parametric modeling in Rhinoceros®
Grasshopper®, where a series of components are de-
Figure 1
Delimitation of
development areas
(i).
veloped in order to generate urban envelopes based
on the 'inverted' method.
Methodology and application
This methodology is intended for urban areas under
development, as the final aim is to create possible ar-
eas for future urban densification based on the gen-
eration of urban open spaces that enhance pedes-
trian outdoor comfort. Therefore, the case study for
the development of the method was the Thälmann-
Park in the centre of the Prenzlauer Berg district in
Berlin. Today this area features public houses as well
as art galleries and a small theatre at the former ad-
ministrative building of a previous gas plant. The in-
terest of investors in the area of the Thalmann Parkes
increased raplidly and projects for the construction of
new living areas are currently under development.
The different procedures explored for the 'in-
verted' approach method are based on generative
and parametric modelling techniques that allow the
systematic creation of new urban forms according
to different solar access and urban boundaries cri-
teria. According to these criteria, the method is di-
vided into three consecutive parts; (i) delimitation
of the development areas, (ii) shadow range gener-
ation, and (iii) selection of recreational urban areas.
At the end, areas for future building densification
are created and extruded in order to generate the fi-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 131

nal building envelopes (possible volumes for build-
ing densification).
The first part of the method aims to establish a
ground-level geometry of the possible building ar-
eas according to the limit distance between buildings
and on the urban block boundaries (city normative).
This is done first by taking the urban block area as
a surface for possible developments. Subsequently,
measuring the minimum distance between building
(city normative) from the current buildings and ex-
tracting this area from the urban block surface. At
this point, an area for possible future developments
is created as shown in figure 1.
The second part is focus on the execution of the
shadow range analysis for summer and winter peri-
ods. For this analyses, the Rhinoceros® Grasshopper®
component called Geco is used. Geco is a custom
component for Grasshopper® that offers a direct link
between Rhinoceros® Grasshopper® models and Au-
todesk® Ecotect®. Ecotect® is an environmental anal-
yses and simulations software intended for architects
to work with highly visual results. It is aimed for early
stages of conceptual designs for the exploration of
environmental factors and interactions. One of the
simulations that can be performed in Ecotect® is the
sun-path simulation based on a particular day of the
year and hour of the day. This feature can also recre-
ate the shadows of the specific time, furthermore,
132 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 2
Shadow range
generation (ii),
(blue) winter
solstice shadow
range and (red)
summer solstice
shadow range.
it can visualise the range of shadows during a pre-
defined day of the year (specifying from which hour
until which hour of the day). All of this simulations
can be used in Rhinoceros® Grasshopper® through
Geco. For this specific case, the shadow range sim-
ulations over the solstice of Winter (21st of Decem-
ber) and Summer (21st of June) where generated.
The purpose of using the two solstice of the year
is to obtain the maximum and the minimum range
of shadow during the year in the particular location.
An additional purpose is to have the possible shad-
ows for the periods of Winter and Summer in order
to make further decisions on the conditions of open
spaces during cold and warm periods of the year.
After the simulation is conducted for each day, the
shadow range result is converted into geometry data
in Grasshopper®. These two separate geometries are
then combined in order to overlay the information as
shown in figure 2 with the blue surfaces as the winter
solstice shadow range and red surfaces as the sum-
mer solstice shadow range.
As a final step, the third part is based on selecting
the specific areas for potential open urban spaces ac-
cording to pre-defined shadowing conditions. These
conditions are based on the desirable characteris-
tics of the urban open spaces for recreational pur-
poses. In this specific case, the conditions are estab-
lished based on open spaces that have shadows dur-
Figure 3
Selection of
recreational urban
areas and
generation of
volumes for future
densification (iii).
ing summer but not during winter. In this way, the
outdoor thermal comfort will be enhance due to the
sunlight penetration in this areas which affect cold
and warm periods of the year. In terms of proce-
dure, the two surfaces generated in the previous part
(shadow ranges of winter and summer solstice) are
overlaid and the segments where the summer shad-
ows without the winter shadows are selected (the red
surfaces). Moreover, a buffer zone is created around
this surface to prevent future building shadowing.
With the combination of the buffer zone and the se-
lected surface, this final area (yellow area in figure
3) is extracted from the urban block surface created
in the first part of the method. The final block sur-
face is then denominated as the areas for future de-
velopment. At the end, this final surface is extruded
according to the city normative (maximum high of
buildings) and urban volumes are created. These fi-
nal envelopes are the possible areas for future build-
ing densification as shown in light blue in figure 3.
The remaining urban spaces are the potential open
areas for recreational purposes like parks and urban
squares. In this way, these areas enhance the outdoor
thermal comfort according to shadow range analy-
ses.
For the first (i) and last (iii) part of this method,
two Rhinoceros® Grasshopper® components are de-
veloped. These custom components are linked to the
Geco components used in the second part (ii) to link
the weather data of the specific location and gener-
ate the geometry of the shadow range simulation.
Grasshopper components
Both components are named SUD, which stands for
"Strategic Urban Densification". The custom com-
ponent for the first part of the method (i) is called
SUD1, and the one for the last part of the method
(iii) is called SUD2. The aim for the development of
these custom components is to reduce complexity of
grasshopper with existing components and to com-
press and simplify the procedure of the proposed
method and make it applicable to other case studies.
The development of SUD1 (figure 4) is based
on the first part of the proposed method explained
above. The input parameters are:
• Buildings: takes building footprints as a first
input parameter.
• Distance: takes the minimum distance be-
tween buildings according to the city norma-
tive. This second parameter creates an offset
of the geometry of the building footprints
• Blocks: takes the urban block perimeter as a
third parameter.
SUD1 first performs an offset command according to
Towards Smarter Cities - Volume 1 - eCAADe 32 | 133

the number given by the Distance input. Then it elim-
inates the overlapping geometry by combining the
different curves and taking the perimeter of the re-
sult. In this way, a single curve is created for each of
the urban blocks.
The output parameter is called surface, but is ac-
tually a curve which later needs to be converted into
surface. This last output represents the area for pos-
sible future developments as shown in figure 1.
The development of SUD2 custom component
(figure 5) is based on the third part of the method (iii)
and takes as input parameters the results (outputs)
of the previous custom component (i) and the Geco
definition (ii). The input parameters are:
• Shadow R1 & R2: takes shadow range 1 and
shadow range 2 from Geco component after
analysis.
• Tolerance: takes tolerance number to reduce
line segments in order to simplify geometry.
• Offset: SUD1 takes real number as a second
parameter to create offset of buildings.
• Buffer V: Buffer vector. Vector from the gen-
eration of Shadows in Geco from Shadow R2.
134 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 4
Grasshopper
custom
component, SUD1
(i).
Figure 5
Grasshopper
custom
component, SUD2
(iii).
A number is linked to the vector in order to
move the geometry to a given distance.
• SUD1: takes resulted surface from the SUD1
component as a parameter.
SUD2 takes input surfaces from Geco component
as the shadow range geometries of the two solstice
of the year. According to the pre-conditions, the
shadows are overlaid and one of the shadows is se-
lected and the overlaid regions created by the second
shadow are deleted. Shadow R1 is the selected one
and Shadow R2 is the one creating the overlaid re-
gion. Eventually Shadow R2 is deleted as well, its pur-
pose is only to created the overlaid region in Shadow
R1. Before going into the next step, these surfaces
have to me simplified by taking input tolerance to re-
duce line segments. Until this point, this area created
from this curve in the first state of the output (urban
open areas for recreation). After the simplification is
done, the component makes an offset of the curves
and then this offset geometry is moved in the same
direction (vector) of the Shadow R2 in order to create
a buffer zone to protect the area created by Shadow
R1 from future Shadow R2 (in this case, the winter
shadows). In a following step, the component cre-
ates a union of all curves created until this point as
show in yellow in figure 3. An overlapping of the final
curve and the input parameter from SUD1 is gener-
ated where the region of intersection is deleted from
the SUD1 surface.
The output parameter is called surface but, as the
output of SUD1, it is actually a curve, which can be
converted to a surface. This final output of the SUD2
needs to be extruded with the Grasshopper® exist-
ing components in order to create the final volume
as shown in light blue in figure 3.
CONCLUSION
As a result, building envelopes are created from so-
lar access analyses in order to explore possible di-
rections for climate-sensitive urban growth. This
method can work as a previous process for the 'so-
lar envelope', as the final urban envelopes from the
'inverted' approach can be used for the process of
controlling solar access for buildings. In this way,
not only the urban spaces but also indoor spaces can
benefit form solar access. Eventually, this method
can contribute to urban planning practices as a
decision-support tool for the construction of climate-
sensitive urban forms.
Additionally, the Grasshopper custom compo-
nents developed for this method can be used to
apply the purposed method to other case studies
around the world, always taking into consideration
the weather data of the specific location. Before ap-
plying this method to other urban settlement, it is im-
portant to consider that this method is intended for
urban areas under development only.
FURTHER WORK
As for the 'solar envelope', different researches have
been taking place towards optimizing the different
design possibilities, for this particular approach the
idea is to use evolutionary algorithms (EAs) for multi-
objective optimization (as this shadow range analy-
ses may encounter contradicting criteria) in order to
explore different optimal solutions for the construc-
tion of new urban envelops.
ACKNOWLEDGMENTS
This research work was developed under the frame of
a Master Plan design project led by GDB Planung, an
architect studio based in Berlin, Germany. A special
thank you to Matthias J. Götz, one of the lead archi-
tects of GDB Planung.
REFERENCES
Calthorpe, P 2011, Urbanism in the Age of Climate Change,
Island Press , US
Chen, L and Ng, E 2011, 'Outdoor thermal comfort and
outdoor activities: A review of research in the past
decade', Elsevier, Cities 29, p. 118–125
Emmanuel, R, Rosenlund, H and Johansson, E 2007, 'Ur-
ban shading - a design option for the tropics? A
study in Colombo, Sri Lanka.', International Journal of
Climatology, 27, p. 1995–2004
Erell, E, Pearlmutter, D and Williamson, T 2011, Urban Mi-
croclimate: Designing the Spaces Between Buildings,
Earthscan, London
Hakim, A, Petrovitch, H, Burchfiel, C.M., Ross, G. W., Ro-
driguez, B. L., White, L. R., Yano, K and Curb, J. D.
1998, 'Effects of walking on mortality among non-
smoking retire men.', N. E. Journal of Medicine, 338(2),
p. 94–99
Howard, L 1818-1820, The Climate of London, Deduced
from Meteorological Observations, Made at Different
Places in the Neighbourhood of the Metropolis., Vol. 2,
London
Ichinose, T, Shimodozono, K and Hanaki, K 1999, 'Impact
of anthropogenic heat on urban climate in Tokyo.',
Atmospheric Environment, 33, p. 3897–909
Johansson, E 2006, 'Influence of urban geometry on out-
door thermal comfort in a hot dry climate: a study in
Fez, Morocco.', Building and Environment, 41(10), p.
1326–1338
Knowles, R and Berry, R 1980, Solar Envelope Concepts:
Moderate Density Building Applications : Final Report,
Solar Energy Information Data Bank , US
Oke, T.R. 1976, 'The distinction between canopy and
boundary layer urban heat islands.', Atmosphere, 14,
pp. 268-277
Robitu, M, Musy, M, Inard, C and Groleau, D 2006, 'Mod-
eling the influence of vegetation and water pond on
urban microclimate.', Solar Energy, 80, p. 435–47
[1] http://en.wikipedia.org/wiki/Climate
[2] http://diva4rhino.com/user-guide/grasshopper/solar
Towards Smarter Cities - Volume 1 - eCAADe 32 | 135
136 | eCAADe 32 - Towards Smarter Cities - Volume 1
Aerodynamic strategy applied in an urban shelter design
Simulation and analysis of aerodynamic phenomena in an urban context

Rafael Moya1 , Daniel Prohasky2 , Simon Watkins3 , Yan Ding4 , Jane Burry5 , Mark
Burry6
1,2,3,4,5,6
RMIT University, Melbourne, Australia
1
s3296513@student.rmit.edu.au
2,3,4,5,6
{daniel.prohasky|simon|yan.ding|jane.burry|mburry}@rmit.edu.au

This paper presents an experimental study on strategies of utilizing wind as an

architectural element, proposing the reconfiguration and projection of wind
patterns to produce vaults of wind as regions of shelter in the outdoor
environment. It shows an aerodynamic analysis and exploration of barriers,
deflectors and porous screens in an existing urban wind canyon for a hypothetical
urban shelter in a tram stop area. Computational Fluid Dynamics (CFD)
software and physical tests in a wind tunnel using microelectronic hot-wire
anemometry are the methods utilised. The experiments involve a comparison
between screens with impermeable surfaces and porous membranes and their
ability to project wind as architecture. The experiments showed that the use of
porous membranes improves the mitigation level of wind speed and turbulence
intensity in the wind vaults regions.

Keywords: Urban aerodynamics, CFD simulation, wind discomfort, wind tunnel

BACKGROUND ing local conditions of the urban configuration, wind

In many cases, urban configuration changes the pat-
tern of wind flow, generating stream flows at the
ground level and accelerating the wind; thus pub-
lic spaces may require mitigation methods for local
wind issues at the ground level (Stathopoulos, 2009).
Even though, existing strategies to change or ame-
liorate pedestrian wind conditions near buildings
are documented in the literature (Cochran, 2004), in
some cases, the excessive use of canopies and trees
difficult a normal incidence of sunlight or pollution
dispersion in the outdoor environment. In general,
it is necessary to elaborate new analysis consider-
speed, intensity and direction. For this reason, tech-
nologies for visualisation and wind analysis have be-
come crucial to gain comprehensive knowledge of
wind dynamics in an urban context to elaborate new
strategies of design (Kim et al., 2011).
The field of numerical simulations (CFD) is under
an intense process of change and improvement while
wind tunnels are a reliable technology. In general, for
architects there is a complementarity of both tech-
nologies in the professional field with their advan-
tages and disadvantages (Salim and Castro, 2012).
However, in the more specific area of the early design

Towards Smarter Cities - Volume 1 - eCAADe 32 | 137


stage, CFD has become a useful tool for architects;
especially, with new generations of CFD programs,
such as Vasari, developed to be used by designers in
the design process [1].
Moreover, porous membranes and permeable
structures are being explored with new morphologi-
cal approaches for designers to provide conditions of
comfort. The current challenge is the adaptation and
optimisation of these designs to be used in an urban
context. In this sense, the studies of Jacques Gan-
demer about aerodynamic features of artificial wind-
breaks (Gandemer, 1979) provided a group of design
rules to elaborate strategies of wind pattern manip-
ulation that can be applied for the designer's explo-
rations. These aerodynamic features might be fins,
gaps, slots, or a graduation of porosity. A combina-
tion of these features can be strategically integrated
into a flat porous screen. This can change the wind
flow dynamics around the screen, increasing the pro-
tection area behind, while reducing the blockage fac-
tor of the wind mitigation feature. By reducing the
geometrical domain of the element, opportunities
arise in exploring wind as an architectural element.
RESEARCH AIM
The research aim is to analyse the differences in per-
formance of impermeable surfaces and porous mem-
branes to create architecture with wind and wind as
architecture.
This research involves the analysis of windbreak
aerodynamic features: fins as deflectors, porous
138 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 1
Sketch of
protection
bubble-regions.
The arches
represent the wind
flow deflected
screens, and slot systems. The approach is to use
a passive strategy to manipulate wind patterns in
an outdoor environment, to generate vaults of wind
as protection regions for pedestrians (Figure1). This
means the exploration is mainly about the wind
aerodynamic patterns rather than to build complex
screen designs. For the visualisation and analysis of
the wind around the configuration of screens, experi-
ments with CFD simulations and a physical wind tun-
nel were performed. In the case of the wind tunnel
experiments, the goal was to verify the architecture
of the wind projected by porous membranes that are
too complex to analyse using Vasari.
EXPERIMENTAL METHOD
The context
Because of the complexity of the conditions involved
with aerodynamic urban phenomena and the limita-
tion of wind measurement instruments, it is neces-
sary to simplify the simulation and to use standard
conditions as parameters for this analysis. Thus, typ-
ical wind phenomena and context are used as the
case for this research: an aerodynamic phenomenon
called Channel Effect (Gandemer et al., 1978) which
produces pedestrian discomfort due to high wind
speed along a narrow street (Erell, Pearlmutter, et al.
2011). This kind of phenomena is very common in the
city and can be viewed as a representation of the case
observed in a tram stop area in the business district
of Melbourne, Australia. For instance, figure 2 shows
the wind measurement in this typical tram stop area
 in Swanston Street in the CBD of Melbourne (Figure
2). The wind fluctuation in this place is over 5 m/s,
which is considered as beyond the discomfort thresh-
old for wind speed in this context (Figure 3).
Figure 2
Tram-stop area in
the business district
of Melbourne
Figure 3
Fluctuation of wind
speed in the tram
stops area (weather
station data)
Thus, because of the street configuration, which pro-
duces this kind of phenomenon, the area analysed
must be defined considering this two spaces: the
street space and a tram stop site. Swanston Street is
orientated from South to North and has 30m width
Figure 4 including a tram track of the Public transport sys-
Configuration of tem. The area is surrounded by tall buildings that
the tram-stop with shape a closed channel. In this street a tram stop was
screens and roof built near the intersection between Swanston St. and
Franklin St. The tram stop has an area of 4.5m width
and 69m length on the west sidewalk of the street.
The site has urban furniture and shelter of two glass
roofs to protect pedestrians. The problem of discom-
fort is produced due to strong wind speeds running
from the north to the south, along the street and near
the corners.
The localised wind phenomena are difficult to at-
tribute to the influence of tall buildings surrounding
the site. But some statistics show a trend of preva-
lent winds from North to South during the second
half of the year (Bureau of M, 2013)[2]. This is coin-
cident with the personal experience of the authors:
recognition that this problem is more frequent in the
winter season. Here, there is no protection from natu-
ral foliage of trees and the wind flows at ground level
with high velocity.
The tram stop configuration
The installations are physical aerodynamic features
based on windbreak designs to generate wind
thresholds. This can be understood as upward
deflections of the wind that produce a chain of bub-
bles of low wind speed (protection regions). There-
fore, the wind mitigation elements must create a
layer of low pressure above the tram stop area. These
elements are gathered in five groups along the tram
stop, including a structure of the roof for the shelter.
With the association of three essential elements: a
porous screen with 20% porosity, a horizontal curved
fin as a deflector with a double slot system and a
convex deflector above the barrier, the main require-
ment is to keep a low porous barrier (1.5m height).
The area of protection is associated with the wind
deflection (wind vault) created by the porous barrier.
The tram stop configuration has five parts, four
vertical deflector components and a roof (Figure 4).
These elements are organised strategically to define
four regions within the tram stop area.
Above each screen, there is a curved surface of 2m
height. This surface is a deflector that creates a wake
region and low pressure layer at the top of the protec-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 139
 tion area. The idea is to produce a vertical lift effect of
deflected wind to increase the height of this region.
The total height of this screen is 5m (Figure 5).
EXPERIMENTAL PROCEDURE PART 1
Numerical simulation using CFD software
The initial design concept was evaluated using CFD
to verify the most relevant phenomena and intro-
duce modifications for further tests. The use of the
engineering CFD software can be difficult for archi-
tects because it requires expertise in numerical anal-
ysis and knowledge in wind engineering. Alterna-
tive easy-to-use programs, such as Vasari, can be
used, but they have limitations in wind parameters
such as the simulation of the atmospheric boundary
layer (ABL) effect. This parameter can produce strong
changes in the CFD analysis outcomes. The software
used in these simulations was ANSYS to take advan-
tage of its extended capabilities to represent wind
phenomena in an urban context i.e. ABL simulation.
But the complexity of the context and aerodynamic
phenomena can be an additional challenge for an ar-
chitect to realise this kind of simulation. Thus, sim-
plification and standardisation of many factors facili-
tates the simulations and the feedback with these ini-
tial experiments. For the numerical simulation, the
urban context was represented as a simple and reg-
ular geometry of a channel. This geometry is a fluid
domain where the channel effect was recreated as a
digital mesh. This volume was 30m wide (similar to
the distance between opposite facades), 16m high
and 69m long. Four faces of the geometry represent
140 | eCAADe 32 - Towards Smarter Cities - Volume 1
the ground, sky and street facades, while two end
faces represent the fluid inlet and outlet zones. A low
grade of roughness is given to these surfaces to sim-
ulate the friction of facades and ground.
Figure 5
The wind profile at the inlet was setup as flow
Porous screen of
with a boundary layer effect, considering the power
20% of density.
boundary layer equation and using a reference ve-
Horizontal
locity relevant to the discomfort threshold - a speed
deflectors with
of 5m/s at 1.5m height (Wisse, 1988). Also, in this
double slot as
case, wind flows only in one direction and the cor-
accelerator. Convex
ner wind effect is omitted. The turbulence intensity
deflector above of
was assumed to be 5%, and other objects were not
the barrier
considered in the area of the tram stop. Inside the
Fluid Domain, a refined digital mesh is built achiev-
ing a more accurate solution of the wind phenomena
such as the wake regions, the chain of aerodynamic
bubbles and the level of wind deflection around the
tram-stop. This refined domain has 10m high, 15m
wide and 46m in length.
CFD Simulation Results
The engineering CFD simulation of these aerody-
namic features showed several strong upward deflec-
tions (wind vaults) and protection regions behind
each porous barrier, and along the area of the tram
stop. This analysis and clear visualisation informs the
designer of the patterns of wind flows for further ex-
periments in the physical wind tunnel. One of the
findings is that the aerodynamic bubble produced at
the leeward side of the screen is higher than a simi-
lar screen without deflectors. This effect is coincident
with the findings of Gandemer's studies, where he
mentioned the use of fins integrated with the barriers
as a method to increase the protection area behind a
screen (Gandemer, 1979). The initial CFD simulation
results confirmed the design concept of fins and slots
at the top to produce a effect of vertical deflection.
based on the initial finding, a new design strategy
with an additional deflector above the screen is for-
mulated. The effect of vertical deflection is more sig-
nificant with this deflector and another of the effects
observed is the regular form of the bubbles in the last
three vertical screens. These screens installed in par-
 allel define a more regular and continuous protection with gaps, the wind speed is lower below these roofs
area (Figure 6). In fact, the first vertical screen after (Figure 7). Based on the engineering CFD simulation
the group of roofs helps to reorder the wind flow that results, for the experiments in the wind tunnel, the
passes through the shelter zone and increase the pro- configuration of the roof was changed to three roofs.
tection region behind the group of roofs.
EXPERIMENTAL PROCEDURE PART 2
Figure 6
Physical simulation using a wind tunnel
Areas of wind
After the CFD simulations, physical tests were con-
deceleration under
ducted in an atmospheric boundary layer condition
the curved roof
within a wind tunnel simulating wind around the
screens and deflectors. The scale models (1:30) of
the screens with modifications were installed along
Figure 7 a panel representing the footpath and west face of
Variation of the the street. The second group of experiments were
wind flow below conducted, changing the roofs composed of sur-
the roofs faces with porous membranes to analyse their per-
formance as deflectors and wind mitigation devices
(Figure 8).
The use of a wind tunnel can be a complicated
task, but the main problem is the difficulty of obtain-
ing accurate and rapid feedback of the simulation
outcomes. To facilitate the visualisation and com-
prehension of the phenomena observed in these ex-
periments several micro wind sensors were installed,
after a process of calibration, to collect and visu-
alise quantitative data in real time. In addition, a
graphic interface was composed to read the data
from the sensors. This platform was developed us-
ing a system of low-cost components, including mi-
cro wind sensors, an Arduino board, Grasshopper,
Firefly and Rhino3D. To know more technical details
about this sensor platform, see the work "wind sens-
Figure 8 ing with real-time visualisations for designers" (Pro-
Model scale for the hasky, 2014).
wind tunnel. Left: The aim of these physical experiments was to
opaque surface. verify (with simple geometries and porous meshes)
Right: porous the level of vertical deflection of the wind behind the
membrane first screen configuration and to measure the wind
speed above the area of roofs. Thus, the tests were
The most interesting observation of this visualisation separated into two stages: the first group of experi-
was the fact that a roof with continuous surface pro- ments with simple geometries of surfaces and a sec-
duces an acceleration of wind speed in the region ond group of experiments with porous membranes.
below. In contrast, if the roof is split in three parts The condition in the wind tunnel considered a

Towards Smarter Cities - Volume 1 - eCAADe 32 | 141

 wind velocity profile of atmospheric boundary layer
at 1:30. Reference velocity at 50mm measured by
petot static tube (3.95m/s). The equivalent refer-
ence velocity at 1:1 scale is located at an elevation of
1.5m. The boundary layer was estimated based on
the power law equation (Aynsley, Melbourne, et al.
1977) (Figure 9).
Experiment with a vertical screen
The first region tested with the group of micro wind
sensors in the ABL wind tunnel experiment was be-
hind the first porous screen and deflector canopy. In
this place, the sensors measured the velocity inten-
sity at 50mm, 75mm, 100mm, 125mm, eight (scale
model 1:30 / 50mm=1.5m). To have a clear idea of
142 | eCAADe 32 - Towards Smarter Cities - Volume 1
the changes in the wind flow pattern three tests were
conducted: no screen and deflector, first screen and
deflector (simple geometries), first screen and deflec-
tor (porous meshes), and a fourth - no screen and
deflector (second post reference).
The results show that the velocity profiles be-
hind both screens are quite similar. Only the sen-
sor at 125mm height detected lower velocity in the
Figure 9
wake region behind the porous canopy than the im-
Wind velocity
permeable surface canopy. But this difference does
profile of
not change the velocity at the other levels (Figure 10).
atmospheric
Therefore, this shows that the vertical upward deflec-
boundary layer
tion of wind produced by the screen, it is not affected
condition at 1:30.
if the canopy is a porous membrane.
Figure 10
Positions of sensors
for four tests and
results of the four
tests with a vertical
screen (porous and
opaque surface)
Experiment with a lateral deflector
A second experiment involves the evaluation of the
first screen configuration and the effect produced by
a lateral deflector with a shorter screen. Two exper-
iments were set up to compare the effect: a first ex-
 periment of a lateral deflection of the wind in a stan-
Figure 11 dard screen and a second test with a shorter screen
Three tests for a with a lateral fin. The distances are taken from the
lateral deflector and vertical wall of the model. The results showed that
results the lateral deflector extended the boundary of the
protection region. A smaller screen with the lateral
deflector provided very similar conditions of wind
mitigation in the whole area of the sensors (Figure
11).

Analysis shelter area

Two sensors were placed in three positions below
the roofs in the shelter area, for the last group of ex-
periments. They were installed to measure the wind
speed at head level to capture the effects of wind rel-
evant to pedestrians and the effects of split roofs and
roofs made with porous meshes.
The results of these experiments showed that the
level of wind velocity in the area of the shelter (pos 2,
3) decreased using porous membranes. Besides, in
the position 3, the turbulence intensity is lower be-
Figure 12 low the membranes, but it did not change at the head
Six tests for opaque level (Figure 12).
and porous surfaces
in the shelter area Results wind tunnel
and results The graphs generated were standardised into non-
dimensional wind velocities relative to the reference
velocities at relative pedestrian level measured from
ABL conditions. Two reference wind velocity profiles
were used to account for reference wind velocity drift
during the experiments. The turbulence intensity
plots are provided in addition to the average wind ve-
locity since the data was available to do so. In subse-
quent papers, the wind sensors will be analysed for
their responsiveness and reliability in measuring tur-
bulence intensities. Though, the nature of the results
seem quite satisfactory after deconstructing the ev-
idence. In general, the wind velocities are reduced
further with the porous deflectors with respect to
the solid deflectors. And turbulence intensities are
also decreased in the same fashion due to the gradu-
ated distribution of pressure differences by filtration
through the porous elements.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 143

 CONCLUSION
The aerodynamic phenomena around buildings and
wind issues in pedestrian areas are an opportunity
for experimentation with new mitigation strategies
involving explorations of wind dynamics and digital
and physical technologies of simulation and visuali-
sation (CFD and wind tunnel). Consequently, the vi-
sualisation and analysis of screen configurations us-
ing CFD allow us to plan a better strategy of anal-
ysis in the physical wind tunnel. This strategy was
focused on specific regions and effects of wind to
evaluate their behaviour, but at the same time, the
questions generated from two methods of simula-
tion (empirical testing with digital sensing and CFD)
were verified through comparison across these two
technologies.
OUTLOOK
The parallel use of numerical simulation and physi-
cal simulation produces a complementary approach
to improve comprehension of complex wind phe-
nomena. For instance, to design the experiments in
a wind tunnel required a general prediction of re-
sults. The position of the sensors must consider the
most significant points to measure reliable data to
derive relevant conclusions. In this sense, the visuali-
sation provided by the CFD software can be useful to
define the areas to test in physical experiments. Thus,
both technologies provide not only data, but rather,
support the facilitation of the design process in the
discoveries and manifestation of the architecture of
wind.
Gandemer, J 1979, 'Wind Shelters', Journal of Industrial
Aerodynamics, 4, pp. 371-389
Gandemer, J 1978, Discomfort due to wind near buildings:
aerodynamic concepts, Dept. of Commerce, National
Bureau of Standards, Washington, D.C., U.S.
Kim, J, Kyu Yi, Y and Malkawi, AM 2011 'Building form
optimization in early design stage to reduce adverse
wind condition, using Computational Fluid Dynam-
ics', 12th Conference of International Building Perfor-
mance Simulation Association, Sydney, Australia
Prohasky, D, Moya Castro, R, Watkins, S, Burry, J and
Burry, M 2014 'Wind sensing with real-time visual-
isations for designers', Proceedings of eCAADe 2014,
Newcastle upon Tyne
Salim, F and Moya Castro, R 2012 'Parallel Analysis of
Urban Aerodynamic Phenomena Using High and
Low-tech tools.', Proceedings of eCAADe 2012, Prague,
Czech Republic, pp. 621-629
Stathopoulos, T 2009 'Wind and Comfort', EACWE 5, Flo-
rence, Italy
Wisse, JA 1988, 'A Philosophy for Teaching Wind in the
Built Environment', Energy and Buildings, 11, pp. 157-
161
[1] http://www.academia.edu/4820747/Evaluation_-
of_simulation_tools_for_assessment_of_urban_form_-
based_on_physical_performance
[2] http://www.bom.gov.au/climate/averages/tables/cw_-
086071_All.shtml

REFERENCES
Aynsley, R.M., Melbourne, W.H. and Vickery, B.J. 1977, Ar-
chitectural aerodynamics, Applied Science Publish-
ers, London
Cochran, L 2004 'Design Features to Change and/or
Ameliorate Pedestrian Wind Conditions.', Structures
Congress 2004 American Society of Civil Engineers.,
Nashville, Tennessee, United States
Evyatar, E, Pearlmutter, D and Williamson, T 2011, Urban
Microclimate. Designing the Spaces between Build-
ings, Earthscan from Routledge.

144 | eCAADe 32 - Towards Smarter Cities - Volume 1

Computing the Urban Block
Local Climate Analysis and Design Strategies

Elif Erdine1 , Evan Greenberg2

1,2
Architectural Association School of Architecture
1,2
aaschool.ac.uk
1
elif.erdine@aaschool.ac.uk 2 evan.greenberg@aaschool.ac.uk

This research develops a method for the analysis, integration and visualisation of
climatic parameters in a dense urban block. In order to test this method, a typical
urban block in Manila, Philippines, is investigated and results are represented
through computational simulation. The translation of latent spatial qualities into
visual data with common tools and techniques allows designers to gain an
understanding of how to design local microclimates, and inhabitants to gain
greater knowledge of the environment. In this regard, this research proposes,
contrary to conventional methodologies, the use of analytical tools as the impetus
to, rather than the outcome of, architectural design.

Keywords: Computation, Urban Design, Environmental Analysis,

Computational Fluid Dynamics, Simulation

INTRODUCTION megacities like Dhaka, Manila, and Hong Kong where

According to the United Nations, the world's popula-
tion will hit 9 billion by 2050 (United Nations 2009),
with 60 million people per year moving to cities (The
World Bank 2013) (Figure 1). This urban migration will
put a huge strain on urban infrastructures, stressing
transport networks, food production patterns, hous-
ing and job availability. Furthermore, the flux in var-
ious environmental conditions will create a change
in the way urban space is used and perceived. A de-
crease in air quality or water availability can lead to
an increase in ambient air temperature and a lack
of physical comfort. The misappropriation of devel-
opable land will lead to the dissolution of public so-
cial space, leaving behind the inherent cultural qual-
ities of the city. As the world's population moves
toward cities, urban density will only increase. In
hyper-density exists, or Karachi, Lagos, and Beijing,
where hyper-density is imminent, there must be an
understanding of how to approach new urban en-
vironments, allowing for both the required density
of inhabitants at the scale of the city, but also pro-
viding comfortable and productive spaces within the
city block. In quickly developing nations such as
China, rapid urban development has been highly un-
successful, leaving entire new cities, designed on a
purely formal basis, devoid of inhabitants (Banerji
and Jackson 2012). Numerous cities, however, have
evolved over a great number of years, dependent
on environmental and cultural aspects in which they
thrive. By understanding the current and future en-
vironmental conditions of these 'Situated Cities' (We-
instock with Gharleghi 2013, p.58), as well as the

Towards Smarter Cities - Volume 1 - eCAADe 32 | 145


cultural contexts in which they reside, new urban
designs can provide a sustainable future for high-
density living. Positioning itself within this frame-
work, this research presents a methodology for the
analysis and visualisation of the local effects of so-
lar radiation and wind flow on the apparent temper-
ature of an urban block. In order to test the devel-
oped methods, the work then describes an investi-
gation on an urban block in one of the densest cities
of the world, Manila, on environmental and climatic
levels, where the resulting data will serve as a foun-
dation for the visualization of latent behaviour and
inherent environmental performance through vari-
ous digital and physical computational processes. In
this respect, the research posited in this paper aims
to address innovative modes of digital visualization
and representation by moving away from the con-
ventional design-analysis-redesign process towards
simultaneous design and analysis methodologies. It
is thus argued that understanding the local effects of
environmental data on the urban block can lead to
the development of more accurate design strategies
with regards to the urban and climatic contexts as
well as the design of localised architectural moments
146 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 1
Population and
population density
of the world's 800
densest cities.
in urban space.
METHODOLOGY
A prototypical city block is selected and modelled
three-dimensionally. The block is then analysed us-
ing computational fluid dynamics software to under-
stand its performance in relation to the effects of
wind velocity within urban space. Within the con-
struction of a computational fluid dynamics analysis
model, a volumetric mesh is generated in order to ac-
curately calculate wind flow, allowing for the obser-
vation of its interaction with the buildings at a cer-
tain time of year, and its effects on air flow through
the urban block. Resulting from this analysis is not
only the wind velocity values, but also the point cloud
generated from the nodes of the volumetric mesh
on which these velocities are plotted. In addition to
computational fluid dynamics analysis, solar analysis
is used to determine the hourly solar radiation on the
building and ground surfaces of the block. Based on
the albedo of the building envelopes' materials, a re-
flected radiation value is determined. Following on
from this, the reflected radiation value is combined
with wind velocity values mathematically to deter-
mine the resulting air temperatures within the urban
block due to the investigated environmental factors.
The results of this computational process provide an
understanding of the differentiated urban space that
is characterised by the flow of energy within a city
block.
Upon completion of these modelling methods,
the data range for a day is simulated visually within
a computational environment, allowing the various
types of data collected to be organised categorically
and compiled into daily representations of urban en-
vironmental flows. The simulation gives the user the
ability to interact with several types of urban data,
aiding in the further understanding of how urban
space is generated and differentiated on a local scale
(Figure 2).
EXPERIMENT 1: MANILA, PHILIPPINES
As an initial test of the proposed design methodol-
ogy, a city block in Manila, Philippines was chosen as
an initial urban scenario. Manila was selected as an
initial test case as it is the densest city in the world.
The city of Manila, in 2010, was reported as having a
population of roughly 1.6 million people, but a popu-
lation density of just under 43,000 people per square
kilometre (National Statistics Office of the Republic
of the Philippines no date). Manila is also located in
a tropical savannah climate, experiencing tempera-
tures well over 25 degrees Celcius during the entire
year. Since it is located on the tropics, the temper-
ature range is very low. The general wind direction
throughout the year, taken from Ecotect data, is dis-
tributed between south-east and west, ranging be-
tween 3 and 8 metres per second.
In addition to the incredible population den-
sity of the city, and the Philippines' tropical cli-
mate, Manila is also one of the most environmentally
stressed cities in the world, with high levels of air pol-
lution, poor sewage systems, water pollution, misap-
propriation of land, and heavy motor traffic (United
Nations 2014). In the framework of this research, par-
ticular attention is given to the air pollution problem.
Similarly with most of the developing cities in Asia,
air pollution in Manila results mainly from traffic con-
gestion and industrial wastes. The concentration of
certain airborne toxic substances emitted by thou-
sands of cars every day contributes to the formation
of smog, increasing the risk of respiratory diseases
for inhabitants (Wallerstein 1999, p. 689). Addition-
ally, the lack of overall urban planning and develop-
ment in the recent decades has resulted in a distinc-
tive division in the urban fabric, where vast areas of
slums can be seen to be located right next to high-
end gated communities. It has been reported that in
2010, an estimated population of 37% Filipinos lived
in slums, and the slum population growth rate was
8% annually in Metro Manila, which is the metropoli-
tan district including Manila (Ballesteros 2010, p. 1).
Therefore, with the climatic and urban environments
presenting huge amount of problems in Manila, the
importance of a clear understanding of the spatial
relationships to the energy flow of the urban block
once again comes into light.
In order to develop a geometrically accurate
three-dimensional model, an urban block in Manila,
at 14°36'34.84"N and 120°58'30.45"E coordinates was
located in Google Earth and imported into Google
SketchUp. By evaluating the shadows cast on the
imported map against the date of year (27 Febru-
ary 2010), building heights were approximated and
modelled three-dimensionally. With the use of com-
putational fluid dynamics (CFD) analysis in CFX for
ANSYS, the wind velocity in the block is simulated.
The computational mesh density generated in CFX
can be controlled in order to both increase the accu-
racy of the simulation, but also the definition of the
point cloud that is then used for further analysis. This
point cloud and its corresponding velocity values are
imported into Grasshopper, Rhino's parametric mod-
elling tool, and visualised three dimensionally on the
block morphology. With this data in the modelling
environment, a series of urban physics calculations
are made in order to accurately describe the effects
of temperature and human comfort within the urban
block condition. Firstly, Ladybug, an environmental
analysis tool in Grasshopper, is used to generate tem-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 147
 Figure 2
Design
Methodology. Wind
velocity data taken
is analysed using
fluid dynamics
analysis and
combined with
solar radiation data,
along with global
temperature and
humidity values in
order to calculate
the ambient air
temperature for
each point on a
point cloud within
perature, humidity and energy data for the specific riod of time, exploring the effects of variations in an urban block. This
site. Reflected solar radiation is calculated by defin- wind velocities, ambient temperatures, and appar- data, collected
ing the albedo of the vernacular building material. ent temperatures. By merging separate sets of data hourly, is then
These radiation values, initial temperature values, which has been created for each hour throughout the simulated
the wind velocity values found in the computational day, it is possible to explore how energy flow is differ- computationally as
fluid dynamics analysis, and general relative humid- entiated for a certain period of time on a local scale a visualisation tool.
ity values are then linked using Steadman's Appar- (Figure 4). It is believed that this tool can be very ad-
ent Temperature equation (Australian Government vantageous for the designer, as it can help in the un-
Bureau of Meteorology no date; originally attributed derstanding of what types of architectural interven-
to Steadman 1994). This mathematical model thus tions can be necessary in improving the environmen-
generates the data for understanding the environ- tal quality of urban space.
mental effects and spatial consequences of urban
block morphologies. During the second phase of RESULTS
the research, the results of the environmental cal- The experiment provided results, on average, of
culations, extracted hourly throughout the summer a three to four degree decrease in apparent tem-
solstice, are transferred as three-dimensional data perature across the points in the point cloud due
into the open source environment Processing (Fig- to wind flow, combined with local solar radiation
ure 3). As an object-oriented programming environ- values and global temperature and humidity data,
ment (OOP), Processing allows for the delineation of modelled mathematically within an accessible three-
each set of environmental data as a class; each class dimensional modelling environment. . The poten-
can then interact according to specific rules in order tial to combine solar and wind flow data at a resolu-
to produce emergent data patterns on a global level. tion useful for architects and designers is significant,
With the incorporation of physics libraries in Process- while the largely successful results from the first test
ing, this imported environmental data become input prove an intelligent and logical methodology. The
parameters for a series of simulations which seam- initial experiment on the Manila block shows the pos-
lessly link the distinct instances into a specified pe-

148 | eCAADe 32 - Towards Smarter Cities - Volume 1

Figure 3
Wind velocity taken
from computational
fluid dynamics
analysis is imported
in the Processing
environment for the
simulation and
representation of
microclimatic data
on an urban block
in Manila,
Philippines.

Figure 4
The change in
temperature is
mapped on the
Manila block,
providing a clear
representation of
the change in
apparent
temperature due to
solar radiation and sibility for Grasshopper, a tool used by a wide audi- understanding of these results has not necessarily
wind flow. ence, to compile climatic data within one model and become straightforward. In this regard, the visualiza-
one environment. Grasshopper allows for the visu- tion of the spatial effects of climatic conditions in Pro-
alization of this data at one specific time, or as an cessing becomes a valuable tool for design (Figure 5).
average over a set time period. While this may be
useful in a large number of scenarios, the visualiza- DISCUSSION
tion of change throughout this time period would The results from Experiment 1 provide some excel-
give a better understanding of climatic conditions lent opportunities for further exploration, along with
and the performance of urban space. Thus, the im- a series of questions and problems that require res-
plementation of Processing in creating a simulation, olution. While the integrated method provides use-
which covers a time span and enables the user to in- ful results, a series of modelling issues must first be
teract with data visualization, becomes incredibly im- resolved. With regard to mathematical modelling,
portant within the scope of this experiment. Even there are two environmental calculations that need
though the inclusion of high-end engineering and development. Solar analysis through the Ladybug
analysis software is becoming more widespread, the plug-in provides accurate data for sun path analysis

Towards Smarter Cities - Volume 1 - eCAADe 32 | 149


and subsequent solar radiation on the building sur-
faces in the three-dimensional model. The reflected
solar radiation is approximated with an albedo fac-
tor, and applied on the surrounding points within
the point cloud, giving a baseline understanding of
the behaviour of reflected energy on surrounding
air temperature, at a resolution useful to designers.
However, the change in air temperature is currently
calculated using the AT formula previously described
on each point within the generated point cloud.
While this may provide an understanding of localised
effects, the precision of the location affected can be
improved by refining the point cloud on which calcu-
lations are based. This would increase the precision
in the correlation between the point of calculation on
the building surface and its coupled point of reflected
radiation, and thus apparent temperature. Addition-
ally, Steadman's Apparent Temperature formula sup-
poses that wind speeds are calculated at 10 meters
above ground; since this research aims to achieve an
understanding of local climatic differentiation, this
seems to be an incorrect calculation. Further investi-
gation is necessary in order to refine this mathemat-
ical relationship.
150 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 5
Visualising climatic
data verifies the
change in apparent
temperature over
time, but also
describes how the
differentiated
change in
temperature
creates varied
microclimatic
effects.
While digital simulation in Processing has helped
in the representation of local effects throughout a
given period of time, it has also uncovered an is-
sue in visualising hourly and daily data within a sin-
gle model. In order to visually understand the re-
lationship between initial and apparent tempera-
tures throughout a daily cycle, the global tempera-
ture range (that is, the lowest and highest values of
the combined initial and apparent temperature val-
ues) has been used, and nodes assigned a colour gra-
dient with reference to this data. However, as the
global temperature range is larger, the difference in
colours is much greater than it would be if colours
were mapped within a local hourly range. Although
this creates a difficulty in representation, it is pre-
ferred to maintain the global range and ensure accu-
rate comparison while developing ways to improve
the data's legibility.
Along with refining the methods and techniques
as indicated within this discussion, future research
hypothesises that taking this computational method
to the physical environment will aid in the under-
standing of the effects of environmental data on the
selected urban block. For this purpose, a scaled phys-
Figure 6
Computational
simulation in
Processing allows
for the visualisation
of numerous data
within one variable
model.
ical model of the urban block will be fabricated and
overlaid with an interactive arrangement which will
carry out the spatialisation of various types of en-
vironmental data, including wind flow, actual tem-
perature, and perceived temperatures. This task will
be realised through the physical computation envi-
ronment Arduino, which will be in continuous com-
munication with the simulation generated in Pro-
cessing. As such, it will be possible to physically
manifest environmental data and its alteration over
time in three-dimensional space by using various me-
dia, including LED's, movable pins, and projection-
mapping, simultaneously disseminating the correla-
tions between separate data sets.
A further observation of this experiment, while
not specific to the architectural scope of this research,
is with regard to the employed computational en-
vironment. Modern computing processing capacity
has rendered it possible to use advanced modelling
and complex simulation tools on a personal laptop.
The fact that various 3D-modelling, analysis, and sim-
ulation tools involving complex physics calculations
can be operated on a laptop presents itself as a major
advantage for the research, as it creates an effective
digital workflow between separate platforms. Fur-
thermore, the ease of using various complex mod-
elling and simulation tools on a laptop also enhances
the interdisciplinary nature of design which has been
improving in the recent decades, making it possible
for a smooth communication between different pro-
fessions involved in design. In effect, this develop-
ment acts as a constituent in one of the major goals
of this research, which is to create innovation in digi-
tal visualization and representation tools by simulta-
neously integrating design and analysis techniques.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 151
 CONCLUSION REFERENCES
The integration of different types of environmental Ballesteros, MM 2010 'Linking Poverty and the Envi-
data inside one tool makes it easier for the designer ronment: Evidence from Slums in Philippine Cities',
to understand the correlations between these vari- Philippine Institute for Development Studies Discus-
sion Paper Series 2010-33
ous data types (Figure 6). The methodology devel-
Steadman, RG 1994, 'Norms of Apparent Temperature
oped in this research is expected to be implemented in Australia', Australian Meteorological Magazine, 43,
as a critical part of conceptual design process. In pp. 1-16
effect, this tool can become an aid in bridging the Wallerstein, C 1999, 'Philippines Tackles Air Pollution',
gap between conceptual design and environmen- British Medical Journal, 318, p. 689
tal analysis, in return helping the designer to make Weinstock, M and Gharleghi, M 2013, 'Intelligent Cities
and the Taxonomy of Cognitive Scales', in Wein-
informed decisions about potential urban interven-
stock, M (eds) 2013, System City: Infrastructure and
tions on a selected site rather than conforming to in- the Space of Flows, John Wiley and Sons, Chichester,
tuitive choices. pp. 56-65
The translation of latent spatial qualities into vi- [1] http://www.un.org/apps/news/
sualised data uncovers significant potential in the story.asp?NewsID=30159&Cr=family
understanding of the performance of the urban [2] http://go.worldbank.org/QHKRLTGH70
[3] http://www.bbc.co.uk/news/magazine-19049254
block, incorporating existing density distributions
[4] http://www.census.gov.ph/sites/default/files/
with effective climatic and productive spatial mod- attachments/hsd/pressrelease/
els. Through the employment of feedback from the National Capital Region.pdf
initial computational model and physical prototype, [5] http://www.un.org/cyberschoolbus/habitat/
urban analysis can be implemented as a methodol- profiles/manila.asp
ogy to anticipate future scenarios for design possibil- [6] http://www.bom.gov.au/info/thermal_stress/
ities. In this respect, the current research methodol-
ogy applied in a specific urban block in Manila will be
advanced in order to investigate a series of additional
urban blocks in the world's most populous cities.
Furthermore, it is argued that the research has
two major future potentials, which are related to the
architect and the inhabitant respectively. From the
architect's point of view, the research outcome can
serve as a design tool allowing the architect to pre-
dict the impact of future design proposals in the ur-
ban context. From the inhabitant's point of view,
the outcome serves as a visualization and representa-
tion tool which enables the user to interact with envi-
ronmental data and fully understand its implications
on the urban scale. As such, "Computing the Urban
Block" proposes, contrary to conventional method-
ologies, the use of analytical tools as the impetus to,
rather than the outcome of, architectural design by
enhancing their potential as a digital design and rep-
resentation tool.

152 | eCAADe 32 - Towards Smarter Cities - Volume 1

Integrative Pedestrian Modelling Techniques based on
Virtual Force Fields
Analysis, Generation and Evaluation in Public Open Spaces

Odysseas Kontovourkis1 , Despo Anagiotou2

1,2
Department of Architecture, Faculty of Engineering, University of Cyprus, Cyprus
1
http://ucy.ac.cy/dir/en/component/comprofiler/userprofile/odykont
1
kontovourkis.odysseas@ucy.ac.cy 2 d.anagiotou@gmail.com

This paper demonstrates a preliminary research methodology towards an

integrative digital design approach for the analysis, generation and evaluation of
architectural proposals in public open spaces based on human movement
behaviour performances. In order to achieve this, various computational
mechanisms that involve the logic of pedestrian modelling are applied, aiming to
be explored in different stages of design. The suggested models follow the idea of
'virtual force' fields, an approach initially introduced in previous work by author.
Based on particle behaviour modelling, this approach examines the interaction
and movement behaviour of individual entities within a system through virtual
effects and specifically through attraction and repulsion forces, influencing
pedestrians behaviour and hence their accelerated movement. Current paper
argues that the idea of 'virtual force' fields can be used not only for pedestrian
simulation but also for the analysis and generation of proposals, aiming on a
holistic design development of public spaces.

Keywords: Pedestrian modelling, Integrative techniques, Virtual force fields,

Public open spaces

INTRODUCTION Within this frame, the human perception that

The meaning of human movement behaviour in pub-
lic open spaces has been examined extensively, es-
pecially in relation bodily motion and perception of
space. Regarding to Spinoza in [1] motion is a transi-
tion situation from one state of motion to another, or
variation of motion/rest and therefore it is a human
behaviour occurred by affordances of external envi-
ronment, which is depended on the intentions, mo-
tivations and knowledge of individual (Lang 2005).
leads to spatial experience is achieved through mo-
tion, going from one vista to another, that is natural
vision as mentioned in (Gibson 1986) followed by the
idea of 'intelligibility', which is defined as the degree
of what we can see from the space that makes up the
system indicating the integration of each space into
the whole system (Hillier 2007). In addition, the con-
cept of 'legibility' defines the visual quality, the ap-
parent clarity of the city that has been introduced in

Towards Smarter Cities - Volume 1 - eCAADe 32 | 153

 (Lynch 1960).
In parallel direction, 'field theory' introduced in
Lewin (1951) described human behaviour based on
force fields, an influential theory in the area of pedes-
trian modelling in micro level. Pioneer work in Hel-
bing et al (2001) introduced a mathematical interpre-
tation of this idea called social force modelling, which
has been mainly used to investigate pedestrian inter-
action behaviour.
The complexity of human movement behaviour,
especially in public spaces, demanded the applica-
tion of simulation techniques and software particu-
larly in regard to the evaluation of existing design
solutions. Spatial analysis examples described the
space utilization and prediction of movement in ac-
tual situations as well as vision based models that
were applied to detect the degree of accessibility and
visibility of space (Bada and Farhi 2009; Woloszyn and
Leduc 2011).
Nevertheless, the movement as a constantly
changing condition and the traces of flow as the mor-
phological result might be considered as space gen-
erators followed by experimentations using agent-
based and social collective intelligence models,
which allow the generation of spatial diagrammatic
results (Ireland 2008; Puusepp and Coates 2008).
Currently, software and computer models mainly
focus their attention on the simulation of pedestrian
movement and the evaluation of solutions after the
design is specified. This paper puts forward the hy-
pothesis that computational mechanisms, algorith-
mic logics and parameters applied for the evaluation
of design solutions might also be part of an integra-
tive process involving mechanisms for analysis and
generation.
VIRTUAL FORCE FIELDS METHODOLOGY
This work aims to introduce an integrative strategy
that is originated in a previous in-house pedestrian
simulation program developed by the author under
the name 'virtual force' model (Kontovourkis 2012).
According to this, pedestrian movement simulation
is based on the idea of particle behaviour modelling
154 | eCAADe 32 - Towards Smarter Cities - Volume 1
and particularly on the modelling of individual enti-
ties (individual entities might be described as pedes-
trians, obstacles, signs, etc), which are influenced by
local rules that accelerate global behaviour. Specif-
ically, the interaction and movement of pedestrians
(a type of individual entity) is achieved by various
behavioural rules that represent influences from the
built environment generating a number of different
effects. Those effects might be described as repul-
sive, attractive, obstacle avoidance effect, etc. For in-
stance, in case of repulsive effect, individual pedes-
trians are subjected to virtual repulsive forces due
to the tendency of pedestrians to keep a distance
from others or from obstacles while their movement
in space. As a result, the interaction and movement
of pedestrians in discrete steps in space is achieved
(Kontovourkis 2012).
Following the idea of particle behaviour mod-
elling and its ability to generate a number of adaptive
results due to the interaction occurred between dif-
ferent individual entities (including pedestrians) and
the built environment, the current methodology sug-
gests a conceptual design framework. This includes
the idea of 'virtual force' as the basic factor for de-
veloping the three different stages of design inves-
tigation, i.e. analysis, generation and evaluation. The
suggested framework describes the adaptive process
as a continuous feedback loop between three basic
parts: input, output and control mechanism. In con-
trol mechanism area, an iterative process between
analysis, generation and evaluation stages is intro-
duced based on the concept of 'virtual force' fields al-
lowing the generation of numerous adaptive design
outputs. Those results are fed back to the evaluation
stage in a continuously repeated process until the de-
sired objectives are achieved (Figure 1).
Analytically, in analysis stage, the different indi-
vidual spatial entities and conditions are codified in
order to transfer the physical behaviour into rules
and parameters based on 'virtual force' concept and
specifically into attraction and repulsion effects. In
generation stage, the parametric logic and control of
outcomes based on 'virtual force' fields are used to
Figure 1
Proposed
conceptual design
framework based
on analysis,
generation and
evaluation stages
incorporating the
idea of 'virtual force'
fields
develop movement flow systems in 2D and 3D envi-
ronment allowing the formation of spatial morpholo-
gies. In evaluation stage, the generated outcomes
are evaluated and the process is verified providing
quantitative results in relation to the spatial utiliza-
tion, maximum density, speed loss, and so on.
Preliminarily, the three stages of the control
mechanism are developed separately using as com-
mon ground the 'virtual force' fields concept (Kon-
tovourkis 2012). The proposed platform includes the
FlowL plug-in [2] associated with parametric design
software Grasshopper (a plug-in in Rhino) [3] and the
commercial pedestrian modelling software SimWalk
[4]. Then, in each stage, results in regard to the out-
put data and findings are discussed together with
theoretical suggestions on how the three different
stages can be integrated and developed further.
CASE STUDY: A PUBLIC OPEN SPACE
The proposed conceptual framework is tested using
the case study of an existing open public space, al-
though the current methodology is expected to be
introduced in other examples attempting to verify
the validity of the idea. The area of design imple-
mentation is the small adjacent square next to the
Eleftheria square, one of the main historical public
spaces in Cyprus. It is located in the old city of Nicosia,
at the beginning of Ledras Street and just after the
Venetian walls. In the surrounding urban fabric there
is a plethora of options for leisure, shopping and
business activities as well as accommodation facili-
ties and significant monuments. For this reason, the
space under investigation is a transition point, highly
accessible by a number of pedestrians. However, an
interest in the appropriation of space from the major-
ity of pedestrians or interaction among them is not
obvious. In contrast, the current spatial configura-
tion shows a separation in space utilization according
to the type of pedestrians, preventing the creation of
a lively space. The current methodology aims to de-
tect the conditions and the degree of their affection
as well as to recognize the spatial qualities in order to
improve the living environment.
The process suggests a constant interaction be-
tween the proposed design stages (analysis, gener-
ation and evaluation) as well as researcher/architect
interventions, whose decisions and observations are
instrumental for the design configuration outcomes.
Briefly, the procedure includes the analysis of the ex-
isting, the generation of 2D flow lines and 3D mor-
phologies by visual algorithms and the evaluation of
the results, all stages based on the 'virtual force' fields
idea. Following sub-sections describe each stage.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 155
 Analysis
The investigation starts by analysing the existing
area. In the process of pedestrian movement be-
haviour modelling, apart from the importance of ori-
gin and destination points that defines the goal of
movement action, other innumerate factors, which
play equal role in this process are visibility (an idea
derived from Hillier 2007 and Gibson 1966) and
pedestrians' collectiveness. Respectively, conditions
related to visibility (advantageous viewing point)
and elements that encourage collectiveness (sitting
areas, shadow areas, etc) are integrated into the
methodology.
Considering that the spatial entities influence
pedestrian movement behaviour by attraction or re-
pulsion, site analysis aims to codify such information
in order to transfer the physical behaviour of pedes-
trians and their interaction with the built environ-
ment into rules and parameters. Spatial elements
and physical conditions affect pedestrian movement,
not only due to their spatial qualities, but also due
to the specificities of individuals. Large impact val-
ues of spatial entities result large attraction or repul-
sion, and hence displacement of the initial pedes-
trians movement. The distance between a pedes-
trian and a spatial element, either attractive or repul-
sive, influences the displacement degree of pedestri-
ans flow lines (Figure 2). In addition, the pedestrian
movement intention (either pass through or stop)
plays a crucial role in the simulation process and can
define the type as well as the degree of local influ-
ences derived from the conditions of the environ-
ment.
Specifically, the proposed pedestrian movement
scenario is analysed as follows: pedestrians move
156 | eCAADe 32 - Towards Smarter Cities - Volume 1
through the public square based on various needs
and interests. The space is perceived differently ac-
cording to pedestrian's destination and background.
As a result, the type of influences and of conditions
in regard to the spatial elements is differentiated in
each case. The site analysis reveals that a pedestrian
who is interested to stop is more attracted by the ad-
vantageous viewing and shadows areas and is less in-
fluenced by areas with great concentration of pedes-
trians. In contrast, a pedestrian who aims to pass
through the public square, he/she is repulsed by con-
centration points while is less influenced by shadow.
This is observed mostly when the pedestrian is famil-
iar with the direction of movement that is based on
destination goal. If the movement is exploratory, for Figure 2
instance tourist's walk, then advantageous viewing Diagrams of
areas attracts the most. pedestrian flow
Briefly, the site analysis indicates four types of displacements
pedestrians: group a: shoppers, group b: immi- according to the
grants/elderly people, group c.: tourists, and group d: intensity and
employees. The conditions that influence the attrac- distance of
tion or the repulsion and hence the pedestrian move- repulsion and
ment are the obstacles, the advantageous viewing attraction
points, the shadows areas and the pedestrian con- behaviour
centration points. The movement behaviour is cod-
ified and analysed based on the four types of individ-
ual pedestrians and is examined by defining move-
ment scenarios and by designing corresponding flow
diagrams in each case.
The pedestrians move through the public square
based on the configuration of the wider urban fabric.
A pedestrian flow diagram within the existing public
square consists of a starting and a destination point.
Each group of pedestrians is related with flow lines
of three possible routes, except the group c that is
related with flow lines of four possible routes. The
flow lines, which correspond to a particular pedes-
trian group, are determined by several points and a
common destination point. For instance, based on
the existing configuration, the initial flow lines that
are set for pedestrians in group a (shoppers) are as fol-
lows: Start point [O2] - Destination point [D1], Start
point [O3] - Destination point [D1] and Start point
Figure 3
Existing spatial
configuration,
possible scenario of
pedestrian
modelling and
generated 2D flow
diagram
[O4] - Destination point [D1]. The starting points [O2,
O3, and O4] refer to shoppers' access points, while
the destination point leads to the main commercial
road of the given area.
In the analysis stage, a preliminary codification of
input data is demonstrated forming the background
where generation and evaluation of solutions will
be developed. In analytical level, the translation of
pedestrian movement and the interaction behaviour
based on repulsive and attractive effects provides
the framework where the three stages will be inte-
grated into a unified methodology under common
input and output principles.
Generation of initial solution
Based on the 'virtual force' fields idea and using the
parametric design software Grasshopper, this stage
focuses on the spatial configuration and generation
of movement flow systems and morphologies in 2D
and 3D environment.
The procedure of morphological generation
starts by digitising the input data, which are derived
from the site analysis. Then, spatial elements are
transformed into charging points of attraction and
repulsion. According to the type of pedestrians, their
behaviour as well as their distance from the point of
influence, the attraction or repulsion effects can vary
creating different displacements of the movement
flow lines. For instance, based on the existing living
environment, in the group a (shopper) the following
effects are applied: Tree - repulsive effect, Bench - re-
pulsive effect, Advantageous viewing point - none,
Shadow area - none and Concentration point - repul-
sive effect.
In detail, the spatial entities named tree, bench
and concentration points repulse the shoppers be-
cause these are perceived as obstacles in space. Since
shoppers are familiar with the environment, aim-
ing to reach their destination without aimless move-
ments, they are not influenced by shadow areas or
advantageous viewing points. The advantageous
viewing point is detected in parametric environment
by calculating the point with the largest viewable
area ('isovist'- See Hillier 2007). This point of influ-
ence, and according to the movement scenario, can
carry repulsive or attractive behaviour.
Initial results show that generated parametric
flow lines that are influenced by the 'virtual force'
fields. The displacement of initial flow lines is con-
figured by the resultant force (or the average charge)
related to the distance between the particular points
of flow lines, the attractive and the repulsive points.
Based on parametric flow lines, a 2D diagram, which
indicates a field of flows related to the movement
tendency, is generated (Figure 3).
Then, by using the parametric tool FlowL (a plug-
in in Grasshopper) that is correlated with the gen-
erated pedestrian movement flow, a 2D diagram of
flow lines according to the average field of charges
is produced. The 3D generation of space is based on
the 2D diagram of flow lines attempting to capture
possible design solutions based on the given move-
ment scenario.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 157
 Figure 4
Scenario A:
Proposed spatial
configuration and
pedestrian
modelling scenario
as well as generated
2D flow lines and
3D morphology

Figure 5
Scenario B:
Proposed spatial
configuration and
pedestrian
modelling scenario
as well as generated
2D flow lines and
3D morphology

Preliminary evaluation and alternative so-
lutions
The 2D flow lines diagrams and 3D morphologies can
be interpreted through the criteria of trend, intensity
and free movement. Observations derived from the
diagrams are correlated with the real situation and
the digital results are evaluated preliminarily to de-
fine and organise alternative spatial configurations
and movement scenarios.
In order to improve the living environment and
to generate a lively space, the existing spatial ele-
ments are preserved and new conditions that attract
more types of pedestrians are suggested. In the con-
text of experimentation, considerable alternative so-
lutions in a relatively short time period are imple-
mented. The proposed experimentation consists of
several charging points (function in real time) with
different degrees of attraction or repulsion and po-
sitioning in the built environment. The following fig-
ures show diagrammatically different stages of gen-
erated procedure based on alternative scenarios A, B,
and C. In the first stage, the conceptual development
is shown, where intentions and movement scenarios
are set. In the second stage, the generation of para-
metric flow lines are demonstrated, and in the third
stage, the generation of the 2D diagrams of flow lines
are shown. In the final stage, the process for generat-
ing 3D morphologies are presented, whereas charg-
ing points and flow lines diagrams are transformed
into morphological and functional solutions (Figure
4, 5, 6).

Figure 6
Scenario C:
Proposed spatial
configuration and
pedestrian
modelling scenario
as well as generated
2D flow lines and
3D morphology

158 | eCAADe 32 - Towards Smarter Cities - Volume 1

Figure 7
Scenario D:
Selected spatial
configuration and
possible scenario of
pedestrian
modelling

Figure 8
Scenario D:
Selected diagram of
parametrically
generated 2D flow
lines and 3D
morphology

In the selected movement scenario D, the spa-
tial configuration and the 3D generation correspond
to the general aim of improving the existing on-site
situation and is the result of a continuously evalua-
tion procedure (Figure 7, 8). The current methodol-
ogy attempts to attract more types of pedestrians by
encouraging stops and spontaneous activities. Also,
it intends to reduce pedestrian's acceleration move-
ment and in parallel to increase the duration of stay
in the public square. The classified functions that
are found to be appropriate in this case are ramps of
movement, shelters and seats/rest areas, which are
formed by grouping the appropriate flow lines and
transforming them into lanes/surfaces.
Through the suggested process of cyclical it-
eration between generation and evaluation stages
a number of new movement flows are generated
based on the alternative movement scenarios, sim-
ulating in parallel corresponding pedestrian move-
ment behaviour. The configuration of lane/surface
depends not only on the function but also on the
range of charge (attractive or repulsive), which is de-
fined in the analysis stage and used as the output
data that are fed back to the generation stage. In

Towards Smarter Cities - Volume 1 - eCAADe 32 | 159

 this process, the user intervention is very important
since he/she is able to determine the final desired
morphology through the feedback loop process al-
lowing the reformulation of analysis and evaluation
scenarios based on the 'virtual force' fields logic.
Advanced evaluation of selected solution
In the next stage, the existing configuration and the
selected solution D are evaluated based on the qual-
itative criteria of free movement, movement ten-
dency and density of flow lines (indicating the inten-
sity of movement) as well as on quantitative results
related with the spatial utilization, maximum density
and speed loss. In order to achieve this, the pedes-
trian simulation software SimWalk is used for evalu-
ating the proposed outcome including the 3D mor-
phological results.
The evaluation process starts by examining the
existing configuration. The aim is to compare the dig-
ital results (diagrams of parametrically generated 2D
flow lines) with the real situation based on the qual-
itative criteria mentioned above. Analytically, it is
detected that the configured digital environment re-
flects the existing spatial utilization and its character-
istics. In real time, the space is characterized by free
movement while the main movement of pedestrians
happens through the centre of space demonstrated
by vectors (using relevant components in Grasshop-
per). Since the 2D flow lines indicate the intensity
of movement, is revealed that there are areas where
the movement flow is nonexistent. Similarly, based
on site analysis, the same areas are used less fre-
quently. It is clear, although, that space is experi-
enced and appropriated not only because of spatial
organization but also because of social economic cir-
cumstances. Nevertheless, the final outcome reflects
the real situation providing a verification of the whole
process consisting of spatial elements' codification
and pedestrians' movement scenarios.
Then, results derived from the simulation of the
existing and proposed configuration (selected solu-
tion D) are evaluated through a comparative pro-
cess taking into account the movement scenarios,
160 | eCAADe 32 - Towards Smarter Cities - Volume 1
the spatial configuration as well as the generated 2D
flow lines and 3D morphology. Quantitative results
related to spatial utilization, maximum density and
speed loss are derived. The spatial utilization graph
shows the number of pedestrians passing through a
specific spatial cell in each time step (P/min), while
density graph shows the number of pedestrians in
the same spatial cell (P/m2). The final graph is related
to the speed loss and it presents the degree of pedes-
trians' speed reduction in each cell (rate of speed re-
duction in %). A pedestrians' trails diagram is also
available, where each colour of trails corresponds to
the destination point of pedestrians.
The simulation is set in three phases of a day (a.
9:00, b. 13:00, c. 20:00) with duration of 45 min-
utes. Also, initial pedestrian settings that include the
time of appearance, attendance, and intentions are
defined. Then, different movement scenarios are or-
ganized and the appropriate start, mid and destina-
tion points are specified (Figure 10).
Movement scenarios are organized based on
the same four types of pedestrians (shoppers, immi-
grants/elderly people, tourists, employees). First sim-
ulation is set at 9:00 am and consists of 401 pedes-
trians (shoppers: 83, immigrants/elderly people: 62,
tourists: 190, employees: 62). Second simulation cor-
responds to 13:00 pm and consists of 327 pedestri-
ans (shoppers: 74, immigrants/elderly people: 41,
tourists: 150, employees: 62). The third simulation
corresponds to 20:00 pm and consists of 181 pedes-
trians (shoppers: 59, immigrants/elderly people: 52,
tourists: 60, employees: 10). Alternative simulations
can be set by modifying the number of pedestrians
and the movement scenarios of pedestrians.
The results of simulation are compared with the
initial intentions/initial goals of spatial organization
scenarios and with the qualitative criteria of trend, in-
tensity and free movement. Observations of the spa-
tial utilization graph in the existing proposal shows
that there is a free movement with large intensity
through the center of space. Figure 10 shows the sim-
ulation set at 13:00 pm. In contrast, results of sim-
ulation related to the selected proposal show that
Figure 9
Existing and
proposed diagrams
of pedestrian
simulation
scenarios (Origin
and Destination
points)

Figure 10
Existing
configuration and
results of simulation
set at 13:00 pm

the spatial utilization is uniformly distributed in the
whole area. Figure 11, 12 and 13 show the simula-
tions set at 9:00 am, 13:00 pm and 20:00 pm respec-
tively.
In relation to the density graph, is observed
that the selected proposal is characterized by high
density through the whole area (Figure 11, 12 and
13). The possibility for large number of interactions
between the pedestrians is provided, showing that
pedestrians' collectivity is encouraged. However,
due to the limited free space and high density, the
phenomenon of collision is more probably to be oc-
curred in real time.
Observations in regard to the speed loss of se-
lected proposal show that there is a high reduction
of movement acceleration through the whole space
(Figure 11, 12 and 13) compared with the results ob-
served in the existing proposal where an uninter-
rupted and relatively fast movement is occurred (Fig-
ure 10). By encouraging stops and by increasing seat-
s/rest time in selected positions, the speed loss target
is also satisfied.
The preliminary conceptual methodology that
integrates the three different stages in a proposed
conceptual framework initially aims to find correla-
tions between different computational techniques
based on 'virtual force' fields. It is obvious that the ap-
plication of such modeling techniques, which accel-
erate pedestrian movement behaviour within a con-
tinuously interactive built environment, can be the
starting point for computational interoperability. In
a preliminary stage, this is achieved by the applica-

Figure 11
Proposed
configuration and
results of simulation
set at 9:00 am

Towards Smarter Cities - Volume 1 - eCAADe 32 | 161


tion of feedback loop logic linking the output results
of the analysis with the input data of generation and
evaluation stages. Also, by using output data derived
from the evaluation as the input information in the
generation stage.
CONCLUSION
This paper attempts to analyse, generate and eval-
uate a public space through a suggested prelimi-
nary integrative pedestrian modelling methodology
based on 'virtual force' fields. The configuration of al-
ternative spatial scenarios provides the possibility for
morphological generation and evaluation of results
in a continuous iterated feedback loop process. Nev-
ertheless, it is obvious that the way space is experi-
enced and appropriated by pedestrians in not only a
result of spatial organization but also a result of other
social-economic conditions. However, the analysis
derived from this methodology might reflect real sit-
uations and the generated outcomes as well as the
evaluation process might offer useful ground for dig-
ital reorganization and examination of spatial out-
comes.
In conclusion, the current study proposes the in-
tegration of various models based on the idea of 'vir-
tual force' fields aiming on the continuous regener-
ation of space with multiple spatial results. The in-
volvement of different techniques based on the parti-
cle behaviour modelling as the programmatic frame-
162 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 12
Proposed
configuration and
results of simulation
set at 13:00 pm
Figure 13
Proposed
configuration and
results of simulation
set at 20:00 pm
work towards analysis, generation and evaluation al-
low a holistic computational approach to be gener-
ated investigating results in all stages of design pro-
cess towards common objectives but different de-
crees of complexity.
Future work will continue towards an in depth
examination of the three different stages, i.e. anal-
ysis, generation and evaluation, aiming on input and
output data specification and algorithmic control of
the proposed methodology towards the interoper-
ability of techniques. This will allow the generation
and investigation of solutions within a holistic design
perspective driven by pedestrian movement perfor-
mances.
REFERENCES
Bada, Y and Farhi, A 2009, 'Experiencing urban spaces:
Isovists properties and spatial use of plazas', Courrier
du Savoir, No 09, pp. 101-112
Gibson, JJ 1966, In the senses considered as perceptual sys-
tems, Houghton Mifflin, Boston
Helbing, D, Molnar, P, Farkas, I and Bolay, K 2001, 'Self-
organizing pedestrian movement', Environment and
Planning B: Planning and Design, 28, pp. 361-383
Hillier, B 2007, Space is the machine, Electronic edition,
Space Syntax, London
Ireland, T 2008 'Space diagrams: The problem of spatial
arrangement and the automatic generation of archi-
tectural plans', Proceedings of eCAADe 2008, Antwerp
Kontovourkis, O 2012, 'Design of circulation diagrams in
macro-scale level based on human movement be-
 havior modelling', Automation in Construction, 22,
pp. 12-23
Lang, J 2005, Urban design: A typology of procedures and
products, Elsevier/ Architectural Press
Lewin, K 1951, Field Theory in Social Science, Harper and
Row, New York
Lynch, K 1960, The Image of the City, MIT Press, Cam-
bridge MA
Peter, A, Newton, B and Wills, C 2009 'Visualising Archi-
tecture: A field study in rural England', Proceedings
of 11th Bogus Conference, London, pp. 12-14
Puusepp, R and Coates, P 2008, 'Emergent Circulation
patterns: Generating movement networks for build-
ings using stigmergy', Proceedings of SFB/TR 8 Spatial
Cognition, Veracruz, pp. 85-89
Woloszyn, P and Leduc, T 2011 'A landscape potential
characterization: spatial template of pedestrian am-
bient fields within the urban fabric', Proceedings of
1e conférence intercontinentale en Intelligence Territo-
riale (IT-Gatineau 2011), Gatineau
[1] http://newschool.academia.edu/EvaPerezdeVega
[2] http://utos.blogspot.com/
[3] http://www.grasshopper3d.com/
[4] http://www.simwalk.com/

Towards Smarter Cities - Volume 1 - eCAADe 32 | 163

164 | eCAADe 32 - Towards Smarter Cities - Volume 1
Wind sensing with real-time visualisations for Designers
An approach to understanding wind phenomena for pedestrian comfort
using low cost wind sensors

Daniel Prohasky1 , Rafael Moya Castro2 , Simon Watkins3 , Jane Burry4 , Mark
Burry5
1,2,3,4,5
RMIT University, Melbourne, Australia
1
http://www.rmit.edu.au/architecturedesign/sial
1,3,4,5
{daniel.prohasky|simon|jane.burry|mark.burry}@rmit.edu.au
2
s3296513@student.rmit.edu.au

The evaluation of a low-tech wind sensing platform for urban aerodynamic

simulations relevant to pedestrian comfort. In this paper, the wind canyon effect
is simulated with two different building morphologies. The platform provides
conceptual knowledge of the dynamics in wind relevant for designers,
architectural practitioners and students of design. Low-cost hot wire anemometry
is utilised for the design of an Experimental Fluid Dynamic (EFD) wind sensing
network interface. This paper explores the validity of the sensing platform for a
new approach for non-wind engineers to gain a better understanding of the
dynamics of wind. The influence of real-time feedback from quantified wind on
the understanding of wind phenomena for non-wind engineers is discussed and
compared with post analysis data. It was found that real-time quantified feedback
from wind intrigues and stimulates the intuitive notion of wind dynamics through
discussion, however post analysis remains critical to evaluate building design
performance.

Keywords: Wind Sensing, Real-time feedback, Experimental Fluid Dynamics,

Hot-wire Anemometry, Atmospheric Boundary Layer

INTRODUCTION tify the wind condition and visualise the effects in

The study and knowledge of urban aerodynamic
phenomena is very important in the fields of sus-
tainability, environmental design and human com-
fort (Boris, 2007). The aerodynamic phenomena pro-
duced by wind in the built environment and its ef-
fects on the level of pedestrian comfort usually re-
quire complex and expensive technologies to quan-
the built environment. Tools such as: Computational
Fluid Dynamics (CFD) or the use of various high cost
and complex wind sensing techniques i.e. multi-hole
pressure probes coupled with Experimental Fluid Dy-
namics (EFD) can be overly complex (Watkins, 2002).
The nature of wind is beautiful and majestic, though
chaotic which creates difficulties when attempting

Towards Smarter Cities - Volume 1 - eCAADe 32 | 165

 to introduce the topics of observation and analysis
of complex wind flow phenomena in built environ-
ments to architects and designers. Architects and de-
signers should have a broader knowledge about the
more important and fundamental concepts involved
in wind around buildings so they may understand the
parameters involved at the conceptual design stage.
Wind phenomena around buildings has strong rele-
vance in urban planning and architectural design and
is imperative for an architect's ability to share com-
mon knowledge with disciplines such as wind engi-
neering in their professional careers in design (Wisse,
1988).
The Wind Sensing Platform
The wind sensing platform proposes an improved
method of phenomenological exploration and ob-
servation of the dynamics of wind by quantifying
empirical experiments in real-time. The platform
provides real-time feedback of the changes in wind
speed in understandable units (metres per second) at
strategic points around a scaled building design with
a direct digital interface to visualise data. The sensing
platform was tested with two scaled buildings in two
different wind flow scenarios to show the capabilities
of the sensing technology to identify the predicted
wind phenomena surrounding the buildings.
Empirical observations of wind dynamics are the
most informative and stimulative way to understand
the dynamics in wind, which is suggested as a good
introduction method for architects (without a wide
theoretical background in fluid mechanics) to un-
derstand aerodynamic phenomena in built environ-
ments. The sensing platform presents an opportu-
nity to add value to and actually quantify the em-
pirical experiment in real-time for on-the-fly discus-
sions and decision making. Rapid feedback about
the dynamics of wind improves a designer's ability
to choose the most effective design based on the
wind speeds around a building (a good indicator of
pedestrian comfort levels) (Gandemer, 1978). It has
been found that using empirical explorations and
digital sensing technology is much faster when iter-
166 | eCAADe 32 - Towards Smarter Cities - Volume 1
ating through multiple design concepts in compari-
son with virtual simulations (Williams, 2013). This is
just as evident in the specific realm of fluid dynamics
where CFD is notoriously enigmatic.
The sensing platform integrates three main tech-
nological approaches: physical simulations of wind
with a wind tunnel, micro anemometer sensors con-
nected to an Arduino board and a digital interface to
visualise data using Grasshopper3d (0.09.0056) with
the Firefly plugin and Rhino5.0 software.
Figure 1
Hot-wire
anemometer
ModernDevice
(revC)
Figure 2
Hot-wire
anemometer
calibration surface
in Rhino 5 using
Grasshopper3d
The adaptation of the wind sensors
The wind sensors were acquired from ModernDe-
vice (see URL in references). These particular sensors
were originally designed for the purpose of recog-
nising if someone was breathing. Five wind sensors
went through a calibration process within the low-
turbulence aerospace wind tunnel at RMIT Univer-
sity. The anemometers are dependent on temper-
 ature changes in the ambient conditions as well as
the temperature difference created by the change in
wind speed (figure 1). The dependency on the orien-
tation of the sensor with respect to the wind direction
was also tested. A surface was created in 3D space
to relate ambient temperature, analogue signal and
wind speed (figure 2).
EXPERIMENTATION IN THE WIND TUNNEL
The low-cost anemometers were utilised in two main
scenarios within the industrial wind tunnel at RMIT
University, Bundoora. Firstly, the sensors were used
to measure the Atmospheric Boundary Layer (ABL)
wind velocity profile at a 1:100 scale. The second set
of experiments involved two different building forms
that would display interesting wind phenomena. A
second building was placed windward to the build-
ing in question to simulate a canyon wind effect in
subsequent experiments.
Wind Velocity Profile Measurement and Cal-
ibration
The first task was to set up a group of sensors to mea-
sure the wind velocity profile of the 1:100 ABL con-
dition (Aynsley, et al. 1977). An existing rig to cre-
ate the 1:100 ABL condition in the industrial wind
tunnel was used (figure 3). The wind velocity pro-
file was measured with the anemometers and plot-
Figure 3 ted against the ABL power function. The wind veloc-
Wind velocity ity profile and turbulence intensity profiles are doc-
profile anemometer umented from previous experiments with a cobra
measurements probe. These results were used as a reference for
(Scale 1:100) the hot-wire anemometers, though detailed analy-
Vertical slats were sis of the hot-wire anemometer performance for tur-
used to create the bulence intensities will be well documented in fur-
ABL condition ther research. The scope of this paper covers an ex-
ploration into techniques to calibrate the wind tun-
nel boundary layer condition for wind speeds in real-
time (figure 4).
The ABL calibration process is usually an arduous
and time consuming task. The hot-wire anemome-
ters used in this wind sensing platform show promise
in this area of wind measurement. Basically, a the-
oretical boundary layer condition should be calcu-
lated and plotted on an elevation verses wind veloc-
ity graph. A reference velocity should be chosen rela-
tive to the height which is of some significance - com-
monly taken as 10m above ground level for urban
conditions. However, these experiments are focused
on the effect of pedestrian comfort around buildings,
so a reference velocity is taken at 1.5m above ground
level (head height). The sensor positions are then
chosen and values along the theoretical ABL power
curve equivalent to their elevation are noted for ref-
erence during the real-time calibration process.
Timber of various sections were used as wind
barriers to configure the wind dynamic to achieve
the ABL. Each effect of each additional timber ele-
ment was measured in real-time. The aim was to
achieve the reference wind speed at each elevation.
Though when one wind barrier is added, the effect
on the wind measurements were dispersed amongst
surrounding measurement points. A major advan-
tage of multiple sensor measurement creates the op-
portunity to observe these de-localised effects and
immediately take action to balance the distributed
effect. The resultant configuration for the 1:30 profile
included one additional slat and a series of 90x45mm
timber sections to alter the 1:100 ABL profile into a
1:30 ABL (figure 4).
The red curve in the Rhino 5 screen capture moves
with the wind along the wind speed axis (figure 5). It
was possible to observe the dynamic fluctuating ef-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 167
 fects of the wind as the curve snaked around its the-
oretical counterpart, the blue curve, which describes
the theoretical ABL wind velocity profile. The val-
ues represented virtually at the location of the sen-
sors were a display of the target speed and real-time
speed measured by the anemometers. It was difficult
to understand the nature of the velocity profile cur-
vature with these measures alone, but proved to be
useful when trying to reach the target velocities dur-
ing the calibration process (while adding or altering
timber elements).
The methods of visualisation had a strong impact
on how one may understand or interpret the quan-
168 | eCAADe 32 - Towards Smarter Cities - Volume 1
tified wind data. For measurement of wind at vari-
ous chainages along a vector, a continuous curve has
shown to be the most useful. ABL measurement re-
quires the measurement of a reference velocity and
should match the ABL power law (Walshe, 1972). Rel-
ative wind velocities or turbulence intensities need
to be quickly compared with respect to one another
- the curvature of the velocity profile is critical since
the results are represented non-dimensionally and
in non-compressible fluid flows the geometry of the
wind flow should not change with respect to the
magnitude of the wind velocity.
The instantaneous ABL curve representing the Figure 4
anemometer measurements appeared to move too Wind velocity
and fro about the theoretical ABL curve. During ob- profile anemometer
servation, discussion and manipulation of the timber measurements
elements it was possible to see the change in the (Scale 1:30) A
wind velocity profile using this curve, however, as combination of
a time dependant visualisation (45.7Hz sample rate vertical slats and
with 20 point weighted smoothing applied). This horizontal wind
removes the abstraction of the static depiction of barriers were used
post analysis velocity profile plots. The decrease in to create the ABL
abstraction of the reality is proportionate to the in-
crease in understanding of the reality. In this sense
the process is valuable to the non-wind engineers Figure 5
whom are interested in observing and understand- An example ABL
ing wind dynamics. wind velocity
It was possible to estimate the wind velocity pro- profile
file of the ABL through real-time observations of the measurement
hot-wire anemometer measurements. The observed real-time
data was strategically logged and quite readily av- visualisation
eraged over a one minute sample to plot the wind (calibration of the
velocity profile. It only took one attempt to achieve 1:30 velocity profile
a very reasonable resultant ABL condition. The en- (Castro et al., 2014))
tire process of calibrating the wind tunnel from a
1:100 profile to a 1:30 profile was achieved in approx-
imately an hour.
Measuring Wind Speeds around Buildings
The following building examples were chosen to cre-
ate some interesting, but clear wind effects and in
some cases quite well known in the wind engineer-
ing literature. Effects, such as: the canyon effect
 and channel effect were created with the two build-
ing morphologies (Penwarden, 1975). It was pos-
sible to check the wind speed at various locations
around each building model and project the quanti-
fied wind values in real-time to the observers. Some
interesting discussions about the wind phenomena
displayed began to change how we approached the
experiment, but also allowed us to reflect directly on
the data which we saw in real-time on the digital in-
terface.
Figure 6
Virtual visualisation
of quantified wind:
Cube building
example
Figure 7
Cube building
example
45(W)x45(L)x45(H)cm
Two scenarios
depicted: 1.
without smaller
building 2. with the
smaller building
The experiment to test the sensor platform was de-
signed to simulate of the canyon effect, an acceler-
ated vortex of wind in the space between both a high
building and lower building. The phenomenon is
produced by a downward wind flow deflected by the
high building facade to the ground level. The wind
effect around a building could have a strong impact
on the pedestrian comfort level. For this reason, it is
relevant to study the relationship between geometry
of a windward facade and the wind speed deflected
from the facade to the ground.
The aim of the experiment was to detect the dif-
ferences of the wind flow produced with respect to
the two buildings, one of a regular shape and another
with a twisted hyperbolic form. The sensors were in-
stalled in specific points in front of the building fa-
cade, at the corners and on the top of the building to
detect the main variations of the wind flow (Penwar-
den, 1975).
Digital Interface and Real-Time Feedback
Using Grasshopper3d with firefly and Rhino 5 it was
possible to develop a digital interface to directly re-
ceive the analogue signals from the sensors, calibrate
them through a surface in 3D space and translate
this data into graphical information. This allowed an
easy comparison between physical and virtual reali-
ties. Additionally, this information can be overlapped
on a 3D digital model of the building shape (figures
6 and 8) or urban configuration to help define zones
of turbulence and wind speed variations.
The anemometers were placed in regions cho-
sen with reference to ratios of the building geom-
etry (Penwarden, 1978). These configurations were
matched with theoretical wind patterns which are
well known in the field of wind engineering. These
were very relevant in the case of the cube building
example, in contrast to the hyperbolic building mor-
phology. So, a smoke machine was used for direct
observation of the turbulent characteristics around
the hyperbolic building (video footage is available).
It was evident that the wind was channelled and ac-
celerated along the facade down towards sensor 10.
Diagonally skewed eddy vortices were also observed
on the leeward hyperbole surface. A similar and
slightly more intense effect was observed in the re-
gion of sensor position 2. Though, this was a combi-
nation of wind channelling and wind shedding about
the acute building edge condition.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 169
 The sensors were strategically placed around each
scale model building. During on-the-fly discussions it Figure 8
was possible to identify the accelerations in wind due Virtual visualisation
to wind shedding on the windward corners of the of quantified wind:
buildings, the amplified effects of the wind within the Hyperbolic building
region between the smaller building and the larger example
(canyon effect) and the differential effects of the wind
on the leading edge of the rooftop. It was also pos-
sible to identify that there are asymmetries in the
wind flow within the wind tunnel. The observation of
the asymmetries in the wind flow became very useful
knowledge for the post analysis process - something
which would not have been readily concluded dur-
ing post analysis. These observations were recorded
through video and audio recording during the exper- Figure 9
iment. A selection of these recordings will be pre- Hyperbolic building
sented at the conference. example Base-Roof:
45(W)x45(L)cm
Post Analysis Results Height: 45(H)cm
An interesting example of the canyon effect has been Twist: 45deg Two
observed in the above results. The canyon effect is scenarios depicted:
created by the differential wind speeds simulated by 1. without smaller
the ABL condition in the wind tunnel. One would as- building 2. with
sume that when one blocks the wind with another smaller building
building the wind speed should decrease at sen-
sor position 3. However, we observe the contrary.
The effect of the ABL and low pressure systems in-
between the two buildings creates an accelerating Figure 10
effect (Penwarden, 1975). Wind speeds at
It is evident that after post processing the col- various locations
lected data, the average speeds surrounding the two
building morphologies expressed very similar pat-
terns to those inferred from the real-time observa-
tions.
This is the first application for the measurement
Figure 11
of turbulence intensities for these particular wind
Turbulence
sensors. So, it is not possible, at this early stage to
intensities at
verify the measurements' accuracy. Though it may
various locations
be within reason to speculate that the measurements
observed are in-fact quite relevant - as is observed
in "Aerodynamic strategy applied in an urban shel-
ter design"(also presented in this conference) (Moya
Castro, 2014). These observations are merely self-

170 | eCAADe 32 - Towards Smarter Cities - Volume 1

reinforcing evidence at this stage, however further
tests will be conducted prior to the conference pro-
ceedings to clarify the accuracy of turbulence inten-
sities and wind speeds measured with the low-cost
anemometers.
The application of the sensor platform for rapid
visualisation and comprehension of physical simu-
lations can improve the observation and analysis of
wind dynamics in the built environment for archi-
tects and designers, potentiating the communica-
tion with other specialists. This tool has the potential
to be used in the study and design of microclimatic
conditions for pedestrian areas near buildings.
CONCLUSION
This research presented a technological platform
which integrated different techniques of simulation
and visualisation of wind phenomena and analysis of
aerodynamic simulations of wind flow around built
environments. This technology has the potential to
help non-engineers to better understand the dynam-
ics of wind around buildings. It is a practical tool to
analyse problems of discomfort produced by wind
in areas near buildings, but also, it is a tool that can
assist architectural practitioners and designers to ex-
plore aerodynamics through quantifiable empirical
observations. It is a cheap, relatively low tech, reliable
and quick method of measuring wind effects around
buildings. This approach to wind visualisation is rel-
evant because architects and designers should share
(at least) a basic knowledge of wind with engineers
to have a discussion about the problems related with
the factor of wind in design.
OUTLOOK
This wind sensing platform is very relevant for use
in the design industry and wind engineering. The
fact that this sensing platform is a low-cost, reliable
and quick feedback tool for the measurement of wind
speeds within reasonable accuracies promotes this
research as an option to adapt the sensors to any con-
ceptual design challenge where the effects of wind
are relevant.
ACKNOWLEDGEMENTS
This research is part of a larger ARC discovery project:
'Integrating architectural and mathematical knowl-
edge to capture the dynamics of air in design'.
REFERENCES
Aynsley, R. M., Melbourne, W. H. and Vickery, B. J. 1977,
Architectural aerodynamics, Applied Science Pub-
lishers, London
Boris, J 2005 'Dust in the Wind: Challenges for Urban
Aerodynamics', Proceeding of the 35th AIAA Fluid Dy-
namics Conference and Exhibit
Moya Castro, R, Prohasky, D. J. and Watkins, S 2014 'Aero-
dynamic strategy applied in an urban shelter de-
sign', Proceedings of eCAADe 2014, Newcastle
Gandemer, J. and National Institute of Standards and
Technology, NIST 1978, Discomfort due to wind near
buildings: aerodynamic concepts, Dept. of Com-
merce, National Bureau of Standards, Washington,
D.C, U.S
Penwarden, A. D. and Wise, A. F. E. 1975, Wind Environ-
ment Around Buildings, BRE Report, London: BRE,
HMSO
Walshe, D. E. J. 1972, Wind-excited oscillations of struc-
tures : wind-tunnel techniques for their investigation
and prediction, HMSO, London
Watkins, S, Mousley, P. D. and Hooper, J. D. 2002
'Measurement of fluctuating flows using multi-hole
probes', Proceedings of the 9th International Congress
of Sound and Vibration, Orlando, Florida, USA
Wisse, J. A. 1988, 'A Philosophy for Teaching Wind in
the Built Environment', Energy and Buildings, 11(1),
p. 157–161
[1] http://moderndevice.com/product/wind-sensor/
[2] http://arduino.cc/
[3] http://fireflyexperiments.com/
[4] http://www.grasshopper3d.com/
[5] http://www.rhino3d.com/
Towards Smarter Cities - Volume 1 - eCAADe 32 | 171
172 | eCAADe 32 - Towards Smarter Cities - Volume 1
Shared Space Navigation
Charles Avis
Princeton University
cavis@princeton.edu

Shared space is a concept of urban planning in which all barriers between cars
and pedestrians, such as curbs and crosswalks, is removed to encourage
heightened awareness of drivers and pedestrians, thus making city streets safer.
The system has been highly successful, but can be highly stressful due to the lack
of rules and signage. Thus, an adaptive feedback system that guides one safely
through shared space could be essential for a shared space on the city scale. This
paper imagines shared space at the city scale, and uses computational strategies
to develop a system of adaptive collision-avoidance. By abstracting the
movement of cars and pedestrians to properties of moving 'agents', collision
detection and adaptive path finding models are developed, and then prototyped in
an immersive environment that experiments with variable visual feedback based
on user interactions.

Keywords: Shared space, movement, visual feedback, traffic, urban

INTRODUCTION things" [4]. However, my hypothesis is that compu-

Shared space is the urban planning concept of get-
ting rid of curbs, crosswalks, and lanes on city streets
to allow pedestrians, cars, and cyclists to travel
through the space at their own risk. The idea is
that each person moving through the space will pro-
ceed with a heightened caution, therefore reducing
accidents and contributing to an even flow of traf-
fic. This system has been implemented most notably
in Northern Europe, but has begun to be adopted
- albeit on a very small scale - in the United States
[1]. This paper confronts, at some level, what the ef-
fects of a city organized completely around shared
space would be, and then sets out to develop a feed-
back system to ease the use of the shared space.
The inventor of shared space, Hans Monderman, said
that, "The trouble with traffic engineers is that when
there's a problem with a road, they always try to add
something. To my mind, it is much better to remove
tationally providing certain traffic feedback and nav-
igation advice - essentially one's own personal road
signage - would be a beneficial for the users of shared
space. This presented two avenues of research: one
was to simulate the movement of agents and their
interactions within shared space and shared space
between built environments, taking particular care
in analyzing patterns of collisions and bottlenecking
around buildings. This is discussed in Section 3. The
second was to design (and speculate) a feedback sys-
tem for users that could provide the necessary in-
formation gathered from the simulation models to
the user, which is discussed in Section 4. The con-
tributions of this paper are in the translation of an al-
gorithm for collision avoidance into a conception of
a virtual environment for safe navigation of shared
space.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 173

 RELATED WORK
To the best of my knowledge, there does not ex-
ist a precedent for an adaptive feedback system for
shared space navigation. A notable work, which in
part inspired the contents of this paper, is the Bjarke
Ingels Group (BIG) project for the Audi Urban Future
conference [3], which imagines a city of shared space,
in which cars are self-driving and their future paths
are displayed on the street by the way of LED lights
embedded the Solar Road product [2]. However, the
project does not confront the actual algorithms of
navigation and avoidance within the shared space,
and is more of an exercise in visualization. What is
useful for our purposes is the use of the Solar Road
as a means of sensing the movement of pedestrian
agents as well as providing visual feedback in an im-
mersive environment. Another important advance-
ment that will factor into the discussion of system im-
174 | eCAADe 32 - Towards Smarter Cities - Volume 1
plementation is the development of augmented real-
ities through current mobile phones (Takacs). Finally,
Anvari et al developed a model of the movement
dynamic between agents in shared space as well as
route finding around buildings. The Anvari model is
based on the Social Force Model (SHM), which deter-
mines an agents movement based on forces applied
to the agent by other agents and barriers in the space.
Although I built my own movement dynamics in the
model simulation, Anvari et al's algorithm is more ro-
bust and will thus be considered useful in further de-
velopment.
ABSTRACT MODEL FOR AGENT INTERAC-
TIONS
This section presents two models for the simulation
of agents in shared space, one based on the SHM
Figure 1
Parameter diagram
for abstract
movement model
 (Anvari), and the other based on my own develop- places of high collision occurrence would reduce the
ment of a geometry-based avoidance algorithm. The number of collisions in the future. The test was con-
first model was useful for observing abstract trends ducted by constructing a high resolution of small
of agent collisions and interactions with built envi- structures, and then deleting structures as collisions
ronments (barriers). The second was a step towards a occurred near them (Figure 3). The noisy pattern
robust route generation through shared space based created reflected a pattern of collisions, yet it was
on current velocities of the agents within a space. found that collisions would still occur in the opened
spaces. Thus, built environments in shared space
Model for Abstract Movements were dropped in favor of a more adaptive, virtual en-
This model incorporates various abstractions in order vironment.
to produce a simulation of agents in shared space.
Figure 2
The agents were imagined as homogeneous - so no
Bottlenecks and
distinction between cars and pedestrians was built in
collisions with
- however each agent was assigned a set of parame-
structures in agent
ters to control its movement. These parameters were
simulation
aggression, sight, steering, and speed. At each time
step in the model simulation, an agent detects other
agents within a cone of sight. This cone of sight has a
radius and angle determined by the sight parameter.
If an agent is detected within the cone of sight, the
Figure 3 two agents compare aggression variables, and the
Generative least aggressive executes an avoidance maneuver to
subtraction of pass behind the agent in its sight. The avoidance ma-
spaces from neuver in this model was to accelerate towards the
predicted collisions. negative velocity vector of the agent to be avoided.
Once the agent executing the avoidance no longer
detects an agent in its sight, it reroutes towards its ini-
tial destination. An avoidance function is also called
if a built environment is detected by the agent, which
similarly accelerates an agent towards a corner of the
structure to allow it to pass. The speed and steer-
Geometric Avoidance
A second model intends to choreograph movements
ing parameters determine how quickly an agent can
more precisely within shared space. This second
accelerate linearly and radially. The parameters and
model is entirely geometric and assumes that agents
avoidance functions are illustrated in Figure 1.
comply fully with its rules, have perfect reaction
This model sets up a framework through which
times, and travel at a constant speed. At each time
to analyze interactions in shared space. The goal
step, all agents project their current velocity vectors
was to introduce a built environment into the shared
forward some integer N times. At each step of the
space simulation, and detect patterns of collisions.
vector projection, the model calculates whether an
As one can see in Figure 2, built environment caused
agents future position, as calculated by the projected
a high concentration of collisions in the intersections
velocity vector, will collide with another agent's re-
(where the paths of agents crossed often) and bottle-
spective future position (Figure 4 left). If a future colli-
necks (where agents were forced into confined path-
sion is detected, then the model deflects the pathway
ways. Another hypothesis was that by opening up
of agent to the right by a magnitude proportional to

Towards Smarter Cities - Volume 1 - eCAADe 32 | 175

 Figure 4
Projection of
velocity vector, the
detected collision,
and the calculated
geometric offset of
path to avoid
collision.

the magnitude of the agent's velocity (Figure 4 right). Figure 5

Both agents in this model deflect to the right, how- Geometric
ever it is possible for the two agents to compare a pa- deflection model
rameter that prioritizes one of the agents, forcing the simulation. Green
less preferred to deflect. This is a scenario I wanted dots represent
to avoid however. What is favorable about this ge- projected velocity
ometric model as a form of shared space navigation vectors, yellow
is that it provides a calculable and unwavering path spots represent
of navigation for the agent. Also, it adapts at each detected collisions
time step, which means that after a deflection, the that forces a
agent changes direction and recalculates the agents geometric
future positions. Figure 5 visualizes the movement deflection. Black
and computations of this model in an environment line represent
of built space and shared space. Thus, although the actual movement of
model deals with 100% compliance with the naviga- agents.
tion system, the system is reliant on the current ve-
Motion Tracking and Immersive Environ-
locity of the agent at each time step, so even if the
ments
The first prototype for a feedback system for moving
agent responds with 75% compliance, the system will
agents was an implementation of motion tracking
keep calculating the agents future collisions and con-
with the Microsoft Kinect and floor projection. The
stantly display a path that avoids collision. This is
Kinect registered the position of users that entered
ideal as a suggestive system of shared space naviga-
into its range, and then a Processing sketch would
tion.
project information onto the ground at the feet of the
user in the space. The system had two primary func-
SHARED SPACE FEEDBACK INTERFACE tionalities, each meant to test the basic conditions
This section outlines one instance of prototyping for of the models described in the previous section with
an immersive environment of motion feedback as an live users:
implementation for a shared space navigation sys-
tem, and also speculates on the scaling up of such a
feedback system to a city scale. • The first was to generate information for each
user based on the shared space conditions

176 | eCAADe 32 - Towards Smarter Cities - Volume 1

 and display the information at the user's feet. the desired path at a constant speed, so that
The information displayed included a trail of the user could follow the agent as it success-
the user's past movements, an arrow pointing fully navigates the space. I speculate in the
to each other user in the shared space, a line next section on possible implementations of
pointing to the user's destination, and a circle a feedback system on the scale of a city.
of "personal space" around the user (Figure 6).
Figure 6 The personal space circle changes from green
Projected personal to red if it comes in contact with another per-
bubble and route son's personal space (Figure 7). The expected
information. result was that displaying such information
would change how one moves through the
space and interacted with others within that
space. What was observed was that users, as
seen in Figure 7, would seek the red warning
circle as a reward for their movements. The ex-
pected result of using the red circle as a warn-
ing was reversed. Therefore, an effective sys-
tem could employ feedback rewards for good
compliance, so that the system would react
positively in the case of good avoidance be-
Figure 7 havior. The other functions of displaying in-
Visual feedback formation were not found to have a significant
through ground effect on movement behavior.
projection for
detected collisions. • The second functionality was to choreograph
movement, as the model in Geometric Avoid-
ance section outlines. Paths were displayed
on the ground, for users in the space to fol-
low. Each path is a straight line from the
user to their respective destination (which, in
this prototype, is randomly generated), unless
their projected velocity vector detects a col-
lision with another user in the space. Since
compliance is not guaranteed with live users,
the path becomes more of a suggestion than Scaled Implementation Speculation
a strict movement-bounding condition. At Scaling up a projection-based immersive environ-
the scale of the implementation, it was hard ment poses numerous problems. It would require
to determine the success of the choreogra- projectors strategically placed at all shared space
phy, as the projection environment is only road systems, there would be interference between
a few square meters. Two further additions the projection and the users, and projection is hard
could produce the desired results: a positive to see in daylight. The Solar Roadway [2] imple-
reward feedback system as discussed above mented as a feedback system would allow for ef-
would encourage compliance, and perhaps fective ocular feedback as discussed in the previous
also generating a simulated agent that takes section, and could be a viable alternative to projec-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 177


tion. It still has aesthetic and functional drawbacks,
though. Mobile augmented reality is another pos-
sible option. A personal projector attached to the
user could project their path in front of them using
technology proposed by Schöning et al, but this still
deals with the problem of luminosity of the projec-
tor. Specific audio headsets could given auditory
feedback, but the results of route finding with the
study done by Vazquez-Alvarez et al reveal low ac-
curacy of user movements given auditory signals.
Mobile phones, however seem to be the most use-
ful tool, as they are widely used and dependable.
Takacs et al have developed an augmented reality al-
gorithm for mobile phone cameras, allowing virtual
images to be displayed over real images. By display-
178 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 8
Visualization of
possible
implementation of
geometric
avoidance system
on a mobile phone
in shared space.
ing the shared space navigation feedback onto a mo-
bile phone screen, the system could function as a
more detailed and precise implementation of current
navigation features on GPS-enabled mobile phones.
GPS-enabled mobile phones also provide a platform
for motion tracking, as the GPS data can inform the
vector calculations of the model. Ocular-centric feed-
back has produced noticeable results in the proto-
type of the previous section, thus translating signage
as a virtual reality onto mobile devices should be con-
sidered in the development of shared space naviga-
tion (Figure 8).
CONCLUSIONS AND FUTURE WORK
Implementable shared space navigation in the form
of adaptive movement feedback requires further de-
velopment to become a robust and scalable op-
tion for making transportation through shared space
more manageable. This paper has outlined a ba-
sic model of geometric collision avoidance, and pre-
sented one possible prototype of integrating live
users into the model via an immersive projection en-
vironment. Such an immersive environment should
not be considered a final design implementation,
however, and other avenues of displaying motion
feedback and navigation should be considered, such
as implementation in mobile phones. Future re-
search should focus on an effective mobile feedback
system that is sleek, user-friendly, and relatively invis-
ible, as well as develop the robustness of a geometric
collision avoidance system.

REFERENCES
Takacs et al, G. 2008 'Outdoors augmented reality on mo-
bile phone using loxel-based visual feature organi-
zation.', Proceedings of the 1st ACM international con-
ference on Multimedia information retrieval (MIR'08),
New York, NY
Anvari, et al, B. 2012 'Shared Space Modeling Based on
Social Forces and Distance Potential Field', Pedes-
trian and Evacuation Dynamics 2012
Schöning, J., Rohs, M., Kratz, S., Lochtefeld, M. and
Krüger, A. 2009 'Map torchlight: a mobile aug-
mented reality camera projector unit', In CHI '09 Ex-
tended Abstracts on Human Factors in Computing Sys-
tems (CHI EA '09)
Vazquez-Alvarez, Y., Oakley, I. and Brewster, SA 2012,
'Auditory display design for exploration in mobile
audio-augmented reality', Personal Ubiquitous Com-
puting, 16, 8, pp. 987-999
[1] http://bettercities.net/article/us-shared-space-
starting-small-13673
[2] http://www.solarroadways.com/main.html
[3] http://www.big.dk/#projects-audi
[4] http://archive.wired.com/wired/archive/12.12/
traffic.html

Towards Smarter Cities - Volume 1 - eCAADe 32 | 179

180 | eCAADe 32 - Towards Smarter Cities - Volume 1
Emergent articulation field in existing urban context
Computational typologies with emergent matter

Peter Búš1 , Lukáš Kurilla2 , Henri Achten3

1,2,3
The Cabinet of Architectural Modelling MOLAB, The Faculty of Architecture,
Czech Technical University in Prague
1
http://www.archa3d.com 2 http://kurilluk.com 3 http://fa.cvut.cz
1,2,3
{buspeter|kurilluk|achten}@fa.cvut.cz

The main purpose of this research is to introduce and verify selected advanced
computational methodologies operating with agent-oriented approach that are
applicable within the framework of emergent urban design strategies in early
stages of zoning proceedings. The aim is to demonstrate that speculative matter
of the phenomenon of emergence in terms of design thought driven by
computational models can also be applicable within real existing urban
conditions. The research concurs previous researches in the field of urban
pre-reconfigurations in the selected city environments. It concentrates on
developing a simulation model as an analytical and generative instrumental
platform for architects in early phases of urban design and synthetizes
accumulated computational advancements based on agent-oriented approach.

Keywords: emergence, agent-oriented modelling, urban agency, colonial

growth, bottom-up strategy

INTRODUCTION In the next phase the research will focus on com-

The simulation model in the previous phase (Buš
2012, 2013) has documented the dynamic processes
in the selected urban place in terms of optimiz-
ing walking distances in distribution routes based
on different intensities of urban activities attrac-
tion which can be interactively parametrically user-
defined. The essential methodological approaches
have been used for the modelling of agent-based
system with Attraction, Path following and Stimergy
methodologies that led to interaction of agents with
selected environment bringing preliminary graphical
outputs (Shiffman 2012),(Spyropoulos 2013)[1].
plementing and synthetisation of other computa-
tional typologies that determine and influence the
behavior of agents in the urban structure and con-
sequently lead to more complex simulation of emer-
gent behavior of structures with characteristics of
non-linearity and unpredictability of reconfiguration
of the existing area based on participant's bottom-up
stimuli (Spyropoulos 2013).
The Aims of Simulation
The main objective of the reconfiguration simula-
tion of the existing environment (selected territory in

Towards Smarter Cities - Volume 1 - eCAADe 32 | 181

 Prague - Southern City with 80,000 inhabitants) is to
gain insight into the possibility of developing an op-
timal urban structure with an emphasis on increas-
ing the quality of residential and public urban envi-
ronments. The current state of the environment has
certain inadequacies that reduce the environmental
quality. These are in particular: strong uniformity of
post-totalitarian urban environment, the lack of avail-
able and city-creating public spaces, approach dis-
tances to places of public activities are insufficient in
some parts, the environment has a large blank areas
that are unrelated and separated by barriers of high-
ways.
The aims of the simulation are mainly:
• Verification of the importance of selected
places in the model and their prospective in-
tensification with new urban activities.
• Strengthening the environment diversity, i.e.
eliminating the uniformity of the areas.
• Verification of approach distances to urban
activities (maximum 10 minutes of walking
time, about 700-800 meters or even less).
• Strengthening and possible reconfiguration
of communication (distribution) flows.
182 | eCAADe 32 - Towards Smarter Cities - Volume 1
• Optimal distribution and incorporation of sur-
rounding green areas - forest parks in the en-
vironment of existing buildings - land over-
growth into the city.
• Extension of public and private activities with
relevant spatial structures with different typo-
logical form (new volumes), i.e. completion
within the framework of undeveloped land
without habitable scale.
METHODOLOGY
For the purposes of the simulation model there were
used advanced computational techniques by means
of Processing language and developed algorithms to
conduct a gent-based system. Simulation model op-
erates with the agent's class from Plethora library [2].
Transcription of the Environment into
Agent-based Model
In the first phase it was necessary to interpret the vec-
tor GIS data of the current site into Processing en-
vironment for the simulation needs into two layers:
the vector flow and the cell grid. The cell grid layer
represents existing activities and volumes of urban
structure, which consists of cell clusters (Batty 2007,
2011). The dynamic processes in the urban environ-
ment are interpreted by the vector flow (see figure
1). The agents will respond to attributes of the envi-
Figure 1
Transcription of the
environment into
the agent-based
model. The
diagram contains
GIS data, the grid
cell layer and the
vector agent flow
layer.
 ronment such as physical barriers or infrastructure by
adopting their behavior to the features of the mod-
eled environment (Stanilov 2012). Both layers are in
mutual interaction.
There are urban activities defined as attractors
and initial positions of agents in the cell grid and
these attractors influence behavior of agents in the
flow layer. The cell grid is apposite to the geomet-
ric characteristics of the area and contains certain in-
formation about the structure (position, height map,
characteristic number of agents received into the po-
sitions).
The territory will operate with a predefined ca-
pacity of the environment for the selected number
of agents, while the population of agents is charac-
terized by its objectives for a specific boundary num-
ber of places for colonization or area reconfiguration
(DeLanda 2011). The environment defined by the cell
clusters will operate with proper basic capacity (ac-
tual statement of the building) and extended capac-
ity in the case of adding new extensions and volumes
(see figure 2).
Figure 2
Existing building
structure is defined
by cell cluster.
Cluster operates
with basic and
extended
capacity-it receives
specific number of
activating agents.
Colonization Simulation - Colonial Growth
The colonization of the area will be conducted by
searching of activities within defined approach dis-
tances and within unpopulated environment using
initial groups of agents - colonizers and then by
adding new geometric volumes within a predefined
grid through a predetermined rule. The capacity of
the agent's population and the environment capac-
ity for a given number of agents are user-defined in
the graphical user interface. The environment is able
to embrace a predefined number of colonizers that
inhabit the territory. The territory is supplemented
and thickened with new areas defined on the basis
of various capacities of agent population and envi-
ronment.
The simulation rules according to colonial
growth will be as follows (see figure 3):
• From the initial positions in a situation (de-
fined centers of individual environment parts,
subway stations) defined number of agents
are released. Agents are representing defined
number of users (participants in the environ-
ment). It is necessary to find level of resolution
of the agents (Leach 2009b).
• The agents seek private urban activities in
specific time period within approach distance
and fill the cluster to its unused capacity (see
figure 3a).
• If agents reach the private urban activities,
their initial position and nature will change.
Agents alter the target of interest and begin
to seek public urban activities and continue
to colonize the environment from different
initial positions represented by already colo-
nized private clusters (see figure 3b).
• In case the basic capacity of the cluster is sat-
urated by the agents, a new extended volume
appears. The cluster attraction is removed
from the original position and located to a
new extension until it is saturated and subse-
quently it disappears (see figure 3b, c).
• In case the agent fails to reach the activity
within the approach distance, a new volume
is established (see figure 3c).
OBSERVATION AND EVALUATION
The developed simulation model has been tested
with several different input parameters and user-
defined parameters interactively during the simu-
lation in order to control agent movement using
different flow algorithms (Path following, Stimergy,
Towards Smarter Cities - Volume 1 - eCAADe 32 | 183

Free Distribution flow, Attraction). Four simulation
cases have been tested with varied number of agents
within 200-2000 iterations (see figures 4, 5, 6, 7).
Model Evaluation in Technological Terms
From a technological point of view, the model is an
open-source platform that can be extended by addi-
tional parameters and algorithms that will comple-
ment its technological potential. The mergence of
the grid layer and vector dynamic flow layer in or-
der to generate solid structure in the grid is relevant
for future work with the system. Such a system simu-
lates the movements and selected distribution flows
in the urban environment in the simplified model.
The method of GIS geometry transcription by using
the developed algorithm is of a major importance for
its application for different environments and spatial
case studies. Grid layer allows architects to work with
varied scales and typologies. Transcript can be exe-
cutable by using any other GIS documents with di-
verse geometrical character.
184 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 3
Simulation rules of
the colonial growth
with initial positions
and activated
cluster capacities.
Interpretation of the Results
The simulated situations must be interpreted and as-
sessed in terms of compliance or non-compliance of
the objectives of the simulation mentioned above.
For optimal evaluation it is necessary to assess ob-
servations separately according to the selected point
scale 1-5 (1 - worst rating, 5 - best rating) for each
stated objective (i.e. the objectives are the evalua-
tion criteria) and within each iteration (200, 400, 800,
1200, 1600, 2000). The environment quality score is
averaged into a single value for each simulated case.
The rating is interpreted in the graph (see figure 8)
where the curves relate to the time evolution of the
model with the quality value of the generated urban
environment. In such a way it is possible to interpret
the validation of the model as a decision-making tool
for the architectural practise.
Figure 4
Observation with
150 colonizers. Free
Distribution Flow
algorithm.

Figure 5
Observation with
300 colonizers. Path
Following
algorithm.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 185

 Figure 6
Observation with
510 colonizers.
Stimergy and Path
Following
algorithm.

Figure 7
Observation with
510 colonizers, high
level of Attraction
algorithm which
leads to decay of
urban structure.

186 | eCAADe 32 - Towards Smarter Cities - Volume 1

Figure 8
Validation of the
simulation model
according to the
quality of the
generated urban
environment.

FUTURE WORK defined pattern formation by specified stand-

It will be essential to test a diverse number of trans-
mitted agents of a particular type (public, private),
the measure of approach distances and the measures
of the population capacities influencing the volume
sizes and environment capacity as well and observe
the differences in structure and compare them. The
simulation would be tested with the same general
inputs and parameters several times. It is also es-
sential to compare models with different input set-
tings to see whether the structure will develop differ-
ently or in the same way. By executing this computer-
based model several times, with different initial set-
tings, we may discern patterns and regularities that
recur in the results (Holland 1998). Architects can ver-
ify several spatial scenarios of the environment de-
velopment by observing the differences between the
results of simulation and look for optimal structure,
which creates a habitable environment and captures
a sufficient volume required facilities.
The future work will include following steps:
• Testing other user-defined urban typologies.
Typology volumes and configurations of ur-
ban growth figures can vary within a pre-
off distance and height of the buildings.
• Testing the validity of the model in various
combinations of parameter values.
• Implementation of new rules, i.e. lighting
requirements, height zoning, mutual stand-
off distances in terms of parametrically user-
defined regulations.
• On-line publication as a standalone applica-
tion.
• A comparative study with other existing city
or cities in terms of pattern formation of urban
settlement.
CONCLUSION AND DISCUSSION
There has been established a spatial non-linear simu-
lation platform of reconfiguration of urban structure
using a combination of several computational strate-
gies that lead to a simulation of colonial growth. It is
necessary to continue with the simulation of recon-
figuration of entire urban activities and observe the

Towards Smarter Cities - Volume 1 - eCAADe 32 | 187

 results of emergent behavior in the model. Its devel-
opment can be interactively influenced according to
user input requirements, features and requirements
of agents that represent certain number of partici-
pants.
It is vital to determine the evaluation criteria for
such a model from multiple perspectives. One of
them is the issue of a real-life application. Simulation
model definitely does not simulate all the complex
dynamic phenomena in urban environments simul-
taneously, however, it is offering a certain system-
atic solution where individual dynamic phenomena
of the environment can be interpreted according to
needs representing, in a particular way, respective so-
lutions that can be used by architects or serve as a
preliminary predicative tool in design process.
By means of different values of the individual fea-
tures of the agents (communication flow speed and
agent population measures, distribution rate and de-
gree of attraction and connectivity) it is possible to
change the nature of the environment and thereby
reach diverse predictions in urban structure config-
urations. Another criterion might be the quality
of the built environment, which can be assessed in
terms of architectural and urban perspective (habit-
ability, spatial quality, diversity of the environment,
aesthetic requirements, etc.). The above computa-
tional model has been providing an insight that can
be characterized in terms of the real and exact ap-
plication to some extent speculative. Anyway, it is
a platform that explores the behavioral relationships
of participants in the environment and therefore it
tends to approximate to the real demands of popu-
lation of the urban environment. The model can be
used also for other spatial case studies.
The agent-based methodology in the existing
urban environment for the creation and reconfigu-
ration of optimized built environment is applicable
within real urban context. It will be useful in a par-
ticular observation of the selected bottom-up partic-
ipant´s stimuli.
It is assumed that the multilevel scenario devel-
opment of the existing environment would be read-
188 | eCAADe 32 - Towards Smarter Cities - Volume 1
able, receivable and intelligible. This assertion is
needed to be verified by extended functional simu-
lation model of the area.
Utilizing the emergence phenomenon as a
paradigm in the design and reconfiguration environ-
ment in the existing urban context is presumed and
characterized as speculative in certain way, but offers
a legitimate insight and a possibility for another un-
derstanding of social and environmental interactiv-
ity in terms of urban design thought using the agent-
based methodology.
Acknowledgement
This work was supported by the Grant Agency of
the Czech Technical University in Prague, grant No.
SGS13/150/OHK1/2T/15.
REFERENCES
Batty, M 2007, Cities and Complexity: Understanding
CIties with Cellular Automata, Agent-based models,
and Fractals, MIT Press, Cambridge
Batty, M 2011, 'Cities, complexity, and emergent order’',
ERA 21, 11(5), pp. 28-30
Buš, P 2012 'Emergence as a design strategy in urban de-
velopment:Using agent-oriented modelling in sim-
ulation of reconfiguration of the urban structure',
Proceedings of eCAADe 2012, Prague, pp. 599-605
Buš, P 2013 'Emergent urban strategies: Rules of city
reconfiguration', Proceedings of eCAADe 2013, Delft,
pp. 283-290
Châtelet, V (eds) 2007, Interactive cities, HYX Anomos, Or-
léans
DeLanda, M 2011, Philosophy and Simulation: The Emer-
gence of Synthetic Reason, Continuum, New York
Holland, JH 1998, Emergence: From Chaos to Order, Ox-
ford University Press, New York
Leach, N 2009a, 'Digital Cities', Architectural Design, 79(4),
pp. 8-13
Leach, N 2009b, 'The Limits of Urban Simulation: An In-
terview with Manuel DeLanda', Architectural Design,
79(4), pp. 50-55
Shiffman, D 2012, The Nature of Code: Simulating Natural
Systems with Processing, Daniel Shiffman, Mountain
View
Spyropoulos, T 2013, Adaptive Ecologies: Correlated Sys-
tems of Living, Architectural Assotiation and the Au-
thors, London
Stanilov, K 2012, 'Space in Agent-Based Models', in Hep-
penstall, AJ, Crooks, AT, See, LM and Batty, M (eds)
2012, Agent-Based Models of Geographical Systems,
Springer Science+Business Media, Dordrecht, Hei-
delberg, London, New York, pp. 253-269
[1] http://natureofcode.com/book/chapter-6-
autonomous-agents/
[2] http://www.plethora-project.com/Plethora-
0.3.0/index.html

Towards Smarter Cities - Volume 1 - eCAADe 32 | 189

190 | eCAADe 32 - Towards Smarter Cities - Volume 1
 Urban body network configurations
Attica

Eirini Androutsopoulou
National Technical University of Athens
iandroutsopoulou@gmail.com

The methodology presented here is grounded on the reconstruction of the urban

body as a network configuration consisting of material and non-material
components (Bateson, 1972). It is based on the assumption that if one can
describe the rules that define the nodes and the connections of the network
construction/urban body, as well as their attributes, then the differentiation on the
relationships between elements, or even a shift from one value to another, would
result in different network constructions, that would produce a time-based
sequence of the self-adaptational and self-organizational reconfigurations
occurring during the mutational procedure. The urban body is defined as the part
of the urban tissue which distinguishes itself from the whole of the urban
landscape, either because of constructed boundaries, or because of the
strengthening of a specific attribute, which would result in a kind of an
immaterial boundary, or, in other words the formation of an identity.

Keywords: Mutation, urban body, visualization techniques, network, data

manipulation

Figure 1 INTRODUCTION
Attributes are The partitions (Blondel et al. 2008) and centrality
assigned at the studies (Newman 2010) made possible through the
elements of the network configuration reveal attributes of the ele-
network. The ments which emerge from the position of each ele-
nodes' colors ment (node) in relation to other elements of the net-
correspond to a work and not from the Cartesian topology. Through
certain type of the application of certain algorithms measuring the
identity or attribute different types of centrality (betweenness central-
assigned on them. ity, degree centrality, closeness centrality) the self-
The network adaptation of the urban body is revealed, on the ba- What is more, being able to represent material
configuration is sis of the alteration of the nodes' connections or the and non-material elements as nodes (Hillier, 2007),
readapted as a differentiation of the attributes and identity of the counter-bodies of mixed proprieties emerge, includ-
counter-body of the nodes themselves. ing physical presence and their attributes. In con-
network is trast to the hierarchical constructions, network con-
introduced

Towards Smarter Cities - Volume 1 - eCAADe 32 | 191

 structions allow for multiple connections between el- firmatively or negatively. The same rule is applied on
ements (Alexander, 1965), therefore being closer to the connections between elements, being connec-
the complexity of the associative forces found in the tions of commerce or civilization (in this case this is
structure of the urban body. written as an adjective), connections with a degree
of accessibility to pedestrians, bicycles etc (numeri-
METHOD cal value) or as connections that satisfy hypotheses
In the case of Attica, inscribed at the Attic peninsula such as safety, environmental comfort etc.
in Greece, spatial, supra-local nodes are selected and
attributes describing their specific qualities are as-
Figure 2
signed (Figure 1). A methodology concerning urban
The network
cohesion, focusing on spatial, social and economic
configuration
attributes of the urban body was followed. The im-
describing
portance of certain nodes, whether well-connected
supralocal nodes
nodes or nodes though which other nodes or groups
and their
of nodes interconnected was evaluated.This study
connections.
is part of the research programme titled "Strategies
for the Networking of Urban Interventions in Athens
Metropolitan Centre" at the National Technical Uni-
versity of Athens and it is directed by Giorgos Par-
menidis which deals with the detection of nodes that
would act as binding elements and the evaluation of
a proposed set of actions that would result in self-
adaptation of the urban body in a way that would re-
evaluate its processes towards a structure of a coher-
ent whole. This research programme was generated
Figure 3
by the region of Attica in Greece (Figure2).
The network
In order to construct the network configuration
configuration of the
of points of interest, rules describing the criteria of in-
map shown in
clusion of nodes and the relationships between them
figure 2. The
are assigned. The urban body is therefore recon-
elements of the
structed in mathematical terms and the mutational
network are
procedure was visualized, based on dissolution of el-
distinguished on
ements and binding forces between them and sub-
the basis of being
sequent reconfiguration of the urban body.
"problematic"
nodes or
NETWORK CONSTRUCTION FOR SPATIAL connections where
NODES AND THEIR ASSIGNED ATTRIBUTES interventions
The network configuration presented here consists of should be made.
elements whose identity depends on their assigned The size of the node
attributes. The characteristics of supra-local nodes depicts the value of
is written on the element's identity, being an adjec- the betweenness
tive describing a spatial node, as a numerical value, centrality
or as a boolean attribute answering an hypothesis af- algorithm.

192 | eCAADe 32 - Towards Smarter Cities - Volume 1

Figure 4 STRATEGIES FOR COHESION OF THE NET-
The network WORK CONFIGURATION FOR THE URBAN
configuration after
the application of
BODY
The methodology adopted here involves the evalua-
the algorithm
tion of the network's nodes in respects to their rela-
measuring Degree
tive position in the network construction and accord-
Centrality
ing to the measurements of betweenness centrality
and degree centrality (Newman, 2010), aiming at a
selection of nodes to be rehabilitated, in a successive
order of phases. It is therefore assumed that nodes
with a greater importance in what concerns their rel-
ative position at the network, would be rehabilitated
in an earlier stage, as opposed to those nodes that
don't contribute much to the cohesion of the urban
body. The term cohesion is used here to describe the
different types of a network construction, depend-
ing on whether they consist of one, connected net-
work or if they consist of two or more separate net-
work configurations (sub-networks). It is assumed
that while sub-networks interconnect to form one,
connected network, the network configuration is re-
garded to "move" from a state of non-coherent form,
Figure 5
to the acquisition of a coherent form of network con-
The network
struction.
configuration after
It is assumed that betweenness centrality mea-
the application of
sure has a greater significance in the re-acquisition
the algorithm
of the coherence of the network construction, due to
measuring Degree
the fact that nodes with a greater degree of between-
Centrality
ness centrality are capable of connecting nodes that
form groups of interconnected nodes and need to
pass their connections through these specific nodes
in order to get connected with the rest of the network
construction.
It is therefore possible to evaluate certain nodes
of the network as being more important, not only be-
cause of the possibility of state interventions, but also
in what concerns their relative position and their sig-
nificance for being a kind of umbilical cord for the
network configuration.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 193

 Table 1
Degree Centrality
Rank Table

Table 2
Betweeness
Centrality Rank
Table

194 | eCAADe 32 - Towards Smarter Cities - Volume 1

Table 3
The resulting values
after the
application of the
different algorithms
at the network
configuration.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 195

 According to the results of the algorithms applied, Figure 6
the nodes KH-TD(12), KP/S(13), POS1, POS2, Thst, Time lapses of the
KIO-AA(14), KIO/P(27), PIR20, Osq, KEO-IO(20), KL- self-adaptations of
K(25), Asq, Ksq seem important for the network co- the network
hesion, due to the high values of degree and be- construction, in
tweeness centrality. Among the nodes proposed accordance to the
for interventions, the nodes KH-TD(12) described as priority given at the
Chamosternas/Tavros Node (c. 12), KP/S(13) de- proposed
scribed as Piraeus/Skilitsi Node (The Gate of Piraeus) interventions. The
(c. 13), KS/K(15) decribed as Siggrou/ Kallirois Node nodes' colour
(c. 15), KIO/P(27) described as Iera Odos/ Pireos Node corresponds at the
(Keramikos) (c. 27), KIO-AA(14) described as Iera priority given, while
Odos/Agias Annis Node (c. 14), SL(9) described as the size of the node
Larissis Station (c. 9), KEO-IO(20) described as Ethniki depicts the value of
Odos / Iera Odos Node (c. 20),KEO-L(21), described the betweenness
as Ethniki Odos/ Lenorman Node (c. 21), KL-K(25) de- centrality
scribed as Lenorman/Konstantinoupoleos Node (c. algorithm. The
25) seem to contribute greatly to the cohesion of the nodes of the
network configuration. category "proposed
for interventions"
NETWORK CONFIGURATION VISUALIZA- are missing.
TION
In order to visualize the network configuration in ac- Figure 7
cordance to the priority given at the proposed inter- The nodes of the
ventions (interventions of a, b and c priority), the fol- category "Priority a"
lowing rules were applied: are introduced.

• initially the problematic nodes are removed

from the network

• at the following stages the nodes are intro-

duced gradually, depending on the priority
given on them. (Figures 6,7,8,9).

More specifically, initially (Figure 6) the network con-

sists of the existing nodes (82.11%) and not those
proposed for interventions.
At the second time-lapse (Figure 7) the nodes be-
longing to category "priority a" are introduced (6.5%;
see table 4). The resulting configurations of the ur-
ban body are based upon the hypothesis that even
though the nodes proposed for interventions exist
spatially, they exhibit a dysfunctional behavior in
what concerns their contribution at the coherence of

196 | eCAADe 32 - Towards Smarter Cities - Volume 1

 the whole. At the network configuration presented
here, it is assumed that if a node doesn't contribute
at the coherence of the network, due to the disconti-
nuity, it creates at a spatial basis and/or at an attribute
basis, it is removed from the network and the result-
ing sub-network is re-adapted and re-organized.
Figure 8
The nodes of the
category "Priority b"
are introduced.
At the third time-lapse (Figure 8) the nodes belong-
ing to "priority b" are introduced (4.88%; see table 5).
Finally, the nodes belonging to "priority c" are in-
troduced (6.5%; see table 6).
The nodes' attributes which refer to their inclu-
sion to categories such as interventions of priority
a,b,c, as well as intervention of a, b and c phase of
completion refer to attributes such as
• the funding potentiality,
• their capability to trigger the development of
compensatory uses,

• their competence with other proposed inter-

ventions,

• their capability to provoke multiscale alter-

ations (concerning the Attica Region, via-
municipal or inter-municipal),

• the ability to substantiate the proposed in-

tervention in discrete phases which enable a
gradual funding accordingly,

• their contribution at the amelioration of the

environmental conditions and the residential
quality.
Figure 9
The nodes of the
category "Priority c" CONCLUSIONS
are introduced. To conclude, the network configuration in the case
where attributes are inscribed as characteristics of
the nodes they refer to supplies us with information
in what concerns the betweenness centrality and de-
gree centrality values, which is an importance mea-
sure of
• the nodes' ability to act as umbilical cords to
the network configuration we refer to

• the nodes being strongly connected with

other nodes in the network

This way, is possible to flourish the criteria of conse-

quent interventions on problematic nodes and there-
fore form a strategy of interventions in different in-
stances (phase a, phase b, etc) which would mean a

Towards Smarter Cities - Volume 1 - eCAADe 32 | 197

 Table 4

Table 5

Table 6

198 | eCAADe 32 - Towards Smarter Cities - Volume 1

selection of certain nodes to be rehabilitated at an
earlier stage, according to their contribution to the
acquisition of a coherent form of the network con-
struction.

REFERENCES
Alexander, C 1965, 'A City is not a Tree', Architectural
Form, 122(1), p. 58–61
Bateson, G 1972, Steps to an Ecology of Mind, University
of Chicago Press, Chicago
Blondel, VD, Guillaume, JL, Lambiotte, R and Lefebvre,
E 2008, 'Fast unfolding of communities in large net-
works', Theory and Experiment, 10
Hillier, B and Vaughan, L 2007, 'The city as one thing',
Progress in Planning, 67, pp. 205 - 230
Maturana, H 2002, 'Autopoiesis, Structural Coupling and
Cognition', Cybernetics & Human Knowing, 9, pp. 5-
34.
Newman, MEG 2010, Networks: An Introduction, Oxford
University Press, Oxford New York

Towards Smarter Cities - Volume 1 - eCAADe 32 | 199

200 | eCAADe 32 - Towards Smarter Cities - Volume 1
A Digital Tool for Customized Mass Housing Design
Ahmet Emre Dincer1 , Gülen Çağdaş2 , Hakan Tong3
1
Karabük University 2,3 Istanbul Technical University
1
aedincer@karabuk.edu.tr 2 cagdas@itu.edu.tr 3 hakantong@gmail.com

Innovative design approaches are needed for mass housing implementations.

Especially increasing interaction between user and designer is major important
in the design decisions of these buildings. For this, it is seriously necessary to
benefit from technological advances in computational designs, because digital
tools like shape grammar, cellular automata, genetic algorithm, l-systems and
agent-based models in this field provide not only to save time and to manage the
relationships but also to generate many different alternatives. Accordingly, a
digital support tool for designers has been developed by using cellular automata
approach and scripts of 3Ds Max software. It produces samples of housing
design plans which is generated by cellular automata approach according to the
data of users' preferences. In this paper the interface and contributions of the
developed model are introduced and discussed.

Keywords: Computational Design, Mass Customization, Innovative Housing

Design, Plugin

INTRODUCTION Mass customization is a process which provides

Housing is where users' basic needs and different
expectations are met and which has private mean-
ings for each user. Mass housing is a large-scale
and dynamic environment where both individual
and collective requirements of many different users
are met. However, mass housing productions which
have been done so far do not fully reflect these prop-
erties of "housing" concept because of the reasons
like mass production and standardization. Users' par-
ticipation has been ignored in the design process.
Also, these productions have been emerged as ordi-
nary and identical building samples. This condition
has revealed a major problem in terms of both users
'satisfaction and urban aesthetic. So, even though it
is in different forms, regarding user preferences is in-
dispensable in housing designs.
users' participation and accelerates interaction of
user-designer-producer (Url-1 2014). It includes
products' properties of breaking into components
and subsystems with modularity and qualifications
like users' selection and method definition with con-
figuration. Pine (1993) points out that its func-
tionality depends on advanced technology applica-
tions like computer aided design, computer inte-
grated manufacturing systems and other computer
technologies. Accordingly, with the impact of de-
veloping digital technologies in architecture, use of
mass customization has been started to increase re-
cently. Mass customization which is used in differ-
ent forms in architecture can be evaluated in four
groups like collaborative, transparent, adaptive and
cosmetic customization (Neimeijer et al. 2010). Al-

Towards Smarter Cities - Volume 1 - eCAADe 32 | 201

 Figure 1
Neighborhood
relations of Housing
blocks in a site plan

Figure 2
Neighborhood
relations of spatial
units in a floor plan

Figure 3
An application of
façade orientations
on a sample of
building model

202 | eCAADe 32 - Towards Smarter Cities - Volume 1

though it is not preferred, adaptive customization
among them is suitable for mass housing designs. In
this model, the production is generally standardized,
but users are given opportunity to make selection
and modifications. This model has been evaluated in
the scope of this study.
A digital tool which generates alternative hous-
ing plans and facade designs in housing blocks by
evaluating user tendencies with cellular automata
approach has been developed for mass housing im-
plementations (Dincer 2014). Cellular automata (CA)
are a set of cells which has emerged over time by the
rules, which is related to local neighborhoods, on a
particular grid system (Wolfram 2002). It is usually
evaluated in the studies like urban design, zoning
and building masses in which social effects can be
simulated with neighborhood relations (Singh and
Gu 2012).
The basic features of using Cellular Automata in
architectural design can be listed as context-sensitive
and bottom-up processes, self-organization, graph-
ical representation, etc. Due to approach of "form
follows function" and its local relationships, CA may
generate many different, coherent and unexpected
solutions in a short time. Although CA have negative
features like autonomy and inability to manage in-
creasing neighborhoods which depends on the size
of context, nevertheless it can be considered to be
adaptable to some stages of design (site and space
planning, facade orientations etc. ) by specific exter-
nal interventions and feedbacks in smaller scales and
to be useful.
In this model, the digital tool is aimed at serving
as a decision-support model that is integrated with
traditional design processes. Users can't be directly
involved in design process of floor plan schemes,
but their tendencies determine which housing type
will be designed. CA is evaluated to manage com-
plex relationships of housing blocks, spatial units
and facade elements with various design criteria and
present many alternatives.
A HOUSING DESIGN MODEL WITH USING
CELLULAR AUTOMATA
The developed model is based on interaction of
Schön's "Reflection in Action" approach, mass cus-
tomization and computational design processes.
With "Reflection in Action" approach it simulates
stages of mass housing designs which consists of
site plan placements, space planning for housing
blocks and façade designs just as in traditional pro-
cesses. Also it provides user/designer to be in-
volved in decision-making processes of each stage.
Clearly, the process of the model works in the form
of "Framing-Evaluating-Reframing" in these stages as
Schön (1985) defines. It evaluates user preferences,
which effect on creation of housing types, in the
space planning stage. Since the preferences are usu-
ally changeable for each space planning, each of the
solutions may differ from the others and it is possible
to get variety of solution which is quite difficult in tra-
ditional designs. Therefore it can be seen as a mass
customization model, Cellular automata which are
the computational component of the model man-
age relationships of housing blocks in a site plan,
space units in space planning and façade orienta-
tions by using neighborhoods and architectural and
urban constraints. The model uses both top-down
and bottom-up processes.
The model uses Conway's "Game of Life"
(Adamatsky 2010) and Wolfram's 2D CA experiences
as samples of CA. The rules which organize neighbor-
hood relations of both housing blocks in site plan-
ning and spatial units in space planning are similar
to that of "Game of life" which includes the concepts
like loneliness, overcrowded, stasis and reproduc-
tion. In site planning, the rules control proximity
states of housing units with each other by legisla-
tive limitations or "shadow cone" (figure 1). In space
planning, the rules are concerned with architectural
topological relations among spatial units (figure 2).
Furthermore in façade designs, whether spatial units
have a cantilever or setback is defined by rules like
Wolfram's 2D CA rules (figure 3). The rules operate
the state list of having a cantilever or setback which
Towards Smarter Cities - Volume 1 - eCAADe 32 | 203

is determined by its downstairs and lateral neigh-
borhoods for a spatial unit. During the generation
process creating, removing and continuity of hous-
ing blocks and spatial units or façade orientations of
spatial units are determined by these rules.
All these rules have a feature of being altered
or being supported by new rules according to de-
signer's decisions. The changes can be made on the
algorithm instead of on interface directly.
The model is mainly configured on space plan-
ning in spite of having three stages. The other two
stages are like a support module. The generated
housing blocks determine boundaries of space plan-
ning in the stage of site planning as different alterna-
tives of housing block placements are generated in
this stage. On the other hand, it is aimed at increas-
ing orientations in third dimension and differentiat-
ing form of the generated housing types in the stage
of façade designs.
The process at the site plan stage starts with the
location of a housing block which has a defined floor
area and whose height can be changeable, within
boundaries of a site. These settlements continue un-
til reaching the value of the total construction area. In
the generation, the neighborhood relations of blocks
with the previous ones are evaluated after the first
settlement. If there is no suitable situation in terms
of the criteria, a selection between newly produced
and existing building blocks can be made and one of
204 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 4
The algorithm of
spatial planning
stage
them can be removed and new building blocks are
located on the left empty spaces of the land.
Space planning stage includes solution pro-
cesses of floor plans for each of the generated hous-
ing blocks in a chosen site plan. At this stage, be-
cause of that plan layouts are same in the blocks and
have 24*24 m square form, the process is carried out
similarly in all blocks. At the starting of the process,
a chosen floor plan is made ready for generation by
dividing into smaller core which is 8*8 m and 4*4 m
in size for creating building grids and space units.
Then sharing area on the floor plan is made for poten-
tial housing types which are determined by means of
user preferences (figure 4).
Spatial units locate on the shared areas in the
framework of their neighborhood relations and they
try to define a housing type together. Housing types
exemplified as A (1+1), B (2+1), C (3+1) and D (4+1)-
double flat models which are determined by num-
bers of their bedrooms and area constraints. They
exist on a floor plan when their minimum area con-
straints are met. Otherwise their area values are
shared among the others.
On a floor plan, generation process starts with
the random placements of "Entrance Hall" units in
the shared housing type regions and these units trig-
ger creation of other spatial units, which are kitchen,
bathroom, bedroom and living room. Similarly each
of created spatial units affects generation, transfor-
 mation or deletion of further units. This process con-
tinues until reaching the criteria of selected housing
types and filling in the blanks of the floor plan (fig-
ure 5). During the process, the spatial units which
can't meet the criteria of selected housing types are
deleted. The empty units which are existed by re-
moving the cells are re-evaluated with new spatial
units in the generation of other or new housing types
and the process is completed and repeated with dif-
ferent user preferences for other floor plans of a cho-
sen block.
A few samples of neighborhood rules of spatial
units, which are used in at this stage, can be given as
Figure 5
Placements and
alternatives of
spatial units and
housing types on a
floor plan
follow:
• "Entrance hall" units which are generated at
starting are placed on the floor plan in case
they have a neighborhood with other spatial
units except "entrance hall" units. Otherwise,
these units are evaluated as "death" and re-
moved.
• "Kitchen" units are usually placed opposite to
"entrance hall" units. In some cases, for an ex-
ample, these units can be created near to right
or left of "entrance Hall" units for A(1+1) hous-
ing types.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 205
 • "Bathroom/WC" units are defined as a single-
cell for same housing types. If they are
more than one, the operation of removing or
changing the function is applied for them.
• If a "bedroom" unit is neighbor with service
core, a "void" unit or different "bedroom" unit
is created opposite of it according to its state
of neighborhood with service core.
• Similarly, if a "living room" unit is neighbor
with service core, a different "living room" unit
is created opposite of it according to its state
of neighborhood with service core.
• If a "living room" unit has neighborhoods with
more than two "living room" cells, they are
evaluated as "death" and defined as a "ter-
race" space.
The algorithm of space planning is given in figure 6.
The stage of facade designs starts after solution
of all floor plans for a housing block is completed. At
this stage, the process progresses as bottom-up from
ground floor plan to upper floor plans and controls
spatial units on the facade are evaluated individually
according to specific procedure. In this procedure, at
first, spatial units are grouped according to its func-
tional properties and these groups are ranked as liv-
ing room, bedroom and kitchen spaces for priority of
generation. Then each spatial unit in the groups is
chosen randomly and their states of facade orienta-
tions are controlled. The controls are done by a set
of basic facade rules and additional rules. In the set
of rules, lateral and downstairs neighbors of a spa-
tial unit are examined with regard to their states of
having a cantilever or not. These states are repre-
sented with colors like "black" and "white" or num-
bers like"1" and "0". Then the sequence of probabil-
ities which these representations form by together
are saved. Additional rules operate specific states of
some spatial units and housing types due to their ar-
chitectural requirements. For example, if a housing
type has more than one "living room" units on the
206 | eCAADe 32 - Towards Smarter Cities - Volume 1
same facade, only one of them can have a cantilever,
even though they are appropriate for a set of basic
facade rules. The generation is made simultaneously
on all facades of a block.
In the set of rules the option of randomness is
partially applied. Besides the rules are organized in
terms of solutions of preventing architectural detail
errors, but these rules can be also developed by dif-
ferent architectural ideas . Finally, as explained be-
fore, facade orientations support the diversity by dif-
ferentiating the created housing types dimensionally
(figure 7).
At the end of the processes for each of three
stages of the model, the results are evaluated and
interpreted by designer and user. Depending on
the results and evaluations operations are repeated.
Furthermore the model generates primitive plan
schemes which can be developed and detailed by de-
signers. Also the structural frame which is chosen for
the model allows flexible solutions and changes on
these created schemes. Thus efficiencies of design-
ers and users are protected.
THE INTERFACE OF THE MODEL
In the model, for the convenience of designers, a sim-
plified data entry interface is designed with 3Ds Max
software. On the interface, there are two rollouts
which are called "Data of Zoning Status" and "Floor
Plan Arrangements".
Starting rollout is the implementation of "Data of
Zoning Status" which allows the generation of hous-
ing blocks in a site plan. Firstly, on the rollout site di-
mensions are defined (figure 8). It is performed by the
way of entering data for a sample of site which has
rectangular or square geometries or selecting a sam-
ple of site which has non-uniform geometries. Also
if it is necessary, definition of landscape direction is
made. Then the heights (min/max values) of housing
blocks and ratio for total construction area according
to data of the chosen region are determined. If there
is a slope in site or custom building heights for spe-
cific region of the site are necessary, slope and height
values are defined for these regions. Finally a model
Figure 6
The algorithm of
spatial planning
stage

Towards Smarter Cities - Volume 1 - eCAADe 32 | 207

 Figure 7
Implementation
stages of facade
generations

208 | eCAADe 32 - Towards Smarter Cities - Volume 1

Figure 8
The interface and
the implementation
of the Site Planning
stage
(shadow cone or legislative limitations) of relation-
ships among housing blocks is chosen for their place-
ments. Optionally, definition of a social area can also
be made by entering its dimensions in the site (figure
8).
After operation of data entry is completed, the
generation process is started. The results are demon-
strated on the main screen of 3Ds Max software. If re-
sults are acceptable then they are saved and second
implementation is passed for space planning of all
generated blocks, else the process is repeated by user
/ designer until desirable generations are obtained.
Another rollout is the implementation of "Floor
Plan Arrangements" which allows the generation of
space planning and façade generations for housing
blocks. In this stage selection of a housing block
which is generated in the site planning is made and
creation process of floor plans on the selected hous-
ing block is started by user/designer. Then, any one
of the created plans is chosen by using "forward" and
"back" buttons on the interface. There is no ranking
for selection among the floors and the generation
processes can be started from any floor. After that,
values of the preferences which potential inhabitants
made for one of A(1+1), B(2+1) and C(3+1) housing
types on the selected floor plan are entered. Also, a
prioritization and starting direction among the hous-
ing types is determined for starting placement and
sharing area on the floor plan. Finally all data is saved
and the generation process is started. If the results
are acceptable, this procedure is similarly repeated
for solutions of the other floor plans according to dif-
ferent user preferences (figure 9).
Another implementation on the rollout is façade
designs which concerned with creating cantilevers or
setbacks on the façades of the selected block. Af-
ter solutions of all the floor plans are completed, the
module of façade designs is activated. After one of
"simple" and "orientation" modes is chosen the gen-
eration is started and results are evaluated.
For space planning, the model is supported by
additional modules like labeling spatial units, detail-
ing the core, demonstrating "housing type" mode or
"spatial units" mode and showing or hiding all floor
appearances. All these modules are considered to
make the model easier and more understandable.
Towards Smarter Cities - Volume 1 - eCAADe 32 | 209
 Figure 9
The interface and
the implementation
of the Space
Planning stage

Figure 10
Practical
experiences in the
chosen site in
Karabuk.

210 | eCAADe 32 - Towards Smarter Cities - Volume 1

RESULTS Singh, V and Gu, N 2012, 'Towards an Integrated Gener-
In this study; mass housing design model, which ative Design Framework', Design Studies, 33(2), pp.
provides integration of mass customization, tradi- 185-207
Wolfram, S 2002, A New Kind of Science, Wolfram Media,
tional and computational design processes, is de-
New York
veloped. This model is thought to be useful as [1] http://www.di.net/articles/the-design-implications-
a design-decision tool which generates alternative of-mass-customization/
housing plans by using the rules, which is deter-
mined and changed by designer, according to inhab-
itants' tendencies. The applicability of the model has
been tested successfully in a site which is located in
Karabuk in Turkey by different user scenarios and lo-
cal legislative constraints. As a result, many useful al-
ternatives have been obtained (figure 10). Further-
more, the stage of space planning in the model is op-
erated in a specific (square) floor plan framework, but
the rules and relationships in this stage can be eval-
uated for other plan schemes in different forms ( e.g.
horizontally expandable forms ). In addition, the con-
tent of the model has an extensive feature by increas-
ing the rules and relationships among space units in
the model. It is considered that future works can be
done in this direction. Finally CA rules have made
quite useful contributions about generation of rich
alternatives in spite of the excess of the rules. Both
adapting CA rules to architectural context instead of
classical approaches and using criteria of random-
ness have been effective on this diversity of alter-
natives. The graphical representation feature of CA
has been helpful and useful in accelerating decision-
making processes and interpretations.

REFERENCES
Adamatsky, A (eds) 2010, Game of Life Cellular Automata,
Springer London Dordrecht Heidelberg, New York
Dinçer, AE 2014, A Digital Model for Mass Housing Design
Using Cellular Automata Approach, Ph.D. Thesis, ITU
Niemeijer, R, De Vries, B and Beetz, J 2010 'Designing
with Constraints', 10th International Conference on
Design & Decision Support Systems in Architecture and
Urban Planning
Pine, BJ 1993, Mass-Customization: The Frontier in Busi-
ness Competititon, Harvard Business School, Boston-
Massachusetts
Schön, D 1985, The Design Studio, RIBA Building Industry
Trust, London

Towards Smarter Cities - Volume 1 - eCAADe 32 | 211

212 | eCAADe 32 - Towards Smarter Cities - Volume 1
Interrogating the Relation between E-Mobility Recharging
Network Design and Drivers' Charging Behaviour
Eiman Y. ElBanhawy1 , Ruth Dalton2 , Emine Mine Thompson3
1,2,3
Northumbria University
1,2,3
{eiman.elbanhawy|ruth.dalton|emine.thompson}@northumbria.ac.uk

In this paper we report on research into patterns of electric vehicle (EV)

commuters’ movement and behavior in a road network. The design of the
charging network is a function of its size and the distribution of the charging
points within a given urban area. It consists of several spatial design qualities,
configuration attributes, travel demand, and users charging patterns. In order to
have reliable recharging facilities (RFs), we need to understand the nature of the
system and eventually plan and design for the current and potential EV
commuters. The recharging experience should not be a worrying matter for EV
drivers. Assessing existing systems helps in picking up the main paradigms of EV
population and system that need to be considered or even adjusted for a better EV
market penetration. This study introduces the spatial configuration of an active
e-mobility system through a case study. The paper investigates the correlation
between the design characteristics of EV recharging infrastructure and its
usability for the given metropolitan area. The usability is a consequential
communication corresponds to the system design. We need to explore the
variations in individual charging behavior within the EV population to
understand the movement patterns in the network. Using data of over 500 EV
drivers charging their cars using public charging infrastructure over a three-year
time, we clustered the EV population based on the charging patterns. Design
configuration analysis is conducted using DepthMap; charging patterns are
captured by the infrastructure service provider while been sorted, tabulated and
analyzed using SPSS. The study outcomes should give a clear insight of how the
use of RFs is affected by the spatial; design features as well as the charging
patterns and profiles of EV real users.

Keywords: Electric Vehicles, Charging Behaviour, Recharging Facilities, Space

Syntax

Towards Smarter Cities - Volume 1 - eCAADe 32 | 213

 INTRODUCTION
Electric vehicles are needed to replace the internal
combustion engines (ICT) for private cars and small
vans. There is a growing momentum of owning a
plug-in electric vehicle (PEVs) as it is promising to cut
oil use and CO2 emissions. To seize this opportunity
and promote the concept of smart and green means
of transport, the users charging behavior and viable
feedback as well as public awareness need to have
more attention. A major impediment to EV main-
stream market is the absence of proactive and con-
structive planning approach for RFs. This is consid-
ered as a major issue especially for full electric car
passengers due to the sole dependency on batteries
as a source of power hence range limits and longer
recharging time (Namdeo et al. 2013).
Embracing the system's psychological dynamics is
important for developing theoretical frameworks
and hypothetical models in order to understand its
nature. Due to the differences in technical and eco-
nomic premises, EV system is not to be treated as
a conventional car network system (Lindblad 2012).
EV system as the vehicle itself, infrastructure net-
work, technicalities, regulations and standards is a
large and complex integrated tapestry that involves
multidisciplinary knowledge. EV market stakehold-
ers should partnership to address the hurdles to con-
sumer sales and also jointly work on ensuring a highly
satisfactory ownership and driving experience (Ac-
centure 2011).
To fully embrace the EV system and realize its real
magnitude, we need to divide and conquer. We need
to investigate the dimensions that constitute en-
ablers and disenables of EV adoption and how poli-
cies can encourage them (Ozaki and Sevastyanova,
2011). The end users' recognition of the system and
perception of its advantages and pitfalls are critical.
This includes both the potential users and the early
adopters who will be referred to as EV advocates in
this context. Studying the charging pattern of differ-
ent users and correlate it with the design parameters
and attributes can show us interesting results.
Sizing and placing RFs problem is such a new topic
214 | eCAADe 32 - Towards Smarter Cities - Volume 1
that recently has been taking some attention. Few
important strides have been made to tackle this
problem particularly in late 2012 and 2013 (Chen
et al. 2013). Several documentations and re-
ports have been published and released containing
phased plans, initiatives and long-term development
of recharging infrastructure; however, these reports
do not share how the presented size and location
of RFs were determined (Wirges and Linder 2012).
Slim literature releases the previous studies which
were conducted to solve the charging stations place-
ment (Lam et al. 2014). This results in the lack of a
strategic approach in the design of publically avail-
able recharging network which could otherwise in-
crease the market penetration level. By improving
the recharging experience, the probability of poten-
tial users to shift to EV market and households loy-
alty to use the EV as their first choice would definitely
soar.
Study Aim and Objectives
Spatial configuration analysis is conducted to ulti-
mately help developing design tools for planning au-
thorities and policy makers. The purpose of the study
is to explore the users' profiles, test the public aware-
ness of RFs, and investigate the relationship between
the design configuration and the user behavior in the
road network. The study introduces the spatial con-
figuration of an active e-mobility system through the
case study NE1, NE4 and NE8, the inner urban core of
Newcastle-Gateshead metropolitan area in the North
of England, see figure 1. It analyses the emergent be-
havior of the crowd and relating and correlating the
design characteristics of the network to the usability
of the charging points and the design configuration
(space syntax measures). This includes a description
of administration procedure of data collection, data
cleaning and description of the data set. It ends with
the use of space syntax as a vehicular movement pre-
diction in the hybrid model.

Figure 1
Newcastle Postal
Districts
Figure 2
Mapping the
Recharging
Network
Research Questions
How and to what extent do the spatial configura-
tion and design parameters of the metropolitan ur-
ban core impact the charging behavior of EV drivers?
Can these factors be identified and quantified?
The paper is structured to start with an introduc-
tion of the available data-set and the approach be-
ing used to sort and group it. For correlation anal-
ysis purpose, all the variables are presented and ex-
plained throughout the article; RFs design qualities,
emergent charging behavior, public awareness and
urban economics attributes. This is followed by the
charging behavior profiles of the 500 EV users. Linear
regression model is developed followed by interpre-
tations.
Newcastle NE1, NE4 and NE8. Tyne and Wear
Metropolitan County contains five metropolitan bor-
oughs: Gateshead, City of Newcastle upon Tyne,
North Tyneside, South Tyneside, and City of Sunder-
land. The case study considers the first two bor-
oughs. In specific, the study area periphery is the in-
ner urban core of their county boroughs which en-
gages three postal districts NE1 (6 KM2), NE4 (14 KM2)
and NE8 (16 kM2), see figure 1. All demographic
and socioeconomic data and information about EV
drivers needed is provided by the city council is pro-
vided by the service operator, CYC Company. An axial
map has been generated; a total number of active 38
charging points have been mapped, see figure 2.
RECHARGING FACILITIES DESIGN QUALI-
TIES
In last decades, space syntax has been known as an
alternative computational language that is used to
spatially pattern the modern cites and analyze the
topological relationships of settlement spaces (Hillier
and Hanson 1984; Hillier 1994). Different shapes of
relations and levels of interactions between spaces
each other and or with society, have superbly shaped
the space syntax notion and its principles in ana-
lyzing the spatial patterns of cities (Jiang and Clara-
munt 2002). Within a given built environment (sys-
tem) and with its axial representation of roads net-
work, the virtue of space syntax theories and tech-
niques can be used to quantify the properties of the
space arrangement and measure its level of integra-
tion (see figure 3) accessibility and connectivity and
depth (Hillier and Hanson 1984). Attraction theory of
vehicular movement is identified as the movement
that is seen as being to and from built forms with
differing degrees of attraction. In this way, the spa-
tial configuration of the urban grid would be basi-
cally the spatial elements through which commuters
Towards Smarter Cities - Volume 1 - eCAADe 32 | 215
 move; streets, squares, alleys and bridges are linked
together to form some kind of global pattern. Move-
ment, configuration and attraction are the three main
pillars of a common-sensed movement within the
built environment. The movement can be justified
by the presence of attraction as well as the configu-
ration properties of the space (Hillier et al. 1993). In
this research, another attribute is added which might
change the equation, the movement of electric ve-
hicles movement and the associated charging pat-
terns. Figure 3 is the integration and connectivity val-
ues represented over the axial map of Newcastle in-
ner urban core.
Measures of Urban Street Patterns
Streets including curvilinear ones are represented as
chains of simple straight-line segments. Street inter-
sections are defined as end points of line segments.
Once there is a choice of paths, nodes occur repre-
senting an intersection or end of road. Roads can be
split into lines at each change in direction and or in-
tersection. This means, a road segment may consist
of many line segments (Peponis et al. 2008). Pepo-
nis et al. (2008) classified the types of movement in
urban street networks into four types: routine move-
ment which denotes the commuter going from fa-
216 | eCAADe 32 - Towards Smarter Cities - Volume 1
miliar origin to familiar destination. Second type is
the new path-making to particular, which describes
originating from home visiting a new place. The third
movement is wandering within a relatively known
area and fourth movement is exploratory as it aims at
understanding an entirely new environment. Given
this format, configuration modeling technique has
proven successful in predicting pedestrians and ve-
hicular movement in simulation modeling (Peponis
et al. 2008). In this research, the problem we are try-
ing to solve is different as the vehicular movement
is associated with other aspects related to EV system.
These aspects are related to behavioral layer of the EV
population, technical issues, and driving EV history.
Figure 3
Space Syntax Measures Integration map of
The above axial representation gives the opportunity Newcastle inner
to measure a particular property of the urban envi- core
ronment; connectivity indexes, control value and in-
tegration are some of these morphological proper-
ties. Connectivity is the measure of how well an axial
line is intersected by others. In principle, there is no
non-intersected line in any urban environment, i.e.
each space is accessible from every other space in the
city. In the mean time, experience tells us that the
length of the axial line has some correlation to con-
nectivity indexes, that is, these are more possibilities
for lengthy lines to be intersected by others. Integra-
tion of a line is by definition a value which indicates
the degree to which a line is more integrated, or seg-
regated, from a system as a whole. The measure is
actually based on a more basic notion called depth.
Depth is more generally a topological distance in a
graph. If two lines are directly connected, then the
distance between them is equal to one, and the dis-
tance of a pair of lines which are not directly con-
nected is the shortest path between them. Integra-
tion is a global measure, as the calculation of inte-
gration is based on the total depth from the current.
However, if a number of depths, instead of all depth,
is considered, then the integration is called local in-
tegration. Thus, global integration is a global mea-
sure describing the relation of each space to the sys-
Figure 4
Datasets and the
selected one for the
study
tem as a whole, while connectivity and local integra-
tion are local measures describing the relationship of
each space to its neighbors.
Statistical Analysis
To carry out the spatial statistical analysis, some
checks have to take place to ensure consistency and
level of detailing. The availability and the quality
of the date, identify the possible analysis that can
be applied and influence the final conclusion that
can be reached. To have meaningful inferences, the
worthwhile objectives of the analysis have to be de-
termined (Coe et al. 1994). Data collection process
is a time consuming and sometimes costly, we need
to make sure that we are benefiting and producing
meaningful observations and have rich data that will
enable objectives (Marshall 2012).
Various statistical analyses can be carried out to
for a given inquiry while using different datasets.
These sets may contain the same information in
forms of different subsets. Values, attributes or vari-
ables regardless their types as continues, categori-
cal, build these sub sets. Hence, sorting, filtering,
and needed level of abstraction, design and form the
dataset which all depends on the desired objectives.
Referring to the research problem, the data sets we
are using contain several subsets. These subtests
were formed to map all the available data, Figure 4
and decide which data set is being used and which
statistical analysis shall be conducted. The analysis
cannot satisfy all the objectives at once and even all
the time. A cluster analysis may take place to exam-
ine the some features of the charging patterns using
a set of data. Whereas, analysis of variance shall be
carried out to investigate the correlation between the
design qualities and the usability of the system using
different data set.
The first check is the number we need to identify
the employed approach. If we are looking with an op-
Towards Smarter Cities - Volume 1 - eCAADe 32 | 217
 timistic approach, we will hope for the best. To inves-
tigate the correlation between the design properties
and the usability of the system, we need to consider
some factors. There are two approaches to analyze
the data: (1) optimistic or (2) Pessimistic. Optimistic
approach is to select a single post in each site.
CHARGING BEHAVIOUR PROFILES
There are five charging behavior types which are ba-
sically formed in respect to the time spent charging.,
technical capabilities of the RF, level of awareness
and confidence level, willingness to spend time in
charging publically and at work and charging pattern
of the real users. Interviewing some of the real users
also gave some indicators and assisted in forming the
following equation which is a function of time, Equa-
tion (1):
 cars in the mornings and in the afternoon. It is

 "T heT opU p" t ≤ 20m

 observed that this charging pattern tends to form

 20m < t ≤ 40m
"GoodEnough"
c(t) = "Superb" 40m < t ≤ 1h



 "F reeT ime" 1h < t ≤ 3h


 clearly is not convenient for users to be spent on On-
"BeyondCharging" else
(1)
The first cluster is “The Top Up”. This cluster con-
tains those drivers who are willing to spend up till
20 minutes in charging their cars over the daily road
trip. This can be down to 10 minutes which is fair
enough to replenish their batteries and let them go
home safely. Having said we are scanning the inhab-
itants and also across-districts commuters, homes are
within 20 miles away as a maximum from the city cen-
tre; however, the majority is within 12 miles. 91%
of this cluster charges their cars for 10 minutes only;
which reflects the reliance on domestic and work-
218 | eCAADe 32 - Towards Smarter Cities - Volume 1
place charging. Almost half of the 10-minute users
charge at the evening time which means after work
and probably on their way home or non-home based
trips. The other half charges their cars with a fairly
equal distribution over the morning and the after-
noon. In a nutshell, this charging pattern tends to
form 20% of the total charging events taking place
over the years of study (2010-2013). This reflects a
low level of awareness that a 10 minute charge can
cover your back home trip. Table 1, explains the cat-
egories of the two behavioral features of the system.
The second cluster is “The Lucky Charge”. This
cluster contains those drivers who are willing to
spend up to 40 minutes charging their cars through-
out their daily road trips. These records were scored
throughout the three years. This means that this clus-
ter barely contributes to the overall charging events.
The fans of this cluster are those who charge their
10% of the total charging events happening over
the years. This reflects another aspect of the behav-
ioral layer of the system. Charging for 40 minutes
Street RFs; however, if it is possible, this charging pat-
tern shall expand the driving comfort zone of the EV
driver.
The third cluster is “The Good Enough”. This clus-
ter contains those drivers who are willing to spend
up to almost an hour charging their cars through-
out their daily road trips. The analysis showed that
this cluster forms another 10% of the total charg-
ing events over the years of study. This means that
this cluster also doesn’t represent a majority of the
users’ charging patterns. The fans of this cluster are
those who charge their cars in the mornings and af-
ternoons, mornings more so than afternoons.
The fourth cluster is “The Superb”. This cluster con-
Table 1
Charging Behaviour
 tains those drivers who are willing to fully charge tential users about their cognitive skills of remember-
their batteries using RFs. This means that they are ing or noticing any of the RF in the area of study. The
so technically oriented and think wisely with respect respondents' feedback was collected, tabulated and
to electricity. This group is also in luck because this analyzed. Each charging point got a score that sums
means they tend to have access to workplace charg- up the number of respondents who are familiar with
ers which brings easy to their charging process. They it and have come across. The RFs which got to hit the
save a lot as they probably don’t mainly rely on do- highest scores of awareness (highlighted in red), Fig-
mestic charging, they plan ahead so that they can ure 5, are the ones which are highly used by current
have full charge which will guarantee a safe daily trip, users. Values of awareness are added as attributes.
and also they will ultimately have the longest life time
Figure 5 for their batteries. The occurrence of this pattern is
Level of Awareness relatively high to other patterns; this pattern tends to
form 50% of the charging events which mainly rep-
resents the workplace charging points. However, in
case of having 3KW charging point, this will only re-
plenish a 50 percent of the battery capacity. Yet, still
is considered as a high level of dependency and re-
liance on public charging services. The mapping re-
flects that the majority of the charging events of this
cluster are more often during the day.
The fifth cluster is “Beyond Charging”. This clus-
ter contains those drivers who are using 3KW charg-
ers or and have the luxury of fully charging their bat-
teries using it. The charging events of this cluster
don’t have a pattern. The occurrence of this pattern
Figure 6 fluctuates; it mainly forms that last 10% of the charg-
Main corridors and ing events. Users tend to charge also in mornings and
urban cores afternoons while less likely in the evenings.

PUBLIC AWARENESS AND URBAN ECO-

NOMICS ATTRIBUTES
EV passenger acceptance depends on being knowl-
edgeable about EV specifications, capabilities, the
benefit of smart transport and how these cars can
fit in their daily life. It also depends on the raised
awareness about the available charging network and
how accessible is the ON street and OFF street charg-
ing service. The attitudes of car drivers towards elec-
tric vehicles were investigated within a sample visual
questionnaire carried out. The aim of the question- Two more independent variables were considered;
naire was to test the awareness about the available the level of closeness from the urban cores and the
charging points in Newcastle-Gateshead inner core traffic count (travel demand) of the nearest main cor-
area. The questionnaire was designed to ask 45 po- ridor feeding the RF site.

Towards Smarter Cities - Volume 1 - eCAADe 32 | 219


From the urban policies, spatial planning reports and
demographic records generated by the Government,
the urban core areas are mapped, blue numbered cir-
cles, see figure 7.
The main corridors are indicated in figure 6. All urban
cores and main corridors are plotted as centers and
axes respectively in figure 7 alongside the charging
points in small circles. Distances to the nearest urban
centers are scaled and calculated, see figure 8.
The Unit of UTMC, the real-time traffic and travel
information service unit in Newcastle upon Tyne
calculates the traffic counts and real time origin-
destination matrix of the main corridors.
Factors Affecting Utilities Profitability
To have a profitable business, we need to look into
the enablers and disenables that influence the mar-
ket. The profitability is a function of several related
220 | eCAADe 32 - Towards Smarter Cities - Volume 1
Figure 7
Distances to main
roads and cores
and unrelated factors; these factors are associated
behavioral, technical and socioeconomic variables of
the EV system. For an example, understanding the ra-
tionale behind the willingness to pay and get a new
car should be taking high priority in automotive and
batteries technologies considerations. The end user's
feedback and preference play as significant a role as
the automotive and batteries technology in affecting
the EV market growth. A better understanding of the
social aspect of the system, a clear depiction of the
changing behavior of the users and ultimately iden-
tify the most influential behavior that affect the us-
ability of the system. The following table summa-
rizes the 12 independent variables (IV) including the
dependent variable (DV), the total energy consumed
(Λ).
Figure 8
Metric distance to
nearest urban
centroid
SPATIAL STATISTICAL ANALYSIS
Linear regression is conducted. Dependent variable
is the total energy used. The method used is EN-
TER. This basically means all the variables are entered
on the same time whereas other methods like step-
wise, remove, backward or forward involved some
sort of wise step regression. The Independent Vari-
ables (Λ). Highly correlated variables should be ex-
cluded as well as a high number of IV in general. Af-
ter conducting several trials trying to mix and match
all the variables and see which set up gives better
results, the following composition was selected, see
table 5. To run the model, the variables are to be
classified into behavioral and spatial elements. Hy-
brid model is developed into two stages 1a and 1b;
different variables are analyzed over 38 levels (ob-
servations) in each stage. In the first stage, the spa-
tial parameters (variables) only are entered. As a first
check, the regression output tables show some val-
ues that specify the model validity, quality and per-
centage of random error. The model validity can be
checked by the significance value (0.049). This means
we can carry on working with the model. The sec-
ond check is the Adjusted R square value. This value
indicates the percentage that the model justifies of
the total variance of the DI. Table 3 indicates that the
five spatial factors explains 17% of the usability vari-
ance in a linear correlation, see table 3 model (1). This
value is quite low as this means there is over 85%
random error if we will use this model to plan for fu-
ture; however, this is not the final model statistics.
In the second stage, the behavior parameters are en-
tered merged with the spatial parameters previously
added in the first stage. The Adjusted R square is im-
proved to 0.871, table 3 and the model significance is
better, 0.000, see table 4.
DISCUSSION
In the first part of the paper we discussed the differ-
ent design, behavioral and urban related parameters
of EV system that we are investigating in this study.
Through Linear regression approach, we developed a
hybrid model that coupled the independent behav-
ioral and spatial variables to explore the relation to
the total energy consumed by the charging points.
Commenting on the regression model output, there
are some meaningful interpretations and important
outcomes. The coupling process of the variables is
fundamental for model validity and reliability. De-
coupling behavior and spatial qualities of a system
would lead to biased analysis. The model is devel-
oped to assist in planning future EV charging net-
work based on real experience and empirical analy-
sis. The depiction of real world population and sys-
tem has to denote and stimulate at a high level of
correlation. In our case, having a system used by hu-
man beings, depicting the behavior layer should not
be overlooked. Spatial urban structure of EV network
does not work independently of charging patterns
and behavior of the EV population. Designing for
EV charging points is similar to petrol stations loca-
tion problem, both can not be dealt as facility or lo-
cation allocation problem. Location allocation prob-
lem can be purely based on spatial and metric cal-
culation tabulated in optimization model, which is
not the case. Effects and intersections of all factors
should be taking into account. To validate this, the
random error value has significantly dropped when
Towards Smarter Cities - Volume 1 - eCAADe 32 | 221
 Table 2
Factors affecting EV
System

Table 3
Hybrid Model
Summary

Table 4
Hybrid Model
ANOVA

222 | eCAADe 32 - Towards Smarter Cities - Volume 1

the behavioral parameters were added in the sec-
ond stage. The model explains 87% of the RFs' us-
ability variance. This high value reflects that the se-
lection of variables was successful and well studied.
In return, this generates a linear equation where the
DV is a function of the influential factors that con-
tribute to a utilized and profitable charging network,
Equation 1. Out of the eleven factors, there are five
most influential variables affect the EV system usabil-
ity. The equation is a simple linear regression where
y = a + (b1x1 + b2x2 + b3x3 + b4x4 + b5x5):
∞
∑ ing pattern indicator (Dependent Variable), the total
f (Λ) = − 864.049 + (
n=1
− .094 Public Awareness (P )
− .120 TrafficCount (T ) (2)
− .220 On/off Street (O)
+ .170 DistanceFromCentres (ι)
+ .159 Integration (I))
CONCLUSION
EV system is a complex system. It is a function
of behavioral, socio-technical and spatiotemporal
paradigms. The lack of a strategic approach in the
RFs design process affects the market penetration
level. By improving the recharging experience, the
probability of potential users to shift to EV market
and households loyalty to use the EV as their first
choice would be definitely raised. A question was
raised about the correlation between design param-
eters of charging network and the charging behav-
ior. In other words, would the influential factors that
affect the EV system usability can be identified and
quantified? The present study addressed this ques-
tion by studying technical, socioeconomic, spatial
and behavioral aspects of the system.
A multi method approach has been employed
to carry out this study. Real users' records spanning
three years of operation were analyzed. The analysis
showed the charging profiles of the users and how
these profiles differ with respect to time of charge
and time spent charging features. The design con-
figuration analysis was applied to measure the level
of accessibility and integration of the roads contain-
ing charging points. 38 charging point were mapped;
Integration values were calculated and tabulated.
Level of awareness was also considered as variables
affecting the public charging network. Through a
questionnaire, the level of awareness was quantified.
Another urban and travel demand attributes were
calculated. The charging patterns of all EV drivers us-
ing the NE1, NE4 and NE8 postal districts' RFs in 2012
were analyzed. After several trials finding the charg-
energy consumed in each charging point over the pe-
riod of study was selected. Linear regression analysis
was conducted to investigate the relation between
the DI and the design configuration values, level of
awareness, and urban and vehicular attributes, Inde-
pendent Variables (IV).
Statistical model developed reveal that mea-
sures of design configuration and spatial qualities,
behavior layer, demographic and travel demand at-
tributes are most strongly associated with charging
events of EVs when planning for future EV integrated
charging network. The study outcomes show that
there is correlation between the EV system design
parameters and its usability and profitability. It also
measured and quantified all factors and identified
the most influential ones. Off street charging sites
increase the odds of public charging when it is lo-
cated in workplaces. On Street with high integra-
tion values, close to main corridors, distant from ur-
ban cores are tend to have high frequency of short
charges over the weekdays. Importantly, the results
presented here also underscore the significance of
the spatial structure of road networks. Integration
plays as significant a role as metric accessibility in af-
fecting the proportion of riders walking for transit.
The spatial structure of street network does not
work independently of urban and behavioral layers
of the given EV system. Based on the standardized
coefficients estimated in regression model, On/off
Street (O) Distance from centers (ι) and Integration
Towards Smarter Cities - Volume 1 - eCAADe 32 | 223
 (I) have comparably high positive impacts on charg-
ing network business profit. Traffic count (T ) and
public awareness (P ) come on the second influential
set of factors that that contributes to profit.
Overall, the analyses presented here confirm the
hypothesis that design configuration values relate to
charging preference. Location-related values explain
17% of the variance of total energy consumed of the
RFs of the network while adding behavior-related val-
ues pulls it up to 78%. The study shows that these
measures: ON street and OFF street location, pub-
lic Awareness, distance from urban centroids, traffic
counts of the nearest corridor (s), and integration val-
ues are likely to capture only 17% of the EV recharg-
ing facilities occupation and usability which is quite a
reasonable ratio for a pilot study. Whereas, the charg-
ing pattern-related attributes: contribute to 84% and
coupling the two categories explains 87% of the sys-
tem usability. Usability is used in this context to in-
dicate the efficiency of the charging points and indi-
cates the system business model. Hence, selection
the total energy consumed is a clear straight indica-
tor of profit in a directly proportional relation. The
better we understand the urban parameters, charg-
ing behavior and EV technical qualities, the higher
chances we are able to positively influence the EV
market penetration. Raised awareness of some RFs
is affects the usability; whereas, the spatial analysis
of the RFs affects the level of awareness. In other
words, designing for EV system is such a complex
prolonged study that requires high level of coordi-
nation between stakeholders, analytical skills, pub-
lic engagement, and interdisciplinary coordination
to embrace the emergent structural properties of the
EV system. Future work should allow for more vari-
ables and more observations. This study is analyzing
year 2012 and specific charging points.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the emobil-
ity NSR-North Sea Region Electric Mobility Partner-
ship e-Mobility NSR (NSR, 2011) project for funding
the PhD Research. Earlier versions of this paper can
224 | eCAADe 32 - Towards Smarter Cities - Volume 1
be found at https://www.linkedin.com/pub/eiman-
el-banhawy/11/316/36a
REFERENCES
Accenture, A 2011, 'Pursuing Electric: Growth Strate-
gies for Automotive OEMs', Accenture, North Amer-
ica, 11(doi: 11-1294/02-2804), p. 10
Chen, T. D, Kockleman, K. M and Khan, M 2013 'The Elec-
tric Vehicle Charging Station Location Problem: A
Parking-Based Assignmnet Method for Seattle.', The
92nd Annual Meeting of the Transportation Research
Board in Washington DC., Washington DC.
Coe, R, Stern, R.D and Allan, E 1993, Objectives and Steps
in Data Analysis, Ilri
Hillier, B 1993, Space is the Machine, University of Cam-
bridge
Hillier, B, Penn, A, Hanson, G and Xu, J 1993, 'Natural
Movement; or, configuration and attraction in urban
pedistrian movement.', Environment and Planning B,
(20), , p. 29–66
Hillier, B, Hanson, G 1984, Social Logic of Space, University
of Cambridge, p. 275
Jiang, B, Claramunt, C 2002, 'Extending Space Syntax to-
wards an Alternative Model of Space within GIS', Pro-
ceedings of the 3rd European Agile Conference on Ge-
ographic Information Science 2000
Lam, A. Y. S, Leung, Y. W and Chu, X 2014, 'Electric Vehi-
cle Charging Station Placement: Formulation, Com-
plexity, and Solutions', arXiv, 2(1310.6925), pp. 1-10
Lindblad, L 2012, Deployment Methods For Electric Vehicle
Infrastructure, Master's Thesis, Uppsala University
Marshall, J 2012, Data Collection and Analysis, Ph.D. The-
sis, Warwick University
Namdeo, A. Y. S, Tiwary, A and Dziurla, R 2013,
'Spatial planning of public charging points
using multi-dimensional analysis of early
adopters of electric vehicles for a city region',
Technological Forecasting and Social Change,
32(doi:10.1016/j.techfore.2013.08.032), p. 10
Ozaki, R and Sevastyanova, K 2011, 'Going Hybrid: An
Analysis of Consumer Purchase Motivations', Energy
Policy, 39, pp. 2217-2227
Peponis, J, Bafna, S and Zhang, Z 2008, 'The Connectivity
of Streets: Reach and Directional Distance', Environ-
ment and Planning B, 35, pp. 881-901
Wirges, J and Linder, K 2012, 'Modelling the Develop-
ment of the Regional Charging Infrastructure for
Electric Vehicles in Time and Space', EJTIR, 4, pp. 391-
416
Design Tool 1
A Model for Sustainable Site Layout Design with Pareto
Genetic Algorithm: SSPM
Yazgı Aksoy1 , Gülen Cağdaş2
1,2
Istanbul Technical University
1
yazbadem@hotmail.com 2 cagdas@itu.edu.tr

In architectural design, computer aided design tools have an important impact on

design process, but still early design stage and sustainable design are problematic
issues. During sustainable architectural design process, the designer needs to
comply with some regulations, which requires calculations and comparisons.
Green building certification systems are developed to assist designers during this
complicated process, but for an efficient sustainable design for different regions,
environmental information and local building codes must be considered with
green building certification system criteria. In this paper, LEED and BREEAM
certification systems are going to be considered as being the most representative
building environment assessment schemes that are in use. As there are conflicting
criteria's according to LEED and BREAM sustainable site parameters, local
building codes and environmental conditions; an efficient decision support system
can be developed by using multi-objective genetic algorithm. This paper presents
an effective site-use multi-objective optimization model that use pareto genetic
algorithm to determine the most efficient sustainable site layout design for social
housing, which could assist designers in the early stage of design process.

Keywords: Sustainable Site Layout Design, Multi Objective Genetic Algorithm,

LEED-BREEAM

INTRODUCTION to the fact that, architectural design process can be

In design industry, as a result of multidisciplinary re-
searches, evolutionary computation is being used at
least for 10-15 years (Bentley, 1999). In architectural
design, advanced computer aided design tools have
an important impact on design process, but still early
design stage and sustainable design are problem-
atic issues. Sustainable building design refers to a
process that begins with selecting the site and opti-
mizing economical and environmental performance
throughout a building's life cycle. This situation leads
regarded as an open process, even though sustain-
able design process needs to comply with some regu-
lations which requires calculations and comparisons.
As Rivard (2006) declared that in order to achieve
a successful sustainable building, particular atten-
tion needs to be paid to the conceptual design stage
when the most important decisions are taken, never-
theless it is the stage with least computer support. As
a result of this, the need for a decision support system
in early stage of sustainable design is clear so that for

Design Tool 1 - Volume 1 - eCAADe 32 | 227

 designers some simulation tools and green building
certification systems are developed to evaluate their
designs according to sustainable building criteria.
There are huge numbers of simulation tools used
for sustainable design; building energy simulation
programs, shell analysis, cost calculation programs,
indoor air quality analysis programs and finally devel-
oped computational fluid dynamics programs to cal-
culate the energy and gas emissions, life cycle analy-
sis (LCA) can be given as examples. Simulation tools
assist the designers during sustainable design pro-
cess. The decision taken in the early stages of de-
sign have the greatest impact on final form, but most
of these simulation tools do not support concep-
tual designs appropriately. In addition, these sim-
ulation tools typically cannot communicate among
themselves and have time-consuming data inputs
and complex interfaces (Rivard,2006).
In addition, green building rating systems and
certification programs play an efficient role in usage
of green buildings. These systems are working ac-
cording to the criteria defined on the basis of scoring.
In this paper, LEED and BREEAM certification systems
are going to be considered, as being the most repre-
sentative building environment assessment schemes
that are in use. Although these certification systems
are used all over the world, the parameters are pre-
pared according to America's and Britain's geograph-
ical, economical and cultural conditions. Other coun-
tries are experiencing difficulties during design pro-
cess so that for an efficient sustainable design for dif-
ferent regions, environmental information and local
building codes must be considered with green build-
ing certification system requirements.
As it consists of several possible conflicting ob-
jectives that need to be considered together, sustain-
able building design process becomes more compli-
cated than traditional building design. At that point,
the reason of preferring multi- objective pareto ge-
netic algorithm in the model, which we will focus
on this paper, will be better understood. The multi-
objective optimization problems consist of the simul-
taneous optimization of several possibly conflicting
228 | eCAADe 32 - Design Tool 1 - Volume 1
objectives and result in a set of non-dominated solu-
tions. In sustainable design, maximum energy con-
servation and utilization of natural light can be given
as an example of two conflicting objectives, so during
sustainable design process pareto genetic algorithm
will be successful to generate design alternatives ac-
cording to conflicting criteria.
To summarize, today some tools developed to
assist designers during sustainable building design
process, address more detailed design stages when
important design decisions already have been taken
in conceptual stage. A successful green building de-
sign can be performed by the creation of alternative
designs generated according to all the sustainability
parameters in conceptual design stage. As there are
conflicting criteria's according to LEED and BREAM
sustainable site parameters, local building codes and
local environmental conditions, an efficient decision
support system can be developed by the help of
pareto genetic algorithm.
LITERATURE REVIEW
Simulation Models Used For Sustainable Ar-
chitectural Design
Many simulation models have been developed to as-
sist designers in green building design. GBTool [5] is a
simulation model that evaluates buildings according
to resource consumption and energy performance
by using a rating system. ATHENA [6] model provides
a convenient platform for simulating processes used
in semiconductor industry, but it is limited with con-
struction materials and installation. All simulation
models try to explore effective ways to assist designer
in complicated design process, but especially during
the earlier stages of sustainable design, there are sev-
eral limitations in practice. As indicated by Harput-
lugil (2010) previous studies about sustainable build-
ing design have some problems during design pro-
cess. The program inputs are very detailed, scientif-
ically there are wide ranges of inputs and the data
are not yet available in the early stages of design.
The outputs of the models sometimes become diffi-
cult to understand and interpret. In addition, most of
the detailed energy simulation models are research-
based and so it requires a long time to learn how to
use them. Sustainable building design criteria needs
to be evaluated from the sketch phase, because the
most important decisions of sustainability are taken
from the settlement of land. For these reasons, it is
clear that there is a need for better decision support
system to support the earlier stages of sustainable
design.
Evolutionary Models Used For Sustainable
Architectural Design
Evolution is a method of searching enormous num-
ber of possibilities to find the solution and in 1950s
and the 1960s it became an optimization tool for en-
gineering problems. The aim is to evolve a popula-
tion of solutions using natural genetic variation and
natural selection operators and genetic algorithms
are the most prominent example (Mitchell,1996).
Genetic algorithm is a population-based search
technique inspired from the biological principles of
natural selection and genetic recombination. It oper-
ates on a population of solutions that are randomly
generated for the first generation (Goldberg, 1989).
Multi-objective problems have multiple objectives,
which leads to the need to obtain a set of optimal
solutions, known as effective solutions. Genetic al-
gorithms is a suitable tool for multi-objective op-
timization problems because it can locate multiple
Pareto optimal solutions in a single simulation run.
Pareto optimality describes a situation in which the
profit of one condition cannot be increased without
reducing the profit of another. In attribute space,
the set of non-dominated solutions lie on a surface
known as the Pareto optimal frontier. The goal of a
Pareto GA is to find a representative sampling of so-
lutions all along the Pareto front (Horn, Nafpliotis and
Goldberg, 1994). A reasonable solution to a multi-
objective problem is to investigate a set of solutions,
each of which satisfies the objectives at an accept-
able level without being dominated by any other so-
lution (Deb, 2001).
In green building design many efforts have been
made, especially integrated simulation environment
provided by tools to evaluate design alternatives
according to different performance criteria. Since
designers rarely consider only one criterion in the
decision-making process, multi objective optimiza-
tion models have been proposed.
Wang, Zmeureanu and Rivard (2005) optimized
the building envelope using multi-objective genetic
algorithm. Variables in the model include the pa-
rameters that are usually determined at the concep-
tual design stage and have critical impact on build-
ing performance. Life cycle analysis methodology
is employed to evaluate design alternatives for both
economical and environmental criteria. A multi-
objective genetic algorithm is employed to find op-
timal solutions. They concentrate on building en-
velope because of its importance in environmental
and economical performance of buildings. The multi-
objective optimization model they developed can
be used to locate the optimum or near optimum
green building designs for given conditions (Wang,
Zmeureanu and Rivard,2005).
In sustainable design, land use is another sub-
ject designers take care of. Zelinska, Church and
Jankowski (2008) present a new multi objective spa-
tial optimization model-SMOLA which minimizes the
conflicting objectives of open space development,
infill and redevelopment, land use neighbourhood
compatibility and cost distance to already urbanized
areas. They examine the applicability of spatial op-
timization as a generative modelling technique for
sustainable land-use allocation. The model uses 400
raster cells, and they generate multiple solutions with
importance of objectives, at the end they evaluate
the patterns produced by the model with multi agent
geo-simulation. (Zelinska, Church, Jankowski, 2008).
These developed models are chosen for using
evolutionary algorithms for solving different scaled
design problems. The plan optimization model for
green building design developed by Wang, Rivard
and Zimeureanu(2006) is functional for considering
material and cost information with building form,
but it is insufficient for early stages of design. Also
Design Tool 1 - Volume 1 - eCAADe 32 | 229
 it is mostly concentrated on energy conservation,
site planning for building complexes is disregarded.
SMOLA model developed by Zielinska, Church and
Jankowski generates land use patterns according to
the building functions. For city scale the model is use-
ful and has visual readability, but it is not convenient
for small-scale problems. The SSPM model presented
in this paper is an integrated model that concentrates
on not only energy efficiency but also wide sustain-
able design criteria; such as green building certifica-
tion systems and local building codes for.
SUSTAINABLE SITE PLANNING MODEL
(SSPM) WITH PARETO GENETIC ALGO-
RITHM
SSPM model is at the intersection of two different dis-
ciplines which are evolutionary algorithms and sus-
tainable architectural design. The SSPM (Sustainable
Site Planning Model) will generate site-planning al-
ternatives for social housing on selected site, accord-
ing to LEED and BREEAM certification systems sus-
tainable site usage criteria, local building codes and
local climate conditions which are accepted as sus-
tainable design objectives. Multi-objective genetic
algorithm which is also known as pareto optimization
will use sustainable design objectives as fitness func-
tions to generate site planning alternatives.
Site Definition Method
The first step of the model is definition of the site to
the computer. Matrix definition technique in Excel is
used to define the site with numbers. The defined
digital site is used as a base to generate site planning
alternatives on by SSPM. Excel program is chosen for
providing quick and easy data input for matrix defi-
nition
The rows and columns of the matrix represent 1
unit =1 m2 of the site. The x, y values of the site cor-
ner points and the centre or corner points of existing
elements are used to define the matrix. Each of the
Excel cell is considered as a point in a coordinate sys-
tem. R1C1 reference style in Excel is used to represent
x and y axes to make data input apparent. The rectan-
230 | eCAADe 32 - Design Tool 1 - Volume 1
gular area which uses the values of maximum x and
y coordinates of site corner points as its dimensions,
is the boundary of the matrix. The cells inside the de-
fined boundary are collared according to their func-
tions so that visual presentation is provided. Each cell
has functions shown below:
0. Empty cells outside the site boundary: They
have no functions only used to define the
boundaries of the site to the model. They are
represented with white colour.
1. Empty cells inside the site boundary: They
can transform to different functions( housing,
green, pedestrian road, vehicle road, car park,
etc.) They are represented with beige colour.
2. Existing reserved tree cells: Their positions are
fixed, cannot be changed. Two neighbour
rows placed around these cells are accepted
as green cells and their positions are fixed too.
They are represented with dark green colour.
3. Housing unit cells: They don't appear in
the excel matrix, they are generated by the
model according to fitness functions. Hous-
ing blocks consist of 1m3 cubes, which are
generated according to floor space ratio (FSR),
floor area ratio (FAR) and maximum building
height value, overlap housing cells. They are
represented with red colour.
4. Vehicle road cells: They are generated by the
model according to fitness functions. They are
represented with grey colour.
5. Car park cells: They are generated by the
model according to fitness functions. They are
represented with dark grey colour.
6. Pedestrian road cells: They are generated
by the model according to fitness functions.
They are represented with brown colour.
7. Existing main vehicle road cells: They are out-
side the site boundary. They are represented
with black colour.
 8. Reserved water cells: Their positions are fixed
cannot be changed. Two neighbour rows
placed around these cells are accepted as
green cells and their positions are fixed too.
They are represented with blue colour.
9. Reserved green cells: Their positions are fixed
cannot be changed. They are represented
with light green colour.
10. Used polluted area cells: They are priority ar-
eas for generating housing cells on. They are
represented with yellow colour.
11. Existing public transportation station cell:
They are represented with the letter of "T".
12. Existing noise origin cells: They are repre-
sented with the letter of "N".
13. Underground cells because of the elevation
difference: They are represented with slightly
transparent grey.
In Figure 1, an example of a piece of land matrix de-
fined in Excel is illustrated.
Figure 1
Piece of land matrix
defined in Excel
Figure 2
Piece of terraced
land matrix defined
in Excel
The selected site must be digitized closed to the
real terrain data without disregarding the topogra-
phy and the natural formations, so that the site is
presented in 3D grid with its slope data. User will
divide the site into different zones according to its
slope data. As a result of this, terracing sites which
have elevation differences will be possible and envi-
ronmentally sensitive solutions will be able to gener-
ated. The SSPM model will generated social housing
cells according to this terraced site-zones.In Figure 2,
an example of a piece of terraced land matrix defined
in Excel is shown.
As it is shown in Figure 2, if an empty cell (pre-
sented with "1") are 5 meters higher than the ac-
cepted zero-level elevation, the cell is presented with
the numbers of 1-5. If it is 10 meters higher than the
zero-level elevation, it is presented with the numbers
of 1-10. In addition, if a reserved tree cell (presented
with "2") is 3 meters higher than the zero-level eleva-
tion, it is presented with 2-5. In this way, the visual
readability of each cell's elevation is provided.
User Interface
After definition of the site with an Excel matrix by
user, the model will use this data to visualize the se-
lected site. At this stage SSPM model needs more in-
formation about the site. User interface is used to
input essential data about the selected site by user.
Processing 2.1 programming language is used to de-
velop SSPM user interface and evaluating software,
for having advanced visual environment and the abil-
ity of working with Windows.
User interface shown in Figure 3 has six values
which will be determined by user:
• Climate type (temperate, warm, cold)
• Direction of prevailing wind
• Direction of view
• Floor space ratio (FSR)
• Floor area ratio (FAR)
• Maximum building height value
Design Tool 1 - Volume 1 - eCAADe 32 | 231
 • Open space ratio
Each cell in Excel matrix is used as a gene of
genetic algorithm. After defining the site with ex-
isting elements and the user chooses climate type,
the direction of sun, wind and view the model will
start producing site layout alternatives according
to the sustainable design parameters. This step is
where the genetic algorithm takes place. The fitness
functions of genetic algorithm are the sustainable
architectural design parameters mentioned below.
Pareto genetic algorithm evaluates each alternative
according to priorities and ranks the each alternative
to find the best sustainable site alternative. SSPM
generates site-planning alternatives on defined cells
that include buildings, green areas, roads, car parks,
pedestrian roads, bicycle roads according to fitness
functions and crossover and mutation operators are
applied to the population to find the pareto optimal
site planning solution.
232 | eCAADe 32 - Design Tool 1 - Volume 1
Evaluation System of the Model: Fitness
Functions
The conflicting parameters will be evaluated accord-
ing to their importance values by the model to find Figure 3
the optimal site planning with the help of pareto ge- User interface
netic algorithm. The sustainable design parameters
that are going to be used in the model coming from
LEED and BREAM sustainable site parameters, local
regulations and local climate conditions can be clas-
sified in four objective functions:
• Site usage: The placement of the building and
its surroundings are determined at this stage.
It is a need to focus on the existing ecological
values on the site. Maximizing open area and
reducing heat islands have priority.
• Building Orientation: According to climate,
wind and view direction housing units that
are placed.
• Building Form: According to climate and local
regulation formulas (FAR-floor area ratio, FSR-
floor space ratio and maximum height) build-
ings get form.
• Accessibility: Model will use neighbourhood
relations and formulas coming from rules to
generate site layout alternatives.
The conflicting parameters will be evaluated accord-
ing to their importance values by the model to find
the optimal sustainable site planning with the help
of pareto genetic algorithm.
Concept 1: Site Usage. The placement of the build-
ing and its surroundings are determined at this stage.
It is a need to focus on the existing ecological values
on the site. The fitness functions for sustainable site
usage are listed below:
• Rule 1.1: Conserve existing natural areas and
restore damaged areas to provide habitat and
promote biodiversity so all existing elements
which are needed to be protected on the site
such as trees, lakes must be defined to the
model. [1]
• Rule 1.2. Design the project to conserve 100%
of all water bodies, wetlands, land within
30.00 meters of water bodies, and land within
15.00 meters of wetlands on the site.[1]
• Rule 1.3. For the reserved trees provide mini-
mum 25 m² green area around, which means
2 rows of neighbour cells must be green cells.
• Rule 1.4. The percentage of the open space
must be more than 25% of the value of open
space given in the land use regulation. [1]
• Rule 1.5. If there is zoning but no open space
requirement, provide open space equal to
20% of the site. [1]
• Rule 1.6. If there is no regulation, the open
space area must be equal to building foot-
print. [1]
• Rule 1.7. If it is impossible to provide the value
of open space ratio , roofs must be green roofs
and they can be added to the value of open
space area.
• Rule 1.8. Reduce heat islands to minimize im-
pact on microclimate and human and wildlife
habitat. [1]
• Rule 1.9. Polluted areas on the site are prefer-
ential to be used as a building placement area.
[2]
• Rule 1.10. If there is an on old building on the
site, 50% of outside area or 20% of the site
area must be used as open space. [1]
• Rule 1.11. Parking areas and car transport in-
terchanges should be far away from the areas
to be protected. [1]
• Rule 1.12. If there is a noise source around,
buildings must be placed 800 meters away
from the source. [2]
Concept 2. Building Orientation. In this context de-
tached building type is preferred so that detached
building regulations are considered and floor height
is accepted as 3.00 meters. For the model maximum
building height is limited up to 60.00 meters.
• Rule 2.1. For detached building minimum dis-
tance of the front and side garden which coin-
cided with the edge of the road must be min-
imum 5.00 meters. Buildings up to 5 storey,
side garden distance is minimum 4mt and
over 5 storey for each floor 0.5 meters is added
to side garden distance.(Istanbul Zoning Reg-
ulations, 2007) [4]
Side garden dist = 4 + [((h/3) − 5)x0.5] (1)
• Rule 2.2. For detached buildings up to 5
storey, backyard distance is minimum 5.00
meters. Over 5 storey, for each floor 1.00 me-
ter is added to backyard distance (Istanbul
Zoning Regulations, 2007). [4]
Backyard dist = 5 + [((h/3) − 5)x1] (2)
• Rule 2.3. Orientation according to the view
has the priority.
• Rule 2.4. For cold climate, wind direction must
be direct to short edge of the building and for
hot climate wind direction must be direct to
the long edge of the building (TS825, 1999,
Rules of Thermal Insulation in Buildings, Turk-
ish Standards Institute). [3]
• Rule 2.5. For hot and dry climate, valley bot-
toms that are influenced by flows of cold air
must be chosen to place. For warm climate
valley ridges and for humid climates valley
slopes are suitable (TS825, 1999, Rules of Ther-
mal Insulation in Buildings, Turkish Standards
Institute). [3]
Design Tool 1 - Volume 1 - eCAADe 32 | 233

Concept 3: Building Form.
• Rule 3.1. The maximum building height value
defined by the user must be considered (Istan-
bul Zoning Regulations, 2007). [4]
• Rule 3.2.For detached buildings, building
depth must be lower than 30.00 meters and
for each building ground floor area cannot be
higher than 900 m² (Istanbul Zoning Regula-
tions, 2007). [4]
• Rule 3.3. Building envelope width should be
minimum 6.00 meters from the front of the
building and should not exceed 30.00 meters
(Istanbul Zoning Regulations, 2007) [4], which
means for housing cells minimum 6 units of
cells should be side by side with a maximum
of 30 cells unit.
• Rule 3.4. FAR (floor area ratio) and FSR (floor
space ratio) values must be considered.
Total number of housing cells
F AR = (3)
Total number of site cells
F AR≤40 (4)
Total number of Housing Cellsx(h/3)
F SR >
Total number of Site cells
(5)
234 | eCAADe 32 - Design Tool 1 - Volume 1
Figure 4
Neighbourhood
relations of the
model
• Rule 3.5. Building form has a huge impact
on building energy performance. Compact
forms must be chosen in cold climate to re-
duce the heat losses. (TS825, 1999, Rules of
Thermal Insulation in Buildings, Turkish Stan-
dards Institute ) [3]
Concept 4: Accessibility.
• Rule 4.1. Public transportation should be
the minimum distance to the entrance of the
buildings. [2]
• Rule 4.2. Maximum 1 car park place for 1 res-
idential must be placed, for cars unit area for
car park is 20.00 m².[1]
• Rule 3.3. Model will use neighborhood rela-
tions shown in Figure 4 to generate site layout
alternatives.
According to the fitness functions of genetic al-
gorithm mentioned above, evolutionary model will
generate site layout alternatives on defined cells that
also include green areas, roads, car parks and pedes-
trian roads.
Algorithm.
1. Defining site and selecting climate type, view
and wind direction by user;
2. Create and visualize the site by using Excel
matrix data prepared by user before, with reg-
ular grid of d dimensions (d = 1mt);
Figure 5 3. Initialization: Randomly generate an initial
Flowchart of the population;
model
4. Evaluation: Calculate the values of the n ob-
jectives for each solution in the current popu-
lation. Then update the tentative set of non-
dominated solutions;

5. Selection: Repeat the following procedures to

select ( Npop - Nelite ) pairs of parent solu-
tions; give points to population according to
fitness functions and show the solutions that
have maximum fitness value points. The user
selects pairs of parent solutions.

6. Termination Test: If the user cannot select and

wants to change, the input data return to Step
1, if selected go Step 5;

7. Crossover and Mutation: Apply a crossover

operator to each of the selected ( Npop -
Nelite) pairs of parent solutions. A new so-
lution is generated from each pair of parent
solutions. Then apply a mutation operator to
the generated new solutions;

8. Elitist Strategy: Randomly select Nelite so-

lutions from the set of non-dominated solu-
tions;

9. Show pareto optimal solution;

10. Termination Test: If the user is not contended

and wants to change the input data return to
Step 1; if stopping condition is satisfied, end
the algorithm;

Testing the SSPM Model

In Figure 5 flowchart of the SSPM model is illustrated.
The SSPM model was tested on a site in Kağıthane
in İstanbul. The reason for choosing Kağıthane was
twofold. First, urban regeneration in the residential
areas of Kağıthane has recently been a central issue of

Design Tool 1 - Volume 1 - eCAADe 32 | 235

 consideration in Turkey. Second, the selected study
area is a realistic example that has reserved water
supplies, green areas and polluted areas to test sus-
tainable social housing units. In Figure 6 selected
area is shown with existing elements. Polluted area,
shown with yellow in Figure 6, is preferential to be
used as a building placement area. Reserved trees,
green areas and water body will be protected and
buildings will land with 30 meters of water body.
The input data coming from user interface are:
• Climate type : Temperate
• Direction of prevailing wind: East
• Direction of view : West
• Floor space ratio (FSR) : 0.25
• Floor area ratio (FAR): 1.2
• Open space ratio: %20
• Maximum building height value: 45.50 me-
ters
236 | eCAADe 32 - Design Tool 1 - Volume 1
In Figure 7 two different generations which have dif-
ferent floor area ratio values are shown. High floor
area ratio provides more building area. The polluted
area shown in Figure 6 has the priority for the place-
ment of the building blocks so that model starts to
generate blocks from the polluted area first.
SSPM model generates housing unit alternatives
according to fitness functions. First generations ac-
cording to selected area information coming from
the user interface are shown in Figure 8. Existing wa-
ter body is protected and neighbour cells in 30 me- Figure 6
ters are accepted as green area, which means place- Selected Site in
ment of buildings are not allowed. Number of gen- Kağıthane
erated buildings depends on FSR, FAR and maximum
height value. Maximum building height value is not
an obligation, it is used to limit the height of the
building. Using floor space ratio limit has the priority
which means low-rise buildings are preferred. At this
run, model produces first population according to
half of the fitness functions mentioned before ( Rule
1.1, Rule 1.2, Rule 2.1, Rule 2.2., Rule 3.1, Rule 3.2, Rule
3.3., Rule 3.4). For future works, direction of view and
direction of prevailing wind are going to be added to
the fitness functions and also pedestrian roads, ve-
hicle roads and car parks are going to be generated
according to all the fitness functions mentioned be-
fore.User will choose each of the parents from gen-
erated alternatives and model will do crossover and
mutation to find the pareto optimal solution for sus-
tainable site planning.
CONCLUSION
This study is in the intersection of two different disci-
plines that are evolutionary algorithms and sustain-
able design. In this paper, it is proposed to com-
bine cellular structures with a multi-objective genetic
algorithm for using its search ability to find Pareto-
optimal sustainable site planning solutions for social
housing complexes.
The green building plan optimization model de-
veloped by Wang, Rivard and Zimeureanu (2006) is
only focused on building envelope in early design
Figure 7
Comparison
between different
FAR values
Figure 8
First generation of
the site planning
layout
stage and as sustainable design parameters , only en-
ergy conservation, maximum daylight and minimum
cost are considered. Furthermore, the SMOLA model
developed by Zelinska, Church and Jankowski(2007)
only generates land use patterns and it addresses the
problem with optimization of such sustainability ob-
jectives like new development, redevelopment, land
use compatibility, and accessibility. This model is
only focused on sustainable large-scale land-use in
2D space. The SSPM model presented in this paper
is used for building scale and the layout of the site is
detailed with roads, lakes, trees and buildings. The
slope of the site and the buildings are presented in
3D space. SSPM is tested on an existing site which is
in Kağıthane in Istanbul to find pareto optimal sus-
tainable site planning alternative for a social hous-
ing complex. First population according to limited fit-
ness functions are generated. In future studies,as fit-
ness functions of the model are widened, generated
alternatives will be more functional and effective for
detailed sustainable site planning including pedes-
trian roads, car parks, vehicle roads, bicycle roads.
This approach would introduce an effective compu-
tational design tool for early design stage of sustain-
Design Tool 1 - Volume 1 - eCAADe 32 | 237
 able design, which does not currently achieved by
current technologies.

REFERENCES
Bentley, PJ 1999, 'Aspects of Evolutionary Design by
Computer', in Roy, R, Furuhashi, T and Chawdrhy,
PK (eds) 1999, Advances in Soft Computing, Springer,
London, pp. 99-118
Deb, K 2001, Multi-objective Optimization Using Evolu-
tionary Algorithms, John Wiley & Sons
Goldberg, DE 1989, Genetic Algorithms in Search, Opti-
mization and Machine Learning, Addison-Wesley
Harputlugil, GU 2010, 'Analysis and Simulation on En-
ergy Performance Based Design', Journal of Megaron,
6(1), pp. 1-12
Horn, J, Nafpliotis, N and Goldberg, DE 1994 'A Niched
Pareto Genetic Algorithm for Multiobjective Opti-
mization', Proceedings of the First IEEE, pp. 82-87
Mitchell, M 1996, An introduction to genetic algorithms,
The MIT Press
Rivard, H 2006, 'Computer Assistance for Sustainable
Building Design', in Smith, IFC (eds) 2006, Intelligent
Computing in Engineering and Architecture, Springer-
Verlag, Berlin, pp. 559-575
Wang, W, Zmeureanu, R and Rivard, H 2005, 'Applying
Multi-objective Genetic Algorithms in Green Build-
ing Design Optimization', Building and Environment,
40(11), p. 1512–1525
Zelinska, AL, Church, R and Jankowski, P 2008, 'Sustain-
able Urban Land Use Allocation with Spatial Op-
timization', Geographical Information Science, 22(6),
pp. 601-622
[1] http://www.usgbc.org/
[2] http://www.breeam.org/
[3] http://www.vitrafix.com/assets/pdf/TS825.pdf
[4] http://www.ibb.gov.tr/tr-TR/kurumsal/Birimler/
ImarMd/Documents/imar_yonetmelik.pdf
[5] http://www.iisbe.org/down/gbc2005/GBC2k2/
GBC2k2_Manual_A4.pdf
[6] http://www.athenasmi.org/news-item/
guide-to-lca-credits-in-green-building-programs/

238 | eCAADe 32 - Design Tool 1 - Volume 1

The generation of possible space layouts
Jesper Thøger Christensen
Research Assistant, Aalborg University, Denmark
jtc@civil.aau.dk

The general scope of this paper is to present the development of an approach for
the generation of possible space layouts in the early stages of design. The
approach sets out to bridge the gap between the free form brainstorming of the
sketching designer and the deterministic parametric model of the computational
designer. A new responsive design tool is developed; applying physical based
modelling techniques to a computational environment. Springs and particles
automates the initial placement and sizing of the spaces, and allows continuous
user and software interaction. Keeping the design process alive, through springs
that connects everything to everything and allows topology to change, opposed to
the strict hierarchy and constraints of traditional parametric design. Thus instead
of the topology being determined by initial constraints, the layout will be
generated by probabilities of spaces to connect. Letting the storyline of potential
users give answers to possible space layouts.

Keywords: Space layout, Digital aids to design creativity, Design tool

development, Spatial quality, Storyline

MOTIVATION - SPACE AS A GRADIENT through space.

The goal is not to generate an optimized geometric
layout, but a range of (m)any possible topological lay-
outs - a narrative landscape - to challenge the archi-
tect's imagination and inspire to investigate the pos-
sibilities further.
Compared to other methods of spatial layout,
the spaces represented in this method will not be lim-
ited to represent functional requirements (kitchen,
living room, bathroom, bedroom etc.) but also spa-
tial qualities in general (light, dark, wide, deep, long,
lifted, heavy, open, closed, private, public etc.)
This rejection of a pure functional division of
space originates from my idea of space as a gradient.
Here functional requirements are not the only driver
of space, but also spatial qualities and the movement
The field of view of the individual and the history
of his movement will determine his next movement.
This continuous revelation of space, Gordon Cullen
named serial vision, and splits the optical viewpoint
into two elements: the existing view and the emerg-
ing view. In separating the existing from the emerg-
ing view, Cullen captures the feelings related with the
experience of the current and the anticipation of the
emerging (Cullen, 1961). Recognizing the role of se-
rial vision one can use it to build anticipation, cre-
ate drama, and invoke feeling. Therefore, space is a
discontinuous landscape without clear boundaries,
where the storylines of the individuals acting in the
space, is just as important as the actual space itself.
For simplicity, this initial paper will only gener-

Design Tool 1 - Volume 1 - eCAADe 32 | 239


ate spatial layouts as a landscape of storylines and
spaces based on either functional requirements or
spatial qualities.
The question at hand in this initial paper is topol-
ogy being more important than the geometry. Thus,
generating solutions for spatial layouts, which can in-
spire the topology of how spaces connect, without
necessarily giving final answers to their geometry.
(figure 1)
THE TOOL
The digital tool is not supposed to replace pen
and paper in the early design process, but give the
designer another resistance, than the unconscious
mind acting through the hand and pen onto a piece
of paper. The resistance being the definition of a set
of parameters, as input to the generation of a story-
line, and the tool's subsequent interpretation of this.
The intention of the tool is to render the spatial
layout as a storyline of spaces, where only the prob-
ability of two spaces to follow each other are prede-
fined. Thus before using the tool, one do not need
240 | eCAADe 32 - Design Tool 1 - Volume 1
Figure 1
The same topology
of spaces can take
on various
geometries.
a conception of the resulting organization, but only
the elements and their probability of connecting. Re-
sulting in a spatial layout that allows perception, in-
terpretation and behavior. Not following, but allow-
ing it to happen.
For these spatial relations to have an effect on
the layout, the tool is set up as a physical system of
particles connected to every other particle with var-
ious types of spring forces. The particles represents
spaces and the spring forces various spatial relations,
one of them being the storyline, another the possibil-
ity of similar spaces to merge and create loops in the
storyline.
How the tool works
The tool is written in Java in and organized in an ob-
ject oriented way.
It extends the open source programming envi-
ronment of Processing, to make use of its render-
ing mechanism and basic classes and methods. Con-
trolP5, provides a GUI (Graphical User Interface), used
to generate the user interface of the tool. NooLab, is
the library that provides access to generation of the
Figure 2
A simple example
of the input for the
generation of a
storyline.
Figure 3
A simple example
of the output for
the generation of a
storyline.
storyline and assigning additional geometric proper-
ties to each state of the storyline.
Input for the tool
The input for the generation of a storyline (figure 2):
• A set of named spaces. Ex. narrow, dark etc.
• A range of return values that specify addi-
tional properties of that space. Ex. a range for
the width of a narrow space.
• A transition matrix of probabilities for the var-
ious spaces to be proceeded by any of the
other spaces in the creation of a storyline. Ex.
from a narrow space (A) there is 10% possibil-
ity to proceed into a low space (C), but from
a low space there is 0% possibility to proceed
into a narrow space.
In terms of programming the storyline is conceived
as a Markov Chain. A Markov Chain is a mathemati-
cal system that undergoes transitions from one state
to another, between a number of possible states. It is
a memoryless probabilistic process, where the next
state depends only on the current states possibility
to connect to another state, and not the sequence of
events that preceded it [1]. These probabilities are set
by the designer and defined in the transition matrix.
The tool generates a storyline
The tool will start out with one storyline, with the pos-
sibility to add an infinite number of storylines.
A random identifying letter defines the first
space of each storyline (figure 3). The transition ma-
trix of the Markov chain generates every proceeding
space. However, before generating the next space in
the storyline, the current space translates to a parti-
cle with a certain size and some additional properties.
Further, it connects to all other particles in the system
through various springs.
With a sufficient long storyline, the output will
represent a string of some of the various configura-
tions that can emerge from the predefined matrix.
The spaces at the two ends of the storyline will only
have one connecting room, thus acting like a dead
end, with only one possible option to proceed move-
ment. In the other spaces with two connections,
movement is predictable, and to proceed one can
only go forward or backward.
The influence of the physical spring system
For a space not to become static and movement pre-
dictable, some of the spaces needs at least three
connections to other spaces. One being the previ-
ously occupied space and the other two, the ones to
choose between for the next step. Doing so, by al-
lowing similar qualities within the same storyline or
across storylines to merge (figure 4).
Separating spaces that are conflicting and con-
Design Tool 1 - Volume 1 - eCAADe 32 | 241
 Figure 4
Not just one
sequence of spaces,
but several
sequences
overlapping. The
letters represent
various spaces.

Figure 5
Springs used to
connect particles

Figure 6
Springs used to
align particles

242 | eCAADe 32 - Design Tool 1 - Volume 1

 necting spaces that are compatible and reinforcing,
ensures that spaces within and across storylines can
interact and cross each other. Practically this is pos-
sible through a set of different springs between the
spaces.
Springs used to connect particles. MarkovSprings
are used to construct a spring between to particles
(space) being created one after the other to make up
the storyline. The length of the springs is set for the
two particles to exactly touch each other (figure 5).
MinDistanceSprings are used between spaces
with different qualities, which are not directly con-
nected with a MarkovSpring, and therefore should be
kept separate with a spring of a minimum distance
(figure 5).
WithinDistanceSprings are created between
spaces that have the same quality, but are not di-
rectly connected with a MarkovSpring. This allows
the particles to merge into one, when they are within
a certain threshold (figure 5).
Springs used to align particles. ChildParticles
around the border of the parent Particle is controlled
by the SoftBody class, which adds a polygon of Child-
Particles to each Particle. The number of polygon
vertices and ChildParticles per side is dependent on
the additional geometric properties of the quality
(figure 6).
Figure 7
A deterministic
model
ColllisionSprings are used to align the SoftBody
polygon of ChildParticles to neighbouring SoftBody
polygons (figure 6).
User interface
The user can interact with the physical system by
dragging spaces around or selecting spaces to prune.
Further the user can turn the display of various
springs and particles on or off.
At any point in the process, the markov chain can
be paused, allowing the user to capture an image of
a promising layout, prune parts of the layout or inter-
act with it, without new spaces popping up.
INVESTIGATING THE OUTPUT OF THE
TOOL
Discussing the effect on spatial layout by changing
the transition matrix.
A deterministic model
One can completely control the output from the tool,
when setting up the transition matrix so that each
space can only connect to the next space in the list
(figure 7).
This transition matrix will always produce the
same spatial sequence and movement will quickly
become predictable. Not making use of the tools
Design Tool 1 - Volume 1 - eCAADe 32 | 243

computational power. Furthermore, circular rings
appear when similar spaces starts to merge.
A corridor emerges
As the transition table is configured and the resulting
layout shows, quality A is dominant, as it links to ev-
ery other quality (figure 8). Further, as the sequence
A-A has a probability of 0.7, A will be followed by A
quite often. This could lead to the perception that A
symbolises a corridor space, whereas B, C and D could
be considered as adjoining rooms.
Again the setup of the transition matrix gener-
ates a quite familiar result, which one could have pre-
dicted beforehand. Thus not taking full advantage
of the tools opportunities to create differentiated an-
swers to the same problem.
However if playing more with the input values
and letting the program run for a longer time, one
would be able to investigate the possibilities of this
setup even further and possibly arrive at new spatial
configurations for such a topology.
A probabilistic model
Spaces with more than one space to connect to, cre-
ate various arrangements each time the program is
run (figure 9).
244 | eCAADe 32 - Design Tool 1 - Volume 1
Figure 8
A corridor emerges
Thus, the resulting configuration is changing
from time to time, but one should be able to find
identical sequences from model to model.
Two storylines
Two storylines, which each have a unique transition
matrix but some similar qualities can merge to form
more complex storylines (figure 10). This crossover
creates completely new sequences that are different
from what initially predicted from the two separate
transition matrices. Thus with an infinite number
of stories and qualities the result will be an infinite
narrative landscape of spatial qualities (a qualitative
landscape) where you, by dissecting the landscape
into smaller parts, can find any organization/building
- built and unbuilt - within the landscape.
This allows one to take full advantage of the tools
opportunities to create differentiated answers to the
same problem. Answers that one could not have
thought of beforehand.
This in turn, also demands that the user of the
tool takes time to interpret the findings, and to create
their own story from the output presented to them.
Just as intended for the tool.
Figure 9
A probabilistic
model

Figure 10
Two storylines
starts to merge

Figure 11
Possible layouts
generated from the
tool

Design Tool 1 - Volume 1 - eCAADe 32 | 245

 DISCUSSION in Space Layout Planning', Proceedings of CAAD Fu-
Only setting up relations between functional require- tures 1999, Atlanta, pp. 245-262
ments for a single storyline will generate a traditional Christensen, J.T. 2013, Architecture as a narrative land-
scape - designing a tool for exploration, Master's The-
spatial layout or bubble diagram (White, 1986), with-
sis, ETH Zürich, Master of Advanced Studies in Com-
out being influenced by spatial qualities and the sto- puter Aided Architectural Design
ryline of other individuals. If one on the other hand Cullen, C 1961, The Concise Townscape, Reinhold, New
would only set up relations between spatial qualities, York
it becomes a more abstract layout of spatial qualities Hiller, B 2007, Space is the machine: a configurational the-
without any functional relation. In both cases limit- ory of architecture, Space Syntax e-edition, London
Hsu, Y and Krawczyk, R 2003 'New Generation of Com-
ing the space of possible solutions.
puter Aided Design in Space Planning Methods - a
However, combining the functional require- Survey and a Proposal', Proceedings of CAADRIA 2003,
ments and the spatial qualities, both will inform each Bangkok Thailand, pp. 101-116
space of the spatial layout, and result in a true com- Hsu, Y and Krawczyk, R 2004 'Space Adjacency Behav-
bination. This will be important for the future work, ior in Space Planning', Proceedings of CAADRIA 2004,
make the method more useful, and further distin- Seoul Korea, pp. 253-264
Lin, C 2005 'Space layout game: An Interactive Game
guish and improve it compared to other methods of
of Space Layout for Teaching and Representing De-
space layout in architecture. sign Knowledge', Proceedings of CAADRIA 2005, New
Delhi, pp. 130-141
CONCLUSION Lobos, D and Donath, D 2010, 'The problem of space
layout in architecture: A survey and reflections', Ar-
The use of several spring systems, for various types of
quitetura Revista, 6(2), pp. 136-161
relationships between the spaces, allows the sketch White, E 1986, Space Adjacency Analysis, Architectural
to develop beyond the relations of each storyline Media LTD, Tucson
while maintaining a balanced composition. [1] http://en.wikipedia.org/wiki/Markov_chain
As it is a probabilistic process, the output will
never be the same. Thus, it is neither a random nor a
deterministic process, neither unpredictable nor pre-
dictable, but somewhere in between. Not optimiz-
ing, but using the power of computation to give dif-
ferentiated answers to the same problem. Answers
that one could not have thought of beforehand. This
in turn, also demands that the user of the tool takes
time to interpret the findings, and to create their own
story from the output presented to them. Ultimately
using the tool as a true partner in design.
Developed further this method of generating dif-
ferentiated answers can implement other design in-
tentions, such as daylight, view etc. Especially in the
phase of determining an initial layout for further ex-
ploration in conventional parametric software.

REFERENCES
Arvin, S.A. and House, D.H. 1999 'Making Designs Come
Alive: Using Physically Based Modeling Techniques

246 | eCAADe 32 - Design Tool 1 - Volume 1

Design ReExplorer: Interactive Design Narratives for
Feedback, Analysis and Exploration
Rodolfo F. Sánchez1 , Halil I. Erhan2
1,2
Simon Fraser University
1,2
{rsanchez|halil_erhan}@sfu.ca

Designers keep a constant record of the design process through their sketches and
notes. In parametric CAD, the record of design moves is implicit and can be
found in the elements upon which the parametric model is built. Current systems
provide designers with limited tools for recording, viewing or analyzing the
design process. We propose a system's approach to capture the design narrative
as an artefact for design. The Design ReExplorer was developed to test ideas on
using these narratives in gaining insights towards how models are built,
exploring alternatives and supporting backtracking and deferral strategies in
design exploration. We evaluate its insertion and viability in real-world scenarios
through an expert panel study. The results of the study are favourable with
positive feedback and multiple suggestions for future work.

Keywords: Parametric computer aided design, design history, design cognition,

design process

INTRODUCTION AND MOTIVATION We developed the Design ReExplorer (DReX)

Hidden within the design process are a myriad of al-
ternatives; some good, some better and more im-
portantly, some that have not been explored and are
waiting to be uncovered and set free. One of the best
methods to understand these alternatives, in par-
ticular those developed using parametric computer-
aided design (pCAD) tools, is through unravelling
the design narrative that enabled (propelled, allowed,
permitted, laid out) their creation. Design narratives
tell the story of the work: how individual actions and
alternatives build upon each other and how different
solutions come into being. They also tell the story of
the alternate paths that were never considered and
the paths, which when reconsidered, become the
correct path towards finding satisficing solutions (Er-
han et al., 2012).
prototype to test these ideas and to evaluate its inser-
tion and viability in real-world design task environ-
ments. DReX is a creativity-support tool envisioned
to extend the capabilities of pCAD systems for explor-
ing design alternatives in past actions. Its goal is to
enable designers to capture an editable visual history
of design alternatives, and to provide visual feedback
on the design process and interactive controls of the
parametric model through directly connecting with
the pCAD systems working models. We aim to seam-
lessly integrate exploration of possible alternative so-
lutions within design modelling tasks.
The prototype is built on previous work on de-
sign narratives (Erhan et al. 2012) and on the Design
Analytics framework (Sánchez, 2014). The approach
has three tenets: (a) Provide feedback through de-

Design Tool 1 - Volume 1 - eCAADe 32 | 247

 sign narratives; a model of the design process built
through the actions and alternatives designers make
and develop in terms of what, how and when; (b) fa-
cilitate a local and global analysis of the design ac-
tions and alternatives across the narrative; and (c)
enable the execution of design decisions within the
DReX interface based on the preceding analysis. We
present a summary of the initial results from the ex-
pert evaluation study we recently completed to un-
derstand the plausibility of integrating our approach
into design task workflows. The results highlight
some challenges as well as future directions for our
research.
BACKGROUND
Designers keep a record of their designs in the form
of sketches, save-as files and notes as part of a
need to collect precedents that will later be revisited
(Menezes and Lawson 2006). Experienced design-
ers are more likely to save and revisit their sketches
than novices (Lawson, 2006; Goldschmidt 2003). As
these records accumulate they become an unstruc-
tured history of their design process (Makkuni 1987).
This is particularly useful when the design process
is as important, if not more, than the solution itself.
There is also an opportunity for uncovering the re-
lationships between the design process and possi-
ble design solutions (Dorst and Cross 2001). Similarly
there is great interest in managing the plethora of al-
ternatives that are now at our disposal through the
use of pCAD tools (Woodbury, 2010). However, we
have yet to see a clear understanding of how these
tools may look like and function in practice.
Almost all current pCAD systems (e.g. Solid-
Works, Revit, GenerativeComponents, and Grasshop-
per) have mechanisms to traverse the action history
to make changes in the current design state. How-
ever, they still work on a single state model providing
designers access to only a single design solution at
any given time. Systems provide features to circum-
vent this problem or allow designers to implement
ad-hoc solutions but don't provide integrated solu-
tions to use the implicit information within the sys-
248 | eCAADe 32 - Design Tool 1 - Volume 1
tem (Woodbury and Burrow 2006). History mecha-
nisms in pCAD systems and research prototypes are
one of the tools designers have at their disposal to
understand and make sense of the design process.
The majority of these history tools are command logs
taken from the command (undo-redo) stack. Their
capacity is usually limited by linear records and don't
allow designers to backtrack without losing subse-
quent work. These mechanisms are not intended to
explicitly capture a design process.
On the other hand, the augmented history tools-
mostly confined to research prototypes as in Chimera
(Kurlander and Feiner 1990)-demonstrated potential
for providing graphical histories on design process.
The main features found in these tools include the
ability to backtrack and edit previous states, and
delegate repetitive tasks to the computer using his-
tory as a guide. To generate the relationship struc-
tures between design states and parametric values,
the prototypes with augmented history capabilities
rely on the parametric engines that run in the back-
ground (Nakamura and Igarashi 2008; Bueno et al.
2011; Edwards and Igarashi 2000; Heer et al. 2008).
These capabilities which are made ad-hoc for the
prototypes are inherent to pCAD systems. This al-
lows pCAD users to build parametric models and
then edit them having changes subsequently propa-
gate downstream updating the current design state.
As the individual actions that build the parametric
model are executed their properties expose a design
process. The history mechanisms can reveal much
more about the designers, their design process, and
in doing so create a tacit narrative for each design
state and final design solution (Erhan et al. 2012).
Although, there are some sub-systems indirectly
enabling working with multiple alternatives, e.g.
configuration management (Krish, 2011), most ex-
plicit solutions capturing design histories are pro-
posed as prototypes for demonstration purposes.
These can fall under three categories based on where
they propose supporting exploration: on the model
through side-by-side editing (Hartmann et al. 2008;
Lunzer and Hornbaek, 2008); as records in a history of
action and states (Heer et al., 2008; Jankun, Ma, and
Gertz, 2007; Kurlander and Feiner, 1991); and on a set
of alternatives created by generative methods (Marks
et al. 1997; Terry et al., 2004). We have yet to know
how these prototypes translate into functional sys-
tems or the combined effect of the three categories.
Although all these categories are important, in this
research we address the second category based on
our experimental findings and on our observations
that potentials of augmented process-centric tools
have yet to be explored. Below we describe the pro-
totype we aim to use as a probe to further understand
alternatives in the design process and in developing
functional tools to support working with them.
THE DESIGN REEXPLORER PROTOTYPE
DReX uses the parametric modelling capabilities of
pCAD systems to generate a design narrative graph
in the background as an integral part of the task flow
(Erhan et al. 2012). Our goal is to allow design-
ers to break down the sequential nature of design
by making it an atemporal process, setting design-
ers free to explore new alternatives unconstrained
by time as exemplified by other prototypes (Edwards
and Mynatt 1997; Rekimoto 1999). The DReX proto-
type experiments with ideas on how designers can
traverse the history of their actions to re-explore, un-
derstand, describe, and analyze the design process
and design alternatives in relation to their design
moves. As a creativity support tool it aims to aug-
ment the designers' creative search through visual-
izing, exploring and reviewing of design actions and
states (Shneiderman 2002).
The information structure used by DReX is the
same as that proposed to develop design narratives
(Erhan et al. 2012). In both cases information is
gathered directly from the actions performed by the
designers on the pCAD system. In DReX the de-
sign moves are captured synchronously to the ac-
tions performed by the designer, and DReX presents
feedback immediately in the form of a design nar-
rative graph. The actual speed of the feedback de-
pends on computational capacity and complexity of
the parametric model.
DESIGN NARRATIVES - VISUAL FEEDBACK
DReX allows designers to look at their design pro-
cess as an artefact in itself as suggested by (Makkuni
1987), which designers now can backup, replay and
recall (Woodbury and Burrow 2006). The interface
is divided by a timeline that separates the two feed-
back types. Top section contains salient parameters
shown as parameter blocks; these contain parameter
controls, thumbnails of previously visited states/al-
ternatives, and a view of the 3D model for a given
combination of values. The bottom contains the de-
sign narratives graph with backtracking edges, these
connect with other blocks as a sign of backtrack-
ing (Fig.1) (Akers et al. 2009). Parameter blocks are
mapped sequentially over time (left to right) to the
actions taken by the user on the base pCAD system
(Erhan et al. 2012). Parameter blocks are preset in the
DReX prototype.
ANALYSIS AND RE-EXPLORATION
In DReX, the designers can navigate the timeline,
complete a basic analysis and edit the parametric
model. Analysis is attainable at two levels: (1) glob-
ally by identifying patterns in the distribution of pa-
rameters, backtracking edges and deferral; and (2) lo-
cally by identifying patterns in the local collection of
alternatives. Re-exploration in response to the analy-
sis or otherwise is achieved through controls on each
parameter block. Local collections contain a thumb-
nail for each alternate CAD model. These design
states are created by editing the parameter's value.
Future implementations would allow edits in both
base system and DReX.
SYSTEMS DESIGN
The prototype is based on the Design Analytics prin-
ciples: feedback, analysis and re-exploration (for de-
tails see Sánchez, 2014). As such, the prototype tries
to approximate real-world scenarios and situations
whenever possible while not trying to be compre-
hensive in terms of the design process. Our objective
Design Tool 1 - Volume 1 - eCAADe 32 | 249
 Figure 1
The Design
ReExplorer UI with 3
parameter blocks,
each with a
collection of
alternatives and
slider and text box
for model editing.
The edges below
signal backtracking
from one node to
another.

was to be able to maintain the level of complexity of other pCAD systems (GenerativeComponents, Solid-
the models we used at a par with the complexity of Works or CATIA) or other deterministic systems.
real-world projects.
An initial design decisions was to decouple CAPABILITIES AND FEATURES OF THE
the prototype from existing pCAD systems; only
PROTOTYPE
providing interaction indirectly through parametric
Through an incremental development process ini-
changes done outside the prototype. Our belief was
tially the most essential features have been imple-
that this would help developers and participants of
mented in the prototype. They allow designers to
any future evaluation to interface with DReX inde-
view a design process by navigating the timeline to
pendently.
identify blocks or areas that are relatively more ac-
The prototype system consists of three compo-
tive or inactive based on the density and location of
nents: the pCAD system; DReX to interact with past
backtracking edges. This allows designers to identify
design moves and multiple design alternatives, and
different stages of the design process, for example
the conjunctive adapter (Microsoft Excel, RhinoScript
detailing, deferral of work or decisions, and slips, er-
and the JavOnet API) that interconnects the system
rors or mistakes (Fig. 2) (Erhan et al. 2012). Addition-
(Fig. 4). The current implementation of the conjunc-
ally, the designers can view a block's local collection
tive adapter is tailored for Rhino Grasshopper (Fig. 5)
of alternatives and identify past alternatives for reuse
but it is relatively simple to create such adaptors for
or values that represent untested parameter values

250 | eCAADe 32 - Design Tool 1 - Volume 1

Figure 2
DReX timeline
showing
backtracking edges
that link parameter
blocks and nodes
set in the present (i)
and (iii). Differences
in backtracking
clustering i) and ii)
allow us to identify
deferral strategies
(Erhan et al. 2012).

Figure 3
Parameter blocks
contain local
collections of
alternatives (i) with
sliders (f ) and text
boxes (d) to explore
the design space
and a larger image
of the local
geometry (h). A
new value can be
explored in a) using
the text box (d). In
b) a thumbnail of
the new geometry
is shown and the
within the collection (Fig. 3). The user can alternately itative study was completed over a period of two
SaveAll toggle (g) is
try different combinations of previously tested alter- weeks and was comprised of 7 participants and 2 pi-
unchecked. The
natives using controls in each block by matching val- lot. Audio recordings of the participants' answers,
Save button (e) can
ues across the prototype. questions and general feedback was made as par-
be used to save. In
ticipants were shown a presentation, a demo of the
c) a new alternative
EVALUATION: SUMMARY OF THE EXPERT prototype and finally answered a set of open ended
is generated using
questions. These recording have been analyzed and
the slider adding to PANEL REVIEW
the results are presented below.
the local collection We have recently completed an expert panel review
The study was divided into four parts: (1) Partici-
of alternatives (i). of the DReX prototype and the ideas behind it. The
pants filled out a questionnaire and signed a consent
goal was to assess the validity and viability of a fu-
form; (2) Participants were shown a PowerPoint pre-
ture, more refined and robust, system and to eval-
sentation to introduce the main ideas and the pro-
uate how the core ideas presented here are met by
totype. Slides provided an introduction to the do-
domain members, both novice and expert. The qual-

Design Tool 1 - Volume 1 - eCAADe 32 | 251

 Figure 4
System and
implementation
design for DReX.

Figure 5
From left to right:
Grasshopper with
Tower definition,
Design ReExplorer
interface and Rhino
showing current
state of the design.
The Grasshopper
main, preliminary studies, the core ideas behind De- STUDY SETUP definition and
sign Analytics (Sánchez 2014) and an introduction to The presentation and interview process was carried
Rhino model are set
the prototype's user interface; (3) Participants were out in person using a laptop with half of the partici-
in the present
shown a demonstration of the prototype using sce- pants viewing the presentation and prototype on an
moment while
narios and use cases. Participants were then encour- additional 23" monitor. The two screens provided a
DReX shows the
aged to try the prototype for themselves; and (4) better viewing experience; other participants viewed
different paths the
Participants were finally asked a set of open ended the prototype by switching between applications or
design has taken
questions. During the course of the study partici- side by side. A standalone digital audio recorder was
before completion.
pants were encouraged to ask questions regarding used to record participant's comments during the
the questionnaire, the presentation and the proto- study.
type. Open-ended questions were pre-defined but The prototype used a basic parametric model of
frequently follow-up questions were used to further a tower developed in Grasshopper. The tower model
understand the participants' response and comple- is a basic pCAD model to demonstrate possible uses
ment their feedback. cases and scenarios that could be completed using

252 | eCAADe 32 - Design Tool 1 - Volume 1

the prototype. The tower example follows a simple
design pattern of transforming basic 2D geometry
into 3D objects though an amassing of features and
operations Fig. 5c. A more complex model would im-
pose greater computational challenges and was be-
yond the scope of this study. The tower model was
created for demonstration purposes only, editing
was limited to editing values attached to Grasshop-
per nodes, and structural changes of the pCAD model
were not implemented.
PARTICIPANTS
Participants were identified within the CAD, compu-
tational design and AEC domain. We identified indi-
viduals with experience in real-world scenarios that
dealt with complex design problems and had expe-
rience managing and teaching other designers. Ad-
ditionally we wanted feedback from computational
designers and developers and the fresh eyes of pCAD
novices. Their academic backgrounds ranged from
engineers and architects to computer scientist. Par-
ticipants were selected based on characteristics that
lay across multiple axes: novice and expert designer,
graduate student/researcher and design studio pro-
fessional, management and non-management and
their knowledge and knowhow of pCAD.
INITIAL RESULTS
The open-ended questions gave us disorganized an-
swers that were spread across multiple questions. In
analyzing the interviews we have grouped together
answers that relate to the same theme. As with any
proof of concept prototype many comments focused
on issues that can be easily solved and provide no
new insight into the future of the tool and have there-
fore been omitted. As expected participants answers
were consistent with the occasional outlier and with
personality, background and experience shifting an-
swers one way or the other.
Initial results are very promising. Participants
were enthusiastic about the prototype and the ideas
behind it. When asked about integrating a future
prototype (FP) into their workflows answers were
positive. There were several reasons for this. The first
is that the prototype transformed the design process
from a unknown black box into an open artefact that
could be visualized onscreen. The second is that it
allowed them to see a true visual history of the de-
sign by seeing all the thumbnails that are automat-
ically saved when changes are made. Some partici-
pants mentioned that the FP would both encourage
you to review your design process and be more intro-
spective; while other described only specific scenar-
ios for its use. Participants were also wary of its impli-
cations within the workplace in terms of competition,
being too self-aware of the processes and overreach-
ing managers. Procrastination was also mentioned.
Participants were asked to assess their willing-
ness to accept automatically generated feedback
based on their design process. This was asked with
the current prototype in mind and a FP in which AI
and pattern recognition would be included. Partic-
ipants were both open to having a computer pro-
vide feedback but always with some hesitation. The
main concern was intrusive feedback so user control
was a common request. Participants were enthusias-
tic about the possibility of having higher level feed-
back than lower level in particular when related to AI
or pattern recognition.
Regarding the design narratives (backtrack-
ing edges between nodes) opinions were less
favourable. The main concern was the visual aspect,
scale and overall usefulness of the design narratives
feedback; even so they all saw its importance when
shown better examples at different scales using the
presentation slides. Relative to this issue participants
suggested several solutions: conventional zoom, se-
mantic zoom and hierarchical structures or simply
turning them off when not in use.
Higher-level usability issues did arise. Propaga-
tion control was the main issue expressed by partici-
pants. The current version of the prototype only up-
dates the thumbnail of the parameter that is being
edited. Running in the background Grasshopper up-
date all parameters but these are not loaded by DReX.
This caused a lot of confusion but was done to lessen
Design Tool 1 - Volume 1 - eCAADe 32 | 253
 the computational load. The solution, shared and
liked by participants, was to include propagation and
update controls that would allow designers to have
more control over the impact a parametric change
would have in DReX. The other issue was complex-
ity and the computational load real-world projects
would impose on a FP. To mitigate these issues hi-
erarchical structures would be added to increase the
abstraction gradient and cede control of computa-
tionally heavy tasks to users. Another issue that re-
quired deeper analysis is how to manage semanti-
cally complex design moves than those currently im-
plemented e.g. copy/paste or merge files. The proto-
type currently is unable to include structural changes
to the parametric model.
There were several interesting UI and interac-
tion suggestions: adding a secondary notation to an-
notate blocks, alternatives and edges; adding split
screen capabilities, responsive design for the time-
line and local collections, have alternatives gener-
ated automatically by setting up minimum and max-
imum ranges and multiple graphs for collaboration.
CONCLUSIONS
From this prototype we have identified several key is-
sues that need to be resolved for this tool to be ef-
fective. We have also been reassured that research in
this area is both promising and relevant to practition-
ers and researchers alike. Our future work is well de-
lineated by the prototype's shortcomings and strong
points and empowered by the confirmation that de-
signers, in particular experts, hold the design process
and the search for alternatives paramount; and that
these experts need tools to support them.
ACKNOWLEDGEMENTS
This work was partially supported through the Boe-
ing AeroInfo Visual Analytics Project and MITACS;
the Graphics, Animation and New Media Network of
Centres of Excellence of Canada; Consejo Nacional
de Ciencia y Tecnología (CONACYT), Mexico; and
JavOnet.com.
254 | eCAADe 32 - Design Tool 1 - Volume 1
REFERENCES
Akers, D, Simpson, M, Jeffries, R and Winograd, T 2009
'Undo and erase events as indicators of usability
problems', Proceedings of the SIGCHI Conference on
Human Factors in Computing Systems, Boston
Bueno, C, Crossland, S, Lutteroth, C and Weber, G 2011
'Rewriting history: more power to creative people',
Proceedings of the 23rd Australian Computer-Human
Interaction Conference, Sydney, pp. 62-71
Dorst, K and Cross, N 2001, 'Creativity in the design
process: co-evolution of problem–solution', Design
Studies, 22, pp. 425-437
Edwards, W K, Igarashi, T, LaMarca, A and Mynatt, E D
2000 'A temporal model for multi-level undo and
redo', Proceedings of the 13th annual ACM symposium
on User interface software and technology, San Diego,
pp. 31-40
Edwards, W K and Mynatt, E D 1997 'Timewarp: tech-
niques for autonomous collaboration', Proceedings
of the ACM SIGCHI Conference on Human factors in
computing systems, Los Angeles, pp. 218-225
Erhan, H I, Sánchez, R F, Woodbury, R F, Mueller, V and
Smith, M 2012 'Visual Narratives of Parametric De-
sign History', Proceedings of eCAADe 2012, Prague,
pp. 259-268
Goldschmidt, G 2003, 'The Backtalk of Self-Generated
Sketches', Design Issues, 19(1), pp. 72-88
Hartmann, B, Yu, L, Allison, A, Yang, Y and Klemmer, S R
2008 'Design as exploration: creating interface alter-
natives through parallel authoring and runtime tun-
ing', Proceedings of the 21st annual ACM symposium
on User interface software and technology, Monterey,
pp. 91-100
Heer, J, Mackinlay, J D, Stolte, C and Agrawala, M
2008, 'Graphical histories for visualization: support-
ing analysis, communication, and evaluation', IEEE
TRANSACTIONS ON VISUALIZATION AND COMPUTER
GRAPHICS, 14(6), pp. 1189-1196
Jankun-Kelly, T J, Ma, K and Gertz, M 2007, 'A Model
and Framework for Visualization Exploration', IEEE
TRANSACTIONS ON VISUALIZATION AND COMPUTER
GRAPHICS, 13(2), pp. 357-369
Krish, S 2011, 'A practical generative design method',
Computer-Aided Design, 43(1), pp. 88-100
Kurlander, D and Feiner, S 1989, 'A Visual Language for
Browsing, Undoing, and Redoing Graphical Inter-
face Commands', in Chang, S K (eds) 1989, Visual Lan-
guages and Visual Programming, Plenum Press, New
York, NY, pp. 257-275
Lawson, B 2005, How designers think: the design process
demystified, Routledge
Lunzer, A and Hornbæk, K 2008, 'Subjunctive inter-
faces: Extending applications to support parallel
setup, viewing and control of alternative scenar-
ios', ACM Transactions on Computer-Human Interac-
tion (TOCHI), 14(4), pp. 1-44
Makkuni, R 1987, 'A Gestural Representation of the
Process of Composing Chinese Temples', Computer
Graphics and Applications, IEEE, 7(12), pp. 45-61
Marks, J, Andalman, B, Beardsley, P A, Freeman, W, Gib-
son, S, Hodgins, J, Kang, T, Mirtich, B, Pfister, H, Ruml,
W, Ryall, K, Seims, J and Shieber, S 1997 'Design gal-
leries: a general approach to setting parameters for
computer graphics and animation', Proceedings of
the 24th annual conference on Computer graphics and
interactive techniques, Los Angeles, pp. 389-400
Menezes, A and Lawson, B 2006, 'How designers per-
ceive sketches', Design Studies, 27(5), pp. 571-585
Nakamura, T and Igarashi, T 2008 'An application-
independent system for visualizing user operation
history', Proceedings of the 21st annual ACM sympo-
sium on User interface software and technology, Mon-
terey, pp. 23-32
Rekimoto, J 1999 'Time-Machine Computing�: A Time-
centric Approach for the Information Environment',
Proceedings of the 12th annual ACM symposium on
User interface software and technology, Asheville, pp.
45-54
Shneiderman, B 2002, 'Creativity support tools', Maga-
zine Communications of the ACM, 45(10), pp. 116-120
Sánchez, R F 2014, Design ReExplorer: Interactive De-
sign Narratives for Feedback, Analysis and Exploration,
Master's Thesis, Simon Fraser University
Tao, C, Wongsuphasawat, K, Clark, K, Plaisant, C, Shnei-
derman, B and Chute, C G 2012 'Towards event se-
quence representation, reasoning and visualization
for EHR data', Proceedings of the 2nd ACM SIGHIT Inter-
national Health Informatics Symposium, Miami, pp.
801-806
Terry, M, Mynatt, E D, Nakakoji, K and Yamamoto, Y 2004
'Variation in element and action: supporting simul-
taneous development of alternative solutions', Pro-
ceedings of the SIGCHI Conference on Human Factors
in Computing Systems, Vienna, pp. 711-718
Woodbury, R 2010, Elements of Parametric Design, Rout-
ledge, New York
Woodbury, R F and Burrow, A L 2006, 'Whither design
space?', Artificial Intelligence for Engineering Design,
Analysis, and Manufacturing, 20(2), pp. 63-82

Design Tool 1 - Volume 1 - eCAADe 32 | 255

256 | eCAADe 32 - Design Tool 1 - Volume 1
On the Expressive Power of Programming Languages for
Generative Design
The Case of Higher-Order Functions

António Leitão1 , Sara Proença2

1,2
INESC-ID, Instituto Superior Técnico, Universidade de Lisboa
1,2
{antonio.menezes.leitao|sara.proenca}@ist.utl.pt

The expressive power of a language measures the breadth of ideas that can be
described in that language and is strongly dependent on the constructs provided
by the language. In the programming language area, one of the constructs that
increases the expressive power is the concept of higher-order function (HOF). A
HOF is a function that accepts functions as arguments and/or returns functions as
results. HOF can drastically simplify the programming activity, reducing the
development effort, and allowing for more adaptable programs. In this paper we
explain the concept of HOFs and its use for Generative Design. We then compare
the support for HOF in the most used programming languages in the GD field
and we discuss the pedagogy of HOFs.

Keywords: Generative Design, Higher-Order Functions, Programming

Languages

INTRODUCTION ming language is Turing-complete, meaning that

Generative Design (GD) involves the use of algo-
rithms that compute designs (McCormack 2004, Ter-
dizis 2003). In order to take advantage of comput-
ers, these algorithms must be implemented in a pro-
gramming language. There are two important con-
cepts concerning programming languages: compu-
tational power, which measures the complexity of
the problems that can be described using the lan-
guage, and expressive power, which measures the
breadth of ideas that can be described using the lan-
guage. The expressive power is directly related to
the human effort needed to describe those ideas in
a given programming language.
It is a known fact that any non-trivial program-
almost all programming languages have the same
computational power. In what regards their expres-
sive power, however, they can be very different. A
given programming language can be textual, visual,
or both but, in any case, it must be expressive enough
to allow the description of a large range of algo-
rithms.
Expressiveness is intuitively used to measure
how easy it is for a programming language to im-
plement complex ideas. There are studies (Felleisen,
1991) that showed that it is possible to rank lan-
guages according to a formal definition of expres-
sive power that captures the intuitive meaning of
the term. In this ranking, a language that supports,

Design Tool 1 - Volume 1 - eCAADe 32 | 257

 for example, user-defined functions, (e.g., FORTRAN),
is more expressive than one that does not supports
them (e.g., BASIC). This does not mean that there is a
problem solvable in FORTRAN that cannot be solved
in BASIC. What it does mean is that there are prob-
lems that require larger implementation efforts from
a BASIC programmer than from a FORTRAN program-
mer.
In this paper, we claim that higher-order func-
tions (HOFs) are an important and powerful program-
ming language feature for GD. HOFs increase the ex-
pressiveness of a language, reducing the program-
mer's effort and improving code reusability.
In the next sections we describe the expressive
power allowed by HOFs, we illustrate the advantages
of its use in GD, and we analyse the level of support
provided by the programming languages currently
being used for GD. Finally, we describe a pedagogic
approach for teaching HOFs in the realm of GD prob-
lems.
HIGHER-ORDER FUNCTIONS
A HOF is a function that accepts functions as argu-
ments and/or computes functions as results. As an
example, consider the derivative operator D(f ), that
accepts a function as argument, such as x → x2 +
3x, and computes another function as result, in this
case, x → 2x + 3. This fact can be written as
D(x → x2 + 3x) = x → 2x + 3. Note that the
functions that we used as argument and result of
the derivative operator did not have a name. These
unnamed function are known as anonymous func-
tions and there is an alternative notation for them:
λx.2x + 3. This notation was proposed by the λ-
Calculus (Barendregt 1984) and, as we will see, it is
adopted by several programming languages. Note,
however, that from the point of view of the deriva-
tive operator, there is no difference between anony-
mous functions and named ones. For example, it is
well-known that the derivative of the sine function is
the cosine function, a fact that we can express using
named functions, namely: D(sin) = cos.
For a different example, consider a typical
258 | eCAADe 32 - Design Tool 1 - Volume 1
summation:
∑
10
i2 = 1 + 4 + 9 + 16 + . . . + 81 + 100 (1)
i=1
Although rarely presented as such, summation is a
HOF: it accepts a function (in the previous example,
disguised as the expression i2 ) and the limits of a
numeric sequence (in the example, 1 and 10), and
computes the sum of the applications of the func-
tion to all the elements of the sequence. Using the
appropriate
∑ mathematical notation, the example be-
comes 10 i=1 (i → i ), where it is now obvious that
2
the summation function accepts another function as
argument. This is also visible in the formal definition
of the summation operation:
{
∑b
0 a>b
f = ∑ (2)
f (a) + ba+1 f a ≤ b
a
Note, in the above definition, that the parameter f is
used as a function, in f (a).
Although ubiquitous in many branches of math-
ematics, HOFs are unjustly considered a complex
topic that, as a result, is frequently not taught prop-
erly and is not fully supported in a large number of
programming languages. However, as we will show,
HOFs are an important tool for GD: they dramatically
simplify scripting (McCullough, 2006) and they are
a very convenient representation for parameterized
geometry (Park and Holt, 2010). Programming using
HOFs is known as higher-order programming.
As an example, consider a balcony in a build-
ing. During the (generative) design process, a de-
signer might be more concerned about the shape
of the building than about the shape of the balcony
and, as a result, he might decide to implement a sim-
plified balcony. This means that the formalization
of the design in a programming language includes
a parametrized definition for the building that de-
pends on the parametrized definition of the balcony,
as follows:
balcony (...) = ...
building (...) =
...
wall (...)
...
slab (...)
...
balcony (...)
...
Later, when the designer turns his attention to the
balcony, he might want to experiment several differ-
ent designs. In less expressive languages, his only op-
tion is to redefine the balcony definition for each ex-
periment. In more expressive languages that support
HOFs, a better solution is to transform the balcony
into a parameter of the building definition, which
then becomes a HOF, as follows:
building (..., balcony ,...) =
...
wall (...)
...
slab (...)
...
balcony (...)
...
It is now possible to simultaneously define different
balcony functions expressing different balcony de-
signs and treat them as plug-ins for the function that
represents the building.
Although it might seem that there is little differ-
ence between these two approaches, only the sec-
ond approach allows the designer to easily create a
building containing different balcony designs. More-
over, both the function that represents the building
as well as the functions that represent the balconies
are now less dependent from each other and can be
reused in different contexts.
In this paper, we claim that, besides balconies,
many other aspects of a design can be transformed
into (functional) parameters of HOFs. We illustrate
our claim by describing the design of a complex
building, more specifically, the Market Hall (Boranyak
2010), represented by functions that accept, as argu-
ments, functions describing the overall shape of the
building, functions for the different elements (posts,
facade, etc.), functions describing sequences of ele-
ments, etc. This representation allows us to gener-
ate not only a building that is identical to the original
concept, but also an infinite number of variations, ex-
pressing different choices for all the function param-
eters, including those that represent other functions.
HIGHER-ORDER FUNCTIONS IN GENERA-
TIVE DESIGN
In spite of the simplicity of the idea of HOFs, a (gen-
erative) designer still needs to think about the best
strategy for its use, which is strongly dependent on
the current design goals. Returning to the previous
example, it is clear that it is not enough to say that
the function that creates the balcony is a functional
parameter of the function that creates the building;
we also need to specify the communication protocol
between both functions, namely, which information
is expected by the balcony function and which infor-
mation does it return to the building function. De-
pending on the programming language being used,
the designer might decide to provide part or the en-
tirety of the expected information as arguments to
the function. Additional information might be pro-
vided in global variables or using other language-
specific mechanisms. For illustration purposes, we
will consider that the balcony function expects the
spatial location of a corner of the balcony and the di-
mensions of the intended balcony. We will assume
that the building function provides this information
each time it calls the balcony function.
Regarding the design of the balcony, the de-
signer might want to experiment different mathe-
matical functions for its shape. This means that it
should be possible to define another (higher-order)
function that accepts the shape function and that
computes a specialized balcony function that follows
that shape. We will now assume the existence of
this function and we will only consider the different
shape functions that the designer might want to ex-
periment.
One possibility is the use of a sinusoidal shape.
To this end, the designer might define a function-
Design Tool 1 - Volume 1 - eCAADe 32 | 259

producing function (again, a HOF), that accepts the
amplitude, frequency, and phase of the sinusoidal
curve:
sinusoidal(α, ω, φ) = x → αsin (ωx + φ) (3)
The balcony produced using the previous function is
visible in figure 1, on the left. So far, this is not very
different from what he could achieve using only first-
order functions. In fact, the advantages of higher-
order functions only become clear when we experi-
ment with different functions. For example, an expo-
nential decay describes a process whose output de-
creases at a rate proportional to its current value, a
fact that can be modeled by the following function:
decay(β) = x → e−β (4)
The designer can now multiply the decay function
with the sinusoidal function to produce a function
that describes a decaying sinusoid. To this end, he
needs to define the product ⊗ of functions, as an HOF
that, given two functions f and g, computes a third
function that is f ⊗g:
⊗(f, g) = x → f (x) × g(x) (5)
As a concrete example, the formula
⊗(decay(0.1), sinusoid(−2, 2, 0)) (6)
describes the shape of the balcony presented in the
center of figure 1.
Many other higher-order operators can be simi-
larly defined. For example, the composition of func-
260 | eCAADe 32 - Design Tool 1 - Volume 1
Figure 1
Balconies
generated using
higher-order
functions. From left
to right, a sinusoidal
function, the
product of decaying
tions f ◦g can be formally defined as: and sinusoidal
functions, and the
◦(f, g) = x → f (g(x)) (7) composition of a
clamping and
Another example is the clamping function, that limits sinusoidal function.
its input to a given range of values:
clamp(inf, sup) = x → min (max (x, inf), sup)
(8)
Using these HOFs, the designer can experiment other
kinds of balconies, like the one represented in figure
1, on the right, that corresponds to the clamped sinu-
soid ◦(clamp(−0.7, 0.7), sinusoid(1, 1, π/2)).
This previous example illustrates the use of HOFs
for a very simple case. The concepts, however, scale
to much bigger cases, as can be seen in figure 2,
which shows a 3D model of the Market Hall building
that was fully generated using programming, partic-
ularly, HOFs. The complete program is implemented
by 332 functions, including 32 HOFs that are used
in 143 different places of the program. An analysis
of these HOFs reveals that these are used not only
to describe an abstract building, in which many of
the building elements are functional parameters, but
also to implement the sampling of parametric func-
tions, i.e., to compute the values of parametric func-
tions for different values of their domain, and to im-
plement mappings of functions over collections, i.e.,
the application of a function to each element of a col-
lection, producing a collection of results.
It is important to note that the generation of the
Market Hall 3D model depends on several HOFs, each
one implementing a particular part of the design.
There are functions that implement the overal shape,
 functions that implement balconies, functions that
implement elements of balconies, etc. Arbitrary com-
binations of these functions can then be created to
implement different variations of the main design.
Figure 3 shows the "same" Market Hall building,
in the sense that it is the result of the same HOF that
was used to produce the model in figure 2, but where
we provided different functions as arguments, in par-
ticular, to describe the overall shape of the building,
thus producing a significant variation. While the orig-
inal building has a shape whose longitudinal evolu-
tion is described by a linear function, the variation il-
lustrated in figure 3 shows, among other differences,
the use of a sinusoidal function to describe this evo-
lution.
Figure 2
The Market Hall
building, entirely
generated using
programming.
Figure 3
A different instance
of the Market Hall
building.
HIGHER-ORDER FUNCTIONS IN PRO-
GRAMMING LANGUAGES FOR GENERA-
TIVE DESIGN
Given the usefulness of HOFs, it is not surprising to
see them supported in a large number of modern
programming languages. There are, however, differ-
ent levels of support among the different program-
ming languages and even among different versions
of the same language. The Java programming lan-
guage, for example, was released in 1995 but its
implementation of anonymous functions appeared
only in version 8, released almost 20 years later.
The Python programming language did not provide
HOFs in its initial implementation, in the late 80's, but
the first official release, in 1994, included anonymous
functions and some list-processing HOFs. Haskell
provides considerable support for HOFs, including
automatic partial application but forces the user
to work in the less conventional lazy-evaluation
paradigm (Hughes, 1989). Finally, there are lan-
guages, such as the original BASIC language, that do
not event support HOFs.
In this paper, we are particularly concerned with
the support for HOFs that is provided by program-
ming languages used in the GD field. The lan-
guages that we will discuss are Python, AutoLisp, VB-
Script, and Grasshopper, but we will also comment
on some additional languages used for GD, such as
GDL, MAXScript, Processing, PLASM, and Racket.
Python
Python is a programming language that is currently
enjoying a considerable momentum in the GD com-
munity, particularly, due to fact that it is one of the
scripting languages of Rhinoceros 3D. Python pro-
vides some pre-defined higher-order functions and
also allows user-defined ones. As an example, the fol-
lowing function implements the exact same summa-
tion function that was described by formula 2:
def summation(f, a, b):
if a > b:
return 0
else:
return f(a) + summation(f, a+1, b)
Similarly, in Python it is possible to provide anony-
mous functions as arguments, using ∑ lambda expres-
sions. For example, the expression 10 i=1 (i → i )
2
becomes:
summation(lambda x: x**2, 1, 10)
The summation function is an example of an HOF that
accepts a function as an argument. The composition
Design Tool 1 - Volume 1 - eCAADe 32 | 261
 of functions is more interesting, as it also returns a
function as result. Its definition, in Python, becomes:
def compose(f, g):
return lambda x: f(g(x))
There is one restriction regarding anonymous func-
tions in Python: they can only contain one expres-
sion, which means that it is not possible to include
statements in the body of the anonymous function.
In practice, this is not a serious limitation, as it is pos-
sible to locally define a named function and use it as if
it was anonymous. Moreover, many of the program-
ming patterns related to the use of anonymous func-
tions, such as mapping, filtering, and reducing lists
can be replaced by list comprehensions.
In practice, this means that Python provides
good support for programming with HOFs and, thus,
can be easily used to implement the GD example that
we described.
AutoLISP
AutoLISP is one of the scripting languages of Auto-
CAD. AutoLISP is a member of the LISP family of lan-
guages (and, thus, provides HOFs) and a widely used
language for GD, with a huge amount of scripts avail-
able on the internet. A cursory look at some of those
scripts show that authors were acquainted with some
of AutoLISP pre-defined HOFs, particularly, the map-
car function.
AutoLISP also allows user-defined HOFs. For ex-
ample, the summation function becomes:
(defun summation (f a b)
(if (> a b)
0
(+ (f a)
(summation f (1+ a) b))))
Unfortunately, AutoCAD is a dynamically scoped lan-
guage, and this means that user-defined higher-
order functions can cause hard-to-debug problems,
particularly, the infamous downward funarg and up-
ward funarg problems (Moses, 1970). The downward
funarg problem occurs when a function provided as
argument has a free variable that is shadowed by
262 | eCAADe 32 - Design Tool 1 - Volume 1
some variable of the HOF. As an example, consider
a function that compute a sum of powers b from 1 to
n. Its definition in AutoLISP might be:
(defun sum -of -powers (n b)
(summation (lambda (x) (expt x b))
1 n))
Unfortunately, if we try any computation involving
the previous function, we will discover that it does
not compute the correct result, because the expo-
nent b used in the anonymous function is shadowed
by the upper limit b of the HOF. This is a serious prob-
lem that can be mitigated but never entirely solved
using some name-mangling techniques that prevent
name collisions.
The second problem, the upward funarg, is more
serious and much more difficult to solve. It occurs
when a HOF returns a function as a result, as it hap-
pens with the function ◦(f, g). Its definition, in Au-
toLISP, would look like:
(defun compose (f g)
(lambda (x)
(f (g x))))
Unfortunately, when the returned function has free
variables that were bound by the HOF, these vari-
ables will lose their current value, thus making the re-
turned function useless. This means that the names
f and g that are used in the anonymous function re-
turned by the function compose will not be correctly
bound and will cause errors. This is a serious limita-
tion of AutoLISP that restricts the expressive power of
the language.
Visual Basic
Visual Basic can be considered a family of languages.
There are two main dialects that are used as script-
ing languages: VBA (Visual Basic for Applications)
and VBScript. VBA is a restricted version of Visual Ba-
sic that is used, for example, in AutoCAD. VBScript
is an even more restricted version that is used in
Rhinoceros 3D, under the name RhinoScript. Most of
the restrictions are related to the execution environ-
ment and are not relevant to our analysis but there
 are important restrictions that directly affect the abil- Grasshopper
ity to define and use HOFs. One is that VBA does not All the languages described previously belong to
consider functions as first-class entities, and, as a re- the category of textual programming languages,
sult, they cannot be passed around just like other val- meaning that programs are written using a one-
ues, such as numbers and strings. However, VBA pro- dimensional sequence of characters. On the other
vides a way for calling functions whose name is de- hand, in visual programming languages (VPLs), pro-
scribed by a string. This is not possible in VBScript grams are "written" in a bi-dimensional representa-
but, starting from version 5.0, it is possible to do dy- tion consisting of iconic components that can be in-
namic code evaluation using the functions Eval (for teractively manipulated by the user. Grasshopper is
evaluating expressions) and Execute (for executing one of the most popular VPLs for GD and some of its
statements). In both cases, the code to evaluate must components can be used for higher-order program-
be contained in a string. ming.
To illustrate the definition of a pseudo-HOF, the Figure 4 shows a Grasshopper program using
following VBScript function
∑ mimics the mathemati- the Evaluate component, that accepts a formula de-
cal definition of the function: scribed as a string and a value or sequence of val-
ues, and computes the evaluation of the formula for
Function Summation(expr , a, b) each value received. By changing the string that rep-
If a > b Then resents the formula it becomes possible to compute
Summation = 0 different behaviors, just like was done, e.g., in the VB-
Else
Script example.
Summation = Eval(expr) + _
Summation(expr , a + 1, b)
Figure 4 End If
A Grasshopper End Function
program that
computes the sum Differently from the previous languages, where func-
of squares of tions were provided as arguments, in the case of VB-
integers from 1 to Script, the function requires a string describe the ex-
10. pression to evaluate: Recent Grasshopper versions can even abstract a set
of components into a cluster, allowing the definition
summation ("a*a", 1, 10) of "functions". Figure 5 shows the same Grasshop-
per program that is presented in figure 4 but where a
Note also that the expression has to explicitly refer cluster was used to abstract the definition of the sum-
the variable of the Summation function that contains mation function.
the correct value to use, which is a serious violation
Figure 5 of the encapsulation principle. Moreover, by repre-
A Grasshopper senting functions with strings, it becomes very dif-
component that ficult to combine functions as this requires the con-
abstracts the struction of a string that, according to the syntax and
summation semantics of the programming language, represents
function. the intended combination of the expressions. In the
general case, this implies writing a compiler. In prac-
Unfortunately, by using strings as representations of
tice, both the VBA and the VBScript dialects cannot
formulas, we end up finding the same limitations that
be considered adequate languages for the definition
were discussed for VBScript, making it very cumber-
and use of higher-order functions.

Design Tool 1 - Volume 1 - eCAADe 32 | 263

 some to use expressions with free variables or to im-
plement HOFs that return functions as results.
Additional Languages
The previous sections analysed the most used pro-
gramming languages in the GD field. However, there
are several other, less used, languages that could
have been included. In this section we briefly review
some of them.
GDL. GDL (Watson, 2009) is the scripting language of
ArchiCAD, a popular CAD application. GDL is a de-
scendant of BASIC but contrary to its siblings, such as,
VisualBasic, GDL did not evolve and has many short-
commings. In particular, subroutines cannot define
local variables, receive parameters, or return values,
thus making the language completely unsuitable for
higher-order programming.
MAXScript. MAXScript (Autodesk, 2006) is the script-
ing language of Autodesk's 3dsMax. In MAXScript,
functions are first-class values and can be provided as
arguments or returned as values to/from other func-
tions. As a result, it is trivial to define HOFs and the
standard library already implements some. Given
that GD scripts tend to have frequent use of mapping
operations over collections of values, MAXScript al-
lows the automatic definition of mapped versions of
normal functions. In spite of using lexical scope, thus
solving the downward funarg problem, each scope
is discarded as soon as the execution flow leaves the
scope, thus suffering from the upward funarg prob-
lem.
Processing. Processing (Reas & Fry, 2010) is a popular
programming language specialized for the produc-
tion of images and animations. The language is a sim-
plified version of Java that suffers from the same lim-
itations of Java for functional programming, namely
forcing the use of some basic design patterns, such as
the Command, for providing first-class status to func-
tions, and the verbosity required for the definition of
even for the most simple higher-order operations.
PLASM. PLaSM (Paoluzzi & Sansoni, 1992) is a func-
tional programming language created for GD with a
264 | eCAADe 32 - Design Tool 1 - Volume 1
strong emphasis in higher-order programming, pro-
viding many pre-defined operators for function com-
bination. PLaSM uses a rather dense mathematical
notation which can be hard to grasp. Recently, a
Python front-end for PLaSM was developed, allowing
the use of PLaSM operators in a more familiar setting.
In spite of allowing the visualization of the generated
designs using some well-known standards, such as
VRML or SVG, it cannot be directly used with CAD ap-
plications.
Racket. Racket (Tobin-Hochstadt 2011) is a modern
member of the LISP family of languages, designed
for pedagogical and practical purposes. Racket is
also one of the languages supported by Rosetta
(Lopes 2011), a programming environment pro-
viding multiple programming languages, such as
Javascript and AutoLISP, and multiple CAD applica-
tions, such as AutoCAD and Rhinoceros 3D. Racket
provides strong support for higher-order program-
ming, with many pre-defined HOFs and unrestricted
user-defined HOFs.
TEACHING HIGHER-ORDER FUNCTIONS
In the last several years we have been teaching a one-
semester computer science course for students of ar-
chitecture that do not have any prior programming
experience. Given the increased expressive power
provided by HOFs, it was our goal from the begin-
ning to explore the topic during the course. In or-
der to present the concept using ideas that the stu-
dents can quickly grasp, we explore the summation
function as a motivating example, but always using
named functions as arguments. When we feel that
students have a good understanding of the concept,
we explain anonymous functions as simplified func-
tions that are created just for an ephemerous use and
then we explain the relation between named and un-
named functions. In order to keep the students inter-
ested in the topic we also solve simple architectonic
problems, such as designing buildings with shapes
defined by functions or computing curves and sur-
faces from their parametric definitions.
In our experience, students can easily under-
 stand the concept of HOF. There is, however, a se- Market Hall entirely generated by a GD program. This
rious problem in the use of those concepts in pro- program was developed by one student and it exten-
gramming languages that were not designed to sup- sively explores the concept of HOF, allowing the sim-
port them. This was the case in the first years, when plification of the code and reducing the necessary ef-
we had to teach computer science using AutoLISP fort to produce it.
(Leitão 2010). AutoLISP is a language that is easy In order to take advantage of HOFs, it is nec-
to learn but, as we explained before, it is severely essary to know the strengths and limitations of
limited in its support for HOFs, both downward (i.e., programming languages, particularly, regarding the
functions that accept functions as arguments) and support for HOFs. Otherwise, the effort to overcome
upward ones (i.e., functions that return functions as the language limitations not only distracts from the
results). This forced us not only to avoid talking main programming goals but might also overcome
about upward HOFs but also to waste some time ex- the benefits of the use of HOFs. The support for
plaining name-mangling techniques that would pre- HOFs and the increased expressiveness they allow
vent the name collisions that, otherwise, would in- are, thus, strongly dependent on the specific pro-
evitable occur with downward HOFs. However, stu- gramming language that is being used. In this pa-
dents never really understood these limitations as, per, we considered the most used programming lan-
from their point of view, there was nothing problem- guages in the GD field and we analyzed their sup-
atic with HOFs. port for higher-order programming. Table 1 synthe-
In order to improve the learning process, the sizes our findings, showing that Phyton, PLASM, and
only option is to use a programming language where Racket are the languages that better support higher-
HOFs are as natural as normal functions. For this rea- order programming.
son, in the last few years we decided to stop teach-
Table 1
ing using AutoLISP and we now teach using Racket.
Programming
This move tremendously simplified the explanation
language support
of HOFs, as they now can be presented without any
for higher-order-
caveats and students can use them without any limi-
programming in
tations.
the GD area. A solid
circle means that
there is effective CONCLUSIONS
support. A The expressive power of a language measures the
half-circle means breadth of ideas that can be described. Through-
that there is some out the history of mankind, we have invented numer-
support but it ous ways of increasing the expressive power of our
might require languages. The concept of HOF is one of those in-
additional efforts ventions that had a transformative effect in our abil-
from the ity to describe nature and, as a result, have been
programmer. A used in many different areas, including genetic pro-
cross means that gramming (Binard 2008), constructive solid geome-
there is no support. try (Davy 1995), and shape grammars (Lewis et al.
2004).
In this paper we explained the expressive power
of HOFs and the potential of its use in GD. We illus- In our teaching experience, the use of languages that
trated that potential by presenting a 3D model of The only provide partial support for HOFs, such as Au-

Design Tool 1 - Volume 1 - eCAADe 32 | 265

 toLISP, forces the teacher to waste precious time ex- sachusetts: The MIT Press, Cambridge
plaining limitations that are difficult to understand Leitão, A, Cabecinhas, F and Martins, S 2010 'Revisiting
and accept by the students. We believe that higher- the Architecture Curriculum: The programming per-
spective', Proceedings of 28th eCAADe, Zurich, pp. 81-
order programming should be considered a natural
88
extension of programming with functions and, con- Lewis, J, Rosenholtz, R, Fong, N and Neumann, U 2004 'Vi-
sequently, teaching the definition and use of HOFs sualIDs: automatic distinctive icons for desktop in-
should be a natural extension of teaching the defini- terfaces', ACM SIGGRAPH 2004, New York, pp. 416-
tion and use of functions. This is only possible, how- 423
ever, when the programming language being used Lopes, J and Leitão, A 2011 'Portable Generative Design
for CAD Applications', Proceedings of ACADIA 2011,
fully supports HOFs, a consideration that made us
Alberta, pp. 196-203
move from teaching with AutoLISP to teaching with McCormack, J, Dorin, A and Innocent, T 2004 'Generative
Racket, allowing us to drastically simplify the expla- design: a paradigm for design research', Proceedings
nation of HOFs and saving time for discussing more of Futureground, Melbourne
interesting uses of HOFs. Moses, J 1970 'The function of FUNCTION in LISP or why
the FUNARG problem should be called the environ-
ment problem', ACM Sigsam Bulletin, pp. 13-27
ACKNOWLEDGEMENTS Paoluzzi, A and Sansoni, C 1992, 'Programming lan-
We would like to thank the architect Rita Fernandes guage for solid variational geometry', Computer-
for sharing with us the GD program for the Market AidedDesign, 24(7), pp. 349-366
Hall project. Terdizis, K 2003, Expressive Form: A Conceptual Approach
to Computational Design, London and New York,
This work was partially supported by Portuguese
Spon Press
national funds through FCT under contract Pest- Tobin-Hochstadt, S 2011, 'Languages as libraries', ACM
OE/EEI/LA0021/2013 and by the Rosetta project un- SIGPLAN Notices, 46(6), pp. 132-141
der contract PTDC/ATP-AQI/5224/2012. Watson, A (eds) 2009, GDL handbook: A comprehensive
guide to creating powerful ArchiCAD objects, Cadim-
age Solutions, NewZealand
REFERENCES
Barendregt, HP 1984, The Lambda Calculus, North-
Holland, Amsterdam
Binard, F and Felty, A 2008 'Genetic programming with
polymorphic types and higher-order functions', Pro-
ceedings of the 10th annual conference on Genetic and
evolutionary computation (GECCO '08), Atlanta, pp.
1187-1194
Boranyak, S 2010, 'Archetype', ASCE, 80(2), pp. 76-79
Davy, J and Dew, P 1995, 'A polymorphic library for con-
structive solid geometry', Journal of Functional Pro-
gramming, 5, pp. 415-442
Felleisen, M 1991 'On the expressive power of program-
ming languages', Selected papers from the sympo-
sium on 3rd European symposium on programming
(ESOP '90), Amsterdam, pp. 35-75
Fry, B and Reas, C 2010, Getting Started with Processing,
O'Reilly Media
Hughes, J 1989, 'Why functional programming matters',
The computer journal, 32(2), pp. 98-107
Kalay, Y 2004, Architecture's New Media: Principles, The-
ories, and Methods of Computer-Aided Design, Mas-

266 | eCAADe 32 - Design Tool 1 - Volume 1

Thermal performance of patterned facades
Studies on effects of patterns on the thermal performance of facades

Mehrnoush Latifi Khorasani1 , Jane Burry2 , Mahsa Salehi3

1,2
RMIT University Australia 3 University of Melbourne
1,2
{latifikhorasgani.mehrnoush|Jane.burry}@rmit.edu.au
3
m.salehi@student.unimelb.edu.au

Skin is the primary shield between our body and its surroundings. It protects the
body from the harmful environmental effects like dehydration and radiation from
intense sunlight. Likewise, the outer layer, or skin of a building has the same
function of protecting its inhabitants against the external elements. This research
is a part of a larger investigation into geometrical patterning and layering of
facades as an effective intervention between the outdoor space and the indoor
environment to regulate the conditions for occupant thermal comfort. This paper
reports on exploration of an approach for measurement, evaluation and feedback
in the design workflow through a mixed digital -physical simulation platform
(MDPS) based on the objectives of the larger study. For this purpose, it
introduces a new way of analyzing thermal performance of double skin facades
by using temperature sensors, Arduino, post visualization with MATLAB and
digital energy simulation. The main aspects of this proposed workflow is the
design of a thermal performance feedback loop as an integral part of the process
of geometrical patterning design for façade.

Keywords: Patterned facades, Thermal performance, Surface temperature, Data

visualization, mixed digital physical simulation

INTRODUCTION with traditional features and elements that draw on

Skin is the primary shield between our body and its
surroundings. It protects the body from the harmful
environmental effects like intense sunlight and radi-
ation. Likewise, the outer layer, the skin, of a building
has the same function of protecting its inhabitants
against external elements.
In harsh environments, like hot and arid climates,
building skins play an important role in moderat-
ing the harsh conditions. Facades can be designed
a climate- and culturally specific vernacular for exam-
ple, Mashrabiyas.These elements control the daylight
and air flow while cooling the air passing through
them (Fathy 1986). In hot arid climates the moder-
ating function of facades, including their geometri-
cal characteristics and features, impact on inhabitant
comfort, reducing direct exposure to the harsh ex-
ternal atmospheric conditions. Geometrical façade
characteristics contribute to the moderation role of

Design Tool 1 - Volume 1 - eCAADe 32 | 267

 the facade, by preventing heat gain through the sur-
faces in part through self-shading. How can de-
signers emulate these effects effectually in early de-
sign in order to develop genuinely climate responsive
screens and barriers to reduce over heating in build-
ings?
This research is going through a proposed plat-
form based on capability of investigation into geo-
metrical patterning of facades as an effective inter-
vention between the outdoor space and the indoor
environment to reduce surface temperature, incident
solar radiation and transmitted heat through the sur-
faces. Design strategies and effective approaches to
measurement, evaluation and feedback in the de-
sign workflow are combined in the proposed plat-
form. Initially such strategies target rapid feedback
to inform early design, enrich the design process, for
getting dynamic feedback, and bring it back to the
design stage for manipulating the skin forms.For this
proposed design loop both physical and digital sim-
ulation are integrated in the thermal performance
feedback.
SKIN THERMAL PERFORMANCE: BACK-
GROUND, OBJECTIVES AND MOTIVATION
Studies in bioscience show that shapes and pat-
terns of Barrel cactuses, including opening patterns,
spines, and ribs, affect their thermal energy ex-
changes (Lewis and Nobel 1977;Nobel 1978).
These studies show how features of the cacti
268 | eCAADe 32 - Design Tool 1 - Volume 1
as examples of desert plants help them moderate
the difficult conditions of hot and arid zones. Lewis
and Noble (1977) observed that the surface tempera-
ture of the cactus' stems increased rapidly during the
summer days when their spines and ribs are removed
.They concluded that the ribs and spines are effective
features in changing the cactus skin's thermal perfor-
mance and characteristics, through shading the exte-
rior surface from solar irrigation (figure 1).
Studies in engineering and architecture demon-
strate that layers of vegetation on walls and double
skin facades can change the thermal performance
of the surfaces in convection, conduction and sur-
face temperature (Eumorfopoulou & Kontoleon 2009;
Hensen, Bartak, & Drkal 2002; Susorova, Angulo,
Bahrami, & Stephens 2013).
Wong et al (2012) in their studies, by comparing
eight different types of greenery wall, concluded that
thermal performance of the greenery wall depends
on various factors such as the shading areas.
These results point to potential strategies for re-
ducing buildings' surface temperature and conse-
quently improving the thermal performance of the
building skin by adding different layers to the facades
or protecting them through self-shading techniques.
Self shading techniques in our research refer to the
ways that parts of surface could protect other parts
from the direct sunlight like a cactus' ribs. These lay-
ers could be vegetation layers as natural elements or
layers of materials and construction elements such as
double skin facades. Both of these layering strate-
Figure 1
Horizontal sections
of a cacti show the
ribs and spines,
effective features in
changing the cactus
skin’s thermal
performance.
 gies can be applied to increase human comfort and mixes physical and digital simulation is one criterion
decrease energy consumption in the buildings; how- for a solution .
ever, vertical greenery systems have limited poten-
tial application in dry climates when water and veg- THERMAL PERFORMANCE EVALUATION
etation are scarce. This leads to the question: What
AND ASSESSMENT
kinds of geometrical patterns and systems could per-
It would be impossible for architects and designers
form similarly and have analogous effects on thermal
to go through the optimization of geometries for im-
behavior of the buildings' elements to improve the
proving thermal performance, without considering
climate modifying function of the building for inhab-
proper tools for assessment and evaluation. Mean-
itants in hot arid conditions? To answer this ques-
while the selection of the right tools depends on
tion, about the effects of geometries on the protec-
the aims and objectives of studies.To select the right
tive function of the façade and designing for it in the
tools, it is essential to define the aims and objectives
early stage of design, finding a good methodology
of the studies and the assessment criteria. In this re-
for assessment and evaluation of the designed ge-
search, following the mentioned literature, we have
ometries is critical. Having a simulation platform that
considered four main parameters affecting thermal

Figure 2
Proposed
framework for
driving the forms
base on drawn
feedbacks from the
mix digital- physical
simulation
platform(MDPS).

Design Tool 1 - Volume 1 - eCAADe 32 | 269

 performance of the façades as below:
a) Incident solar radiation
b) Pattern surface temperature
c) Temperature behind the surfaces
d) Area of self-shading
Time is considered as a critical factor in select-
ing the proper tools as in the early stage of design
designers need to make decisions in a limited time.
These are derived from the background research and
lead to the detailed design of a mixed digital -physical
simulation platform as below.
MIXED DIGITAL -PHYSICAL SIMULATION
PLATFORM
According to the background studies, we proposed
a mixed digital-physical simulation platform (MDPS)
based on two main stages: a) data collection b) data
visualization and analysis including a series of physi-
cal and digital simulation ,data collection and visual-
izations (Figure 2).
The output from the conceptual design process
enters as an input to the MDPS platform for the ther-
mal performance evaluation and assessment by con-
sidering the defined criteria to form the next genera-
tion of the patterned screen based on the feedbacks.
The analyzing output data from this mixed digital -
physical platform are:
• Colored images based on incident solar radia-
tion (ladybugs+Grasshopper)
• Thermal images for surface temperature
(thermal camera images)
• Plotted charts of temperature of individ-
ual sensors including 16 individual dia-
grams for the Test Box and Monitor Box (Ar-
duino+temperature sensors+Matlab) and a
chart for mean temperature of Box A and Box
B.
• Colored images based on transmitted heat
through the surfaces (Arduino+temperature
sensors+Matlab)
270 | eCAADe 32 - Design Tool 1 - Volume 1
• Shadow movies and images, illustrating self-
shading areas(Autodesk Vasari+Ladybugs)
Data fusion plays an important role in the post
processing as we have various types of extracted data
in the performance loop stage. Therefore, we de-
cided to use Matlab as the post processing visualiza-
tion tool, which can be used to extract knowledge
from the data for the future studies.
DATA COLLECTION
The data collection from the MDPS platform in-
volves analyzing in the digital energy simulation soft-
ware. Advanced computational methods offer vari-
ous tools for design exploration, evaluation and as-
sessment during the early stage of design. Based on
our studies' objective,Vasari and different plugins of
grasshopper like Ladybugs, Diva and Geco exist as
good tools for energy simulation .They are very user-
friendly tools and their combination with other plug-
ins like Galapagos in grasshopper offers the opportu-
nity of undertaking complex optimization. However,
they are for the two criteria only: incident solar radia-
tion on the patterned surface and self-shading areas.
Getting accurate results based on the surface
temperature and temperature behind the surfaces
involves fluid flows dynamics into the simulation
platform, offered by computational fluid dynamics
(CFD) software packages. However, the simulation
process with CFD is very time-consuming for the
early design stage, especially when the aim is op-
timizing different generations of the forms to have
a comprehensive overview, and considering the dy-
namics of heat phenomena within the study do-
mains. Consequently, the physical analogue simula-
tion has been considered as an alternative solution.
Many practitioners use physical simulation as a
tool for thermal analysis.For instance, Wong et al
(2012) used a physical set up and field measurement
to study the effects of greenery layers on reducing
the radiation on the surfaces.Two similar labs have
been constructed in a one-to-one scale set up named
"thermal lab", in the Department of Architecture of
Texas, Austin, where students and researchers can
Figure 3
The designed
physical platform.
test their designed geometries by attaching them to
the glass walls of the lab [1].However,it has the limita-
tion of the study being narrowed to only one location
and orientation(South façade in Texas). Another ex-
ample is "Thermal Reticulation cluster of Smart Geom-
etry 2013" in which the series of sensors and thermal
camera were used for capturing the heat movement
inside of a box. In this case they used digital sim-
ulation packages beside the analogue test platform
(Burry et al 2013).
With physical Simulation tools it is possible to
test and evaluate different configurations while in-
creasing the environmental perceptions by creating
more tangible platforms. Moreover, through experi-
encing the physical models of designed motifs that
are tested in the platform we could study their aes-
thetics in physical space including dynamic aspects
of shadows as hidden secondary geometries. Hence,
a physical platform was designed.
PHYSICAL PLATFORM
Two similar boxes are designed, one to be used as
a control during experimentation and the other as
the test domain out of 90*60cm MDF panels, rep-
resenting the wall of an existing building that de-
signed screens are attached to them externally. The
designed screens are applied as a second protective
building skin and the impact on the permanent pro-
tected inner surface is measured. This measurement
gives feedbacks about heat transmission by the pat-
terned screens to consider the effects of first screens
on the second facade.
The key novel contribution of this platform is
using two identical rigs for evaluating the designed
screen pattern, in which certain design iterations are
evaluated against different design generations. The
'control box', named "Box A", is designed to monitor
the effects of changing temperatures though a previ-
ous generation of the façade. The second alternative
of designed façade geometry could be applied exter-
nally to the second box, which is named "Box B or
Test box". There are 32 digital one wire temperature
sensors (DS18B20) which are embedded in the 3mm
MDF test box and monitor box (16 sensors on each
box) to measure the amount of heat that is transmit-
ted through the patterned surface.
The data collection in this stage involves data
gathering by temperature sensors and a thermal
camera. The sensor arrays can measure the tempera-
tures of the grid on the second wall through connect-
ing to an Arduino board.The thermal camera, which
provides thermal images of the different design sur-
faces, is mounted in front of the domains to capture
the thermal images of the external surfaces.These im-
ages show the mean temperature of each domain (in-
cluding Monitor and Test Box)(Figure 3).
Design Tool 1 - Volume 1 - eCAADe 32 | 271

The extracted data from the sensor measurements
will be visualized using MATLAB, which enables re-
searchers to gain insight into the extracted data .This
stage enables the researcher to get feedback on their
designed models visually and give them the oppor-
tunity of comparing the extracted data for detailed
assessment processes. Moreover, MATLAB is used
for mapping the temperature sensors as they are ad-
dressed randomly by the Arduino board. A sample
of the visualization is illustrated in Figure 4 in which
sensor B10 and A7 were tested.
ADVANTAGES OF THE PHYSICAL PLAT-
FORM
Having a portable laboratory gives the opportunity
of doing the test both indoors and outdoors. Al-
though the indoor test could be repeatable by con-
trolling the domain conditions in the laboratory,
these two platforms would be helpful for outdoor
tests according to dynamic atmospheric conditions
and enables the users to compare thermal analysis
of two designed screens at the same time in similar
conditions. Moreover having two similar test boxes
in the platform makes it possible to go through the
optimization process by comparing two similar ge-
ometries that differ just by changing one parameter,
for instance porosity or 3D pattern shapes.
The other contribution of this platform is having
272 | eCAADe 32 - Design Tool 1 - Volume 1
Figure 4
Sample of data
visualization of
temperature
sensors with
MATLAB.
a plan for a design process and bringing the feed-
back to the design stage, which is beneficial to make
better-informed design decisions.Moreover, it is ca-
pable of visualizing both Surface temperature and
the gradient mean temperature behind the screen.
The size of this platform is adjustable, which al-
lows users to make the various scale models for test-
ing and do tests even in the real site location. Further-
more, by using Arduino, it is possible to add different
kinds of sensors like light and humidity sensors, to
explore other natural phenomena that may have an
influence on the thermal performance of facades un-
der simulated conditions. It depends on the aims of
the studies.However getting very precise and accu-
rate results needs the engagement of other sciences
with the platforms.
CASE STUDY
A series of prototypes has been implemented out of
white polypropylene sheets to illustrate the potential
and limitations of the MDPS platform for early stage
of façade design.
Reflecting on a series of prototyping techniques
like layering and casting, we used origami folding as a
good technique for this stage of the studies, offering
potential of rapid prototyping without needs of addi-
tional material like glue. Consequently, this method
offers homogeneous surfaces for the heat studies
The design started with ideas of having two clus-
Figure 5
First and second
clusters of patterns
were studied in the
mixed digital -
physical platform.
ters of patterns for a scenario of designing a sec-
ond screen for the north façade of a building in Mel-
bourne :
a) Smooth surfaces with opening
b) Patterned or 3D shape surfaces without open-
ings (Figure 5)
For the first cluster, 'smooth surfaces with open-
ings', several smooth patterns with different shapes
and size of openings were studied. The results of this
stage were too complicated to inform the next gen-
eration of the design. Therefore, it was concluded
that for minimizing this complexity we should have
same area of opening for the next generation of stud-
ies. Reflecting on the second cluster of studies, it was
concluded that one origami shape is capable of gen-
erating at least two different patterns with different
thermal characteristics (figure 5).
The result from the second cluster in the physi-
cal platform led to the question: how could we have
variation of forms with different thermal character-
istics just by rotating the screens? This question in-
formed the next generation of forms designed to pro-
vide variation out of one shape. Through this explo-
ration, it was possible to play with one module to vary
the screen outcomes by rotating. Therefore, the next
generations of forms were designed by combination
Design Tool 1 - Volume 1 - eCAADe 32 | 273

of both opening and patterning for façade composi-
tion.
Eight different shapes were generated out of one
studied module through rotation and flipping the
module (Figure 6).The patterns evaluated and ana-
lyzed in terms of the defined criteria including self-
shading , temperature behind the surfaces , surface
temperature and incident solar radiation. The results
are summarized in figure 6. The results of studies
in the mix digital and physical simulation platform
showed the capability of having one module with
different thermal characteristics. By testing the de-
signed patterns in the MDPS platform the third gen-
eration were designed based on this capability of
controlling more heat in the hottest days and block-
ing less heat during winter time(Figure 7) .
The final outcome was a screen consisting of
horizantal louvers capable of rotating 180 degree for
changing the shape based on the season. The studies
showed that one side is good for winter as it receives
more heat while the other side would be better for
summer because it could block more solar radiation
and provide more selfshading areas.
274 | eCAADe 32 - Design Tool 1 - Volume 1
Figure 6
Shows eight
different shapes
that were
generated out of
the initial module in
the third cluster.
CONCLUSION AND FUTURE WORK
This paper reports on exploration of an approach for
measurement, evaluation and feedback in the de-
sign workflow through a mixed digital -physical sim-
ulation platform (MDPS) based on objectives of the
study. For this purpose, it introduces a new way
of analyzing thermal performance of double skin fa-
cades by using temperature sensors, Arduino con-
trol system, post visualization with MATLAB and en-
ergy simulation with grasshopper plugins and Vasari.
The main aspects of this proposed workflow, in which
both physical and digital simulation are integrated, is
the design of a thermal performance feedback loop
as an integral part of the design process to facilitate
driving forms based on better understanding of geo-
metrical patterning thermal performance. This leads
to the design of a physical platform to provide qual-
itative and quantitative information. From the re-
search that has been carried out, it is possible to con-
clude that having such a mixed Digital-Physical sim-
ulation platform enriches the feedback loop during
the form finding process. Each of them covered dif-
ferent aspects of the studies based on their capac-
ities. Although there are some limitations and dis-
advantages for both physical and digital simulations,
Figure 7
Two samples from
the third cluster
with more details.

Design Tool 1 - Volume 1 - eCAADe 32 | 275

 by considering both of them in the early stages, we
could get comprehensive results. Such a platform
is user friendly for investigators and they can get
rapid feedback on their designed surfaces just by at-
taching their sample models to the existing platform,
that may lead to an increased ability of understand-
ing the thermal performance of the design modules
while applying user-friendly digital software in par-
allel .Having physical exploration with real time do-
main beside digital simulations improved our under-
standing of heat as a dynamic phenomenon. All of
these efforts are for exposing environmental param-
eters to the design process of the early stage.
For the case study series, the designed surfaces
were simulated to study their thermal performance,
with the aim of optimizing and refining patterns to
minimize the mean surface temperatures, incident
solar radiation and the temperature behind the sur-
faces during hot summers of Melbourne.
Future work will involve using additional sensors
and developing the platform for more accurate tests
and results, including outdoor and indoor tests.
Lewis, D and Nobel, P. S. 1977, 'Thermal energy exchange
model and water loss of a barrel cactus, Ferocactus
acanthodes', Plant Physiology, 60(4), pp. 609-616
Nobel, P.S. 1978, 'Surface Temperatures of Cacti--
Influences of Environmental and Morphological Fac-
tors', Ecology, 59(5), pp. 986-995
Susorova, I., Angulo, M, Bahrami, P and Stephens, B 2013,
'A model of vegetated exterior facades for evalua-
tion of wall thermal performance.', Building and Envi-
ronment, 67, pp. 1-13
Wong, N. H, Kwang Tan, A. Y, Chen, Y, Sekar, K, Tan, P.
Y and Wong, N. C 2010, 'Thermal evaluation of ver-
tical greenery systems for building walls. Building
and environment', Building and environment, 45(3),
pp. 663-672
[1] http://www.soa.utexas.edu/csd/research/experimental-
research

ACKNOWLEDGEMENT
The authors would like to thank their colleagues
Daniel Prohasky and Mani Williams for technical as-
sistance.

REFERENCES
Burry, J, Salim, F, Williams, M, Pena De Leon, A., Sharaidin,
K, Burry, M and Nielsen, S 2013 'Understanding heat
transfer performance for designing better façades',
ACADIA 13: Adaptive Architecture, pp. 71-78
Eumorfopoulou, EA and Kontoleon, KJ 2009, 'Exper-
imental approach to the contribution of plant-
covered walls to the thermal behaviour of build-
ing envelopes', Building and Environment, 44(5), pp.
1024-1038
Fathy, H 1986, Natural energy and vernacular architecture:
principles and examples with reference to hot arid cli-
mates, University of Chicago Press
Hensen, J, Bartak, M and Drkal, F 2002, 'Modeling and
simulation of a double-skin façade system', ASHRAE
Tansactions, 108(2), pp. 1251-1259

276 | eCAADe 32 - Design Tool 1 - Volume 1

Fuzzy computing for layout design in ill-defined, uncertain
spaces
Asli Cekmis
Istanbul Technical University
http://akademi.itu.edu.tr/cekmis/
cekmis@itu.edu.tr

Layout design has been supported by some computational tools, where fuzzy
systems have been approved as an appropriate method to handle uncertainty in
the early design stage. In this paper, a new mathematical model depending on the
fuzzy logic and sets theory is proposed to assist in layout design. The model
distinctly deals with spatial uncertainty in open planned designs, where there is
no clear layout configuration or definite patterns of usage. The model calculates
the possibility of occupancy according to space, function and user related
parameters and logical rules. It also visualises the architectural plan as being
comprised of sub-spaces formed by the distribution of those possibilities.
Sub-spaces are characterised as "Fuzzy Architectural Spatial Objects" (FASOs).
As a result, layouts are represented as an accumulation of FASOs showing a
certain inhabitation pattern. Various layouts can be generated within the identity
of FASOs. Architects can evaluate the layouts and propose new ones by
organising the FASOs on the plan and considering their relations. After
describing the model the paper demonstrates an application which aims to design
a proper layout for a major exhibition hall in Istanbul.

Keywords: Spatial uncertainty, Open-plans, Inhabitation patterns, Layout

design, Fuzzy Architectural Spatial Objects (FASOs)

INTRODUCTION and relevant morphological theories, agent-based

Spatial uncertainty is prevalent in fluid and contin-
uous spaces, where an entire space is occupied in
a holistic way, and multi-functionality plays an im-
portant role. It is experienced in free and open plan
designs that are not divided into cell-like rooms by
doors or walls. This lack of clear spatial configuration
makes the task of modelling inhabitation patterns
complex and difficult. The few equivalent methods
to handle spatial uncertainty include space syntax
models, simulations of gaze dynamics, behaviour
maps, and GIS-tracking and GPS. Recently, a new ap-
proach depending on fuzzy computing was intro-
duced by Cekmis et al. (2014) to predict and sim-
ulate occupational behaviour in response to higher
levels of uncertainty. Seeing open plans as an ensem-
ble of various temporary 'sub-functions' and their
virtual 'sub-spaces,' "Fuzzy Architectural Spatial Ob-
jects" (FASOs) was devised to represent sub-spaces.

Design Tool 1 - Volume 1 - eCAADe 32 | 277

 FASOs were characterised as fuzzy subsets within
complex environments. A FASO was a spatial en-
tity that is comprised of persons whose possibility of
presence changes from 100% (completely present)
to 0% (completely absent). By using fuzzy reason-
ing, the proposed model was capable of calculating
and visualising the FASOs on plans according to cer-
tain parameters; thereby mapping the possible spa-
tial usage in ill-defined, uncertain spaces. The model
was applied to analyse the inhabitation patterns of
the "Dali Exhibition" in "Tophane-i Amire Exhibition
Hall," Istanbul, and its efficiency was tested against
observed behaviour.
The capacity of the model showing how an open
planned space will be occupied can be used in the
preliminary design stage to provide feedback to the
project at hand. Accordingly, the aim of this research
is to suggest the model as a layout planner in the de-
sign process. The model presents architectural plans
as the scatter of the FASOs to assist in layout design.
The model can generate numerous layouts for total
spaces based on usage prediction. Architects could
decide on one of them by evaluating the sizes and re-
lations of the FASOs, for any aspect such as functional
requirements. This property of the model is exercised
through a case study. The layout of the "Palladio Ex-
hibition" in "Tophane-i Amire Exhibition Hall" was de-
signed for the installation of the seventeen wooden
architectural models of the buildings by Andrea Pal-
ladio, the famous 16th century Italian architect.
Considering the uncertainty and ambiguity in
the early design stage, fuzzy modelling was used for
the floor plan designs. Koutamanis (2007) proposed
a digital equivalent of analogue sketching which al-
lows designers to register and manipulate imprecise
and uncertain information. The fuzzification of crisp
forms created more flexible and adaptable shapes by
expressing degrees of tolerance, incompleteness and
uncertainty. Bayraktar and Cagdas (2013) proposed
a digital design tool which allowed creating sketch-
like footprints. The user could create layouts made
of bubbles for the spaces needed. The bubbles were
moving dynamic parts which had fuzzy boundaries.
278 | eCAADe 32 - Design Tool 1 - Volume 1
The vague definition of the placement of the spaces
brought more opportunities in the design process.
Those approaches focused on not losing the uncer-
tain or creative language of sketches and concept
models in the preliminary design stage. The abstract
level of the layout idea was converted into a digi-
tal representation. However, the layout planning by
the model introduced in this paper does not follow
the architects' intentions but the patterns of inhabi-
tation. The model does not deal with uncertainty in
the way floor plans are designed but the spatial un-
certainty inherent in the floor plans of total spaces. In
that sense, the model creates 'layouts' for open plans,
where architects intentionally leave the interior con-
figuration unplanned, without sketches or any other
descriptive material, in order to create more adapt-
able and flexible systems for changing environments.
The paper first describes the notion of the FA-
SOs and introduces the fuzzy computational model.
Later, the layout planning capacity of the model is ex-
plained through an application.
THE MODEL AND THE FASOS
The novel approach uses the fuzzy logic and sets the-
ory as the appropriate method to deal with uncer-
tainty; in both the expression of sub-spaces as FASOs
and in assigning the possibilities of people involved
in those sub-spaces.
Fuzzy systems are ideal for expressing spatial
identity of open-plans, since fuzzy logic is based on
the notion of partial (relative) truth, and fuzzy sets
have imprecise boundaries different from crisp sets,
thereby allowing gradual membership. A total space
could, hypothetically, be seen as a fuzzy set where
each sub-space (and its sub-function) is a fuzzy sub-
set. The whole, two-dimensional space is 'the uni-
verse of discourse' of the fuzzy set. The smallest unit
of this space, which can accommodate a person, is an
element or member. Spatial units (centroids) are ob-
tained when the architectural plan is divided using a
50 by 50 cm grid. Centroids can reside in more than
one sub-space with different membership degrees (0
≤ x ≤ 1); consequently a sub-space can be comprised
Figure 1
A FASO
Figure 2
Topological
relations
of members with various differing degrees of be-
longing. This is the fundamental attribute of spatial
uncertainty, where no physical boundaries of sub-
spaces exist and people can fulfil multiple functions
within different sub-spaces simultaneously.
The sub-spaces in the model are defined as FA-
SOs, adopted from the concept of "Fuzzy Spatial Ob-
jects" (FSOs) which are used by geographers to ap-
ply the value of the multiple membership capacity
into the spatial entities. This conceptual tool is pro-
posed to identify architectural spaces with imprecise
or vague spatial attributes. The fuzzy objects are
comprised of fuzzy points, lines and regions. As the
most relevant, a FASO-region has three zones com-
prised of spatial units with different membership val-
ues in the fuzzy set: "the core" (x = 1), "the indetermi-
nate boundary" (0 < x < 1) and "the exterior" (x = 0);
which describes the possibilities of being wholly or
partly involved in the activity (sub-function), or being
totally disengaged from it. These zones of a region
are virtually constructed in the space. The borders
between the zones are not demarcated physically
but through conceptual separations that are flexible
and permeable to members. They support a free flow
of possibilities of involvement or engagement in the
space. So, the FASOs in space are neither stable in
size nor in shape; they are able to enlarge, dimin-
ish, move, merge, separate or disappear. They may
also transform into each other. The topological rela-
tions of the FASOs, including disjoint, touch (meet)
and overlapping, help to make deductions about the
use of space (Figure 1 and 2).
The model serves to calculate and visualise the
membership values in spatial units thereby reveal-
ing the FASOs on an architectural plan. Both phases
are run in MATLAB. Membership degrees are calcu-
lated through a fuzzy inference system of pre-set pa-
rameters and logical postulates which conceptual-
ize possible occupational behaviour. The method of
fuzzy reasoning is used, since spatial uncertainty is
not related to frequency or randomness (which re-
quires probability theory) but to the inherent com-
plexity and indeterminacy of human usage of space.
The parameters affect the membership degree
of a centroid by either increasing or decreasing the
possibility of a person's presence in that spatial unit.
They can be grouped into three areas: space, func-
tion and user. These are the main factors shaping the
formation of a FASO. These parameters are the inputs,
and predicates on which the logical postulates in the
calculation process are based.
The calculation side of the model is undertaken
Design Tool 1 - Volume 1 - eCAADe 32 | 279
 Figure 3
The FIS process

using a fuzzy inference engine; including a fuzzy in- major touring exhibitions (Figure 4). It has a rectan-
ference system (FIS) and, input data file which is the gular symmetrical plan scheme of 1150 m², which,
whole list of parameter variables of all the spatial despite its pillars, can be regarded as an open plan.
units. The FIS takes the input data which it pro- The Dali Exhibition contained 121 drawings hung on
cesses using fuzzy reasoning and set theory through display walls (Figure 5).
a series of control rules to calculate the output data;
Figure 4
the membership degree. The FIS involves five steps:
Tophane-i Amire
fuzzification, rules, inference process, aggregation
Exhibition Hall
and defuzzification. Figure 3 shows the flow of the
process of FIS.
When the overall calculation process is com-
pleted, a list of outputs of all centroids is produced
as a precursor to visualisation. After the outputs are
arranged on a matrix that corresponds to the spatial
units on the architectural plan, two-dimensional in-
terpolation is applied in order to provide a robust vi-
sualisation in terms of sensitive colour gradients and
higher image resolution. The image process of the
model makes the FASOs appear by representing their
core, exterior and intermediate values on a heat map.
In practical terms the colours describe the likelihood
of people being present in a sub-space. When the vi-
sualisation process is completed, the total space is il-
lustrated as an accumulation of the FASOs with dif-
ferent sizes and shapes indicating the probability of
In this application, 2849 centroids of both the habit-
inhabitation.
able and uninhabitable dots on the 37x77 grid were
The model mapping spatial usage provides a
identified. Four key parameters were selected; space
framework, as it is open to the addition of new rules
related parameters: the distance from the viewing
and parameters to suit specific conditions. As a case
point, the distance from the order line point, visual in-
study, the Dali Exhibition (from 23 December 2011 to
tegration, and a user related parameter: population.
26 February 2012) in Tophane-i Amire Exhibition Hall
Some parameters were deliberately not chosen; like
was analysed by using the model. Tophane-i Amire
the function related inputs; "the number of exhibits"
is a historical building in Istanbul and houses various
since it was fixed for the duration of the exhibition,

280 | eCAADe 32 - Design Tool 1 - Volume 1

 and "popularity" since all the pieces of art in the col- jects in functional terms, as three or more people
lection were prominent and on a par with each other. were involved in the same sub-function. Also, the
The fact that people chose to attend the exhibit was queuing zones blocked the view when they formed
also taken as a basis to ignore personal preference around display walls and interrupted the visitor flow
parameters. on the main circulation paths. By using photographs
and video recordings, the efficacy of the models was
Figure 5 tested. The predictive maps broadly replicated the
The Dali Exhibition observed usage of space. The model captured well
what was happening in the real space.
Figure 6
The results of the
Dali Exhibition

FIS was built as a four-input, one-output, and forty

five-rule analysis problem. For each centroid, it trans-
formed the crisp inputs to fuzzified inputs (fuzzifica-
tion: input to inputmf ), then to fuzzy outputs (evalua-
tion of the fuzzy control rules: inputmf to outputmf ),
and finally to the single crisp output (aggregation
and defuzzification: ouputmf to output). The list of
2849 membership degrees proceeded to visualiza-
tion. The visualisation processes were operated ac-
cording to the guideline issued earlier.
The exhibition space was then represented as
being comprised of FASOs showing the possible in- In sum, the pattern of inhabitation pertaining to
habitation pattern of that specific condition. The pro- higher density levels could be foreseen (and poten-
cess was repeated three times for the three visitor tially avoided) by modelling the FASOs and evalu-
population density values; low, middle and high (Fig- ating their topological relations. The disadvantages
ure 6). The other inputs remained constant, since the caused by the increased population could be elimi-
display layout did not change. As the result of three nated through minor adjustments, such as hanging
usage maps, the spatial arrangement of the Dali Ex- paintings with a wider gap between them or simply
hibition in Tophane-i Amire was efficient and effec- taking the fake column away to remove the corridors.
tive for a lower population density, but it was less Alternative layouts could also be modelled and eval-
so for much higher densities. The crowd began to uated before the installation. Following this notion,
undermine the proper engagement with the art ob-

Design Tool 1 - Volume 1 - eCAADe 32 | 281

 the next section will focus on the layout planning ca-
pacity of the model by analysing a different exhibi-
tion in the same hall.
THE MODEL FOR LAYOUT DESIGN
In this case study the fuzzy computational model is
advanced to support design decision on layout plan-
ning. The Palladio exhibition (from 29 November to
31 December 2010) in Tophane-i Amire was a collec-
tion comprised of seventeen wooden architectural
models (Figure 7). This exhibition was intentionally
chosen for this research. Unlike pictures that need to
be hung, the total space was maintained without any
additional separators or dividing walls. The objects
could be freely moved and arranged in the exhibition
space. So, spatial uncertainty was clearly apparent,
also the occupational behaviour became more com-
plex and difficult to foresee. However, the FASO of
each of the seventeen sub-functions is expected to
282 | eCAADe 32 - Design Tool 1 - Volume 1
be read clearly, while their influence on layout plan-
ning can be easily understood.
The creator proposed a spatial layout for the Pal-
ladio exhibition, that the architectural models of dif-
ferent sizes were placed in the entire space except Figure 7
the four corner areas where the screen equipment The Palladio
was located (Figure 8). The display objects seemed Exhibition
to be arbitrarily distributed on the plan, but the No17,
which was the largest exhibit, was solely placed in a
big axial partition. Additionally, the entrance was de-
signed at the opposite side from where the entrance
for the Dali Exhibition was.
In this application, the 42x97 grid (50 by 50 cm)
on the plan contained 4074 centroids. Three key
parameters were selected. The first one was the
space related parameter: the "Distance from the sub-
function" (D). The relative proximity to the object
would raise the possibility of a person being present
in the sub-space. This parameter formed the basic
and the same FASO for all exhibits without the ef-
fects of other inputs. "Visual integration" (V), the sec-
ond parameter, was also a space-related parameter.
Depending on the syntactic property of the space,
the pattern of integration showed how much a per-
son could see from certain points, those where move-
ment was more exploratory and desirable. So, the FA-
SOs residing in highly integrated zones of the space
would be much bigger in order to support a higher
possibility of presence. The last parameter: "Model"
(M) was a qualitative factor from the function related
set. Since the architectural models' sizes and details
were all different from each other, the 'length of time'
people spent involved in an activity and their pres-
ence in the relevant spatial unit would change. Some
parameters were deliberately not chosen. For exam-
ple, "Population" was not used this time, because the
objective was to create the FASOs for layout planning;
to see their abstract self-forms and relations, not how
they would change when the number of people in
the space increased.
The calculation side of the model was comprised
of FIS and input data file. The data file was the list of
parameter variables of all the centroids. The distance
Figure 8
The layout of the
Palladio Exhibition

Figure 9 • If D is near then membership degree is mf7

The VGA (high)

• If V is min then membership degree is mf1

(low)

• If M is less then membership degree is mf1

(low)
from the sub-function was identified as the distance
Figure 10 from a certain exhibit and measured by using Ar-
The membership cMap software. To obtain integration values Visibility
function plots Graph Analysis (VGA) was run in the UCL Depthmap
(MATLAB) application (Figure 9). The quality values of the ex-
hibits were scaled from the level 1 to 17; and the cen-
troids within the circle d = 300 of a certain exhibit
were affected by its quality.
In FIS, each input was defined as a fuzzy set
which was characterized by its membership func-
tions (MFs); this was the subset. As seen in Figure
10, the MFs for "D" were near, middle and far; for "V"
were min, average and max; and for "M" were less,
mean and more. The output was also translated into
the MFs from the lowest membership (mf1 = 0) to
the highest membership (mf7 = 1). 27 rules in total
were constructed by the combination of the individ-
ual logical postulates between each input and output
as seen in the following examples:

Design Tool 1 - Volume 1 - eCAADe 32 | 283

 After the overall calculation process was completed,
4074 outputs proceeded to visualisation. The listed
membership degrees of centroids were mapped to a
matrix that corresponds to the grid cells on the archi-
tectural plan. So, the fuzzy subsets (the FASO sets)
were numerically represented in the space. Later,
grid-based interpolation was applied to create inter-
mediate values between cells. The complete set of
values was assigned colours according to the mem-
bership degrees; 0 ≤ x ≤ 1.
First, each FASO was individually modelled to see
the sub-spaces independently (Figure 11). The FASOs
were demarcated by a circle d = 300; outside of this
284 | eCAADe 32 - Design Tool 1 - Volume 1
line was defined as "the exterior" where the member-
ship degree was x = 0. The zones of the core and the
indeterminate boundary within the circle were deter-
mined by the parameters: visual integration and the
quality of the exhibit. For example, the lowest and
highest ranked exhibits No.6 (rank 1) and No.17 (rank
17) have different FASO-regions within their limits.
The bigger area was 0.5 ≤ x (reddish) in No.17 FASO,
whereas it was x ≤ 0.5 (bluish) in the No.6; which
demonstrated the possibility of occupancy on those
spatial units. Visual integration also affected the for-
mation of the FASOs. For example, although No.7
(rank 14) and No.16 (rank 13) had very similar quality
Figure 11
17 FASOs
 values, the zones of the regions changes dramatically
due to their location in the exhibition space. The dif-
ference can also be read at the structure of a FASO
itself; like in No.7 and No.8, where the parts of the
FASOs coincided with a higher integration value red-
dened. Additionally, some FASOs such as the objects
belonging to No.13 and No.14 were cut since the ex-
hibits were very close to the walls and pillars.
Figure 12
17 FASOs altogether
Figure 13
The result of the
Palladio Exhibition
Figure 14
A new layout
In terms of spatial reserve of the sub-function
these exhibits were restricted. In Figure 12, the in-
dividual FASOs were gathered by decreasing their
opacity and converting into grayscale to understand
the initial topological relations. It was mostly seen
that the FASOs were overlapped, even to an extent
that the outside edge covered the other exhibit. For
example, the FASOs of No.13 and No.14, also No.8
and No.9 showed this relation. It was rarely seen
that the FASOs touched (met), like between No.6 and
No.10. In addition, none of the FASOs stood sepa-
rately; in a relation of disjoint.
To express the overall result, the integrated heat-
map of all seventeen FASOs was produced as seen in
Figure 13. The parameter "V" was constant for all cen-
troids. However, "D" was accepted as the shortest dis-
tance from any nearest sub-function. And in the pa-
rameter "M," the centroid affected by more than one
exhibit was assigned the value of the highest ranked
exhibit. The result shows that the cores of some FA-
SOs merged including No.13 and No.14; also No.8
and No.9. The penetration at the high values of the
indeterminate zone (the red and ochre areas) is much
more than the penetration at the low values of the
indeterminate zone (the blue and turquoise areas).
It can be concluded that there is a dense layout of
cramped spatial arrangement. Especially the middle
axis of the space is expected to be occupied without
allowing a proper circulation path. The dark blue ar-
eas, which are proposed as hardly inhabited, are clus-
tered rather than sprawling reasonably among the
FASOs.
As an alternative, a new layout for the Palladio Ex-
hibition was devised by keeping the main design ap-
proach proposed by the creator. Some of the exhibits
were moved slightly (such as No.4 and No.13) and
some of them were rotated (such as No.7 and No.12).
The aim here was to show how the minor changes
would result in different patterns, which could not
be realized without the FASOs map. In this new lay-
out (Figure 14) the FASOs didn't overlap on the cores
but on the lower values of the indeterminate zone as
the yellow and bluish connections were prominent
Design Tool 1 - Volume 1 - eCAADe 32 | 285
 on the plan. The density was reduced by the spacious
rearrangement of the exhibits, which would provide
a more comfortable movement in the space. Thus,
the dark blue vacant areas mostly disappeared. Since
the exhibits such as No.13 and No.14 were carried
away from the walls and pillars, their FASOs were re-
covered.
It is also possible to model and evaluate diverse
layouts. For example, a layout with a deterministic
itinerary can be proposed via a clear circulation path
in the exhibition space. Or another layout can be pro-
posed to balance the "V" and "M" parameters for the
FASOs; as putting the smaller architectural models
(low M) in the middle of the exhibition space (high V)
and vice versa. In sum, by changing the places of the
exhibits numerous FASO layouts that show possible
inhabitation can be produced. The creator or archi-
tect can select and apply one of them according to
certain criteria.
the designed FASOs. So a reverse process is initiated;
from FASOs housed in open plans to kinetic and flex-
ible spaces and buildings shaped by the FASOs.
REFERENCES
Bayraktar, Mehmet Emin and Cagdas, Gulen 2013 'Fuzzy
Layout Planner: A simple layout planning tool for
early stages of design', Proceedings of eCAADe 2013,
The Netherlands, pp. 375-382
Cekmis, Asli, Hacihasanoglu, Isil and Ostwald, Michael J.
2014, 'A computational model for accommodating
spatial uncertainty: Predicting inhabitation patterns
in open-planned spaces', Building and Environment,
73, pp. 115-126
Koutamanis, Alexander 2007, 'Fuzzy Modelling for Early
Architectural Design', International Journal of Archi-
tectural Computing (IJAC), 5(4), pp. 589-610

CONCLUSION
This paper introduced a fuzzy computation model as
a design decision tool for layout planning in uncer-
tain or ill-defined spaces. Layouts are conceptualised
by using fuzzy architectural spatial objects (FASOs)
which represent the probability of a sub-space be-
ing occupied, as a function of the features, objects
and people within the total space. That the model is
structured to be open-ended, to accommodate addi-
tional parameters and logical rules as required, vari-
ous layouts can be generated. The evaluation of in-
habitation pattern layouts is expected to assist ar-
chitects to develop more feasible and rational argu-
ments as well as to contribute to their design knowl-
edge.
This research advanced the FASOs as generative-
syntactic components in the early design stage,
which could be arranged for spatial layouts. In a vi-
sionary sense, these FASOs can also be utilized to
design the borders of a space; the outer shell of a
building. Instead of that the FASOs are delimited by
the walls of the building (as they cut in the exhibi-
tion space), the built form will transform according to

286 | eCAADe 32 - Design Tool 1 - Volume 1

A 3-Dimensional Architectural Layout Generation
Procedure for Optimization Applications : DC-RVD
Ioannis Chatzikonstantinou
Yaşar University
i.chatzikonstantinou@yasar.edu.tr

A procedure for generating 3-dimensional spatial configurations for optimization

applications, termed Dimension Constrained Rectangular Voronoi Diagram
(DC-RVD), is presented in this paper. The procedure is able to generate a
non-overlapping configuration of spatial units in 3-dimensional space, given a
string of real values. It constitutes an extension and adaptation of the
Rectangular Voronoi Diagram generating procedure, found in the work of Choi
and Young (1991). An extensive description of the procedure, with the relevant
pseudocode is included in the paper. The procedure is tested in a stochastic
optimisation-based decision support environment. Testing is done using a case
study of a medium-sized family house. The result indicate promising performance.

Keywords: Optimization, Layout, Representation

INTRODUCTION pre-existing rules of thumb and heuristics (Loemker

The problem of spatial configuration is concerned
with finding suitable locations for a set of interrelated
objects that meet design requirements and maxi-
mize design quality according to design preferences.
It is a general problem that applies in many fields
of science and engineering. Specifically for architec-
tural design, the problem is of central significance.
The spatial configuration of a building or complex is
a design aspect that crucially determines it's success.
However, the problem of spatial configuration
is very challenging. In order to tackle it, designers
have up until recently relied on heuristics and rules-
of-thumb, or guidelines developed through personal
experience, which helped them produce suitable lay-
outs in the most common design scenarios. While
such approaches may be helpful in practice, they do
not really address the problem adequately. The rea-
son is that the problem is irreducible to following
2006). As such, different, more general approach is
required.
Computational optimization techniques have
been applied to layout problems, as an alternative to
the traditional methods. A popular category of algo-
rithms are Stochastic Optimization (SO) algorithms.
SO algorithms have gained a lot of traction in the re-
search community, because of their flexibility and ef-
ficiency. SO Algorithms are able to deal with prob-
lems that we have very limited specific knowledge
about, are able to explore solution spaces that are
highly irregular and fragmented, and to handle mul-
tiple, conflicting goals and multiple constraints. This
features make them a good candidate for tackling de-
sign problems related to spatial layout, since it can
be proven that the objective space of layout prob-
lems is highly fragmented and fractal-like (Cagan et
al. 2002). On the other hand, the computational

Design Tool 1 - Volume 1 - eCAADe 32 | 287

 requirements of SO algorithms are high. However,
with today's increase in computing power, the use
of SO algorithms is justified if the problem at hand
is complex. Stochastic optimization includes cate-
gories of algorithms such as Genetic Algorithms (GA),
Simulated Annealing (SA), Partice Swarm Optimiza-
tion (PSO) and more.
Many SO algorithms work by iteratively improv-
ing upon a single or multiple solutions at the same
time. This happens through a continuous loop of
altering decision variables and evaluating resulting
solutions. While evaluation is performed on the ac-
tual design itself, optimization algorithms perform
variations by altering values in a serial representa-
tion of the solution, also known in Evolutionary Com-
putation terminology as Genotype. Each solution is
then generated according to these values, resulting
in what is frequently termed Phenotype. The na-
ture of the representation and the generative process
plays a significant role in the efficiency of the opti-
mization process. A representation language's ade-
quacy, as Cha and Gero mention, is a critically impor-
tant factor (Cha and Gero 1998).
AIM OF THE STUDY
Recognizing the importance of the layout-
generating procedure in spatial configuration prob-
lems, this study proposes a new procedure for gener-
ating 3-dimensional architectural layout configura-
tions, which is suitable for use in computational opti-
mization scenarios. The proposed method is termed
Dimension Constrained Rectangular Voronoi Dia-
gram (DC-RVD). The procedure allows generating 3-
dimensional non-overlapping configurations based
on rectilinear volumes, using an encoding compris-
ing of a series of real-valued parameters.
The proposed method extends and comple-
ments the procedure presented in Choi and Kyung
(1991) which has been termed Rectangular Voronoi
Diagram (RVD), and which itself is an adaptation of
the well known Voronoi subdivision method. The
RVD method tackles one issue that occurs frequently
when dealing with the spatial configuration problem,
288 | eCAADe 32 - Design Tool 1 - Volume 1
namely that of overlapping spaces. Instead of treat-
ing overlaps as a constraint, which turns out to be a
difficult problem to solve, RVD re-distributes the con-
tested space, allowing the overlap to be resolved.
The proposed procedure leads to a smooth ob-
jective function landscape that may be easier for an
optimization algorithm to traverse. Decision vari-
ables represent design parameters such as positions
and widths, which correspond to meaningful real-
world quantities. This allows meaningful interpreta-
tion and analysis of data resulting from an optimisa-
tion process.
The paper is structured as follows: In Section 3.
We present a review of approaches to the Building
Configuration problem, focusing on how layouts are
represented and generated. In Section 4., the RVD Al-
gorithm from Choi and Kyung is briefly outlined. In
Section 5., the adaptation of the algorithm for appli-
cation in building configurations is discussed. In Sec-
tion 6., the revised algorithm is presented. In section
7., the case study is presented. Section 8. discusses
the results and Section 9. concludes the study.
EXISTING WORKS
There exist a wide variety of works on the problem
of automatically identifying optimal architectural lay-
out configurations. The work of Lobos and Donath
(2010), presents a good overview of the most promi-
nent approaches. Here we are going to briefly men-
tion a few, focusing on the layout/configuration rep-
resentation.
Jo and Gero (1998), present a binary encoding
scheme, which encodes placement of spatial units
within a predefined boundary, by sequentially en-
coding their positions along a path. The method is
used to solve the Liggett problem (Liggett, 1985) of
locating departments within a multi-storey building.
The encoding they propose uses two bits to encode
the possible movements of a "cursor" to neighboring
cells in a grid. The spaces are placed sequentially as
the cursor moves on the path specified by the encod-
ing.
 Elezkurtaj and Franck (1999) make use of a
Genetic Algorithm to search for suitable ground
plans for architectural applications. The process of
genotype-phenotype translation is as follows: First,
the outline of the building is specified. Second, the
list of rooms to be fitted into the outline and the
proportions preferred are entered. Third, the func-
tional scheme of organization and access is specified.
While their work is novel, no details are given as to the
specifics of the encoding scheme.
In the work of Michalek and colleagues, (2002),
stochastic optimization is interweaved with user in-
put, in order to accommodate second-order criteria
and preferences. In the layout representation they
describe, rooms, hallways, doorways (Accessways),
and boundaries are all represented as combinations
of orthogonal rectangular Units. Units are repre-
sented as a point in space and the perpendicular dis-
tance from that point to each of the four walls. This
model has more variables than necessary to describe
the shape; however, it allows an optimization algo-
rithm to change the position of a Unit independently
without affecting its size. Although this model in-
creases the problem dimensionality, it offers a lot of
flexibility to make the best design moves at each step
of the optimization (Michalek et al. 2002).
Loemker (2006), has formulated the layout prob-
lem as a constrained optimization problem, with
the objective of maximizing areas, subject to con-
straints regarding satisfaction of spatial relations be-
tween units. The principle of the geometric model
adopted was the representation of rooms as rect-
angular units. The concept was similar to that of
Michalek et al. (2002), but differing in that the
representation adopted describes a rectangular unit
through a reference point, a length and a width di-
mension.
Kamol and Krung (2005) have used Mixed In-
teger Programming techniques to address the 2-d
architectural layout problem. They formulate func-
tional and dimensional constraints (such as size,
proximity and overlap) into linear functions, and
make use of a mixed integer programming solver,
GLPK (GNU Linear Programming Kit), in order to iden-
tify optimal solutions. They conclude with promising
results, however the requirement for linear functions
is constraining the expression of complex objectives.
The encoding of layouts is simply based on x,y origin
point coordinates, and widths and heights of spaces.
In the study of Yeh and colleagues, the architec-
tural space assignment problem has been addressed
by the use of Annealed Neural Networks. Yeh used
an Annealed Hopfield Network, that combines the
speed of convergence of a Hopfield net and the
global search characteristics of Simulated Annealing.
They apply their method on the problem of locating
a range of facilities in a hospital building. The prob-
lem includes multiple objectives and a constraint, but
these are handled in a single-objective fashion, as
such, a considerable amount of parameer tweaking
(e.g. weights, constraint penalties etc.) is required.
They conclude that while the algorithm is efficient,
future research is needed to address the issue of ob-
taining suitable parameter values. (Yeh et al. 2006)
THE RECTANGULAR VORONOI SUBDIVI-
SION METHOD
Choi and Kyung have discussed the Rectangular
Voronoi Diagram (RVD) method, and applied it to
the problem of digital VLSI circuit design, yield-
ing promising results (Choi and Kyung, 1991). The
method takes it's name partially from the well-known
Voronoi diagram. A Voronoi diagram is a method of
subdividing space into regions, where each point of
a region is closest to one of several predefined seed
points. In this method, the resulting regions may
have arbitrary shape, depending on the distribution
of the seed points. While the Voronoi diagram is be-
ing routinely used in architectural design to derive in-
teresting forms and structures, it's application in de-
riving architectural layouts is relatively limited, due to
the irregularity of the shapes resulting from the pro-
cess.
The characteristic difference of RVD from a clas-
sical Voronoi diagram is that, as a result of the RVD
process, regions of space of exclusively rectangular
Design Tool 1 - Volume 1 - eCAADe 32 | 289
 shape are produced. This is an important property,
because it makes it much more applicable to build-
ing layout design than the Voronoi diagram, which
produces arbitrary shapes.
The steps of the RVD method can be summarized
as follows:
1. A set of points P is considered in 2 dimensions,
each of which represent the center of mass for
a corresponding rectangle.
2. For each point in p, two subsets are created,
Sx and Sy, each of the remaining points ps is
placed in one of them according to the follow-
ing rules:
{
Sx if |px − pSx | > |py − pSy
(1)
Sy otherwise
3. The rectangular areas corresponding to
points are derived by determining the middle
line between each point p and the points clos-
est to it in all four directions. For edge points,
a predefined outline may be considered as
a border. As an example, the corresponding
rule for determining the left edge of all rect-
angular spaces may be staed as follows:
repeat for each element r {
Lr = Loutline
repeat for each element e in Sx{
Lm = left_midline(r, e)
Lr = max(Lr , Lm), if Ce < Cr
} design. It makes use of a compact, continuous vari-
} able encoding; it offers parameters with high seman-
where Lr, Le symbolize the left edges and Ce,
Cr the centers of the spaces.
The result of the process is a series of rectangular re-
gions that form subdivisions of space. The resulting
configuration needs to be bound by a rectangle in or-
der to obtain the limits of the outer regions. An ex-
ample is shown in figure 1. Essentially, the way the
RVD method functions eliminates the need for using
290 | eCAADe 32 - Design Tool 1 - Volume 1
overlap constraints for ensuring configuration feasi-
bility. The method instead constrains spatial units to
avid overlapping completely. This, of course, has a
constraining effect on the area of the spatial unit, but
such a constraint, is claimed in this study, is easier to
satisfy.
Figure 1
An arrangement
resulting from the
application of the
RVD procedure. The
cross symbols
represent the
centres resulting
from the x,y
coordinate
encoding.
In the study of Choi and Kyung (ibid. 1991), a
force-directed packing process was used to optimize
the layout with respect to the positions of the seed
points. In the present study, this process is substi-
tuted in favor of a Stochastic Optimization algorithm.
As can be seen from he results, this offers the required
flexibility for dealing with the complex objectives of
a building configuration scenario.
ADAPTATIONS AND EXTENSIONS
The RVD method presents several advantages that
make it attractive for application in building layout
tic value; and it requires few calculations for generat-
ing the configuration from the encoding. However,
there are also drawbacks to the method, namely the
possibility of gaps appearing within the arrangement
and the inability of generating configurations that
do not follow a rectangular outline, which should
be addressed if it is to be applied in architectural
design. Furthermore, the RVD method works in 2-
dimensions, but for architectural layout problems, an
 approach that considers 3-dimensional layout would of required decision variables for the description of a
be preferable. spatial unit by two, it's length and width. Reiterating
To address these shortcomings and make the the above example of finding the left border, with the
RVD method more applicable to architectural layout new rule:
problems, a new method, resulting from adaptations
repeat for each element r {
and extensions of the original, is proposed.
Bd ,r = Bd ,original
repeat for each element e in Sd {
Figure 2 Bd ,m = midline(r, e)
An example of Bd ,r = max(Bd ,r, Bd ,m), if Ce < Cr
applying the }
DC-RVD procedure, }
using spatial unit
dimensions to It should be noted here that constraints are not sym-
constrain the unit metric for each side of the space, as such it's position-
size. In this case, ing may be implicitly altered by the constraining pro-
encoding includes cess. An example of the procedure applied in 2-d is
x,y, coordinates of shown in figure 2. An example of overlap resolution,
the centres, as well including spatial segmentation, is available in figure
as width and height 3.
of the shapes. Secondly, the algorithm is modified to support
generation of three-dimensional configurations. The
Figure 3 intuition behind the proposed procedure is to con-
Spatial overlaps, RV strain each of the volumes only for those of it's neigh-
regions and bors with which a vertical intersection is present. By
resolution of doing so, the space is only constrained as much as
overlaps in the required in order for it not to overlap with it's neigh-
DC-RVD procedure bors. To achieve that, we create separate sets of
neighbors for each unit and perform the constraint
operation there:
Figure 4
Neighbor relations repeat for each element r {
in a simple case of Bd ,r = Bd ,original
vertically repeat for each element e in Sr ,d {
if vertical_intersection (r, e) {
distributed spatial
Bd ,m = midline(r, e)
units Bd ,r = max(Bd ,r, Bd ,m), if Ce <
,→ Cr
}
}
As a first modification, predefined element dimen- }
sions are incorporated in rule #3. The new rule will
only constrain spaces' dimensions if the closest mid- where vertical_intersection would be a function to
line to a neighboring space is found to be closer that detect whether the two spaces share some interval in
the space's original value for that dimension. The in- the z (vertical) axis. This way of extending the RVD al-
clusion of dimensioning information extends the set gorithm ensures that spaces of arbitrary dimensions
(albeit still rectilinear) may be accomodated, e.g. ver-

Design Tool 1 - Volume 1 - eCAADe 32 | 291

 tical circulation spaces, such as staircases, double-
height spaces etc. The formation of neighbour rela-
tionships during the process is illustrated in figure 4.
Thirdly, the issue of leftover space within the ar-
rangement is addressed. The issue may be either ex-
plicitly treated (by employing some post-processing
step), or left to be treated implicitly as part of the op-
timization process, considering that leftover spaces
negatively affect the performance of a configuration.
With explicit treatment, leftover spaces may be re-
distributed to neighboring spatial units. However, in
the case study that follows, leftover spaces are left to
be treated implicitly during the optimization. It turns
out that such an implicit treatment offers satisfactory
results.
OVERVIEW AND PSEUDOCODE OF PRO-
POSED ALGORITHM
Having summarized the extensions to the RVD algo-
rithm, we may now write down the complete pro-
posed algorithm in the form of pseudocode:
repeat for each element r {
repeat for each element e excluding
,→ r {
Sr ,l += e, if rx - ex > |ry - ey|
Sr ,r += e, if ex - rx > |ry - ey|
Sr ,b += e, if |ex - rx| < ry - ey
Sr ,f += e, otherwise
}
} Figure 5
repeat for each direction d (among
,→ left (l), right (r), front (f) and
,→ back (b)) {
repeat for each element r {
Bd ,r = Bd ,original
repeat for each element e in Sr ,d
,→ {
if vertical_intersection(r, e) {
Bd ,m = midline(r, e)
Bd ,r = max(Bd ,r, Bd ,m), if Ce
,→ < Cr
} lengths of the spatial units. Separate function assign-
} ment was not used. The vertical position of the spa-
} tial units was encoded as a real-valued variable, so
}
292 | eCAADe 32 - Design Tool 1 - Volume 1
After the algorithm has finished, we end up with a
configuration of non-overlapping spaces with areas
equal or less to the predefined ones. As a final step,
assignment of functions to spaces may be allowed,
based on, e.g., sorting according to a separate list
of real-valued decision variables, one for each space.
In this case, the number of decision variables is in-
creased by one for each space. Function assignment
allows in some cases radical alterations to take place
in the design (Merrell et al. 2010).
APPLICATION IN A STOCHASTIC
OPTIMIZATION-BASED DECISION-
SUPPORT SYSTEM
In order to investigate applicability and performance,
the proposed mehod was used to tackle a design
problem: that of designing a simple detached family
house, with basic design goals. The brief includes a
total of seven spaces: a living room, a kitchen, a bath-
room, a hall and three bedrooms. Site constraints are
minimal. The site area is 15x15 meters. The objec-
tive is to select appropriate positions and sizes of the
spatial units, so that their area is maximized, subject
to constraints regarding connectivity and proximity
of the spaces, as well as minimum areas. The list of
spaces, desired areas and connectivity requirements
are shown in figure 5.
Labels, minimum
and maximum
areas and
adjacency matrix
for the spatial units
included in the case
study
A Genetic Algorithm was used for optimisation.
The decision variables were the positions, widths and
 they could be positioned anywhere between floors. The results that were obtained did correspond to
Three objectives were formulated. The first one is to reasonable configurations. The final set of nondom-
maximize the area, prescribed by the problem defini- inated solutions included mostly variations of a sin-
tion, for three main spaces: Living Room, Maser Bed- gle configuration; i.e., not the whole spectrum of fea-
room and Kitchen. The second one was to minimize sible and well-performing solutions was attained. In
cost. The third one to maximise either proximity or order to improve the diversity of solutions, a different
separation between desired spaces. strategy was employed: Instead of a single run with
The system has been modelled in the Grasshop- a large population and many generations, the results
per parametric design program. DC-RVD was mod- of shorter runs of smaller populations, with restarts
elled using simple mathematical and geometrical between them, were combined, i.e. an archive of
components. The objectives and constraints were non-dominated solutions was established from the
also modelled in the same way. A diagram of the results of multiple runs. Indeed, in this case, di-
parametric model is available n figure 6. For opti- verse spatial configurations complemented the non-
mization, the NSGA-II Genetic Algorithm was used. dominated set. Three of the resulting configurations
NSGA-II is a multi-objective constrained genetic al- from different runs, can be seen in figure 8.
gorithm, developed by Deb and colleagues (Deb et
Figure 6
al. 2002). It is a well established optimization algo-
Overview of the
rithm that has seen widespread use in a wide spec-
implementation of
trum of applications. An implementation of NSGA-II
the procedure in
for Grasshopper, Lotus, implemented in the Chair of
Grasshopper. A.
Design Informatics in the Faculty of Architecture, TU
Parameters, B.
Delft, by I. Chatzikonstantinou and Dr. M. S. Bitter-
Generation of
mann was used. Details of the implementation are
rectangles, C.
discussed in the work of Chatzikonstantinou (2011).
DC-RVD procedure,
The model was left to execute on an Intel Core i5 PC,
D. Evaluation of
with 4GB RAM, running the Microsoft Windows 7 Op-
performance
erating System.
metrics
Figure 7 DISCUSSION
The set of non The optimization process was initially carried out
dominated with a population of 300 individuals, for a predeter-
solutions after 120 mined amount of 200 generations. It was observed
generations, using that after about 100 generations, no significant im-
NSGA-II with a provement of the solutions can be observed. The op-
population of 300 timization run starts with infeasible solutions for the
individuals. Larger first 15-20 generations. After this point, feasible solu-
dots indicate better tions start to appear. As a result f the process, a Pareto
performance in front was generated about 60 generations after start-
terms of spatial ing. Smaller improvements to the solutions were be-
relations. ing made after this point, but no radical changes oc-
cur. A graph showing the distribution in objective
space of the non dominated solutions after 100 gen-
erations can be seen in figure 7.

Design Tool 1 - Volume 1 - eCAADe 32 | 293

 Figure 8
Layout diagram of
two of the resulting
solutions

Figure 9
Layout diagram of a
solution with four
half-floors

294 | eCAADe 32 - Design Tool 1 - Volume 1

Figure 10
Volumetric study,
resulting from the
arrangement in
figure 9
In order to further evaluate the potential of the pro-
cedure, a comparison was made with a simple proce-
dure of placing unmodified spatial units, defined by
their positions and size, and including their overlap
as an extra constraint, alongside the ones descibed
earlier. The optimisation process was confiured with
the same parameters, i.e. population of 300 individ-
uals. It was observed that in the latter case, the opti-
mization process was not able to converge to a feasi-
ble solution at all, after 80 generations. In contrast to
that, the proposed procedure was able to reach the
first feasible solution on average at the 24th genera-
tion. Not only that, but in the case of the unmodified
procedure, the solutions converged to a very specific
area of the objective function space. This gives a hint
that, even if a feasible solution were to be found, it
would not be possible to improve it much further.
We observed that distribution of spatial units in
the resulting solutions is generally according to the
requirements, and in some cases demonstrates in-
novative arrangements. However, it should also be
noted that in some cases, the results after several
generations are unsatisfactory, leading to the con-
clusion that in those cases, the process is trapped in
local optima. An example of an interesting result is
shown in figure 9. It represents a design with four
half-floors. In this solution, the Master bedroom and
bathroom take the upper level, together with one of
the bedrooms. The living room and kitchen take the
mid-level. The other bedroom takes the lower level.
Since in this solution no requirement for proximity
between bedrooms was specified, it was considered
as a good solution. A volumetric study based on this
solution was performed as an indicative next design
step, resulting in the model of figure 10.
An optimization run of 100 generations with a
population size of 150 individuals last approximately
12 minutes. However, we should note that the pro-
cess, including generation and performance evalua-
tion, was modelled using components found in the
Grasshopper parametric Design program. This type
of implementation is rather slow compared to code.
As such, with a code-based implementation the exe-
cution time would drastically drop.
CONCLUSION
In this paper, a novel procedure for generating 3-
dimensional architectural configurations for opti-
mization applications has been proposed, termed Di-
mension Constrained Rectangular Voronoi Diagram
(DC-RVD). The method produces 3-dimensional, non-
overlapping configurations of spatial units, based on
an array of real-valued variables. It is based off an
adaptation of the Voronoi subdivision, namely the
Rectangular Voronoi Diagram (RVD). The procedure
is using a string of values that represent real-world
layout parameters, such as positions and dimensions.
This facilitates elaboration of the knowledge gener-
ated by optimisation processes.
DC-RVD has been tested in an Evolutionary
Design Tool 1 - Volume 1 - eCAADe 32 | 295
 Computation-based decision support environment, Jo, J and Gero, J 1998, 'Space layout planning using an
built in Grasshopper. As a case study, the design of evolutionary approach', Artificial Intelligence in Engi-
a medium-size detached family house has been ad- neering, 12 (3), p. 149–162
Kamol, K and Krung, S 2005 'Optimizing Architectural
dressed. The algorithm produces interesting results,
Layout Design via Mixed Integer Programming', Pro-
although in many cases the presence of local optima ceedings of the 11th International CAAD Futures Con-
is evident. The 3-dimensional representation allows ference, pp. 175-184
some novel features to be present, such as vertical Liggett, RS 1985, 'Optimal spatial arrangement as a
circulation cores and spaces with vertical relation- quadratic assignment problem', Design Optimiza-
ships, as well as innovative configurations, interme- tion, 2, pp. 1-40
Lobos, D and Donath, D 2010, 'The problem of space
diate level placement of public spaces.
layout in architecture: A survey and reflections', Ar-
At it's present state, the tool can be used as a de- quitetura Revista, 6, pp. 136-161
sign aid for inspiration, as it allows insight into so- Loemker, TM 2006 'Designing with Machines: solving ar-
lutions that otherwise may not have been consid- chitectural layout planning problems by the use of a
ered. There are certainly numerous perspectives for constraint programming language and scheduling
improvement in this work. Improvement is possible algorithms', Proceeding of The Second International
Conference of the Arab Society for Computer Aided Ar-
in the evaluation of configurations, so that soft de-
chitectural Design, pp. 88-106
sign aspects related to e.g. perception, such as, e.g., Merrel, P, Schkufza, E and Koltun, V 2010, 'Computer-
those described in the work of Bittermann (2009) may generated residential building layouts', ACM Trans-
be included. Finally, there is ample room for more ex- actions on Graphics, 29 (6), p. 1
tended trials, with building configuration problems Michalek, J, Choudary, R and Papalambros, P 2002, 'Ar-
of higher complexity. chitectural layout design optimization', Engineering
optimization, 34 (5), pp. 37-41
Yeh, I 2006, 'Architectural layout optimization using an-
REFERENCES nealed neural network', Automation in construction,
Bittermann, MS 2009, Intelligent Design Objects (IDO): 15, pp. 531-539
a cognitive approach for performance-based design,
Ph.D. Thesis, TU Delft
Cagan, J, Shimada, K and Yin, S 2002, 'A survey of com-
putational approaches to three-dimensional layout
problems', Computer-Aided Design, 34, pp. 597-611
Cha, MY and Gero, J 1998, Shape pattern representation
for design computation, Working Paper
Chatzikonstantinou, I 2011, Evolutionary Computation
and Parametric Pattern Generation for Airport Termi-
nal Design, Master's Thesis, TU Delft
Choi, SG and Kyung, CM 1991 'A floorplanning algo-
rithm using rectangular Voronoi diagram and force-
directed block shaping', IEEE International Confer-
ence on Computer-Aided Design Digest of Technical
Papers
Deb, K, Pratap, A, Agarwal, S and Meyarivan, T 2002,
'A fast and elitist multiobjective genetic algorithm:
NSGA-II', IEEE Transactions on Evolutionary Computa-
tion, 6 (2), pp. 182-197
Elezkurtaj, T and Franck, G 1999 'Genetic algorithms
in support of creative architectural design', eCAADe
1999 Conference Proceedings, pp. 645-651

296 | eCAADe 32 - Design Tool 1 - Volume 1

Cellular Automata as a learning process in Architecture
and Urban design
Mads Brath Jensen1 , Isak Worre Foged2
1,2
Department of Architecture, Design & Media Technology, Aalborg University
1,2
www.create.aau.dk
1,2
{mbje|iwfo}@create.aau.dk

This paper explores the application of cellular automata as method for

investigating the dynamic parameters and interrelationships that constitute the
urban space. With increasing aspects needed for integration during the
architectural and urban design process with the relations between these aspects
growing in parallel, complexity of the design process and design solution
increases. Additionally, aspects and relations are of a transformative character in
that they change over time and therefore construct a time-based condition for
which problems are presented and solutions are sought. An architectural
methodological response to this situation is presented through the development of
a conceptual computational design system that allows these dynamics to unfold
and to be observed for architectural design decision taking. Reflecting on the
development and implementation of a cellular automata based design approach
on a master level urban design studio this paper will discuss the strategies for
dealing with complexity at an urban scale as well as the pedagogical
considerations behind applying computational tools and methods to a urban
design education.

Keywords: Computational design, Cellular automata, Education, Design

exploration

INTRODUCTION proving not only the quality of the students' design

Understanding the evolving complexities of the tem-
porary city is a crucial challenge facing architectural
and urban designers. Being able to deal with and re-
spond to the many dynamic parameters and inter-
relationships that constitute the urban space could
enhance the final design concepts of the designer.
Strengthening such skills could be achieved by im-
plementing computational design tools in the edu-
cation of architects and urban designer thereby im-
projects but also creating a more transparent learn-
ing process and a heightened understanding of the
urban design field.
Managing urban growth and handling the dy-
namic mechanisms of an urban site has put the tra-
ditional master plan under great pressure revealing
its inadequate formal layout and inflexibility (Popov
2001). This has lead to increased research in the field
of computational design where generative design

Design Tool 1 - Volume 1 - eCAADe 32 | 297

 systems has been explored due to their ability to deal
with mass amounts of information and parametric re-
lationships. Most work in this field has focused on
complex systems either demanding users to have a
high level of scripting or programming knowledge or
ending up with a limited level of user-customization -
both extremes often with design solutions as the pri-
mary goal and applicable only to students with a cer-
tain knowledge and expertise in the field of compu-
tational design.
The research forming the basis for this paper
has adopted a design approach based on John
Conway's two-dimensional cellular automaton (CA)
called the 'Game of Life'. Integrating this discrete
model in the user-friendly parametric environment of
the Grasshopper plug-in for Rhinoceros 3D has per-
mitted the construction of a generative design sys-
tems that can deal with explicitly described design
rules and allow design students to investigate time-
based urban design concepts.
CELLULAR AUTOMATA
Cellular Automata (CA) invented by the mathemati-
cians John van Neumann and Stanislaw Ulam and
continuously explored and developed by another
mathematician Stephen Wolfram contains the un-
derlying framework for these studies in that a compu-
tational system can be devised which heavily utilizes
computational parallism, considered a very powerful
computational technique to generate different time
based solutions from different input rules and axiom
conditions. John Conway in his 'Game of Life' ex-
plored the development of CA as a continuously up-
dated system, which illustrate the dynamics of inter-
acting and interdependent actors in a system.
CA has for its general system setup, but pow-
erful processing abilities, been applied to very dif-
ferent scientific inquiries from simulation of fire de-
velopment, biological modelling, disease outbreaks
(Huang, Sun, Hsieh, & Lin, 2004) to urban sprawl
(Batty, 1997). The latter has illustrated different ap-
proaches to simulating complex urban development
through the DUEM projects, generating various large
298 | eCAADe 32 - Design Tool 1 - Volume 1
city developments.
Architectural propositions using CA as a method-
ology has been presented among others by the office
Kokkugia and Object-E, with both addressing large
scale organisations, with what appears to be classi-
cal two-state CA, defined by 1 and 0, also described
as On and OFF or lastly as LIVE and DIE in the context
of Conway's 'Game of Life'.
The CA framework
The CA framework constructed through this research
is based on a Moore-neighbourhood of r = 1 dimen-
sion, meaning it registers the eight cells touching the
center cell either on its four sides or in its four cor-
ners, see figure 1 & 2. Additionally it surveys the en-
tire cell space to detect if a described global thresh-
old is met. The reason for this contextual awareness
organization is twofold, 1) to only address immedi-
ate local actors for reasons of design rules with lower
complexity compared to neighbourhoods of higher
r-dimensions, and 2) to address if some states are
overly dominant allowing a negative feedback func-
tion stabilizing the design system at a given desired
condition set by the designer.
Figure 1
Moore
neighborhoods for
ranges r=0, 1, 2, and
3 with number of
neighbors in
parenthesis.
The 2-dimensional layer in which the develop-
ment occurs is extended vertically, by a method
where CA rule-sets are described as a 'vertically ex-
tended Moore neighbourhood', creating a rule set
described as a 3-dimensional organization, but com-
puted in 2-dimensions. This allows spatial constructs
and diverse descriptions of states such as higher
building mass, higher vegetation, e.g. trees, rather
than lower plants. The development of a spatial or-
ganization is integrated in the rules described by the
designer.
Figure 2
a) Moore
neighbourhood, b)
vertically extended
Moore
neighbourhood
and c) cell space
The model is developed with a system boundary
condition that terminates states if they are in direct
contact. This is chosen over a 'toroidal shape' method
in which a cell is wrapped to the first cell in the other
end of the grid making the model spatial continu-
ous. This makes the definition of the cell space im-
portant and closer aligned with real life conditions of
contained spaces.
The complexity of the model can be described as
a relation between the number (s) of cells including
the centre cell, and the number (k) of states available,
in the following equation, kˆkˆs, lifting each value to
the power of its previous. With a 10 state system, with
9 cells, the following CA solution space can be de-
scribed as 10ˆ10ˆ9.
Constructing an intuitive computational system
that enables design students without prior program-
ming experience to explore a recursive rule-based
system implies for educational considerations and
the establishment of a new educational framework.
Due to its intuitive and open-ended configuration
Grasshopper® (GH), a graphical algorithm editor for
Rhinoceros, was utilized as the platform in which to
construct a CA-based design tool which were to be
handed to the students at the commencement of the
design studio. Before building the definition in GH it
was imperative that the design tool could facilitate
an immediate adaptation into the explorative work
of students with little or no experience in GH or com-
puter programming. As a result the final definition
was constructed in such a way so that the student
needed only to interact with a few groups of key com-
ponents: the initialization of the cell space (includ-
ing cell space resolution, selection of initial cells with
corresponding states, and establishment of bound-
ary curves), constructing the rule sets (using if/else
conditional statements to ex. change the state of
a cell depending on the states of its neighbouring
cells), and lastly to extract desired data concerning
the performance of the system (count of different cell
states, increase/decrease based on previous steps,
etc.). Due to the generic configuration of a rule set
based on only a few simple components in GH it was
possible to re-use this rule set configuration to estab-
lish a large variety of simple rules all acting on the re-
sults of the previous rule sets.
Grasshopper doesn't by default allow for recur-
sive feedback loops but with the introduction of the
Opensource Hoopsnake component, created by Yan-
nis Chatzikonstantinou, it is possible to escape its
otherwise linear data flow. As an advantage for the
exploration and understanding of the evolution of
CA the Hoopsnake component allows the user to
control the completion of each successive loop in
a stepwise fashion while simultaneously storing the
data for each recursion for use in further analytical
work and cross-population comparisons.
METHOD
In order to utilize the constructed CA framework a
method was established that would allow design stu-
dents with none or little experience in computational
Design Tool 1 - Volume 1 - eCAADe 32 | 299
 design to use the CA-system as an open exploratory
playground allowing experimental studies in urban
and architectural design.
The main goal of the design method was en-
abling students to use the computational system as
a learning process with emphasis on detecting aris-
ing design problems rather than searching for design
solutions. To achieve this, the method supported a
transparent process where the outcome of each iter-
ation was visible to the design student while simul-
taneously allowing the implementation of complex
and integrated mechanism for design exploration.
Adopting the CA framework also required a con-
version from a programming terminology dealing
with binary data in a two-dimensional matrix to an ar-
chitectural terminology generating a volumetric ma-
trix for spatial organisation of urban programmes -
a critical step in making the method accessible to
architectural students and letting them explore and
unfold the architectural language and diversity into
a binary data structure. The conversion allowed for
an examination of design rules that adopted the ter-
minology of the loneliness/overcrowding rule-set of
traditional CA (Shiffman 2012) enabling students to
gradually construct and implement contextual pa-
rameters by setting up simple rule-sets and to study
the developed relations and complexities of the de-
sign project. To further inform the design system the
rule-sets in this method have been expanded from
those of the traditional CA systems, where an anal-
ysis of very local criteria (the state of the cell's eight
neighbours) form the basis of future actions, to also
deal with global thresholds rendering it possible to
inform the system when a certain design vision has
been met.
Applying the constructed design system to an
urban design workshop that explores design con-
cepts evolving over a fixed 20-year time span made
the time-based evolution of traditional CA directly
applicable and enabled the design students to
explore time-based design proposals allowing for
strategic planning across the fixed time period. Rule
sets were described by the designer, applying in
300 | eCAADe 32 - Design Tool 1 - Volume 1
each cell a given programmatic state, such as vege-
tation, void, infrastructure, building, but also meta-
programmatic states, such as investor and creative
class, understanding not only a state as a physical
object, but also as a conditions that influence other
states by their potential actions. Additionally, during
the continuous step by step process, all activities for
each time step is recorded and described in graph re-
lations detecting the development of the spatial or-
ganization in order to understand consequences of
a devised design rule set and the axiom conditions
from which the development initiates from.
RESULTS
The generative design system and supporting design
method described in this paper allowed for a trans-
parent process where the student describes all the
design rules through a visual programming environ-
ment supporting a gentle learning curve. Although
technically simple and following an iterative progres-
sion where new rules were established from evalu-
ation of the systems previous performance, the cre-
ation of the rule-sets showed to be a difficult task.
Educated to describe and develop design solutions
this method instead requested that the students be-
gan describing design problems and challenges, sub-
sequently refining their rule-set to explore new de-
sign outputs within this altered multidimensional so-
lution space. However, the system showed to sup-
port design thinking over a given time period and
thereby made the students capable of involving the
dynamics of the urban site. Instead of drawing a
static "ideal" solution the students focused their de-
sign implementation towards how design aspect are
related and how they unfold over time - this shifted
their way of thinking design from the absolute design
solution towards the potential and agent-based de-
sign solution, creating more focus on actors and rela-
tions (Figure 3).
Figure 3
Example of student
work showcasing
the cell matrix and
table visualization
derived from the
design system.
Work by students
Anne Marie Menå It is clear from the projects developed by the stu-
Heltborg and Stine dents using this CA framework and corresponding
Juul Christensen. design method that a rich and challenging set of in-
teractions occur. The level of predictability is largely
dependent on the control intentions inserted into
the design rules by the designer. This means that a
preconceived and desired spatial design can be ap-
proximated in the design rules for which the applica-
tion of the method has little use. However, if the de-
sign rules are described with sufficient diversity and
without conditions in which one state will terminate
the others immediately, a rich and time-based evolu-
tion develops, to which a designer can read, not only
the spatial programmatic propositions, but also how
the different states act against each other allowing
the method and framework to become an architec-
tural system which exhibits and informs the designer
about interdependencies between the design states
over time. While this will not be a 'truth' model for
how the urban environment will evolve over a de-
scribed time period in real life, it grasp the complex-
ity, possibilities and limitations of a design develop-
ment during the initial design phases as a support to
design decision taking.
DISCUSSION
Reflecting on the work performed in both the devel-
opment of the CA-based design framework and the
corresponding design method, as well as the explo-
rative work of the design students utilizing this gen-
erative design tool, there are a number of topics that
stand out. One is that of control. When working with
a CA-based system the machine computes the de-
sign solutions and the designer can only inform or
control the outcome through the rule-set defined be-
fore the initiation of the iterative development pro-
cess. Reflecting on the work of the involved design
students this showed to be a difficult task and often
ended in rule-set configurations leading to very dic-
tated design solutions. One conclusion to this chal-
lenge could be that besides requiring an understand-
ing of the design system, including the theory be-
hind CA, working with the system also requires an en-
hanced focus on design intent and not design solu-
tions.
Another topic is evaluation. How to evaluate so-
lutions when there are no concrete optima and when
all the steps in the design process are as important
as the solution in the final step? One aspect that
became very important when evaluating and under-
standing the complex datasets was visual feedback,
which in the case of this design system, comprised
of both a three-dimensional color-coded cell matrix
and a variety of data graphs. In a teaching aspect this
visual feedback allowed for an extended understand-
ing of relations and solutions as a pedagogic/didac-
tic learning method allowing the user to explore rule
sets with different and multiple objectives.
Any system described inserts a series of bound-
ary conditions, not only spatially, but also logically.
Through the working with the model and develop-
ment of the method, it is evident that understand-
ing of the preconditions for the CA's spatial organi-
sations needs to be well defined and understood in
order to make the method truly instrumental beyond
its abilities to generate formal and spatial proposi-
tions. Thus, it can be argued that a CA system in this
framework has its strength in informing a designer
about the complexity, time-based relations, bound-
ary conditions and variable relations more than an
ability to be a direct design tool for completed de-
sign solutions. However, when above has been un-
derstood, solutions and problems can be generated
quickly with multiple elements and factors inserted
simultaneously allowing for a deep integrative archi-
tectural process overcoming some of the drawbacks
of common linear design processes, in which prob-
lems are handled successively rather than in paral-
Design Tool 1 - Volume 1 - eCAADe 32 | 301
 lel. The higher the number of states presented in
the model, the higher the integration and informed
architectural models. But it also appears important
that the cell space remains large enough for inter-
actions to be computed. The smaller cell space re-
duces the level of complexity, and thus the ability to
evolve truly surprising and informative solutions be-
yond what the designer could propose without the
model.

REFERENCES
Batty, M 1997, 'Cellular Automata and Urban Form: A
Primer', Journal of American Planning Association,
63(2), pp. 266-274
Huang, CY, Sun, CT, Hsieh, JL and Lin, H 2004, 'Simu-
lating SARS: Small-World Epidemiological Modeling
and Public Health Policy Assessments', Journal of Ar-
tificial Societies and Social Simulation, 7(4), pp. 100-
131
Popov, N 2011 'Generative sub-division morphogenesis
with Cellular Automata and Agent-Based Modelling',
Respecting Fragile Places – 29th eCAADe Conference
Proceedings, pp. 166-174
Shiffman, D 2012, The Nature of Code, Creative Commons
Attribution-NonCommercial 3.0 Unported License

302 | eCAADe 32 - Design Tool 1 - Volume 1

CAAD Education
New Approaches in Architecture Education
Conclusion

Birgul Colakoglu
Yıldız Technical University
colak@yildiz.edu.tr

The explosion of computationally-oriented content into architectural design and

making has generated curricular changes in architectural education. This paper
describes an educational model developed following this changes. One of the
important goals of the educational model explained in this paper is to introduce
the knowledge of geometric assembly/disassembly planning to the field of
architecture encouraging students of architecture to integrate this knowledge into
the design process.

Keywords: computational design, , design for disassembly

Computation in all scientific and design fields
made radical implications in the ways of reasoning.
It caused the most radical shift of mental processes
since modernism. The new ways of reasoning require
a new epistemological approach. With "Computa-
tional design" there is no anymore talking about con-
ventional design aided with computers but a com-
pletely new way to perform design activities, from re-
search to evaluation, communication and education
affirming that there is a new culture of design activity.
The integration of this new culture of design
activity into architecture curricula has been inves-
tigated by many design educators through educa-
tional models. This paper will summarize one of this
models conducted in "Computational Design" grad-
uate program.
The educational model includes two modules
with two courses in each. The first module run in fall
semester includes Shape Grammar and Introduction
to Computational Design courses. The second mod-
ule run in spring includes Design Systems and De-
signing the Design courses.
The students enrolled in graduate program usu-
ally have good computerization skills. They are good
in using the tools however, are not aware of compu-
tational logic and thinking involved in computational
design making. The aims of the first module courses
have been: to introduce design students with com-
putational concepts utilizing shape grammar, and to
create intellectual foundations for this new design
culture. Due to their inherent algorithmic structure
patterns are used as tool for teaching these new con-
cepts of design. (Colakoglu, Atawula, Alkhoudari,
2014)
Once the students have learned to communicate
and manipulate their designs using this new design
language, than they are asked to apply them on 3D
surface considering sun movement on defined loca-
tion. Here, the angle between sun light and surface
normal has been the determining factor of the pat-
tern variation on the surface (Figure 1).
The first module has been prerequisite for the
second module courses run in spring semester. The
aims of the second module have been; first, to make

CAAD Education - Volume 1 - eCAADe 32 | 305

 students search for possible design systems that can
optimize computational design making and then to
develop designs considering the principles of these
systems. Design systems course introduced "Design
for Disassembly" (DfD) approach in design and con-
struction. Designing the Design course engaged
students with computational design making consid-
ering DfD strategies through real design problem.
Here, culture pavilion for Kadıköy in İstanbul was
given as design task. The design brief emphasized
that pavilion design strategy should correspond to
some of the principles of DfD.
DESIGN SYSTEMS-DESIGN FOR DISAS-
SEMBLY
Design for Disassembly (DfD) is a new design method
that introduces new design criteria. DfD was first ap-
plied in product design. It has been driven by well es-
tablished regulations that cover sustainability issues.
It is a new concept for the design and building com-
munity and is an important contributor to sustain-
able design. In this new design method a building
is composed of different layers having different life-
times. It may be expected that similar regulatory ap-
proaches applied in product design can soon be ex-
306 | eCAADe 32 - CAAD Education - Volume 1
tended into the building industry. Although, build-
ings are not"product," they are composed of mate-
rials, components, and connections that are assem-
bled.
Design for Disassembly is a method of design
that supports future change; it allows buildings to
transform and adapt either specified or emerging
needs. This method of design simply creates condi-
tions to recover, reuse, rebuild and reconfigure com-
ponents and materials. It is the most promising
method of optimization through; assembly, compo-
nent connections, and systematic design. It changes
craftsmen based building construction process to
Figure 1
one of assembly. Kieran S. And Timberlake J. Point
Application of
out that "Assembly differs from construction in that
patterns on 3D
it requires very little skill; it does not rely on infor-
surface. The
mation passed on through experience or develop-
openings on the
ment through apprenticeship. Assembly comes from
surface are
a hierarchical understanding of groups of assemblies
designed
that all connect through series of steps." The new de-
considering sun
sign culture based on computational designing and
movement.
making have changed craftsmen based building con-
struction process to one of assembly. DfD returns
to architect responsibility for the craft and quality of
the construction through computational processes
in construction and design. These new tools allow
architects to explore and construct buildings digi-
tally with extreme precision. Architects can simu-
late assembly processes prior anything is actually is
build. By simulation model of a building architect can
decide which components could be pre-assembled,
and which should be assembled on site.
DfD systems provide the opportunity for a build-
ing to be dismantled (in part or whole) both during
and at the end of its lifecycle giving building an abil-
ity to quickly adapt.
Design systems course elaborated DfD model of
design and construction for architecture. It inves-
tigated development of design through parametric
model integrating DfD principle applicable for archi-
tecture.
 DESIGNING THE DESIGN
In the second module two courses are run in par-
allel. Design system is required course for the stu-
dents that have taken Designing the Design course.
In the first one, the students elaborated DfD method
of design and construction, and in the second they
applied this new approach in real design problem -
culture pavilion design for Kadıkoy bay area (Figure
2).
Figure 2
Kadıköy bay area
Kadıkoy is one of the 32 districts of Istanbul. It
is located on the Asian side of the city, consists of
28 neighborhoods. Kadıköy is ranked as the cul-
tural centre of Asian side of Istanbul with, its theatres,
cinemas, culture centers distributed throughout the
neighborhoods.
However, the analyses conducted in Kadıköy
showed that there is more need for culture spaces
especially for temporary culture activities in Kadıköy
waterfront and bay area. Designing the Design
course is structured around this real design problem.
The students are asked to develop culture pavilion
designs for Kadıkoy bay area considering DfD princi-
ples.
The key to successful DfD lies in maintaining flex-
ibility within assemblies, easy component separation
and easy access to parts. The design brief required to
cover some basic DfD principles integrated with ar-
chitectural requirements:
Architectural requirements:
• Multifunctional pavilion design, (trans-
formable at least into two different functions)
• Environmental responsive design, (sun move-
ment, wind direction, ect.)
• Architectural expression,
• Structural, spatial, technical transformability,
Some basic DfD principles:
• Simplified connections between components
for easy assembly and disassembly,
• Minimized components in assembly, (without
compromising the structural integrity or func-
tion of the design)
• Minimized material types in assembly,
• Separated components into modular sub-
assemblies
• Minimized number and type of fasteners in
the assembly,
• Standardized fasteners,
• Easy accessible fasteners,
• Easy separable components,
First, context and environmental factors of
Kadıköy are analyzed. Kadıköy old market and the
streets surrounding the market area are famous with
street art and various cultural events that are taking
place on the streets. In order to locate these cultural
events in specified areas throughout Kadikoy, loca-
tion map for pavilions is defined. In this map, specific
location in Kadıkoy bay area has been chosen for cul-
ture pavilion.
The foot print of the culture pavilion is developed
from Fibonacci spiral (Figure 3). It is generated by
drawing quarter circles inside the squares that are
generated by drawing small squares on top of each
CAAD Education - Volume 1 - eCAADe 32 | 307
 Figure 3
Fibonacci spiral and
nature examples

Figure 4
Pavilion is created
by gradually
increasing the
height of the spiral.

Figure 5
Figure 5: Kadıkoy
culture pavilion Sun
and wind analyses.

Figure 6
Grasshopper model
of pavilion

Figure 7
Panels,
joints-connections

308 | eCAADe 32 - CAAD Education - Volume 1

Figure 8
Grouping the
panels
other and adding a sequence of growing squares
counter clockwise. The length of the square sides in
the order generates sequence 1,1,2,3,5,8,13,21, of Fi-
bonacci numbers.
These numbers are find in many spiral constructs
in nature (Figure 2). The pavilion is created by grad-
ually increasing the height of the arcs that generate
the pavilion. In other words by extrusion of spiral in
3D. Here the height of starting arc is 1.2m and ending
arc is 5m (Figure 4).
The environmental conditions, wind direction
and sun movements of Kadıköy bay area is ana-
lyzed utilizing Vasari software in design development
phase (Figure 5).
These two environmental factors are considered
as challenge during the development of design con-
cept. After the design concept and methodology is
selected, grasshopper is used to generate the design
solution space and to calculate the openings of the
façade according to the angle of the sun with surface
normal (Figure 6). Some generated design solutions
are examined in Vasari for solar movement and wind
analysis.
Culture pavilion is designed for two functions; as
an info desk during the day and as performance stage
at night.
One of the design purposes of the project was
easy assemble and disassemble of design compo-
nents. To realize this aim, pavilion structure is pan-
elized and divided into 20 groups, each one defined
as stripe, consisting of 16 panels (Figure 7). During
the assemble process, these groups are put together
separately, then connected to each other to build
the pavilion. The disassemble process are the oppo-
site. Mechanical joints were used rather than chemi-
cal joints to make the process of reuse and reassem-
CAAD Education - Volume 1 - eCAADe 32 | 309
 Figure 9
Culture pavilion
model construction.

Figure 10
The openings on
the pavilion are
constructed
according sun
movement in
Kadikoy.

310 | eCAADe 32 - CAAD Education - Volume 1

Figure 11
İllustration of
Culture pavilion on
site. Design work:
Düzgüneş.Ç, Özen
G, Yönter U,
Atawula E,
Alkhoudari L.

CAAD Education - Volume 1 - eCAADe 32 | 311

 ble more efficient. The use of dry connection de- challenge. The main constrain was luck of integrative
tails made the dismantling process and reconstruc- approach (integration of architectural program with
tion possible with limited waste. design and construction system) in students design
Each panel of the pavilion is different; therefore thinking process.
they are numbered before fabrications according to In today's climate of constant change and relent-
the group numbers for easy assemble. less competition, the key skill of an architect is to
Joints are made according to the angles between know how to lead its profession. This needs integra-
panels. The joint element located in the middle of the tive thinking. Integrative thinkers
intersection line of the panels also defines the open- embrace complexity, tolerate uncertainty, and
ing direction of the panel. During the assembly pro- manage tension in searching for
cess, the joints and panels are separated in to groups creative solutions to problems. Architects of fu-
and assembled separately (Figures 8, 9, 10, 11). ture in order to lead their profession need to be edu-
In this educational model, a new process which cated and trained in this manner. This is a challenge
constructs wooden panels with mechanical fasten- in education of architecture that forces it to change.
ers is explored. The process creates joint details in
members and generates the assembly sequence. It REFERENCES
introduces the fundamental knowledge of geomet- Colakoglu, Atawula, Alkhodari, B, E, L 2014 'Facilitat-
ric assembly planning to the field of architecture. This ing Computational Thinking Through Pattern Gen-
knowledge is essential to the research of automated eration', VIII. Mimarlıkta Sayısal Tasarım Ulusal Sem-
assembly process in architecture. pozyumu, Izmir
Bhamra, D 1996, 'Design for disassembly’, Co-Design',
The Interdisciplinary Journal of Design and Contextual
CONCLUSION Studies, 5 and 6, pp.28-33
Profession of architecture combines multiple disci- Ching, T 2014, Design for Assembly: A Computational
plinary skills into a single mind-set that requires in- Approach to Construct İnterlocking Wooden Frames,
tegrative thinking. Master's Thesis, MIT
Kieran, Timberlake, S, J 2003, Refabricating Architecture
The most significant pedagogical challenge in
McGraw-Hill Professional, New York
explored educational model was to make students
integrative thinkers. The educational model con-
structed in two modules. First run in fall, second in
spring semester. Both modules consisted of two par-
allel run courses. The students were required to im-
plement learning's of one into another. At the end
of first module the students gained comprehensive
understanding of operational logic of computation in
design. They had good grasp of procedural structure
of computational approaches in design.
The second module that was implemented first
time has been partially successful. The students were
required to grasp many new subjects and integrate
these new subjects in real design problem. Although
their design solution full filed most of the require-
ments of design brief related with DfD, integrating
all subjects in a sounding architecture remained as

312 | eCAADe 32 - CAAD Education - Volume 1

Digital Curation for CAAD Curricula
Bridging Mainstream and Speculative Design Procedures to promote
curatorial competences for Architecture

Ruggero Lancia1 , Ian Anderson2

1,2
Humanities Advanced Technology and Information Institute, Univ. of Glasgow
1,2
{Ruggero.Lancia|Ian.G.Anderson}@glasgow.ac.uk

As the design processes in Architectural practices switch toward entirely digital

workflows, architects are gradually required, because of their legal and
commercial liability, to provide for both a relatively long term curation of their
own digital products and the deposit of authoritative data. But, despite being the
sole curation actors for their data, architects receive little education or training
in either pertinent competences nor agreed and established procedures to comply
with these duties.In this paper, the design of Digital Curation courses within
CAAD Curricula will be discussed against the investigation results of the
DEDICATE project, an AHRC funded project hosted at the Humanities Advanced
Technology and Information Institute of the University of Glasgow.

Keywords: CAAD Curricula, Digital Curation, Generative Design, Digital

Fabrication

INTRODUCTION ical Archives by architectural practices (Lancia 2011).

This paper discusses the need for Digital Curation and
Information Management competences in the Archi-
tect profession and proposes a Digital Curation mod-
ule for CAAD curricula detailing its learning objec-
tives. This need is based on collaborations with com-
mercial practices and HEI partners undertaken during
the DEDICATE (Design Digital Curation for Architec-
ture) project, an AHRC funded initiative led and su-
pervised by the authors.
EMERGING NEED FOR DIGITAL CURATION
IN ARCHITECTURE
There is evidence that statutory project documenta-
tion is mainly adapted to be printed and kept in phys-
But, the request for producing and depositing au-
thoritative digital data is an emerging phenomenon
across many international Building Control authori-
ties and Public Authorities that are urging architec-
tural practices to implement consistent data man-
agement procedures.
For example, in Netherlands, since November
2011, the Rgd BIM Norm obliges design contractors
involved in public building projects to produce and
deliver their products in Building Information Mod-
elling (BIM) formats following the policies of the Rgd
BIM standard (Van Rillaer et al. 2012). In the same
year, the United Kingdom Cabinet Office announced
in the Government's new Construction Strategy that
all public works will require BIM documentation in

CAAD Education - Volume 1 - eCAADe 32 | 313

 both native and IFC4 formats by 2016 (Kelly 2011).
Professional Indemnity insurance in the UK when us-
ing BIM will depend on Information Management
policies agreed between the parties, including, for
example the definition of the trail of information cus-
tody (Bamforth 2013).
The commercial liability of architects extends
these emerging requirements, introducing in the
profession the need for relatively long periods of
reliable digital assets retention. For example, in
United Kingdom, the architects' professional liability
period amounts to 12 years post practical completion
(Speaight and Stone 2005) - a very demanding reten-
tion period for digital information to be held in infor-
mal repositories such as those found in architectural
practices. Where workflows involve complex Digi-
tal Design procedures, such as Parametric and Gen-
erative Design, Collaborative and Concurrent prac-
tices and Building Information Modelling, the man-
agement and preservation of assets are especially
complicated by the multiplication of dependencies,
the use of IPR encumbered file formats and the short-
lived platforms adopted for the data production. Fur-
ther, the rapid obsolescence of CAD packages and
their common lack of documentation are long stand-
ing recognised problems of CAD data curation and
preservation. This is also a threat for practices imple-
menting more traditional Digital Design procedures,
such as 2D drafting.
Digital Curation, like its cognate disciplines Data
Curation, Digital Preservation, Digital Archiving, and
Digital Stewardship, is commonly associated with
archival routines and research institutions' policies
and objectives. Though, the investigative and op-
erative repertory of Digital Curation is not uniquely
represented by research data or other akin controlled
productions. Despite this most research institutions,
such as the Digital Curation Centre [1], have given
prominence to e-science data and other research dig-
ital outputs as repertory of their activities. Individ-
ual investigators from this discipline have defined
more widely the Digital Curation objects of inter-
est as the entire production of digital assets in our
314 | eCAADe 32 - CAAD Education - Volume 1
society, including personal digital collections (Bea-
grie 2008). Digital Curation both extends and spe-
cialises the scope of Information Management as it
both concentrates on the information assets in the
digital domain and approaches the management of
data through a life-cycle model spanning informa-
tion production to its disposal. The 'digital curator' is
both a competent interpreter of the products stake-
holders' needs and objectives, and an active member
of a specific community of practice.
Borrowing from the definition of Digital Curation
by Neil Beagrie as "the actions needed to maintain
digital [...] materials over their entire life-cycle and
over time" (Beagrie 2008, 4) and combining these
novel management requirements, it is arguable that
architects will soon be expected to take curatorial
responsibilities over their data. This renewed scope
on the professional services expected from architects
urges also a novel approach to the commercial ex-
ploitation of designs based on the legal and persis-
tent management of its Intellectual Property Rights
(IPR).
The workflows ramifications and the still lim-
ited legal pressure on Architecture practices for both
implementing consistent data retention procedures
and depositing authoritative data in statutory repos-
itories, actually prevent both Digital Curation profes-
sional from approaching this field and architects from
appealing to their services. As a result, most practices
resort to commercial services for information man-
agement that do not solve crucial curatorial issues,
such as long term data reuse, accessibility, IPR man-
agement and authoritative retention.
In such a context, post-hoc data curation pro-
cedures, such as those implemented in institutional
archives, are neither feasible nor effective. Instead,
as Alistair Miles proposed for the project ImageStore,
(part of the Digital Curation Centre SCARP initia-
tive) curatorial attention when integrated in the very
workflows defining the datasets lifecycle, enhances
both the data quality and their preservation viability
in informal retention contexts [2].
Curatorial activities should be highly integrated
with architectural design procedures to address with
informed decisions the potential threats to data in-
tegrity and long-term legal management. As a result,
the professional digital curators' interventions on this
repertory should be substituted with a competent
management by the stakeholders themselves.
Higher Education and Vocational Training are
generally lagging behind these new professional re-
quirements. But we might admit as a justifica-
tion for such a slow response that some of the is-
sues highlighted are still implicit in the technologi-
cal trends and political initiatives we have indicated
above. Nevertheless, in a rapidly shifting landscape
one would expect Higher Education to adapt to new
phenomena by addressing their future impact, that is
shaping it through future generations of profession-
als.
AUDITS AND ANALYSIS WITHIN DEDICATE
Within the Design's Digital Curation for Architecture
(DEDICATE) project the development of a Curation
module for CAAD Curricula was especially investi-
gated.
DEDICATE was funded by the Arts and Human-
ities Research Council (with the ref. AH/J008265/1)
and was hosted by the Humanities Advanced Tech-
nology and Information Institute (HATII) of the Uni-
versity of Glasgow. Running for nine months, this
project was aimed at investigating the policies, re-
quirements and procedures to build a sustainable cu-
ratorial framework for the Built Environment related
digital assets, minimising their loss risks and maximis-
ing their reusability and interoperability within their
stakeholder community.
The digital assets, production and management
procedures, and infrastructures of a select group of
Scottish architectural firms and engineering consul-
tancies were audited using affirmed curatorial tools,
original scripted analysis and unstructured observa-
tions.
The project partners repositories were first au-
dited adopting methods and tools from three recog-
nised Digital Curation tools: DAFD (Data Audit
Framework Development) which provides an audit
methodology and online tools to support and fa-
cilitate organisations to establish an overview of
their data holdings, policies and practices against
best practices and new risks [3]; DRAMBORA (Digital
Repository Audit Method Based on Risk Assessment)
which offers an audit methodology, complemented
by a computer-aided audit software, addressing the
assessment of risks implied by the policies adopted
by the repositories [4]; and Planets (Preservation and
Long-term Access through NETworked Services) that
offers a testbed to experiment with the effects of cu-
ratorial actions on digital assets, such as format mi-
gration [5]. Subsequently, the files and folder struc-
ture of a selection of the project partners' projects
were ingested in a drive on a UNIX server and anal-
ysed with UNIX and PERL scripts to extract the data
population distribution in terms of formats, size and
typologies. The same files were scanned with Jhove2
to test both the efficiency of automatic characterisa-
tion routines on built environment related digital as-
sets and the syntactic integrity of data files against
their format documentation. A selection of these as-
sets was then transferred to the Planets testbed to
experiment both the effects and the long-term fea-
sibility of the management actions most frequently
adopted by the project partners. The observations
on the partners' workflows and the interviews with
the practices' staff recorded both the specific con-
textual information attached to the projects selected
for analysis and all the procedures and tools imple-
mented by the partners producing, sharing and hold-
ing their digital products. These observations fo-
cused as much on the production of data as on its
management and in both cases compliance to stan-
dards such as the Uniclass or the ISO 9001:2008 were
considered.
The information gathered through these analy-
ses exhaustively documents the problems and the
specific needs and requirements associated with the
curation of digital assets in small and medium sized
architecture practices.
Despite salient examples from this frame-
CAAD Education - Volume 1 - eCAADe 32 | 315
 work having been reported in the past decades at
the Education and research in Computer Aided Ar-
chitectural Design in Europe (eCAADe) conferences,
speculative and experimental procedures, such as
the workflows often adopted by students for Gener-
ative Design and Digital Fabrication, the results have
scarcely explored Digital Curation and Information
Management initiatives. To delve into this subject,
under the auspices of DEDICATE, two academic col-
laborations were initiated - the first with the Labo-
ratory for Applied Building Science of the Graduate
School of Architecture, Planning and Preservation at
Columbia University, and the latter with the Master of
Research in Creative Practices of the Glasgow School
of Art.
GSAPP's Paris-Atelier 2013
Thanks to Phillip Anzalone, director of the Labora-
tory for Applied Building Science, project staff vis-
ited the Paris-Atelier 2013 Summer School during its
conclusive week to document the digital workflows
adopted by the course's students and explore both
their procedures and issues in managing data files
and their understanding of the future impact of in-
formation management on the architect profession
[6].
The Paris Atelier is an advanced academic pro-
gram of the Avery Digital Fabrication Laboratory, part
of a five years program co-founded by GSAPP & Asaf
Gottesman, that endeavours to provide a platform to
explore the role of the Architect as a critical maker,
that is as a designer both involved in the cultural
debate on architecture and active in the actual con-
struction of his/her design. Workshops led by indus-
try partners are offered during the 8 weeks course to
develop new means working with computational de-
sign, material processes, and assembly of spatial con-
structions. The course culminates in an exhibition at
the Palais de Tokyo in Paris of the students' work. The
students presented as a final work a large installation
designed in collaboration with artist Tomas Saraceno
- an ethereal pavilion built by arranging a field of in-
flated mylar balloons to describe two NURB surfaces
316 | eCAADe 32 - CAAD Education - Volume 1
splitting the Palais de Tokyo exhibit space into three
distinct volumes (see figure 1).
The morphology of the pavilion was researched
by the students adopting different scripting strate-
gies in Grasshopper and its final version was agreed
through direct discussion on the proposed scripted
routines within a staged collaboration process.
The students decided to manage their collabo-
ration through a shared folder in the file hosting ser-
vice DropBox. They autonomously created a folder
structure differentiating individual researches, the
knowledge base (such as handbooks, publications
and teaching material), and the design, presentation
and fabrication of the pavilion. In this latter sec-
tion a system of agreed naming conventions records
the versioning of both models and deliverables files
whilst the granularity of the folder structures closely
matches the changes in direction, the final decisions
and the contextual information of the design process
rather than a general categorisation protocol.
Figure 1
The Paris Atelier
pavilion at the
Palais de Tokyo.
Image courtesy of
the course students.
'Hamlet Waiting for Godot' a research from
the MRes in Creative Practices of the GSA
The collaboration with the Master of Research in Cre-
ative Practices of the Glasgow School of Art exclu-
sively addressed the assessment of the workflows
adopted in digital fabrication, from early conception
to module assembly, and took place through assist-
ing a Master student, Gaetano Chianese, in the reali-
sation of his final work - an eight cubic meter pavilion
composed of EPS sheets shaped with a Heiz High-Z S-
 1000-T, a large format CNC router.
Figure 2 This project analysed the role of the script in a
The HWFG pavilion theatrical performance giving prominence to its im-
at the Light House, plicit physical participation on the scene. The pavil-
Scotland's National ion was developed as a demonstration of this inves-
Centre for Design tigation's results, gathering together text and perfor-
and Architecture, mative actions in a reclusive physical environment. In
Glasgow. this way, its visitors were considered both spectators
Installation of the and active characters of the play. Then, the physical
pavilion with appearance of the text was substituted by QR codes
superimposed embossed on the external facade of the pavilion link-
render of the whole ing each face of the object to both acts of the play
structure. and multimedia contents hosted in an original web-
site [7].
The Design of the pavilion started with a poser
application, DAZ, the actions expected to be enacted
by the visitors on human body models. From these
posed characters, the negative space of their move-
ments was reconstructed with Rhino creating a sur- OPERATIVE FRAMEWORKS IN ADVANCED
face tangent to their shapes. Then, again in Rhino, DIGITAL DESIGN AND TRADITIONAL
this volume was extracted from the mass of the pavil-
PRACTICES
ion exterior geometry. With the same process, the
These initiatives, were documented and assessed
technical voids to host cables, speakers, player de-
against curatorial criteria and an operative frame-
vices and poles for a reinforcement structure were
work that resulted paradigmatically divergent from
added to the pavilion. Through Grasshopper and
those of the audited architecture practices, especially
RhinoScript, these geometries were interpreted as
for the prevalence of 3D models centred procedures
horizontal crosscuts, optimised for the fabrication
of design representation as opposed to the commer-
and exported to the DWG format. Subsequently,
cial abundance of both documents representations
these files were used in AutoCAD to produce DXF
and relative CAD datasets.
files respecting a set of layer conventions. Eventu-
The data files population reported by the analy-
ally, adopting VCarve, these shapes were elaborated
sis of the file repository used by the ParisAtelier stu-
to deliver GCODE commands ready to be ingested by
dents confirm the central role of Rhino/Grasshopper
the router.
in both their Design process and the creation of vi-
For the fabrication of the pavilion, 168 EPS70
sualisations. Opposite to what is recorded within
sheets (2400 x 1200 x 25 mm) were cut and assem-
small and medium Architecture practices, the data
bled in modules with vinyl glue in approximately 2
files population hints at the prevalence of a small
weeks at the MAKLab, the first open access digital
group of models used as basis to generate deliver-
fabrication studio in Scotland. The pavilion was even-
able materials. This is inferable by comparing the
tually on display at Scotland's National Centre for
number of file types to their cumulated size, i.e. 61%
Design and Architecture, the Lighthouse, during the
of total file counts are JPG but their total size is only
GSA Graduate Degree Show 2013 (see figure 2).
34%. 3DM Rhino models files account for 7% of the
count of repository files but their size is the 20% of
the total repository's size (see figures 3 and 4).

CAAD Education - Volume 1 - eCAADe 32 | 317


The analysis of the data repository of the HWFG
project demonstrates the proliferation of intermedi-
ate products to accomplish a digital fabrication ob-
jective. In this case, the most numerous data files, the
30% of the total files population, are in DWG format.
The DXF files are the 13% of the total files number and
318 | eCAADe 32 - CAAD Education - Volume 1
whilst, being closely related in the digital workflow to Figure 3
create the GCODE routines, CRV and NC files account Percentages of file
for 18% of the analysed data. The 3DM files are just types recorded
6% of the audited files but their cumulated size rep- within the
resents 46% of the total repository size (see figure 5). repository shared
Because of the implemented Rhino settings, by the students of
3DMBAK files show the same demographic figures of the Paris Atelier
their parent 3DM files. Because of the strong empha- Summer School.
sis in this project on producing the physical pavilion,
a small amount of digital products were intended for
the project dissemination, for this reason, PSD and
raster images files are almost absent - their number
does not reach 1% of the total files number and their
cumulated size do not exceed 2% of the repository
size (see figure 6).
These latter figures demonstrate an individual
workflow, as opposed to the collaborative practices
of the ParisAtelier courses, the results are also op- Figure 4
posite to the procedures implemented by SME prac- The data
tices. These features of the HWFG project represent a population of the
pattern shared by the 'maker culture' initiatives and Paris Atelier
linked to the data access and reuse practices com- students' repository
mon to both this and the ampler movement for open distributed per size
source data sharing - in fact, deliverables and visuali- of the file types.
sation are almost irrelevant in the sharing of informa-
tion in these contexts.
An aggregate and simplified rendering of the
data population recorded in the repositories of SME
practices confirm the difference of their operative
framework as compared to advanced digital design
techniques (see figures 7 and 8).
As DWG files amount to the 22% of all the files
population and PDF files are at 24%, the design pro-
cedures implemented in these practices are visibly
based on the production of digital deliverables corre-
spondent to 2D CAD drawings - the PDF format sup-
ports U3D models but in the audited repositories no
file in this format was found we can assume the PDF
analysed do not hold 3D information.
Further, since the values of size and number of
DWG and PDF files are very close, we can affirm that
in this context, collaboration is based most proba-
bly on the sharing either of editable CAD files or of
Figure 5 their static digital surrogates. These metric evidences
Percentages of file confirm the observations of the procedures of these
types recorded practices: the information shared for collaboration
within the with external contractors is not intended to be edited
repository of the but integrated in other workflows.
HWFG project. Meanwhile, the deliverable files addressed to
clients and Building Control authorities follow the na-
tional standards for the architectural documentation
and mirror consolidated practices of managing draw-
ings printed on paper. In such a professional environ-
ment, both the communication between parties and
the office suites' documents records are associated
to CAD data to document and manage the project -
DOC files are 22% of the data analysed in SME prac-
tices and formats associated with email correspon-
dence, such as EML and MSG, amount to 5%.
Despite the extensive access granted for inves-
tigation to the practices' digital assets, the acquisi-
tion of unprecedented quantitative information on
Figure 6 the digital workflows adopted, these results may be
The population of more representative of the local context of Scottish
the data files Architecture practices.
recorded within the Furthermore, analysis of the experimental activ-
repository of the ities in Advanced Digital Design conducted by stu-
HWFG project dents recorded procedures implemented without re-
distributed per specting any professional requirements, although
aggregated size of there is a growing community of Architects and De-
the file types. signers, nationally and internationally, exploring a
variety of Digital Design procedures within their pro-
fessional practices.
Once merged, these opposite frameworks result
in a set of curatorial challenges, needs and require-
ments that demand a common curatorial approach
for their digital assets and hint at the future conver-
gence of Digital Design procedures for Architecture.

CAAD Education - Volume 1 - eCAADe 32 | 319


DIGITAL CURATION EDUCATION IN ARCHI-
TECTURE
As professional requirements and integrated Digital
Design workflows demand authoritativeness, inter-
operability, safety, efficiency and legality in manag-
ing digital assets, this project stresses that Architects
should be able to expertly and autonomously hold
320 | eCAADe 32 - CAAD Education - Volume 1
long-term control over their digital products. From Figure 7
this new competence would result numerous oppor- File types recorded
tunities for professional development such as satisfy- within the the
ing the political expectations on the redevelopment repositories of the
of the Construction sector, understanding the oppor- Architecture
tunities for enhanced commercial exploitation of dig- practices
ital assets and facilitating the engagement in coop- collaborating with
erative and concurrent Advanced Digital Design pro- the DEDICATE
cedures. Moreover, this expertise enables designers project as project
to both understand and implement curatorial frame- partners. This is an
work updates as the technologies and the tools for aggregated and
both curation and design evolve, and to adapt them simplified rendition
to the legal, commercial and creative needs of his/her of the actual data
own business. population
At the beginning of the 1980s, Information Man- recorded during the
agement was first introduced within CAAD educa- project.
tion in the MIT Master of Science course in Computer-
aided design and at the Carnegie-Mellon Univer- Figure 8
sity, in a four-week module in the professional ar- The population of
chitectural programme (Purcell 1980). As idiosyn- the data files
cratic CAAD systems development was superseded recorded within the
by commercial package solutions, the interest on this repositories of the
topic was exhausted and, to date, there is no evi- DEDICATE project
dence of other CAAD courses held in Architecture partners distributed
Schools addressing the management of design data. per cumulated size
Advanced Digital Design practices, such as concur- of each file type.
rent and collaborative practices, Digital Fabrication,
Generative Design and BIM are now giving unprece-
dented emphasis to the role of CAAD applications in
both the offer of professional services and their docu-
mentation. As a result, native digital assets produced
through CAAD software have a growing professional
importance demanding numerous curatorial actions.
For this reason, this project recommends that Digital
Curation education should be incorporated into the
CAAD curricula.
To define the curatorial competences attached to
both the production and management of these com-
plex assets repertory the project borrowed from the
DCC lifecycle model its abstract categorisation of the
data lifecycle (Higgins 2008). As this model proposes,
seven stages in the assets production and manage-
ment can be identified to specify curatorial actions to
be undertaken at each of them, that is the data pre-
production, data creation, data appraisal, data inges-
tion, data preservation and storage, data access and
data transformation stages.
The curatorial competences required at each of
these stages can be expressed as learning objectives
of a prospected Digital Curation module in CAAD cur-
ricula.
Architects should be able to both plan and im-
plement consistent curatorial procedures along the
digital design workflows and to formulate data ap-
praisal and selection criteria against a set of econom-
ical and professional objectives to formalise informa-
tion disposal procedures. For example, the Informa-
tion Requirements framework proposed by the CIC
BIM Protocol (Croft 2013) is an early proposal for in-
corporating in the contractual planning of the roles
and relationships between parties and details of the
Information Management procedures, including in-
formation exchange, access and IPR management. In
particular, this protocol, supporting BIM implemen-
tation working at Level 2, provides for the appoint-
ment of an Information Manager in charge of the pro-
cesses for information exchange, outputs prepara-
tion and data assessment. Although this role has not
design related duties, which are instead associated
with the BIM Coordinator, the protocol suggests the
Design Lead or the Project Lead to perform this role.
In so doing, this protocol is including curatorial com-
petences within the Architects' tasks.
Architects should have also the competences to
manage the ingestion of digital assets according to
agreed curatorial policies to ensure data authorita-
tiveness, persistence and accessibility.
To implement preservation procedures on the
assets held in the repository, Architects should be
able to establish preservation policies according to
professional and legal needs. The preservation of as-
sets should be then be guaranteed by managing the
persistent feasibility of data storage.
As professional liability and authoritativeness of
data ought to be documented, Architects might be
able to monitor and restrict privileges for data ac-
cess and reuse according to professional and legal re-
quirements.
To ensure the digital management and produc-
tion procedures can document long-term rights and
licenses associated with data, designers need to have
competences to plan and implement procedures to
track the data reuse and transformation according to
good practices in IPR management.
These activities must be rooted on a set of techni-
cal skills in the management of digital infrastructures,
the implementation of standards and the production
of Digital Assets.
First of all, architects might have a proficient
knowledge of formats, data structure and digital de-
sign computing routines to take expert decisions on
the digital infrastructure and the procedures for data
production to be implemented in their practices.
Then, an advanced knowledge of both metadata
standards and data quality assessment parameters is
requested to appraise the quality of the digital assets.
In particular, Architects should be able to understand
both the purpose and the originating digital work-
flow of the data.
Managing their own repository, Architects
should be experts on the preservation routines, func-
tions and strategies and the repository architecture's
available options.
To implement a granular access control to the
data held in the practice's repository, Architects
should understand the techniques and procedures
for privileges based data access.
Managing the data transformation and the IPR
associated with the assets, Architects should have
knowledge of data watermarking, cryptographic
techniques options and format migration issues.
CONCLUSIONS
The rising importance of Digital Curation compe-
tences in the Architect profession demands updating
the education of the future generations of Architec-
ture professionals. Further, the affirmation of the dig-
ital cultural market and especially the closer integra-
tion of Digital Design with Manufacturing and Con-
CAAD Education - Volume 1 - eCAADe 32 | 321
 struction are also redefining the role of the Architect Lancia, R 2011, First Step Award Final Report - Project
as a content provider and demanding from profes- DIDECU (Digital Design Curation), University of Glas-
sionals a granular control over their data IPR manage- gow, Glasgow
Purcell, P 1980, 'Computer Education in Architecture',
ment. As these recent developments also facilitate
Computer-Aided Design, 5(5), pp. 239-51
the commercial exploitation of digital products, such Van Rillaer, D, Burger, J, Ploegmakers, R and Mitossi, V
as for example promoting the reuse of elements in 2012, Rgd BIM Standard, Rijksgebouwendienst
BIM workflows (Waterhouse et al. 2013), the integra- Speaight, A and Stone, G 2005, Architects Legal Hand-
tion of curatorial activities in data production can re- book. The Law for Architects, Architectural Press, Ox-
sult in numerous opportunities for architects' profes- ford
Waterhouse, R, Philp, D, Malleson, A, Sinclair, D, Hamil,
sional development.
S, Northwood, C and Burrel, E 2013, National BIM Re-
Despite the frameworks resulting from the tradi- port 2013, NBS, London
tional use of CAAD solutions in Architecture practices [1] http://www.dcc.ac.uk
and the Advanced Digital Design procedures have [2] http://alimanfoo.wordpress.com/2007/06/27/zoological-
different needs and requirements, we foresee the im- case-studies-in-digital-curation-dcc-scarp-imagestore
pact of digital innovation in the construction sector [3] http://www.data-audit.eu
[4] http://www.repositoryaudit.eu
will push these frameworks to converge, especially
[5] http://www.planets-project.eu
to support the persistent retention and accessibility [6] http://www.arch.columbia.edu/programs/paris-
of digital assets across domains. atelier
The Digital Curation module for CAAD curricula [7] http://hwfg.altervista.org
discussed in this paper will offer architecture stu-
dents the competences to both practice their future
profession within this integrated framework and ful-
filling political and commercial expectations on the
development of the Architect role.
At present, the results of the DEDICATE investiga-
tion are being adopted to create a first Digital Cura-
tion teaching module to be tested in different archi-
tecture curricula. A partial application of this teach-
ing project is expected to be within the forthcoming
Master in Built Heritage Conservation of Strathclyde
University organised by Cristina Gonzalez-Longo.

REFERENCES
Bamforth, S 2013, Best Practice Guide for Professional In-
demnity Insurance When Using Building Information
Models, Construction Industry Council, London
Beagrie, N. 2008, 'Digital Curation for Science, Digital Li-
braries, and Individuals.', International Journal of Dig-
ital Curation, 1(1), pp. 3-16
Croft, A. 2103, CIC BIM Protocol, Construction Industry
Council, London
Higgins, S. 2008, 'The DCC Curation Lifecycle', Interna-
tional Journal of Digital Curation, 1(3), pp. 134-40
Kelly, S 2011, Government Construction Strategy, Cabinet
Office, London

322 | eCAADe 32 - CAAD Education - Volume 1

Integrating Sustainability in the Architectural Design
Education Process
Taxonomy of Challenges and Guidelines

Ahmed Sarhan1 , Peter Rutherford2

1
Anglia Ruskin University 2 University of Nottingham
1
ahmed.sarhan@anglia.ac.uk 2 Peter.Rutherford@nottingham.ac.uk

The last decade have seen substantial calls and increasing pressure for
developing an integrated design teaching framework, where sustainability is an
imperative priority. This paper focuses on presenting a taxonomy of the main
challenges encountered within the educational domain, in the attempt to reach an
effective integration. The paper also presents a set guidelines to address and try
to resolve the noted challenges. As the use of Building Performance Simulation
(BPS) applications is a central approach in this process aiming to reach energy
efficient buildings, the paper focuses on the shortcomings noted as a result of the
use of these applications in the design studios, with particular emphasis on the
thermal and lighting aspects of the simulation. The taxonomy presented is a
summary of the findings from literature review, as well as the surveys results
which were part of the author's research project discussed in the paper.

Keywords: Environmental Design, Building Performance Simulation,

Architectural Design Education

INTRODUCTION approach for design; one where students incorpo-

Introducing sustainability measures to the conven-
tional architectural design process for the purpose of
reaching energy efficient buildings, has led to a re-
markable evolution, and subsequently to the intro-
duction of a new multitude of variables that should
be accommodated and resolved within the design to
be approved. Architectural design students are sub-
sequently facing mounting challenges; adding an-
other dimension to attain conceptual building de-
signs that can perform favourably from a sustain-
ability perspective. The notable paradigm shift has
demanded and/or imposed a rather more holistic
rate more performative measures, and demonstrate
solid comprehension of the implications these mea-
sures brings to the design and the forces that fos-
ter or impede it (Sarhan, 2012). Schools of architec-
ture are ultimately compelled to embrace the new
paradigm, and develop an integrated design teach-
ing approach, aiming to prepare new generations of
architects with high awareness and comprehension
of sustainability and the measures it inflicts on the de-
sign and decision-making process.
This paper presents an analysis of educating the
design students how to integrate environmental de-

CAAD Education - Volume 1 - eCAADe 32 | 323

 sign measures, through the use of Building Perfor-
mance Simulation (BPS) tools, focusing mainly on the
thermal and lighting aspects of simulation process.
This leads to a taxonomy of the challenges encoun-
tered in the process and guidelines on how to over-
come these challenges, and improve the students'
environmental design experience. The study rep-
resented in this paper is part of a research project
(Sarhan and Rutherford, 2011), aiming to introduce a
new approach to overcome the noted challenges and
to facilitate developing an integrated design teach-
ing context. While the study focuses on BPS applica-
tions as a direct tool for integration, it does not fail to
acknowledge the merits and values that these tools
add to the architecture, engineering and construc-
tion professions.
EDUCATING THE DESIGNER
A new 'Decade of Education for Sustainable Develop-
ment' has been initiated by the UN in 2005, envisag-
ing the development of new approaches for devel-
oping structured curricula that can evidently accom-
modate 'learning for sustainable development'. One
of these approaches was EDUCATE (2012), which em-
brace key directives for repositioning sustainability at
a higher 'priority', and enriching the knowledge and
skills needed for its integrating in architectural edu-
cation. These directives seek to promote a holistic in-
tegrated approach of thinking and learning, as well as
creating an 'inter-disciplinary dialogue between con-
ventional cognitive domains'. In doing so, these di-
rective embrace key concepts; including 'Experiential
Learning', 'Reflection' in and on action, and 'Motiva-
tion' though active learning. For these directives and
their derived concepts to be effective, they should
adopt methods for adequately meshing these princi-
ples into the cognitive creative design context. This
includes careful consideration to the architect's ap-
proach to learning, thinking, and problem solving.
Rutherford and Wilson (2006) explain that archi-
tects normally tackle the design challenge with a
solution-focused rather than a problem-focus strat-
egy. This strategy was defined by Gelernter (1988) as
324 | eCAADe 32 - CAAD Education - Volume 1
'Cognitive Schemata' approach for problem solving,
which is an iterative methodology where students
expand their knowledge through sequence of design
analysis and criticism that is mainly focused on the
solution being put forward, rather than the method-
ology that can be applied. In other words, students
will adopt an approach that most likely focused on
solutions to satisfy design challenges, rather than
embracing a more analytical approach for a critical
analysis of this challenges' complexity leading into a
resolution (Altomonte, 2009).
In essence, the aspiration of achieving an effec-
tive integrated design context is not one without no-
table challenges, particularly in relevance to the two
contrasting aspects of the design process; the cog-
nitive and analytical aspects. The concept of 'Con-
jecture Analysis' has been highlighted as a mean of
accommodating these aspects (Hillier et al. within
Genlernter, 1988), based firstly on 'conjecture', where
students use "extra-rational and artistic procedures of
analogy, metaphor, sudden flashes of insight, and dis-
placement of concepts to create new ideas" address-
ing the cognitive schemata. The concept is also
based on 'Analysis', where students adopt an ap-
proach of solid rationale and critical assessment, thus
analysing technical consequences of concepts and
theories adopted. Understanding the concepts and
approaches to the integration can aid analysing the
subsequent barriers for effective application of the
integrated design methodology, which is discussed
in the next section.
CHALLENGES SURROUNDING ENVIRON-
MENTAL DESIGN EDUCATION
Building Performance Simulation (BPS) applications
offer designers the capability to generate solid as-
sessment models for different environmental enti-
ties including thermal, acoustics, lighting, and ven-
tilation. BPS applications also present students with
an opportunity to expand their skill set and exper-
tise, and gain constructive analysis proficiency with
regards to complex building physics phenomena
(Charles and Thomas, 2009). Calls for integrated de-
sign framework has been reinforced in particular in
the last decade, mainly as a result of the compelling
demands raised by governments and international
bodies, and the use of advanced computing process-
ing capabilities and algorithms that can support gen-
erating extensive data-sets as a result of complex sim-
ulations.
Recent years have seen notable increase in the
number of surveys, research projects, and case stud-
ies (including the study and survey carried out as
part of this research), that have been carried out with
the clear intent of evaluating BPS tools' suitability
and appropriateness within the architectural design
educational context. These studies were based on
some key criteria that fall under either the usabil-
ity and information management of the BPS appli-
cation interfaces, or the efficiency of the tools as a
design decision support system. The outcomes of
these studies were very useful; highlighting the main
shortcomings and pragmatic issues surrounding the
use of such tools within the architectural design con-
text. Analysisng these outcomes can provide op-
portunities to bridge the gap and address any con-
cern, which is done through presenting guidelines
for adopting a more 'Architect Friendly' approach for
simulation and analysis.
This section discusses a compilation and taxon-
omy of the outcomes of these studies, which is based
on three key questions; "Why use BPS tools", "How to
use BPS tools", and "What to do with the simulation
data".
Motivation for using BPS Tools
In this research project's survey, one section intended
to probe the participants' motivation to incorporate
environmental design concepts and measures. Re-
sponses indicate that the majority (79%) of partici-
pants' motivation for that matter is that 'they have to
do it'; highlighting the pressure/obligation from gov-
ernments and accreditation bodies, or -on a smaller
scale- from the instructors as part of the assessment
process to satisfy specific learning outcomes. Al-
though the participants acknowledge that this inte-
gration is the ethical approach for reducing energy
needs and preserving natural resources, they still ad-
mit that it wouldn't have been in their highest prior-
ities without the pressure. This section investigates
more into the factors affecting the designers' moti-
vation for integrating environmental measures.
Cognitive Creative Nature of Design. The call for in-
tegrating environmental measures and concepts into
the architectural design process has left the design-
ers with a set of new variables to manage and in-
corporate. Many studies, including that by Srivas-
tav et al. (2009), indicate that these variables were
not deeply welcomed. They explain that most ar-
chitects in general do not place environmental vari-
ables/measures at higher priorities, compared -for
example- to design aesthetics. Designers are inclined
more towards 'conjecture', where they think of visual,
spatial, relational, proportional entities of their de-
sign elements. On the other hand, simulation data
are perceived as series of mathematical analytical cal-
culations that require extensive effort to familiarise
with and make sense of; and that do not naturally
fit the solution focused approach designers adopt for
problem solving (Rutherford and Wilson, 2006).
In this sense, many design students develop dif-
ferent interpretation of the 'priorities'. For example,
students may attempt to create a sense of indoors-
outdoors continuity, with no consideration for the ef-
fect this decision can make on solar gains. The out-
comes from the studies indicated that the benefits
offered by an 'integrated design approach' can be
missed by some students. On one hand, students
may perceive the new variables as being 'unimpor-
tant', whereby they could be simply ignored. On the
other hand, some may deem them 'vital', leading to
impractical highly deterministic alterations 'just to
make the numbers look good'.
Suitability to Early Design Stage. In building de-
sign, the earlier stages are the most critical in the
process, where most of the key decisions are made.
Any uninformed decisions at this stage can have se-
vere consequences on all subsequent decisions, and
in turn on the building capacity/performance and
CAAD Education - Volume 1 - eCAADe 32 | 325
 energy consumption throughout its lifecycle. How-
ever, making informed decisions at this early design
stage -where the available data are mostly intricate
and inadequate- is quite challenging. Consequently,
it is very common for designers to revisit these deci-
sions after gaining more details and solid data in later
stages of design.
A central method for students to improve their
comprehension of environmental design measures
and various cause-effect relationships; ultimately en-
riching their design decisions, is through the use of
BPS applications. However, reaching this level of
awareness and understanding, as argued by Bam-
bardekar and Poerschke (2009), requires the gather-
ing of a huge amount of preliminary data and pro-
jected building details, just to be able to start the
simulation process. The required level of details is
normally unavailable -and the information is thus
insufficient- at the earlier stages of design, where
there is constant alteration and revision to the build-
ing's geometry and configuration. In this regard, it
is understandable that some students can see the
use of BPS application to evaluate their building's
design and performance at these early stages -to
an extent- aimless, frustrating, and wasting valuable
time and effort. This in turn have the capacity to ex-
pand the psychological gap between the two par-
allel processes of design and analysis. Designers in
general, according to Mahdavi (2005), do not tend to
prepare this entailed level of details until later stages
of design, and only then a meaningful BPS analysis
can be carried out. He added that this analysis will -
most likely- not be carried out by the architects them-
selves, but rather by specialists and service engineers.
Preference of Experience and Guidebooks. Ac-
cording to Pedrini and Szokolay's (2005) findings,
mathematical simulation models are considered the
least popular in the architectural design community.
These findings rank the BPS applications in the lower
ranks within a list of eighteen design decision sup-
port techniques. The main justification for this rank-
ing is the technical analytical nature of these appli-
cations, which contradicts with the cognitive, reflec-
326 | eCAADe 32 - CAAD Education - Volume 1
tive, deductive nature of the design process. Their
research findings also noted that the designers' rely
more commonly -particularly in early design stages-
on their personal experience and intuition in decision
making, with generic design guidelines and rules-of-
thumb following that also in a higher ranking. These
findings generally fall in line with earlier discussion of
the designers' learning approach, as well as Hillier et
al. (1984) depiction of the design as a "cyclic reflec-
tive deductive process".
The simplicity and usability of design guide-
books -as an efficient reference source- is the main
justification for the high level of inclination shown
by the designers towards them. This can be broken
further down to the fact that these guidebooks are
rather generic (with no particular building specifica-
tion), thus making them simple to navigate and ex-
plore. More importantly, they do not require thor-
ough level of details at any stage, making guidebooks
comparatively more time and effort effective. How-
ever, the main factor that makes guidebooks easy to
use can itself be a big hindrance. The generic nature
and lack of building-specific details and zones' inter-
operability can deem these guidebooks rather pas-
sive with minimum level of interaction, and subse-
quently detached from the design specification. In
other words, guidebooks can be helpful in suggest-
ing what can be done on a generic basis, rather than
highlighting a problem and reflecting on possible
causalities in a specific design case.
Complexity of BPS Tools
Schmid (2008) indicates that design students in gen-
eral are neither attracted nor keen to use the current
environmental performance evaluation tools; mainly
owing to the relatively complex and technical nature
of these tools. Recent years have seen various at-
tempts and approaches to make these tools more
architect-oriented; owing to meet the designers' ap-
titude and expectations However, there are still some
raised issues relating mainly to the BPS interfaces' us-
ability, information management, and lack of guid-
ance throughout the simulation procedure (Attia et
al., 2009). This section investigates in more details
these highlighted issues.
Steep Learning Curve. The designers' effective use
of the BPS tools, according to (Schmid, 2008), en-
tails a rather steep learning curve, where they are
primarily required to develop a solid understanding
and knowledge base of the underlying science and
building physics; these comprise a huge set of con-
cepts and technical terminologies. Designers also
need to relate this knowledge base to the various
features and functionalities implemented in the BPS
tools, aiming to prepare the relevant settings and
supply the required preliminary geometric and non-
geometric design details to be able to run the simu-
lation process.
From an architectural perspective, the interface
of most BPS applications, as argued by many re-
searches including Punjabi and Miranda (2005), can
be deemed rather too technical, complex, cumber-
some, and uneasy to learn. Attia et al. (2009), relate
this to the fact that most of these tools were built
with an engineering-oriented directive with mini-
mum consideration to the architectural community.
More recent research and surveys, however, indi-
cated that much work has been done towards this
matter, leading to noticeable improvements in the
usability of some BPS tools, in terms of adaptation
to the architects' mentality and experience. One
of these tools that are becoming more architect-
friendly is Ecotect, which is gaining increasing pop-
ularity due to its comparatively simpler interface,
highly visual representations, and interactive analy-
sis mechanisms.
Simulation Process Procedure. Warburton (2003)
argues that the effective implementation of the 'Con-
jecture Analysis' method for integrating sustainabil-
ity in the design curricula - discussed earlier in this
paper-, entailed the effective amalgamation of 'reve-
latory activities', which are responsible for enhancing
the students' learning experience through assisting
them to 'ask the right questions'. Bambardekar and
Poerschke (2009), confirm that statement, explaining
that this is accomplished through effective interpre-
tation of the theories and underlying science intro-
duced in the lectures into profound set of tasks to be
carried out within the simulation process. They noted
that "Architects are usually familiar with environmen-
tal concepts, but often do not clearly understand how
to translate the design and performance inquiries into
simulation tasks and evaluate them using ESP's [Energy
Simulation Programs]" (pp. 1307). Schmid (2008) also
indicated that it is not difficult for students to prop-
erly understand theories and building physics, but
what is really challenging for them is to formulate
meaningful procedure that will efficiently help them
assess and evaluate their building's performance.
Design students, in this respect, should be
guided and/or made aware of the set of assessments
involved in the simulation, which is currently quite a
challenge. Even for experienced users, Marsh (2006)
argues that preparing a simulation model on the
grounds of basic requirements targeting the analysis
process, is still quite hard and demanding. It is thus
evident that there is growing need for novel methods
that can bridge the gap between design and simula-
tion, aiding the students throughout the process and
assisting them to ask the correct questions, leading
subsequently to informed design decisions.
Simulation Data Visualisation and Analysis
One of the most crucial aspects in the building simu-
lation process is communicating back the results and
feedback to the designers in a simple manner they
are familiar with. There is no doubt that the cur-
rent BPS applications exploit highly advanced visual-
isation and data representation methods, however,
many challenges and shortcomings are still being
highlighted in relation to communicating simulation
outcomes to the architects. Srivastav et al. (2009), for
example argue that one of the key reasons for these
shortcomings is the complexity of the resulting data,
and thus the challenge to correlate the data to de-
rive causalities and design decisions. This section dis-
cusses the challenges associated with data analysis
and representation techniques utilised by BPS appli-
cations.
CAAD Education - Volume 1 - eCAADe 32 | 327
 Design vs. Simulation Visualisation Techniques.
The rapid advances of current CAD and BIM tools
had great benefits that go far beyond the sole expan-
sion of productivity scale. These advances also sup-
ported the production of highly interactive and im-
mensely graphical 3D design visuals. Visualisation is
a fundamental aspect in the architectural design pro-
cess, as it is a representation of the designers' con-
ceptual models that directly relates to their cognitive
schemata. On the other hand, simulation data rep-
resentation techniques do not follow the same line;
being mostly sophisticated and highly technical, as
they are based upon empirical models that can fail to
"reconcile the relationship between design actions and
performance outcomes" (Toth et al., 2010, pp.315).
It can thus be argued that the analytical nature
of the simulation and the resulting outcomes lack
sufficient consideration to the existent norms domi-
nant within the design realm, thus incapable of inter-
preting the resultants (as valuable as these are) into
meaningful decisions. The survey carried out within
this project supported that; indicating that the ma-
jority of participants indicated that the current data
representation mechanisms are not as problematic in
terms of defining performance problems as translat-
ing this definition into possible causalities and sub-
sequent design decisions (Figure 1). Current BPS ap-
plications, according to Attia et al. (2009), are thus
328 | eCAADe 32 - CAAD Education - Volume 1
falling short in terms of relating the design to the
performance outcomes, which can compromise the
effectiveness of these tools as decision support sys-
tems, and ultimately reduce the value of the "inte-
grated simulation".
Spatiotemporal Dimension of Simulation Data.
The simulation data, as argued by Yan and Jiang
(2005), is comprised of two essential dimensions;
special and temporal, which should both be consid-
ered carefully in the means of analysis and represen-
tation. This, however, is quite challenging; the spa-
tiotemporal 4D attribute of the simulation data has
a rather convoluted nature, which is problematic to
visualise and interpret. In the design community,
whereby dealing with only three dimensional mod-
els is the norm, introducing an additional temporal Figure 1
dimension has the capacity to increase the level of Participants'
uncertainty and confusion, and thus affecting the re- responses for
sulting decisions. It is thus essential to utilise a visu- evaluating BPS
alisation technique that can accommodate the tem- applications'
poral data and simplify the representation process. effectiveness in
Most of the current BPS tools utilise visualisation ap- defining
proaches that fall short in dealing with the complex- performance
ity of the nature of multidimensional simulation data. problems, as well as
A key factor in dealing with and effectively rep- identifying possible
resenting multidimensional data is through expand- causalities and
ing the degree of interactivity offered to the users. making subsequent
One medium that has proven very effective in terms decisions.
of incorporating high levels of interactivity and data
visualisation is Virtual Reality (VR). This medium, ac-
cording to Prazeres and Clarke (2003), has the capac-
ity to 'bring alive the informational domain', and of-
fer novel techniques to visualise information that is
not relatively as simple in other techniques. Provid-
ing design students with additional levels of inter-
action, and opportunities to explore various multidi-
mensional aspects of their design can assist in break-
ing down the complexity of the imbedded informa-
tion, and thus help in the knowledge extraction and
comprehension of the presented materials.
 THE PROPOSED METHOD AND EXPERI- implementation process, the proposed method and
MENT the game demo were tested to stand upon the key
Based on the taxonomy of challenges presented ear- advantages and areas of improvement for this ap-
lier, this research proposed a new method to resolve proach. The testing and evaluation sessions involved
these challenges, and to facilitate integrating envi- 28 design students and 11 instructors from three UK
ronmental measures in the architectural design cur- universities. The testing session was a combination
riculum. The proposed method is an environmental of questionnaires and structured interviews aiming
design e-tutor game (Sarhan and Rutherford, 2011), to gather both quantitative and qualitative feedback
which presents an interactive narrative 3D virtual ex- that can support the assessment of this method con-
perience for the design students, to examine their de- ceptual basis. The participants were also presented
sign geometric and non-geometric parameters. The with the e-Tutor game as part of the testing session,
method is based on the utilisation and integration and were subjected to the different features imple-
of Multi-Agent Systems and Data Mining Techniques mented and discussed earlier. Figure 2 presents a
to create additional software modules on top of the screenshot of the proposed method's demo game
basic game engine technology (C4 Game Engine in that was part of the testing sessions.
Figure 2 this project). These additional modules are responsi-
Abstract illustration ble for communicating with the BPS application and
for the proposed pulling data from it, and later formatting and storing
method, depicting the data in its own Data Warehouse. Once the game
the outputs of each is run, the design model is analysed and coupled with
key node the stored simulation data to form an informational
(simulation and ED model, which will be the basis for the communication
Game), as well as and feedback with the design student.
the tasks that are A Reporting Agent (part of the implemented
automatically MAS), is responsible for creating a dialogue with the
performed, and students; presenting them with initial findings after
those that require analysing the simulation data (using DM and knowl-
manual edge extraction mechanisms). This dialogue can ei-
intervention from ther be feedback in the form of reports, or 'interroga-
The feedback provided from this experiment -
the students (Image tion routines' that attempt to fill in the missing gaps
although only indicative; representing the small sam-
from Sarhan, 2012). in the informational model and get feedback from
ple involved in testing and evaluation- was quite
the student according to their experience and pref-
valuable and informative. One of the main high-
erences. The Reporting agent can finally present the
lighted advantages of the proposed approach is its
student with a set of zone-specific design guidelines
positive effect in motivating the students, through
that can resolve any raised issues/problems in the
offering a higher level of interactivity within a 'game'
design, and ultimately improve the building perfor-
context; one that their generation is well accustomed
mance. The student can then manually update the
to and familiar with, and thus can be more engage-
design and the CAD model, and thus the process can
ment in. Another noted advantage was the capac-
be repeated to assess the effect of the new design up-
ity to assist students in asking the right questions
dates. Figure 1 presents an abstract overview of the
through encouraging them into more in-depth inves-
main entities in the proposed method, with the flow
tigate more of the simulation data. The method also
of information ans main outputs of these entities.
attempts to act as a building-specific guidebook; of-
After the initial survey, the development and
fering simple guidelines to improve the design and

CAAD Education - Volume 1 - eCAADe 32 | 329

 performance. The method also exploits the cause-
effect relationships; depicting the consequences of
the strategies and decisions that the students choose
to make. Finally, another advantage mentioned was
that the method offers a tutoring system that the stu-
dents can exploit in their own time and convenience.
However, there were also some highlighted dis-
advantages and areas for further development and
improvement. One of the main shortcomings (ex-
pressed mainly bu the instructors) is that the method
offers only abstract level of information, without
enough links to the underlying building physics, to
allow students to exploit the information in more
depth. Another noted shortcoming was that there
is no compilation of information provided during the
game; the students mainly wanted the problems and
guidelines compiled in a printable format acting as a
checklist for design updates. Learning new interface
(the game engine) and having to build 3D models
were also considered to be risks in consuming time
and effort required before running the game. Finally,
the transformation of data between the applications
involved in this method is considered cumbersome
and rather problematic in terms of the time and effort
required, as well as jeopardising the data. These ad-
vantages and disadvantages were considered in for-
mulating the set of guidelines presented in the next
section.
330 | eCAADe 32 - CAAD Education - Volume 1
GUIDELINES FOR RESOLVING THE CHAL-
LENGES
A set of key guidelines can now be presented, based
on the taxonomy of challenges discussed earlier This
set is also based on this research survey experiment
and findings, and recommendations and conclusions
presented in number of research projects and case
studies. These guidelines present some basic speci-
fication for developing new methods and technolo-
gies for integrating environmental measures in the
design curriculum.
• The approach should adopt a 'revelatory'
nature (Warburton, 2003); guiding students
through the simulation process, and allowing
them to 'ask the right question' through ex-
ploring different attributes of their design.
• The method should support the conjecture
analysis approach through constant reports
to cause-effect relationships, and highlight-
ing any pragmatic data patterns that can af-
fect the building performance. This should
be carried out while highlighting any possible
causalities for problems, and linking these re-
ports to the underlying science and building
physics.
• The method should offer constant communi-
Figure 3
cation with the students through Q&A rou-
A screenshot from
tines and feedback reports. In so doing, the
the proposed
students' decisions and their impact on the
method's demo
building performance can be rapidly assessed
game (Sarhan
and criticised. This communication will also
2012); presenting
enrich the students' interactive narrative vir-
the initial basic
tual learning experience.
reports and
interrogation
• The directive of being an effective decision
routines generated
support system should be rigidly accom-
by the Reporting
modated, through the presentation of zone
Agent.
and building-specific design guidelines. This
should be done while presenting clear ratio-
nale behind these suggestions in light of the
theories presented in the lectures. This ap-
 proach has the capacity to offer instant dy-
namic analysis; offering simplified feedback
and outcomes similar to those of generic
guidebooks, but with higher levels of interac-
tion and specification.
• The method should adopt advanced data
representation visualisation mechanisms (for
both geometric and non-geometric design
parameters); that can accommodate the con-
ceptual visual nature of design and the mul-
tidimensional spatiotemporal nature of the
simulation data. One medium that has been
proven effective in this accommodation is the
3D interactive narrative virtual context, like
that of the 3D games.
• A 'layered' approach like Prazeres and Clarke's
(2005) "Integrated Performance Views" can be
adopted, which presents information in a sim-
ple hierarchical style. It starts with a rather ab-
stract level, which can be further investigated
by students according to their decisions. Such
approach can overcome overwhelming stu-
dents with excessive information at the initial
stages, which can be a clear barrier in compre-
hending and analysing this information.
• The method should focus on 'simplicity'
and 'abstraction, while offering a reasonable
level of interactive interrogative mechanisms.
Higher levels of control and interaction in the
learning context can ensure greater sense of
control, and more freedom for investigating
various aspects of the problem. This in turn
can have a positive effect on the students mo-
tivation and engagement in the learning pro-
cess, and subsequently on the level of com-
prehension and decision making.
• The method can dynamically provide any ini-
tially required data using default values, to
reduce the preparation time needed before
starting the simulation. In this case, it is im-
portant for the students to be well informed
about these default values and their implica-
tions. Reducing the preparation time can in-
ject more confidence to run the simulation,
and resolve some of the motivational barriers
noted earlier.
• The method should be of instructive nature,
acting like a personalised e-Tutor. It should
be able to analyse the massive amount of sim-
ulation data and present abstract knowledge
back to the students. In order to achieve
this, utilising techniques like Data Mining for
data analysis and knowledge extraction can
be very effective in uncovering and simplify-
ing details that can otherwise be difficult to
uncover, investigate, and act upon.
• A modular built-in Data Warehouse should be
incorporated, where it can pull in all the re-
quired simulation data. In so doing, a reliable
informational model can be dynamically con-
structed and updates, and a flexible structure
for storing, arranging, searching, and retriev-
ing data can be attained.
• The method should address the huge de-
mand for smooth transition of information
between the design and simulation, which
can be accomplished through creating a more
centralised informational model that can hold
all the information related to the building.
Currently, the huge progress in BIM applica-
tions can address this guideline where the in-
formational model can work in conjunction
with the BPS tools.
CONCLUSION
Integrating sustainability in the architectural design
process is an emerging imperative. Subsequently
educational institutions are challenged to prepare
new generations of architects that can effectively ac-
commodate sustainability in their design and con-
cepts. This integrated design education framework
has proven to be rather pragmatic, and its attainment
CAAD Education - Volume 1 - eCAADe 32 | 331
 depends on addressing and resolving number of bar-
riers and challenges. The paper presented a brief
outline of the authors' research project, which pro-
posed a framework for facilitating an effective inte-
gration of environmental measures in the design pro-
cess. The paper focused mainly on presenting a tax-
onomy of the challenges in light of the study and sur-
vey carried out as part of this research project. The
paper also presented guidelines for adopting new
methodologies/approaches for effective integration
of sustainability, taking into consideration the archi-
tects' mentality, expectation, and approach to learn-
ing and problem solving. The research is currently
in the process of developing an updated version of
the environmental design eTutor, acting upon afore
mentioned criterion.
REFERENCES
Altomonte, S 2009, 'Environmental Education for Sus-
tainable Architecture', Review of European Studies,
1(2), pp. 12-24
Altomonte, S, Reimer, H, Rutherford, P and Wilson, R
2013 'Towards Education for Sustainability in Uni-
versity Curricula and in the Practice of Design', PLEA -
Sustainable Architecture for a Renewable Future, Mu-
nich
Attia, S, Beltran, L, De Herde, A and Hensen, J 2009
'Architect friendly: a comparison of ten different
building performance simulation tools', Eleventh In-
ternational IBPSA Conference, Glasgow, Scotland, p.
204–211
Bambardekar, S and Poerschke, U 2009 'The architect as
the performer of energy simulation in the early de-
sign stage', Eleventh International IBPSA Conference,
Glasgow, Scotland, p. 1306–1313
Charles, PP and Poerschke, U 2009 'Building Performance
Simulation in Undergraduate Multidisciplinary Ed-
ucation: Learning from Architecture and Engineer-
ing Collaboration', Eleventh International IBPSA Con-
ference, Glasgow, Scotland, pp. 212-219
EDUCATE, White Paper 2012 'White Paper-Sustainable
Architectural Education', [Online], Available:
http://www.educate-sustainability.eu/white-papers
[08 May 2014]
Gelernter, M 1988, 'Reconciling Lectures and Studios',
Journal of Architectural Education, 41(2), pp. 46-62
332 | eCAADe 32 - CAAD Education - Volume 1
Pedrini, A and Szokolay, S 2005 'The architects approach
to the project of energy efficient office buildings in
warm climate and the importance of design meth-
ods', Building Simulation, Montreal, p. 937–944
Prazeres, L and Clarke, JA 2003 'Communicating Build-
ing Simulation Outputs to Users', Building Simula-
tion, Eindhoven, pp. 11-14
Rutherford, P and Wilson, R 2006 'Educating environ-
mental awareness: creativity in integrated environ-
mental design teaching', Proceedings of the 40th an-
nual conference of the Architectural Science Associa-
tion ANZAScA, Adelaide, pp. 261-269
Sarhan, A 2012, Utilising Games Technology for Environ-
mental Design Education, Ph.D. Thesis, University of
Nottingham
Sarhan, A and Rutherford, P 2011 'Environmental De-
sign eTutor: Utilizing Games Technology for Envi-
ronmental Design Education', Proceeding of the 29th
eCAADe Conference, University of Ljubljana, pp. 699-
708
Schmid, AL 2008, 'The Introduction of Building Sim-
ulation into an Architectural Faculty: Preliminary
Findings', Journal of Building Performance Simulation,
1(3), pp. 197-208
Srivastav, S, Lannon, S, Alexander, D and Jones, P 2009
'A Review and Comparison of Data Visualisation
Techniques used in Building Design and in Building
Simulation', Eleventh International IBPSA Conference,
Glasgow, Scotland, pp. 1942-1949
Warburton, K 2003, 'Deep Learning and Education for
Sustainability', International Journal of Sustainability
in Higher Education, 4(1), pp. 44-56
Yan, D and Jiang, Y 2005 'An overview of an inte-
grated building simulation tool - designer’s simula-
tion toolkit (DEST)', Building Simulation, Montreal, p.
1393–1400
Digital Recipes
A diagrammatic approach to digital design methodologies in
undergraduate architecture studios

Miguel Paredes Maldonado

University of Edinburgh
http://www.eca.ed.ac.uk/
miguel.paredes@ed.ac.uk

This paper is concerned with the development of a digitally-informed teaching

methodology for architectural design courses within the context of undergraduate
education. Besides situating this methodology within the broader theoretical
framework of current architectural discourses, this paper will also discuss the
practical implications of its implementation within the context of the second year
architectural design course at the University of Edinburgh, which will be
examined as an ongoing case study.

Keywords: Diagram, Generative Design, Fabrication, Recipes, Case Study

THE DIGITAL IN ARCHITECTURAL DESIGN
The notion that the massive adoption of digital tech-
nologies has played a key role in the advent of a radi-
cal paradigm shift in architectural production during
the last two decades is nowadays a hardly contested
fact. However the clarification of what this paradigm
shift does exactly amount to -in strictly disciplinary
terms- has been the subject of many diverging theo-
retical approaches.
Perhaps one of the most compelling early visions
of the influence of the digital in architecture is Bart
Lootsma's idea of architecture presenting itself as a si-
multaneous, three-fold investigation on topological
geometry, robotic material production and genera-
tive, kinematic space - all of which could, in his mind,
be united transversally under the umbrella of a hy-
bridized notion of space (Lootsma 1999).
These three themes - the articulation of com-
plex, fluid relationships, the possibility of architec-
ture becoming industrially manufactured, and the
autopoietic capabilities of digital processes for spa-
tial production - have progressively become well-
established fields of contemporary design research,
which often overlap one another.
Another approach to the possibilities of the dig-
ital is contained in Peter Eisenmann's writings from
the early nineties. For Eisenman, the paradigm
shift triggered by the digital in architecture revolved
mainly around the question of vision or, more specif-
ically, of the dramatic possibilities for the reinven-
tion of projected vision (Eisenman 1992). Eisenman
argued that the organizational qualities of architec-
ture were determined by the constrains of its own
representational tools, and therefore he explored the
notion of folding as an operating framework that -
while remaining representational - could potentially
overcome the limitations of Cartesian descriptions of
space.

CAAD Education - Volume 1 - eCAADe 32 | 333

 John Frazer also signalled two very interesting
fields of operation that would, according to his view,
characterize the progressive implementation of dig-
ital technologies in architectural production. On the
one hand, Frazer noted the possibility of architecture
becoming an organ of interaction with the global
informational environment, therefore rendering vis-
ible the intangible qualities of the omnipresent dig-
ital networks that are juxtaposed to our physical ex-
istence. On the other hand, he also noted how the
cultural emphasis in design had progressively trans-
ferred from the product to the process (Frazer 1995).
If we collate all these views together, we could
summarize that a methodological approach to ar-
chitectural design that is digital in nature would be
based on the articulation of two aspects: On the one
hand, a continuous productive environment based
on the exploration of relational organizations. On the
other hand, the capabilities of such organizations to
generatively sculpt spaces that are articulated -and
therefore represented- through the tectonic ethos of
their associated fabrication processes.
IMPLEMENTATION IN UNDERGRADUATE
STUDIO TEACHING
Considered as a whole, the visions outlined above
can provide a sufficient general overview of the dis-
ciplinary aspects of architectural design that have
been most substantially transformed by digital tech-
nologies in the last two decades. However, it is still
unclear whether the depth and breadth that this
radical shift of focus anticipated twenty years ago
has been successfully implemented in design edu-
cation. Whereas an increasing number of graduate
and postgraduate programs have been able to articu-
late a consistent pedagogic framework that incorpo-
rates the particularities and capabilities of the digital
world, this task has proved to be much more difficult
in the case of undergraduate studies.
On the one hand, many younger architecture
students struggle to develop their own individual de-
sign processes in a manner that guarantees a suffi-
cient degree of spatial and functional sophistication.
334 | eCAADe 32 - CAAD Education - Volume 1
Hence, they often become frustrated by the shallow-
ness of their own design arguments. On the other
hand, undergraduate students also tend to be inex-
perienced in the use of digital tools, which -to make
matters worse- are often understood as simple rep-
resentational skills to be acquired rather than as de-
sign tools. Because of this, the incorporation of dig-
ital thinking into the learning experience often has
the undesired effect of hampering the already prob-
lematic development of the students' design process
rather than leveraging it.
One of the reasons for this state of affairs is
that instructors often stress the importance of digital
skills for architectural production, but rarely outline
a methodological model that facilitates a consistent
approximation to the design process while being dig-
ital in nature.
DIAGRAMS AND THE DIGITAL WORKFLOW
Therefore, this paper intends to address this peda-
gogic challenge by putting forward a series of digi-
tally informed teaching strategies for the design stu-
dio. These strategies are primarily based on the de-
velopment and dissemination of a series of digital
procedures for generating rigorous geometric orga-
nizations, whose spatial qualities possess a high de-
gree of architectural potential. Furthermore, this pa-
per argues that these geometric procedures can be
read as diagrammatic processes, and therefore can
be articulated by tapping into the rich theoretical
sources that have addressed the notion of the dia-
gram -both from outside and from within the disci-
plinary framework of architecture-.
In terms of architectural design, a diagram is sim-
ply a visual, non-representational device that is used
as an organizational tool. In essence, diagrams con-
stitute tools for compressing and assembling infor-
mation in the form of techniques, tactics, situations
and functions. As such assemblages of information
can support a multiplicity of readings a diagram can
potentially give rise to a very large number of differ-
ent organizational effects (Van Berkel & Bos 2010).
Diagrams are often confused with maps, which
establish a codified relationship of correspondence
between two systems -one being the initial source,
the other being the map itself- and therefore are
purely representational. As opposed to this, dia-
grams are generative in the sense that they can re-
verse the operation produced by mapping by be-
coming the starting point for generating a real con-
dition (Solomon 2007).
This understanding of diagrams as generative
systems is strongly connected to Deleuze and Guat-
tari's work (1987) on the distinction between topo-
logic and metric spaces. For Deleuze and Guattari,
metric systems are characterised by notions such as
'length' and 'area', which remain unaltered through-
out any geometric operations taking place within
them, whereas in topologic systems the aspects that
remain invariant are the number of dimensions and
their degrees of connectivity (Deleuze & Guattari
1987). Therefore, topologic systems have no scale
as such, but are rather organised around a network
of connections and relations between points. Hence
any architectural diagram with generative properties
is necessarily a topologic system which, rather than
producing a single, specific form, determines an ab-
stract organizational structure that can be materi-
alised through a potentially infinite number of in-
stances (Umemoto & Reiser 2006).
Once diagrams are conceptualised as holding
both topologic and generative qualities, three addi-
tional characteristics emerge:
Firstly, the development of diagrams is necessar-
ily deployed over time, in a context in which matter
is considered as an animated, dynamic flux which is
permanently activated (Kwinter 2002). The tempo-
ral dimension of diagrams is, precisely, what allows
them to be generative. Once animated by this tem-
poral flux, a diagram can trigger formal processes of
progressive differentiation. In such processes, topo-
logic and relational invariants remain but are assem-
bled with different sets of metric properties, there-
fore generating multiple material instances of the
original topologic diagram (Umemoto & Reiser 2006).
Thus, diagrams simultaneously belong to the realms
of the real and the virtual. Their products are gener-
ated as material actualisations of their abstract, for-
mative principles, and emerge through interlinked
processes of progressive differentiation (De Landa
2010).
Secondly, the generative process that can poten-
tially be deployed by a given diagram only acquires
metric qualities -such as scale and size- once it is ap-
plied to specific material assemblages. However, this
process is regulated by non-metric qualities such as
differences of intensity between points, which de-
fine the topologic form of the diagram (Umemoto &
Reiser 2006). Therefore, diagrams can be understood
as abstract, vector-based organizations that, rather
than numerical data, contain relational information
that is qualified by differences of potential between
points.
Finally, being animated by both material and
temporal fluxes, and organised by differences of po-
tential, generative diagrams are in a continuous state
of production. As a consequence of this, they give
rise to a multiplicity of progressively differentiated
outputs. Such outputs reflect not only the topolog-
ical distribution of their originating diagram, but also
our intervention as designers through the selective
introduction of mutations, folds and local connec-
tions. This combinatory productivity fosters a pro-
gressive formal synthesis that is defined within the
framework of what could be defined as a 'reproduc-
tive community' (Umemoto & Reiser 2006). Hence,
the roles or the topologic designer are related to the
choice of a material substrate for the deployment of
the diagram, the intentional filtering of the dynam-
ically generated instances, and the selective modu-
lation of local mutations and relational connections
into the original diagram.
DIGITAL DIAGRAMMATIC PROCESSES IN
THE DESIGN STUDIO
At this point, it is clearly possible to consider the
qualities and capabilities of a diagrammatic design
practice -outlined in the previous paragraphs- as par-
ticularly well attuned to the mechanisms and pro-
CAAD Education - Volume 1 - eCAADe 32 | 335
 cesses of a digital approach to architectural produc- as some of these conditions. In clear resonance with
tion. This is reflected in aspects like the primacy of John Frazer's ideas previously mentioned in this pa-
an abstract, topologic organization over the form of per, the course handbook also emphasizes the need
Cartesian metric space, the emphasis on generative, for architecture to be approached as an instrument
time-driven processes of progressive formal explo- of mediation between the scale of the individual and
ration, and the understanding of design outputs as the scale of collective culture.
parts of a larger continuum of production and manu-
facturing. Therefore, it can be argued that any design DIGITAL RECIPES AS PROCEDURAL DIA-
practice that is digital in nature (at least considering
GRAMS
the way in which such practice has been outlined at
As a vehicle for channelling this set of interests and
the beginning of this paper) can be articulated as a
ambitions, the course proposes the development of
diagrammatic process.
a mid-sized library project in the city of Madrid as
Taking these considerations as a starting point
its central theme. Considering that the cohort size
and in my role as a Course Organiser for the sec-
for the 'Any Place' studio is relatively large (120 stu-
ond year undergraduate architectural design studio
dents), young and culturally diverse, the main chal-
at the University of Edinburgh, I have attempted
lenge of the course is that of articulating a digitally
to articulate a pedagogical strategy that relies on
informed, diagrammatic approach to design in a way
the capabilities of three-dimensional modelling soft-
that was both extremely clear and easy to dissemi-
ware to control generative, diagrammatic design pro-
nate.
cesses. In doing so, it is expected that students will
Therefore, the first half of the course was struc-
see the development of their individual approaches
tured around the development of a series of spatial
to design facilitated and articulated within a frame-
and organizational explorations for a site-less library
work that is rigorous, transferable and consistent
program, which were elaborated by students using a
while simultaneously allowing for ambitious spatial
set of seven open procedural diagrams as points of
and organizational explorations. The following is a
departure. These diagrammatic instructions were in-
description of the particularities of this pedagogical
tentionally referred to as 'digital recipes' in order to
strategy as it is currently being implemented.
clarify the methodological approach that they were
The second year studio in which this experimen-
embedding.
tal approach to architectural design pedagogy is be-
Hence, and using the metaphor of cooking
ing undertaken in entitled 'Any Place' and runs dur-
recipes, 'digital recipes' were introduced to students
ing the second semester of the academic year. Fol-
as procedural diagrams for the transformation and
lowing the general guidelines set by the four-year
assemblage of matter. As in cooking books, these
BA/MA Architecture degree it is part of, 'Any Place'
recipes are articulated as a series of interconnected
is aimed at exploring how architecture can be in-
steps or actions that orchestrate the interaction be-
debted to things beyond its own vocabulary. In do-
tween diverse components. Variations in the param-
ing so, and whereas previous studios within the de-
eters that inform such interactions have the potential
gree explore the ways in which immediate contex-
to significantly alter the final outcome of the process.
tual inputs could inform the design process, 'Any
Moreover, there is a certain expectation that, once
Place' specifically deals with the exploration of archi-
the basic interactions of a given recipe are under-
tectural design conditions that extend beyond the
stood, the cook (or, in our case, the designer) can in-
level of the local. The course handbook - compiled
troduce substantial modifications in the process that
by me- lists function, programme, spatial experience
will yield alternative outputs.
or insertion within a broader socio-cultural context
Consequently, 'digital recipes' outline general

336 | eCAADe 32 - CAAD Education - Volume 1

Figure 1
Digital Recipes 2
(Smooth, Punctured
Space) and 3 (Space
of Continuous
Strips)

design procedures for generating certain spatial and
organizational assemblages, whereas it is up to the
students to propose their own versions, modifica-
tions and alterations of such protocols in ways that
fulfil both the requirements of the design brief and
their own design inclinations.
The seven 'digital recipes' (figure 1, 2) used in
the Any Place course are based on the Rhinoceros
modelling software due to its combination of acces-
sibility, data processing power and geometric rigor-
ousness. Each recipe was presented under a name
that summarised the spatial properties it could po-
tentially generate, therefore giving rise to the follow-
ing set of diagrammatic abstractions, listed in order
of increasing complexity:

Figure 2
Digital Recipes 4
(Carved, Sculpted
Space) and 5 (Space
of Structural
Frames)

CAAD Education - Volume 1 - eCAADe 32 | 337

 1. Recipe for a flowing, repetitive space
2. Recipe for a smooth, punctured space
3. Recipe for a space of continuous strips
4. Recipe for a carved, sculpted space
5. Recipe for a space of structural frames
6. Recipe for a connective tiling space
7. Recipe for a folded, tiling space
From a formal point of view, 'digital recipes' were
presented to students as detailed, step-by-step il-
lustrated guides to generate each particular spa-
tial effect using Rhinoceros commands. Each recipe
was also associated to at least one contemporary
built precedent that exemplified these spatial effects.
From a methodological point of view, the 'digital
recipes' implemented a set of geometric operations
that contained exclusively topologic relationships or,
in other words, connective relationships between
components. No considerations of dimensions or
scale were introduced, and hence the procedures de-
scribed did not contain any metric constrains. By
demonstrating or testing the recipes, students in-
troduced such metric constrains into their diagram-
matic structure, together with additional sets of con-
nective relationships that complemented those of
the original topologic assemblage.
STAGE 1 - INITIAL EXPLORATIONS
Recipes were distributed and demonstrated during
the first three weeks of the course, and by the fourth
week students were asked to serially develop and
present at least two different spatial studies for a site-
less library building using two 'digital recipes' of their
choice as points of departure.
This emphasized the diagrammatic qualities of
the process: On the one hand, the consideration of
additional constrains -such as the scale and organi-
zation of a library program, or their possible arrange-
ment and differentiation- led students to progres-
sively incorporate them into their recipe-based spa-
tial experimentations, therefore producing multiple
338 | eCAADe 32 - CAAD Education - Volume 1
iterations of the same spatial study over time, each
one introducing an additional degree of complexity.
Students willing to expand and deepen their explo-
rations were invited to consider aspects such as ma-
teriality, tectonic organisation and possible articula-
tions of social interactions within the framework of
the library considered as an urban public space.
On the other hand, the introduction of metric in-
formation and additional constrains, as well as the
manipulation of the recipe procedures led these spa-
tial studies to quickly transcend the static, generic
qualities of the original recipes, effectively mobiliz-
ing them as generative diagrams and demonstrating
their ability to be materialised as a potentially infinite
number of dynamically generated instances.
Students were asked to document the devel-
opment process from the generic 'digital recipe' of
their choice into a conceptual library prototype and
present it as a consistent design argument, using di-
agrams, axonometric drawings, plans, sections and
elevations. Students also produced multiple physi-
cal models experimenting with different geometric
variations and arrangements, therefore introducing
additional constrains related to the logics embedded
into the digital fabrication of such models, which had
the potential to suggest possible tectonic arrange-
ments for future developments of their projects.
Figure 3
Student work
image: James
Duffield
Figure 4 STAGE 2 - PROJECT DEVELOPMENT
Student work After the first stage of exploration was completed,
sample: James students travelled to Madrid and were assigned spe-
Duffield cific sites to implement their library projects. This led
to a remobilization of the diagrammatic process ini-
tiated in the previous stage, in which students were
asked to test the viability of their initial exploratory
schemes against the specific conditions of the con-
text in which it was going to be implemented.
The remainder of the course has been spent pro-
ducing and testing as many design iterations as pos-
sible, in an attempt to articulate a continuum of re-
sponses to the initial design problem that would pro-
gressively incorporate an increasing number of con-
strains and considerations. As a general summary,
the following inputs have been introduced:
Firstly, metric considerations of scale and propor-
tion, related to the specific conditions of the site. Sec-
ondly, relational qualities such as the organization
of the massing -attuned to site characteristics such
as orientation, topography, boundary geometry and
boundary conditions. Thirdly, other non-geometric
constrains, such as those concerned with both the ar-
ticulation of social spaces and the response to spe-
cific social and cultural conditions at a metropolitan
scale.
In order to emphasize the productive, accumu-
lative value of this design process, students were
currently expected to materialise each of their de-
sign iterations in the form of a combination of Carte-
sian representations, physical models and three-
dimensional digital models. This effort of materiali-
sation played an important role in the design process
by constituting an extremely efficient source of feed-
back on the validity of the proposed solutions, which
in turn initiated further design iterations.

ANALYSIS OF WORK SAMPLES

At the time of writing this paper, the project devel-
opment stage had ended and students had submit-
ted their final portfolios for assessment. At this stage,
most projects have a substantial level of resolution,
and therefore this paper will conclude by analysing

CAAD Education - Volume 1 - eCAADe 32 | 339

 the design processes that were undertaken by three Figure 5
selected students. Student work
James Duffield took the recipe for a 'space of sample: Lynn Lin
structural frames' as a base to trigger an investigation
on the progressive sectional variations of an elon-
gated structure. Branching, compressions and ex-
pansions of space constituted the spatial vocabulary
of his first proposals, which also incorporated pro-
grammatic and organizational aspects. Further in-
vestigations were concerned with the serialised vi-
bration that this formal approach could potentially
offer, especially in regards to the articulation of both
shading and lighting. Once confronted with the
physical constrains of the Madrid project site, James'
spatial strategy was extended to the ground floor
plane. This led to a final design solution that or-
ganises the library program as a single, continuously
flowing space that is diverted into various functional
branches with distinct sectional qualities, each artic-
ulating local relationships with the different bound-
ary conditions of the site (figure 3, 4).
Lynn Lin's work initially tapped into the recipe for
a 'smooth, punctured space' to develop an open-plan
library prototype in which different strategies for ar-
ticulating spatial relationships between floor slabs
were tested. Her initial studies used the deforma-
tion along the vertical axis of an otherwise regular
grid to deploy a topographic field of courtyards, folds
and ramps, along which functional areas were laid
out in the form of overlapping patterns rather than
enclosed spaces. This resulted in a spatial experi-
ence that emphasised continuity and flow. Once con-
strains present in the Madrid site were introduced,
Lin's design response was articulated through the
complex functional relationships of two compressed,
horizontal spaces: a ground-level organic topogra-
phy that occupies the whole site and an elevated,
punctured slab with a strictly rectangular perimeter
and subtle sectional variations (figure 5).

340 | eCAADe 32 - CAAD Education - Volume 1

Figure 6 Edda Steingrimsdottir blended together the
Student work recipes for a 'carved, sculpted space' and a 'space of
sample: Edda continuous strips' to organize a system with the abil-
Steingrimsdottir ity to penetrate existing urban fabrics while simulta-
neously remaining highly organised from a program-
matic point of view. The structural, tectonic and ex-
periential aspects originated by the interaction be-
tween the resulting series of differentiated functional
bands where the subject of further design research.
After returning from Madrid, Edda was able to consol-
idate her strategy for 'cutting through' urban fabrics
by proposing the colonization of the derelict interior
of a typical building block. Her original system of par-
allel functional bands was formally adapted to fit the
boundary conditions of her intervention, and walka-
ble roof topographies emerged as a reaction to ex-
isting differences in height. Edda's final proposal re-
oriented public flows by opening up the traditionally
enclosed blocks and articulating functional spaces
with an urban scale within them (figure 6).

CONCLUSIONS
Digital techniques of spatial exploration constitute
an excellent vehicle to articulate pedagogic ap-
proaches to design that are based on the develop-
ment of open diagrammatic processes. Introducing
these methodologies in undergraduate design stu-
dios has two main effects. On the one hand, stu-
dents can be liberated from the misunderstandings
and anxieties normally associated with the use of
digital technologies in architecture. On the other
hand, this approach can significantly leverage the
rigour, depth and breadth of the architectural design
enquiries being conducted -especially in the early
stages of undergraduate studies- therefore consti-
tuting a major improvement to the overall learning
experience. Finally, since this methodological ap-
proach unashamedly emphasizes process over final
results, it is particularly well attuned to contemporary
sensibilities that consider the materialisation of archi-
tecture as the continuously animated fabrication of a
progressively differentiated milieu of spatial and tec-
tonic assemblages.

CAAD Education - Volume 1 - eCAADe 32 | 341

 REFERENCES
Van Berkel, B and Bos, C 2006, 'Diagrams', in Mark, G
(eds) 2006, The Diagrams of Architecture, John Wiley
& Sons, London, pp. 222-227
Deleuze, G and Guattari, F 1987, A Thousand Plateaus:
Capitalism and Schizophrenia, University of Min-
nesota Press, Minneapolis
Eisenman, P 1992, 'The Affects of Singularity', Architec-
tural Design, 62, pp. 42-45
Frazer, J 1995, 'The Architectural Relevance of Cy-
berspace', Architectural Design, 64, pp. 76-77
Kwinter, S 2002, Architectures of Time: Toward a Theory of
the Event in Modernist Culture, MIT Press
De Landa, M 2010, Deleuze, History and Science, Atropos
Press, New York
Lootsma, B 1999, 'Emergent dimensions, information
technologies and evolutionary architectures', in Zell-
ner, P (eds) 1999, Hybrid Space, Rizzoli, London, pp.
8-16
Reiser, J and Umemoto, N 2006, Atlas of Novel Tectonics,
Princeton Architectural Press, New York
Solomon, J 2007, 'Seeing the cities from the trees', in
Abruzzo, E, Ellingsen, E and Solomon, J (eds) 2007,
Models, 306090 inc.

342 | eCAADe 32 - CAAD Education - Volume 1

Augmented Reality in architectural studio learning:
How Augmented Reality can be used as an exploratory tool in the design
learning journey

David Morton
Department of Architecture & the Built Environment, Northumbria University at
Newcastle, Newcastle-upon-Tyne, England UK.
http://www.northumbria.ac.uk/sd/academic/ee/staff/davidmorton
david.e.morton@northumbria.ac.uk

The boundaries of augmented reality in the academic field are now being
explored at an ever increasing level. In this paper we present the initial findings
of an educational project focusing on the use of augmented reality in the design
process of an architectural student. The study seeks to evaluate the use of AR as a
tool in the design stages, allowing effective exploration of spatial qualities of
design projects undertaken in the studio. The learning process is guided by the
exploration and detection of a design idea in both form and function, with the
virtual environment providing a dynamic environment (Mantovani, 2001). This is
further reflected in the constructivist theory where the learning processes use
conceptual models, which are used to create incremental stages that become the
platform to attain the next [Winn, 1993]. The additional benefit of augmented
reality within the learning journey is the ability of the students to visually explore
the architectural forms they are creating in greater depth.

Keywords: augmented reality, pedagogy, learning journey, exploration

AIM their pedagogic journey.

Research into the use of augmented reality in the
learning environment is relatively emergent in the
academic arena. However, the possibilities for explo-
ration of spatial, geometric environments abound.
As the pedagogical landscape is now changing in
terms of cohort requirements, as a reflection of new
societal goals, the pedagogy of 5 years ago has al-
ready begun to change. What is expected in to-
day's pedagogic environment is now different, with
leaner's requiring change and a deeper learning to
In this paper we present the initial findings of
an educational project focusing on the use of aug-
mented reality in the design process of an architec-
tural student. The study seeks to evaluate the use
of AR as a tool in the design stages, allowing effec-
tive exploration of spatial qualities of design projects
undertaken in the studio. The learning process is
guided by the exploration and detection of a design
idea in both form and function, with the virtual en-
vironment providing a dynamic environment (Man-

CAAD Education - Volume 1 - eCAADe 32 | 343

 tovani, 2001). This is further reflected in the con-
structivist theory where the learning processes use
conceptual models, which are used to create incre-
mental stages that become the platform to attain the
next [Winn, 1993]. The additional benefit of aug-
mented reality within the learning journey is the abil-
ity of the students to visually explore the architectural
forms they are creating in greater depth. Tradition-
ally, drawings, sketches were transferred to the 3D
format of models. This transfer has allowed the co-
hort to better understand and comprehend spatial
and constructional issues with their designs in a more
detailed and clearer manner than traditional learn-
ing, via drawings and models. The ability to move
seamlessly from sketch idea to augmented model
has emphasised a step change in the pedagogical
landscape in the architectural studio. The interface of
augmented reality allows for faster comprehension
of design ideas, the visualisation allowing for intuitive
and instructive learning to occur.
Design thinking and the journey we take to-
wards a solution takes what Cross. refers to as mid-
dle path and is primarily concerned with appropri-
ateness, understood as that fragile quality which is
achieved when the best of human intentions are re-
344 | eCAADe 32 - CAAD Education - Volume 1
alized within the constraints of reality (Cross, 2001).
The current dichotomy emerging in studies on the
use of visualisation in the design process would sug-
gest there remains confusion as to whether design
is thinking or doing? The study attempts to observe
and correlate the learning 'loop' of thinking and do-
ing, which forms a critical part of the self learning
and reflective feedback that is fundamental to ac-
quiring knowledge. The study assesses the poten-
tial, within the architectural studio, to combine think-
ing and doing into 'design thinking' where the aug-
mented reality models are used to explore and raise
questions, that the learning journey may never have
anticipated, with the use of augmented reality. The
emerging routes of learning and their occurrence
in the learning journey are also noted and assessed
in relation to the traditional studio based expecta-
tions of and the qualitative differences when choos-
ing augmented reality as an exploration tool in the
architectural studio environment.
What is Augmented Reality, what is virtual real-
Figure 1
ity?
The Pedagogic
It is important for this study to ensure that these
Iceberg - The
immersive realities are understood, both augmented
Known knows and
reality (AR) and virtual reality (VR) are often taken
unknown unknows.
as the same. Whilst both are considered immer-
sive, it is only virtual reality that is deemed truly 'im-
mersive' as it allows to user to only see and expe-
rience the virtual world, with no 'connection' to ex-
ternal stimulus. Whereas, augmented reality allows
the user to experience both the 'altered world' and
the existing real world environments simultaneously
with one 'superimposed' over or within the other.
The use of AR in teaching occurred more recently,
in 2000, when Shelton and Hedley used the 'tool' of
AR in the teaching of their undergraduate students.
This teaching included simple rotation of images and
shapes that represented the planets and their rela-
tionship to the sun. The teaching allowed for the im-
ages to be moved and experienced from many dif-
fering viewpoints almost simultaneously by the stu-
dents. This used AR to allow for new levels of learning
that were outside of those studied and formalised by
Schon (1983) and Bruner (1961). The learning expe-
rience was autodidatic, in that the student using AR
learns about their design during their journey with-
out the formal interaction of the studio tutor, in re-
ality the student becomes 'self-taught'. The use of AR
within the learning journey that occurs within an aca-
demic architectural studio yields a learning that has
elements of autodidactism. Where the learning that
occurs is self-directed and related towards informal
learning and which occurs in a series of successive ab-
sorptive and contemplative occurrences. Mantovani
(2001), stated that these successions within the learn-
ing journey occur through discovery via exploration
in a natural manner as an apportion of the process of
learning.
Re-abstraction and re-alignment: AR as the new
convention for design process
There are few recent academic papers that have
continued the exploration and study of creativity in
the archietctural academic arena (Neely, D. 2008, -
BIM 101 overview - 'From hand drawings to CAD and
now BIM', Ambrose, M. A. 2012- 'Agent provocateur
- BIM in the design studio', Ibrahim, et al 2010- 'Com-
parison of CAD and manual sketching tools for teach-
ing architectural design'). The precursor to research
into creativity and the creative process were the stud-
ies concluded by Finke, Ward and Smith (1992) posit
the hypothesis that the creative process is one of a
multiple stranded series of processes that together
evolve towards a combined phase of 'creative insight
and discovery'. Intriguingly these multiple occur-
rences converge only having moved through a dis-
tinct phase termed 'pre-inventive' where the 'struc-
tures' of the artefact are brought together from men-
tal representations of the prior creative occurrences.
Interestingly Cross and Dorst (2011), concluded
subsequently to Visser (1995), that the creative jour-
ney is also a series elements termed activities that
also occur in succession and result in a 'novel event'.
This novel event could be aligned with the notion
of 'creative leap' or eureka moment, when the pro-
cesses that have occurred prior to that moment allow
an alignment or focus to the thought processes that
enable the production of the 'artifact' or design solu-
tion. However, Cross ( 2001) and Dorst (2011) do not
suggest that the processes or stages to arrive at the
artefact have to be part of a cognitive model in or-
der to materialise. In fact they concluded that such
events can be random in nature or 'non-routine' and
the occurrence of the artefact is merely one activity
that is different to all others due to the emergence
of a 'considerable or unanticipated' happening. This
study uses the hypothesis of Visser (2004) who pro-
poses that design activity is comprised of a series of
procedural stages that conclude in the production of
what he calls the 'creative artefact'. The use of AR can
allow for enhanced reflection of their designs and the
ability to create iterative steps that employ progres-
sional advances of multiple layers that are fully im-
plicit and part of the students learning journey.
THE PEDAGOGIC PUSH - PULL: AR INTE-
GRATION IN THE ARCHITECTURAL STUDIO
It is now understood in academia that the pace of
technology in the learning environment is impact-
ing of the differentiation between digital physical-
ity and physical digitality. These previously separate
domains are now becoming increasingly integrated
and the clear lines between the two are now blurred.
There is concern amongst many teaching in archi-
tectural studios within academia that this integra-
tion will bring with them a new era of exploration
within studio design, that of synchronic exploration
of building form using both traditional and increas-
ingly digital methods, this goes against the long held
traditional norm of non-digital methods of pedagogy
in the studio environment.
What was once a dialogue that stemmed from
pen to paper to thought and back to pen, in an
ordered and responsive learning and design jour-
ney. The students within today's architectural studio
can utilise the integrated and immersive approaches
of Building Information Modelling (BIM) and aug-
mented reality (AR). The student still 'models' their
building, its form and spatial expression, however via
digital modelling, an immersive exploration can be
CAAD Education - Volume 1 - eCAADe 32 | 345
 experienced. One that allows the embedded data
and information about space, structure and materials
to be synchronically manipulated in a blended and
deeper learning journey.
The augmented models that are used allow for
the synergy of both physical and digital modelling
and exploration to occur seamlessly. This readily
accessible and seamless environment allows for a
greater accuracy and finer levels of manipulation to
be achieved. There is a new pedagogic classification
occurring when we use BIM and AR modelling in ar-
chitectural teaching, that of 'depth of learning'. The
study indicated that when using the BIM model, the
cohort were able to explore more 'what if's' in rapid
succession which, inturn, enriched the final learning
journey and its pedagogic outcomes. Using AR the
learning was deeper still; the connection with the
model was far more immediate, seeing the model in
a real world setting seemed to communication be-
tween student to student or student to tutor dia-
logue (Clayton 2010, Dorst 2011). Both perception
and interpretations of the models studied were richer
in detail and created a new critique dynamic.
Modelling is a key tool in architectural studio
teaching, using this method of exploration allows
higher order cognitive skills to be used by the cohort,
such as spatial, planning and relationship of form.
Using AR to explore the model(s) allows for a criti-
cal synthesis of these cognitive skills, where interac-
tions with the model create immediate design itera-
tions that move the student forward through the de-
sign journey. Levels of detail can be explored simul-
taneously and in parallel. This studio process should
be re-structured to allow for a greater transparency
for the cohort to realise that the ability to explore,
amend and reflect is a iterative loop that can exist at
any point in the design journey. The academic studio
should also empower their cohorts with 'tools' such
as AR that allow for enhanced reflection of their de-
signs and the ability to create iterative steps that em-
ploy progressional advances of multiple layers that
are fully implicit and part of the students learning
journey. "The discussions at the end of the course
346 | eCAADe 32 - CAAD Education - Volume 1
clearly indicated to me that the students now con-
sider the development of programs as their own per-
sonal expression of an idea, that CAD systems could
be used to investigate ideas and not only document
deci¬sions already made. They began to understand
the feedback their rules created and how it could be
used to clarify concepts" (Krawczyk, 1998).
Figure 2
Evolution of the
iterative 'loop' with
problem and
solution space
within the iteration.
The base iterative 'loop' shown in Figure 2, has
evolved, via the emergence of a digital domain in the
studio, to include the concept of problem and solu-
tion space proposed by Maher (1996). Maher's con-
cept suggested the idea of movement from problem
space to solution space as the learning occurred. This
study suggests that utilising AR allows the problem
finding - problem solving phases of the design jour-
ney to be far more acquiescent, as the exploration of
both scale, spatial and structure amongst many other
design layers can be viewed and experienced with
greater levels of detail at the identical point in time
within the journey.
This plurality is conceived of a new construct rep-
resented a fundamentally altered medium from the
tradi¬tional representation in contemporary prac-
tice of architecture. However, this altered approach
brings with it many benefits and a richer appreciation
of the very design be¬ing formulated within schools
around the UK. If AR brings such benefits why is it
not fully integrated into the architectural curriculum?
As the architectural industry is now approaching full
assimilation of design functions and seamless inte-
gration for those that use BIM and AR packages, why
are we not educating our architectural graduates to
stream into such working environments with seam-
less curricula? As with integration of previous tech-
nology in the academic studio, there is current con-
cern and general resistance that the rapidity of trans-
formation in digitial technology will result in an in-
creasing 'gap' between those who teach in the archi-
tectural academia and the mastery that such digitial-
ity requires. In order to direct the cohorts towards
a greater understanding of the capabilities proposi-
tioned by such technology it must first be embraced
and embedded. This study demonstrated that the
bottom up attitude of using AR and computers in the
academic studio first reported by Martinez and Vigo
in 1999, it still as challenging today, some 15 years
later.
This paper will investigate the reasons behind
the complications of AR implementation in curric-
ula by using both an initial forming survey and then
structured interviews of current graduates from the
Department of Architecture and Built Environment,
June 2014. The paper is structured as follows: first,
we have an overview of technological change liter-
ature will be presented and will introduce the con-
cept of AR in curricula; next, we will describe the how
the data is collected and analysed in this study.; then
we explain the results of this study; lastly we will con-
clude the findings.
METHODOLOGY
The ongoing research follows a heuristic approach to
evaluation, which consists of an iterative process of
analysis, design and re-test phases. This approach
has been adopted for this research as there are no
current guidelines as to how AR should be tested
and what data should be achieved in both type and
amount. The cohort were introduced to AR using ba-
sic, freely available software from AR Media. This ap-
proach was used as it allowed for rudimentary meth-
ods of interfacing with their building models. The
data was then captured in two stages, seminar ses-
sions were used to review and explore the building
model via the use of AR, the findings made by the co-
hort and the dialogue of these sessions was captured
and analysed. The same cohort were also asked to
capture their learning journey, finding and how the
model was navigated and explored. The study fo-
cuses on the use of augmented reality in a collabo-
rative learning context. The studio format for learn-
ing architectural subjects within academia is a funda-
mental influence in the discipline. Cohorts learn both
within lectures and also socially, whilst in the studio,
by peer to peer learning and conversation and dia-
logue that become reflections which inturn are em-
bedded into futher design work or learning. Such in-
teractions have been historically studied in this genre
by academics and theorists such as Dewy and Schon.
This exploratory study has been carried out in a field
of research that is still emergent. Although, in the
last 3 years there is a notable increase in interest
and a growth of research in the application of AR in
the learning environment to enhance erudition and
scholarship. Particular focus will be given to the use
of augmented reality in architectural studio learning.
The study is considered as an exploratory review and
survey of architectural students currently studying at
Northumbria University and completing their archi-
tectural design module - BE1341 Virtual Project. This
module is studio based and contains a high level of
technical challenges (such as continued resistance by
the academic studio to digital explorative methods of
design and learning by abstraction).
The specific objective of this study was to es-
tablish those challenges and the potential benefits
of using augmented reality within the learning jour-
ney of the cohort studying this module. The study
included a questionnaire to the 45 students study-
ing the Virtual Project module. The final submitted
number of questionnaires totalled 41. The survey
contained 10 questions relating to the use of aug-
mented reality and how this was used within the
learning journey of the student(s). Three questions
gathered information on the development of the de-
sign and what was learnt at key stage of its devel-
opment using traditional methods and AR within the
studio environment. The subsequent four questions
related to the use of augmented reality, the software,
it ease of use or adoption as part of the learning, ex-
ploratory process of the students journey through
the project. The concluding questions related the use
CAAD Education - Volume 1 - eCAADe 32 | 347
 and perceived benefits of augmented reality within
this learning journey. Focusing on the proficiency
of the students to visually explore the architectural
forms they created as the output for the module. The
questions were generated to allow for identification
of the types of inferences that would best inform a
process of change in the future pedagogic format of
learning and delivery within the architectural studio.
The cohort has an average of 2-3 years' experi-
ence of the design process and have used various
design software packages prior to commencing on
this degree course. So far, the results have indi-
cated some robust themes that occur across the data.
Therefore, although the results are partial, they can
inform future study in this area of pedagogy.
AUGMENTED REALITY: CATALYST FOR RE-
THINKING DESIGN EDUCATION
The study of design activity has grown progressively
over the last 30 years. During this period a rather
small number of studies have been undertaken. The
early pedagogic studies of Bruner (1961) and Schon
(1983) set the grounding for more digitally focused
studies by Cross (2001), followed by Winn (1993) with
a marked increase in studies and research by Am-
brose (2009), Clayton (2010), Ozener (2010), Farias
(2010) and more recently Decker (2013). These stud-
ies were often small in scale and remain untested, as
their findings were not verified in repeat studies. AR
allows the problem finding - problem solving phases
of the design journey to be far more acquiescent, as
the exploration of both scale, spatial and structure
amongst many other design layers can be viewed
and experienced with greater levels of detail at the
identical point in time within the journey. Cross
(1999), concluded that within design cognition there
are distinct phases of 'finding' and 'solving'. These ac-
tivities included significant problem structuring and
problem solving (Cross, 1999) cycles. Cross formu-
lated three cognitive phases from his early studies;
(1) Formulation, (2) generation and (3) process. These
phases allow for an identification of design as a pro-
cess in the context of academia.
348 | eCAADe 32 - CAAD Education - Volume 1
AUGMENTED REALITY AND ITS INTEGRA-
TION: EXPLORING THE UNBUILT
The process being observed in this study is the pro-
cess of design, the creativity that drives the progres-
sion and the activities that are involved. Creativity it-
self is often stated as an event of singular occurrence.
However the 'creative leap' is often debated in aca-
demic theories and academic research as a manifes-
tation of a series of smaller events that culminate in
the 'leap'. This point in the learning journey is ac-
knowledged as a significant observation point of the
journey, where the idea becomes clear and better un-
derstood. Such activity is known as 'design cogni-
tion' with pioneering studies undertaken by Eastman
(1969). The design output from the academic studio
can take on many forms, varying at different stages
of the learning journey.
Figure 3
Mutliple Iterations
occurring within
learning journey
evolution.
These variations are often due to a need to com-
municate rapidly and effectively which in turn allows
for interaction from others (namely the tutor in this
context) externalizing the thoughts of the architec-
tural student. This need to switch from design jour-
ney to what is essentially 'presentation' of the idea(s)
 is a quick side step, the question is, using AR, could - problem solving structure. With 73% of the cohort
this side step become a forward moving process. This using AR in the exploration of their designs to formu-
is not to state that discussions with the tutor are not late solutions for spatial or structural relationships.
a forward movement in the design journey. Simply The findings would suggest that the use of AR
that, the need to 'produce' in order to communicate allowed these students to confront complex tasks
an evolution seems somewhat unnecessary. AR and within their design and manage the complexity by
other technology available to the cohorts of today's seeing the relationships of their intentions at both a
architectural academic studios have a chance to up- macro and micro scale. The resultant design change
date this process. could then be assessed and accepted or assessed
and rejected, so the 'task' could be managed at
Figure 4
the smaller scale, but with a clear understanding of
Occurance of AR in
what the implication of that would be for the over-
the learning
all scheme. The finding replicates the relationships
Journey, overlaid
found between talking and drawing in studies of ar-
onto the Schon's
chitectural cohorts by Schon and Wiggins (2006). In
Design Process of
this study the act of 'seeing - moving - seeing', is sub-
Problem Setting
limated by AR into a series of small iteration cycles
and Problem
that contain the same analysis, synthesis and evalua-
Solving
tion stages.
The cognitive steps taken in 58% of the co-
Figure 5 hort studied follow a linear progression similar to
The Co-Evolution Piaget's constructing operational thinking findings
Diagram of Maher (1979). However, although the progression is lin-
et al (1996) (Image: ear using AR, the progression is a series of paral-
D.Morton) lel linear progressions. The student creates abstract
concepts and articulated intuitive progressions from
which they begin creating constructs that arch be-
tween these parallels. The learning route is not au-
tonomous but in fact manifold in its pedagogic struc-
ture. The arches or leaps between the parallel learn-
ing seem to form the reflective abstraction. The leaps
from linear routes create an extensive mapping of for-
malisation of conceptualisation and transformation
of the design. Such constructs of abstraction and re-
constructs can be found in studies of Piaget (1979).
In these studies the journey of learning in also linear,
however the journey is a series of successive stages,
FINDINGS FROM STUDY : CONJECTURES the findings of which have been critised for being res-
AND REFUTATIONS IN A DIGITAL CONTEXT olute.
Finding 1: Developmental theory of cognition The study demonstrates a pedagogic journey
and the act of seeing moving seeing. The study undertaken by the cohort using AR that reflects a
demonstrated a pedagogic journey undertaken by problem finding - problem solving structure. With
the cohort using AR that reflects a problem finding 73% of the cohort using AR in the exploration of their

CAAD Education - Volume 1 - eCAADe 32 | 349

 designs to formulate solutions for spatial or structural
relationships.
The findings would suggest that the use of AR
allowed these students to confront complex tasks
within their design and manage the complexity by
seeing the relationships of their intentions at both a
macro and micro scale. The resultant design change
could then be assessed and accepted or assessed
and rejected, so the 'task' could be managed at
the smaller scale, but with a clear understanding of
350 | eCAADe 32 - CAAD Education - Volume 1
what the implication of that would be for the over-
all scheme. The finding replicates the relationships
found between talking and drawing in studies of ar-
Figure 6
chitectural cohorts by Schon and Wiggins (2006).
Mahers Concept of
Problem- Solution
Occurrence of Manifold Evolution Phases
Space, blended
In this study the act of 'seeing - moving - seeing', is
with the continuous
sublimated by AR into a series of small iteration cycles
overlay of AR:
that contain the same analysis, synthesis and eval-
Creating a series of
uation stages. The cognitive steps taken in 58% of
'itteration loops'
the cohort studied followed a linear progression sim-
that becomes
ilar to Piaget's constructing operational thinking find-
cycles of
ings (1979). However, although the progression is
assimilation and
linear using AR, the progression is a series of paral-
accommodation.
lel linear progressions. The student creates abstract
concepts and articulated intuitive progressions from Figure 7
which they begin creating constructs that arch be- The AR influenced
tween these parallels. The learning route is not au- Iteration Loop :
tonomous but in fact manifold in its pedagogic struc- Occurrence of
ture. The arches or leaps between the parallel learn- Manifold Evolution
ing seem to form the reflective abstraction. The leaps Phases.
from linear routes create an extensive mapping of for-
malisation of conceptualisation and transformation
of the design. Such constructs of abstraction and re-
constructs can be found in studies of Piaget (1979).
In these studies the journey of learning is also linear,
however the journey is a series of successive stages,
the findings of which have been critised for being res-
olute.
Finding 2: Parallel learning streams: Interactive
cyclical learning journeys. The results suggest that
as the design journey evolves there is a formation
of concept(s). The formation of these concepts are
generated from the interactive nature of exploration
of the model in AR. The study findings indicate the
process as being highly interactive, mirroring stud-
ies of Vygotsky (1978). The study indicated that 68%
of the cohort confirmed that the process of using AR
allowed their exploration to be a series of adaptive
and evolutionary stages, as the 'what if's' were ex-
plored. A higher figure of 81% stated that their de-
sign journeys were dialectic in formation arriving at a
final concept formation.
 The interactive cycles indicated in the findings
of this study show distinct alignment with Piaget's
theory of cognitive development, when using AR in
the design journey. Piaget's theory focused on two
processes, which were titled 'assimilation' and 'ac-
commodation'. Assimilation is a building element of
learning. From a series of these 'blocks' we assem-
ble our own unique and personal relationship to how
we understand the world around us. These learning
blocks are then used by us to add to and incorpo-
rate our understanding of the world and enhance our
existing knowledge. These associative connections
then build upon one another to deepen our knowl-
edge base. The process of accommodation, accord-
ing to Piaget, is the process by which the new ex-
perience or knowledge gained changes or replaces
the existing in your schema. These two processes
are organised in our knowledge as 'schemas'. These
schemas can be thought of as sections in a library.
The schemas allow us to readily access and categorise
the knowledge acquired from assimilation and ac-
commodation.
In the results of this study, 72% of the cohort
agreed that use of the VR model allowed for better
perception of the developing building. In particular,
the spatial interrogation and structure were recog-
nised in terms of their attributes. A room was no
longer simply a space of three dimensions. These el-
ements of the design were queried in terms of their
location on plan, the relationship to section and how
all of these constituents were recognised in terms of
impact of the final design. During the exploration
of the model in AR, the assimilation and accommo-
dation processes are in continuous 'loop' as the de-
sign is created by abstraction. In the study the co-
hort, used the AR to test 'what if's' in a series of as-
similation processes. Many of these did not result in
overall amendments to their designs, but were part
of the learning journey. The occasions where these
tests of the design resulted in amended their design
or part of the design, this is in effect modifying the
schema of the model (in Piagets terminology) and
from this the new information is created and added
the overall knowledge of the model and the learn-
ing experience, accommodation took place. After 8
weeks of designing their projects for BE1341, the co-
hort showed no defined 'pattern' to the timing of ac-
commodation in their learning journeys. However,
their was a considerable increased in general accom-
modation in the earlier stages of 67% of the cohort in
the study.
Finding 3: The act of abstraction and concept de-
velopment. According to Arnheim (1998), the pro-
cess of abstraction in the design journey in part of
what is termed 'productive thinking'. He believed
that the abstraction of a design is significant in un-
derstanding the underlying concepts that are within
a form or object. The observations of this study indi-
cate that as the cohort explored and moved around
their designs in AR, the immersive environment al-
lowed for informed levels of abstraction. These en-
compassed relationships of form and spatial qualities
that could be readily experienced by the AR environ-
ment. The significance of these discoveries where
that they indicated an assimilation underlying char-
acteristics of the model elements being investigated.
The unifying route of investigation parallels the idea
of true abstraction in Arnheims 'productive thinking'
theory.
Finding 4: The AR stages of design exploration:
From metaphoric to metonymic.. The use of AR in
the design journey allows, as discussed in previous
findings of this study, for exploration by abstraction.
The understanding of exploration in the metaphoric
and metonymic forms have been observed and anal-
ysed for over 50 years. In 1968, Arnheim first con-
sidered the concept of visual reasoning in architec-
tural design. Three decades later, in 1995 and 1999,
Goldschmidt continued these observations into rea-
soning and visual relationships of design elements.
More recently Oxman (2002) continued the enquiry
into design discourse, with research indicating that
domain knowledge can be employed in virtual de-
sign, within the architectural context of design. Ox-
man has established that digital media has made
significant impact on design processes and meth-
CAAD Education - Volume 1 - eCAADe 32 | 351
 ods of pedagogic discovery. Oxman's latest paper tions in the abstraction of the learning journey con-
'The Digital in Architecture' is the culmination of her tained the spatial, functional and structural elements.
research and applies new definitions to theoretical However, as they were perceived simultaneously, the
and methodological structures to the field of digital design elements were globally conceptualised.
methods of design. The current proposition of her Finding 6: Learning and the acquisition of archi-
studies is the idea of morphogenesis. This incorpo- tectural syntax.. The research results Indicate that
rates both form finding and parametric design as part the use of AR in the learning journey created the
of the exploratory process of using processes such as paradigm of increased conviction in design decisions
AR in design exploration. during iterative abstraction of design ideas. When
This study observed iterative methods of design questioned in the studio during the study, the co-
learning and abstraction that work based around the hort stated that the use of AR allowed them to under-
students using the AR to explore and augment forms, stand the design in a clearer and more accurate man-
and adapting elements from previous iterations of ner. The decisions they made during iterative design
their designs. These gradual analogous relationships changes contained both scale, form and function,
to form abstraction replicate the morphogenesis that they were not merely sketched ideas, but something
occurs in nature. However, these are more digital in more concrete. Although the iterations were concep-
their development. tual in abstraction they contained a syntax that was
Finding 5: Exploration of design using AR: From real world. The predicted outcome of these iterations
analogous connections to functional variation.. was therefore to some extent known, giving the co-
The exploration of the building as its component hort a high level of knowledge transmission to their
parts is a route that has been historic in architectural design than approximated outputs in non digital for-
pedagogy as the process of 'partii'. This process de- mat, such as sketching.
fines the building form as a syntax of elements that
Figure 8
form the whole. The partii is both explicit and im-
The impact of AR in
plicit in the buildings final form. That is, a completed
the solution-
building design could be formally analysed and 'de-
exploration cycle of
constructed' to functional, spatial and structural ele-
the learning
ments. The partii not only encompasses all of these
journey.
elements but also includes the interrelationship of
these without design conjecture.
The use of AR to generate elements of the par-
tii and abstract multiple elements to explore relation-
ships of function and aesthetics was also replicated in
the journeys of 79% of those students studied. How-
ever, the remaining 21% of the cohort studied did not
include this governing principle in their design jour-
ney. This set of journeys contained analysis of the
design thought, abstraction and representation that
occurred in parallel but without any perceivable sin-
gular linear progression. The design partii was con-
ceived simultaneously and processed in a seemingly
complex interactive cognition that may well be inves-
tigated in a future study. These analogous connec-

352 | eCAADe 32 - CAAD Education - Volume 1

 Finding 7: Conjectures and refutations: The re-e- focus is on the exploration of individuals using tech-
mergence of Poppers Method in AR.. The mode of nology in settings (Crook 1994). The findings corre-
abstraction that emerged from this study aligns itself spondingly indicate a similar model to that proposed
with the method Popper suggested (1953) in that the by Maher et al, which is based upon a both a problem
use of AR in the learning journey allowed the cohort and solution space that occupy a 'co-evolution' space
studied to explore their design theories whether they when the cohort use augmented reality to visualise a
were singular or multiple in nature (ie. Exploration design iteration (Maher et al, 1996). These elements
of a design element such as spatial or structural, or co-exist and efficiently allow a more effective inter-
combination of these or more elements), this allowed change in the design process.
them to be tested or exposed to refutation as part of
Figure 9
the generation of ideas from the design process itself.
The Cradle
However, as conjecture requires a reliance on the
Principle: Initial
prior concept of knowledge for its intelligibility. As
Loop: Problem
knowledge is based on the premise that it is an idea
Finding -
proffered without proof, many iterative steps in the
Abstraction.
design journey could be conceived as comprising this
key component. The abstraction that occurs within
the learning journey at various stages (these were un-
fixed and varied in occurrence) allowing the cohort
to have prior concept of knowledge within their de-
sign. The iterative abstraction allows conjecture to
become a concrete knowledge of the design at any
Figure 10 particular point in time. The study findings suggests
The Cradle that for 39% of the cohort expressed that the use of
Principle: AR allowed for a series of visual operations in the de-
Secondary Loop - sign process that did not rely on them fully under-
Exploration Cycle. standing the 'architecture' in precise detail, but allow
for clarification of a design element.

RESULTS & CONCLUSION

Christiaans studied industrial design students with
the study focused on the search for creativity in de-
sign (Christiaans 1992). The research presented in
this paper develops this work with students further,
by extending a similar research methodology into
studies of the architectural studio context and the
learning journeys that occur within. The results indicate that when using augmented
This study has indicated a simple but influen- reality in the academic architectural studio context
tial system for supporting the existing learning jour- it becomes a new foundation from which to view
ney in the architectural studio. Students are allowed the possible design decisions and from which to ex-
effective exploration of spatial qualities of design plore new iterations. The use of augmented reality
projects undertaken in the studio. The whole is then in facilitating both pedagogic and didactic learning
greater than the sum of its parts so the investigative in the use and application of this exploratory tool

CAAD Education - Volume 1 - eCAADe 32 | 353

 within the architectural studio. This in-turn enables
the cohort to more effectively explore design pos-
sibilities and the impacts of these design decisions
visually and more coherently. The learning experi-
ence is more located in the exploration arena of the
students learning, the 'what if's' are more effectively
navigated towards a deeper understanding of the de-
sign that would otherwise be difficult to obtain in the
more traditional methods within the studio. The au-
thor plans to create a template from this module that
could be used as shared resource.
FINAL THOUGHTS: AR IN THE ACADEMIC
STUDIO
It is now understood in academia that the pace of
technology in the learning environment is impact-
ing of the differentiation between digital physicality
and physical digitality. These previously separate do-
mains are now becoming increasing integrated and
the clear lines between the two are now blurred.
There is concern amongst many in architectural stu-
dios within academia that this integration will bring
with them a new era of exploration within studio de-
sign, that of synchronic exploration of building form
using both traditional and increasingly digital meth-
ods.
What was once a dialogue that stemmed from
pen to paper to thought and back to pen, in an
354 | eCAADe 32 - CAAD Education - Volume 1
ordered and responsive learning and design jour-
ney. The students within today's architectural studio
can utilise the integrated and immersive approaches
of Building Information Modelling (BIM) and aug-
mented reality (AR). The student still 'models' their
building, its form and spatial expression, however via
digital modelling, an immersive exploration can be
experienced. One that allows the embedded data
and information about space, structure and materi-
als to be synchronically manipulated in a blended
and deeper learning journey. The augmented mod-
els that are used allow for the synergy of both phys- Figure 11
ical and digital modelling and exploration to occur The Cradle
seamlessly. This readily accessible and seamless envi- Principle:
ronment allows for a greater accuracy and finer levels Synchronic Loop -
of manipulation to be achieved. Exploration Cycle
There is a new pedagogic classification occurring using digital tools.
when we use BIM and AR modelling in architectural
teaching... that of depth of learning. When using
the BIM model, the cohort were able to explore more
'what if's' in rapid succession which enriched the final
learning journey and its pedagogic outcomes. Using
AR the learning was deeper still; the connection with
the model was far more immediate, seeing the model
in a real world setting seemed to enhance communi-
cation between student to student or student to tu-
tor dialogue. Both perception and interpretations of
the models studied were richer in detail and created
a new critique dynamic.
Modelling is a key tool in architectural studio
teaching, using them allows higher order cognitive
skills to be used by the cohort, such as spatial, plan-
ning and relationship of form. Using AR to explore
the model(s) allows for a critical synthesis of these
cognitive skills, where interactions with the model
create immediate design iterations that move the
student forward through their design journey.
LIMITATIONS
The study was programmed to evolve over an 8 week
duration, with the final results programmed being
captured after this time. The analysis of the findings
thus far from the study data, indicating the value of
augmented reality as an exploration tool within the
learning journey of an architectural student. There
are high levels of practical value in the findings thus
far, with findings indicating the pedagogic advan-
tages for application of the augmented reality 'tools'
and how the pedagogic interface of augmented re-
ality allows for increased comprehension of design
ideas. In turn how the visualisation allows for intu-
itive and instructive learning to occur. During this
study the availability of free augmented reality soft-
ware increased 240%. The advances in better in-
teroperability will also be a potential future study
as augmented reality systems within the studio may
become ever more feasible and translate into peda-
gogic efficacy.
REFERENCES
Ambrose, MA 2006 'Plan is dead: To BIM or not to BIM,
that is the question', Computing in architecture / re-
thinking the discourse, Proceedings of the Second In-
ternational Conference of the Arab Society for Com-
puter Aided Architectural Design,, Sharjah, United
Arab Emirates
Ambrose, M 2009 'Agent Provocateur — BIM and the De-
sign Studio: Questioning Roles of Abstraction and
Simulation in Design Education', ACSA Annual Con-
ference: The Value of Design, p. 85
Ambrosini, V and Bowman, C 2001, 'Tacit knowledge:
Some suggestions for operationalization', Journal of
Management Studies, 38, pp. 811-829
Arnheim, R 1988, 'Visual dynamics', American Scientist
(reprinted. In: Swede, G. -Ed.,The Psychology of Art: An
Experimental Approach -1994, Canadian Scholars, 76,
pp. 585-591
Bruner, J 1961, 'The Act of Discovery', Harvard Education
Review, pp. 118-2002
Christiaans, H 1992, Creativity in design, Ph.D. Thesis,
Delft University of Technology
Christiaans, H and Venselaar, K 2005, 'Creativity in Design
Engineering and the Role of Knowledge: Modelling
the Expert', International Journal of technology and
Design Education, 15(3), pp. 217-236
Clayton, MJ, Ozener, O, Haliburton, J and Farias., F 2010
'Towards Studio 21: Experiments in Design Educa-
tion Using BIM (K. Goosens. and L. Agudelo. -
eds)', SIGraDi 2010 Proceedings of the 14th Congress of
the Iberoamerican Society of Digital Graphics, Bogota,
Columbia
Crook, C 1994, Computers and the collaborative experi-
ence of learning, Routledge, London
Cross, N 1997, 'Descriptive models of creative design:
application to an example', Design Studies, 18(4), p.
427–455
Cross, N 2001, 'Designerly ways of knowing: design disci-
pline versus design science', Design Issues, 17(3), pp.
49-55
Cross, N and Clayburn Cross, A 1988, 'Expert designers', in
Frankenburger, E, Badke- Schaub, P and Birkhofer, H
(eds) 1988, Designers—the key to successful product
development, Springer Verlag, London, UK
Decker, M 2013, 'New Materials Compositions', in
Rashida, N and Patel, S (eds) 2013, Performative Ma-
terials in Architecture and Design, Intellect, University
of Chicago Press, Chicago
Dorst, K 1997, Describing Design: A Comparison of
Paradigms, Ph.D. Thesis, Delft University of Technol-
ogy
Dorst, K 2002, 'Design and use as plans: an action theo-
rectical account', Design Studies, 23(3), pp. 303-320
Dorst, K 2011, 'The Core of Design Thinking and its appli-
cation', Design Studies, 32(6), pp. 521-532
Eastman, CM 1969 'Cognitive Processes and Ill-defined
Problems: a case study from design', International
Joint Conference on Artificial Intelligence, Washing-
ton, DC
Eastman, CM 1970, 'On the Analysis of Intuitive Design
Processes', in Moore, GT (eds) 1970, Emerging Meth-
ods in Environmental Design and Planning, MIT Press,
Cambridge, Ma
Goldman, G 2012, 'Digital Media and the Beginning
Designer', IEEE Computer Graphics and Application,
32(2-March/April ), p. 14–21
Goldschmidt, G 2006, 'Variances in the impact of visual
stimuli on design problem solving performance', De-
sign Studies, 27(5), pp. 954-569
Goldschmidt, G and Weil, M 1998, 'Contents and Struc-
ture of Design Reasoning', Design Studies, 16(2), pp.
189-210
Ibrahim, R and Pour Rahimian, F 2010, 'Comparison of
CAD and manual sketching tools for teaching archi-
tectural design', Automation in Construction, 19(8)
Krawczyk, H, Canetti, R and Bellare, M 1998 'A modular
approach to the design and analysis of authentica-
tion and key exchange protocols', 13th Annual ACM
Symposium on Theory Computing
Maher, ML, Poon, J and Boulanger, S 1996, 'Formalis-
ing design exploration as co-evolution: a combined
gene approach', in surname missing, initials miss-
CAAD Education - Volume 1 - eCAADe 32 | 355
 ing (eds) 1996, Advances in formal design methods for
CAD, Chapman and Hall, London, UK, pp. 3-30
Mantovani, F 2001, 'VRLearning: Potential and Chal-
lenges for the Use of 3D Environments in Education
and Training', in Giuseppe, R and Carlo, G (eds) 2001,
Towards CyberPsychology: Mind, Cognitions and Soci-
ety in the Internet Age, IOS Press, Amsterdam
Oxman, R 2006, 'Theory and Design in the First Digital
Age', Design Studies, 27(3), pp. 229-265
Pasquarelli, G 2013 'Closing Keynote', ACSA 101: New
Constellations/New Ecologies, Annual Conference
(https://vimeo.com/63902910), San Francisco, CA
Piaget, J 1970/1977/1979, Psychology and epistemology,
Harmondsworth: Penguin
Popper, K 1953, 'Prediction and Prophecy in the So-
cialSciences in . 2nded. NY: Harper & Row, [1948]
1965a, pp. 336-46.', in no editors given 1953, Conjec-
tures and Refutations -2nsd Ed., Harper & Row, New
York, pp. 336-46
Roussos, M, Johnson, A, Moher, T, Leigh, J, Vasilakis, C
and Barnes, C 1999, 'Learning and Building Together
in an Immersive Virtual World', PRESENCE, 8(3), pp.
247-263
Schon, DA 1983, The reflective practitioner: how profes-
sionals think in action, Basic Books, New York
Shelton, B and Hedley, N 2004, 'Exploring a cognitive ba-
sis for learning spatial relationships with augmented
reality', Utah State University, Old City Publishing, 1(4),
pp. 323-357
Visser, W 1995, 'Use of episodic knowledge and infor-
mation in design problem thinking', Interacting with
Computers, 6(3), pp. 235-274
Winn, W 1993, A Conceptual Basis for Educational Ap-
plications of Virtual Reality, Technical Report TR 93-
9, http://www.hitl.washington.edu/publications/r-
93-9/
Zarzycki, A 2009 'Dynamics-based Tools: An Unusual
Path to Design Integration', Asia Proceedings, Yoko-
hama, Japan

356 | eCAADe 32 - CAAD Education - Volume 1

Teaching and Designing for Augmented Reality
Andrzej Zarzycki
New Jersey Institute of Technology
www.njit.edu
andrzej.zarzycki@njit.edu

This paper discusses ways emerging interactive technologies are adopted by

designers and extended into areas of design, education, entertainment, and
commerce. It looks, in detail, at various project development stages and
methodologies used to engage design focused students into, often complex,
technological issues. The discussion is contextualized through a number of case
studies of mobile and marker-based augmented reality (AR) applications
developed by students. These applications include an app for a fashion based
social event that allows participants to preview recent collection additions, an
info-navigational app for the High Line elevated urban park in New York City, a
marker-based maze game, and an interior decorating interface to visualize
various furnishing scenarios. While a number of case studies will be discussed
from a developer perspective, the primary focus is on the concept and content
development, interface design, and user participation.

Keywords: Augmented Reality, AR, Gamification, Mobile Culture

INTRODUCTION mulated by students as part of an augmented reality

Let's pretend you're on your way to Manhattan to buy
some new clothes. Maybe you're looking to impress
someone on a date. Maybe you need an outfit to ace
that interview. Maybe you're looking to change your
style and try something new. Whatever the case may
be, you know New York fashion will not disappoint.
You arrive on 54th Street on Fifth Avenue early in
the afternoon. There's a ton of different stores in that
five-block radius. There's high-end fashion retail and
some typical name-brand stores. Some stores aren't in
your price range, but you might be interested in what
they have to offer. You're not sure where to shop first.
Decisions, decisions...Luckily, there's an app for that!
(students: Philip Caleja, Nick Haby, & Daniel Schittone)
The above excerpt from one of the scenarios for-
(AR) project proposal explored ways mobile AR ap-
plications can enhance the fashion retail experience
by taking advantage of location-aware services and
electronic social networks. These applications also
address other, less tangible needs, such as the sense
of social happening and the excitement associated
with emerging technologies.
This paper looks into ways that emerging inter-
active technologies are being adopted by designers
and extended into the areas of tourism, education,
entertainment, and commerce. It discusses in de-
tail the project development stages and methodolo-
gies used to engage design-focused students into
often-complex technological issues. The discussion
is contextualized through a number of case studies

CAAD Education - Volume 1 - eCAADe 32 | 357

 of mobile and marker-based AR applications devel-
oped by students. These applications include an app
for fashion-based social events that allows partici-
pants to preview recent collection additions, an info-
navigational app for the High Line elevated urban
park in New York City, a marker-based maze game,
and an interior decorating interface to visualize vari-
ous furnishing scenarios.
While a number of case studies are discussed
from a developer perspective touching on technical
intricacies, the primary focus is on content develop-
ment, interface design, and user interaction consid-
erations. The paper also discusses pedagogy, con-
cept formation, and broader social and spatial narra-
tives.
AUGMENTED REALITY ENVIRONMENTS
Traditionally, AR environments employed two dis-
tinct types of data overlay. Marker-based environ-
ments employed distinct markers and, more recently,
images (image targets) to locate virtual data within
the physical world. A second approach associated
with mobile AR involved GPS, digital compass, and
accelerometer sensors to position users and the vir-
tual content around them.
AR-based applications provide an opportunity
reconnect and better realign virtual and physical
worlds through location awareness, enhanced data
overlays, and user-focused content. Unlike more
static forms of digital media, augmented reality, with
its interactive and context-aware functionalities, en-
gages users in more direct and meaningful ways. This
is evident not only in academia but also, and perhaps
primarily, in commerce and advertising. The interac-
tive print approach popularized by Layar or AR Lego
models [1] , associated with mainstream toy products
are successful because they extend the level of con-
sumer engagement. They provide additional infor-
mation and enticement (enticement) for consumers.
The same lessons of consumer engagement are di-
rectly applicable to design education and design ser-
vices.
While AR technology is routinely employed in
358 | eCAADe 32 - CAAD Education - Volume 1
the form of data overlays providing supplementary
information for physical objects that are visible with
the unaided eye, it is also increasingly used to vi-
sualize less tangible structures and concepts such
as historical events, cultural phenomena, and scien-
tific processes. This can be seen in a number of re-
search projects and mobile apps that help users to
learn history and facilitate explorations of urban envi-
ronments. (Niedmermair, Ferschin, 2011) TimeWarp
(Herbst, et al, 2008) a mobile edutainment applica-
tion designed as an AR game situated in Cologne,
Germany, focuses on virtual reconstruction of historic
buildings by superimposing virtual imagery over cur-
rently existing structures. The application not only
shows no-longer-existing buildings as they originally
appeared but also visualizes design changes that oc-
curred over time to still-present structures. Along the
same lines, the Immersive Experience of Cultural Her-
itage project (Kim et al, 2009) uses an AR tour ap-
proach to provide tourists with a more realistic expe-
rience by placing virtual characters within historical
structures. Visitors to the heritage sites of Sajeong-
jeon and Gangnyeongjeon in Korea can use their mo-
bile devices to access additional facts associated with
the showcased physical content. While a similar ap-
proach is routinely used by many museums, this par-
ticular project does not rely on AR markers such as QR
codes. It implements visual camera tracking of the
rectangular display space to position its virtual actors
without a need for visually intrusive markers.
Figure 1
Augmented Reality
(AR) environment
as social and design
activism and urban
games.
 Virtual environment allow for explorations of in-
accessible or not-yet-materialized designs. They can
be precursors of future physical urban spaces and po-
tent drives in their realization. This is the case with AR
environments (fig.1) developed by Tremont Under-
ground Theater Space (TUTS) initiative [2]. This initia-
tive is using AR gaming media not only to popularize
ideas of the adaptive reuse of the abandoned pub-
lic infrastructure but also to build social constituency
and connect with general public (fig.2) The shifting
focus from virtual-reality (VR) environments toward
mixed-reality and AR frameworks indicates the reex-
amination of earlier visions of separated physical and
digital worlds. The emerging picture fuses both di-
mensions into a single continuum.
Figure 2
Mystery Spaces, a
map with POIs
arranged in the
form of the game
play.
CASE STUDIES
FashNYC
The FashNYC application helps its users make smarter
shopping choices while informing them about the
current fashion scene. Through the app, users have
access to online videos from each store's current col-
lection, watch interviews with designers, and access
garment information including sizes, colors, materi-
als, prices, and availability. The app is geared to-
ward those interested primarily in high-end fashion
designers. FashNYC brings awareness to the fashion
industry, connects with seasonal events, creates an
exciting new shopping experience, and establishes
a presence in today's heightened mobile application
culture.
The app was developed for the Layar AR browser
using PHP programming and MySQL as a database.
The student team focused not only on creation of in-
dividual assets and associated Web pages (fig.3) but
also on the overall packaging, user experience (fig.4),
and final layout for the navigational map (fig.5). Since
the seamless connectivity to various social networks
and fashion-based websites was key here, the app
user interface became a critical part of the overall ef-
fort.

Figure 3
Virtual changing
room with FashNYC

The newfound physical context adopted by AR

games encourages players to push the boundaries of
social conventions and accepted public behavior.
Unlike more passive forms of entertainment such
as reality TV or even active-yet-confined console-
based games, the AR framework incorporates phys-
ical activities and social interaction as well as en-
couraging exploration, learning, and discovery. Fur-
thermore, as activities integrate digital media culture
within the built environment-cities-these games pro-
vide an insight into our physical-digital selves and
better understanding of ourselves and our commu-
nities.

CAAD Education - Volume 1 - eCAADe 32 | 359

 While the student team developed a fully func-
tional prototype [3], they also proposed the next level
of functionality that went beyond the scope of the
class and student technical competencies.
This was an important part of the overall design
strategy, where each team was asked to develop the
project concept initially and later rework it into a
more formalized business proposal with future devel-
opment stages indicated. One of the proposed fea-
tures that could be implemented with Layar's interac-
tive graphics was garment (image) recognition. The
proposed app functionality was expressed in the fol-
lowing statement.
Let's say you're looking for a new blue suit. You see
one you like in the window of Versace. To be quite frank,
360 | eCAADe 32 - CAAD Education - Volume 1
you can't afford it, but you really like the color and the
style. You can take a picture of the suit and have the ap-
plication search for it. As the application searches for
it, you take the time to watch a video of the Fall/Winter
2013 runway show where the suit was first featured. You
Figure 4
notice you like the way it looks when on the runway, but
Mobile interface for
the blue one that was shown looks even better. Once the
FashNYC
suit has been found in the search, you can try to find sim-
ilar garments in nearby stores. You're in luck! That ap-
plication located a similar blue tight-fitting suit in Zara
across the street for half the price! Bam! Success. (stu-
dents: Philip Caleja, Nick Haby, & Daniel Schittone)
As with any new app or a product relying on so-
cial interconnectivity and input, the key issue is to de-
velop a critical mass of active users who would prop-
agate its virtual life. This is a major challenge facing
many new media products including an AR commu-
nity like the one proposed by students. The strategy Figure 5
to address this impediment and help with the future App navigation
commercialization of the app was to tie it to a partic- map
ular event that is highly localized with a defined time
frame. The student team proposed to connect it to
the Fashion's Night Out (FNO) event or the New York
Fashion Week. While this was not implemented yet, it
provides a feasible strategy for launching an AR app
product that is highly contextualized with its theme,
location, and timing. This is also an approach used by
other mobile AR games such as Comfort of Strangers
[4] that rely on a critical mass of participants for their
success.
Highline Tour
Highline Tour [5] is a navigational and informational
mobile AR app geared toward visitors to the High
Line, an urban park in New York City (fig.6).
It provides users with historical and current in-
formation as well as plans for future developments.
Its location-aware functionality allows for sorting and
positioning data in relationship to the urban context.
It shows year-around activities with imagery of var-
ious plants and foliage reflecting seasonal changes
occurring in the park. Users of the app can look at
a particular section of a project and freely navigate
Figure 6 through historic photographs and future proposed
Highline designs (fig.7). The Highline app utilizes a Layar AR
navigational AR app browser that is available for most mobile platforms.
After initial time spent on understanding Layar SDK
environment, students focused on gathering geo-
location data for individual points of interest (POIs)
(fig.8) and setting up online databases.
Since this particular section of the course was
made up almost exclusively of architecture and de-
sign students, teaching faculty had to provide ini-
tial help with basic PHP programming and MySQL
database setup. As part of the development process,
students participated in hands-on workshops orga-
nized by teaching faculty and on some occasions re-
Figure 7 ceived a skeletal prototype of an app. This helped
The diversity of to stage the progress of the project in such a way
assets developed that at any level of its development, students had a
by students fully functional prototype ready for testing with vari-
ous numbers of features and assets. The focus of the
student design team was on gathering relevant infor-
mation, imagery, and outside references. The second
stage involved population of the database, interface
design, and development of Web page links with ad-
ditional information.
Since many of the assets were Web pages related
to the app content, students had to consider designs
that were both desktop and mobile browser friendly.
This quickly became a challenge on its own, consid-
ering the diversity of mobile devices (phones and
tablets), their screen resolutions, and horizontal/ver-
tical layouts. Once all the assets were in place, the
Figure 8 design team focused on overall packaging, user ex-
Points of Interest perience, and interface design.
(POIs) for Highline
app

CAAD Education - Volume 1 - eCAADe 32 | 361

 Augmented Interiors
Figure 9
The goal of the project was to enhance communica-
A student working
tion between interior designers and their clients, and
with a
to empower consumers to experience the impact a
marker-based AR
particular design or set of furniture may have on their
application
home. Traditional home decorating is done by imag-
ining what a space would look like with the new fur-
niture or other design features without having a true
sense of scale or color gamut. Most commonly, cus-
tomers would measure the space in a house and see
if a new piece of furniture would fit within. Let's con-
sider another scenario.
You're looking through a furniture catalog and find
a piece that you like. But how would it fit into your liv-
ing room? Now you can find out without leaving your
couch, or wherever you are. Take the marker attached in Figure 10
the catalog, place it on the desired location, download A student
the AR app, and look through the display of your mobile interacting with the
device camera. The piece of furniture you're considering Maze Game
is there for you to see in the context of your own living
room.
The marker-based AR application associates
each marker with a piece of furniture, material color,
or design features. The combination of markers al-
lows for a high number of variations of possible de-
signs. This app would allow ordinary people to take
design into their own hands and see exactly how a
new furnishing would look like in the context of their
home before they buy it. The approach does not re-
quire a digital rendering of the entire room. Instead,
it overlays a product in real time over the image of the
existing space. The applicability of AR in this project
is appropriate; it achieves its desired effect with very
few resources, could be easily commercialized, and
has the potential of reaching a broad consumer pop-
ulation.
This AR application was developed with the Pro-
cessing platform utilizing the ARToolKit library, which
provided a relatively streamlined production process
with good online support. The actual prototype
(fig.9) was developed for a regular PC laptop, not a
mobile device.

362 | eCAADe 32 - CAAD Education - Volume 1

 The major limitation of this platform was a low forms were introduced. Presently, the latter would be
resolution of AR markers-16 by 16 pixels-allowing for the platform of choice from the physics engine, eas-
a small number of possible marker variations. This iness of development environment, and mobile out-
provided a major limitation to the size of the project put perspectives. Additionally, the Unity3D game en-
and the number of assets. Furthermore, even with gine would provide a more effective and streamlined
mid-size projects, some markers were easily mistaken environment for graphic user interface (GUI) devel-
when they looked too similar to each other. opment, particularly in tracking game scores and en-
hancing user interactions.
The Maze Game
This marker-based AR game involves navigation of a CLASS OBSERVATIONS AND LESSONS
virtual ball through a virtual maze by physically mov-
LEARNED
ing around and tilting the AR marker (fig.10). Move-
AR technology is entering a new stage where it is no
ments and adjustments of the marker in the physi-
longer the domain of technology-oriented individ-
cal world are translated to the virtual space and in-
uals with heavy involvement of computer program-
terpreted with physics-based interaction (gravity and
ming and other software tools. Products such as Vu-
collisions).
foria, Qualcomm's plugin for the Unity3D Game En-
While this is a relatively straightforward design,
gine, delivers a highly functional tool that can be
it involved a wide range of problems to be resolved
easily integrated into academic teaching and profes-
and thus multiple software toolkits. Four main toolk-
sional practice. The choice of a game engine like
its that contribute to this game's functionality. FLAR-
Unity3D further makes the commercialization of AR
toolkit (Flash/ActionScript port of ARToolkit) deals
technology easier and more imminent. The ability
with camera and marker detection. Papervision3D
to integrate physics and other modules already exist-
(open-source real-time 3D engine for Flash) deals
ing in game engines simplifies the development pro-
with the construction and placement of maze walls.
cess and reduces the need for technology savviness
JiglibFlash (open-source ActionScript 3D physics en-
from the creative team. This does not mean that the
gine) provides the collision detection between the
development is completely effortless as far as cod-
ball, the floor, and the maze walls as well as gravity to
ing is considered-scripting is always required for ef-
propel ball movement. Finally, FlashDevelop (open-
fective game engine implementations-but it signifi-
source code editor for ActionScript 3) compiles all the
cantly eases the learning curve, leading to democ-
layers of code and runs the game application. The
ratization of digital creative tools. This transition
maze walls are built and placed based on X, Y, and
from technology heavily involving tools to designer-
Z coordinates, and then the gravity is directed in-
oriented technology was visible in class projects with
ward from the Z axis. The floor is the plane with colli-
over time, shifting from Processing and Flash-based
sion detection, preventing the virtual ball from falling
development environments (Interior Decoration and
down.
3D Maze projects) to Vuforia and Unity3D. This di-
Another variation of this game, proposed but not
rectly facilitates the content and the user becom-
realized, could utilize a mobile device instead of a
ing the primary drivers for the future of AR. This
computer. It would use a stationary marker with the
also suggests that the climate is ready for design
device functioning as a display and a virtual maze. In
schools to embrace AR technology as a new creative
this case, the device's accelerometer, compass, and
and information visualization medium. Through the
tilt sensors would provide the rotation and slop infor-
AR projects and courses discussed above, students
mation. This is one of the earlier projects developed
are becoming aware of new modes of visual and
for the course before the Vuforia and Unity3D plat-
data-based thinking. Concepts such as location and

CAAD Education - Volume 1 - eCAADe 32 | 363

 context awareness form an important framework for
dealing with the over-supply of information and nav-
igating the current, almost ubiquitous data jungle.
While teaching AR-based courses, early lectures
and discussion are usually heavily involved with the
mechanics of AR technology, which often overwhelm
students. However, this initial technology shock
quickly evaporates, with projects' focus shifting to-
ward design, user experience, and content. Projects
connect with other disciplines and uses that respond
to a broad range of social and cultural needs. Stu-
dents perceive AR technology, even more than other
modes of computer graphics, as highly transpar-
ent, without a strong technological footprint. Thus,
this technology naturally transitions them to explore
diverse content-based topics. There was very lit-
tle "technology for the sake of technology" attitude
among students, who naturally gravitated toward
the multitude of ways to connect AR technology with
design, cultural, or social needs.
Furthermore, many of the team projects were de-
veloped by the multidisciplinary teams with students
from various programs across the campus, including
digital design, computer science, and information
technologies, but also architecture, interior design,
product design, and media communications. Some
of the most successful teams were in fact a random
combination of competent and passionate strangers
who just met during the first class when prospective
semester projects were discussed and teams were
formed. The fact that students brought to the team
their own distinct competencies and were able to re-
alize their expertise provided a successful base for
their team projects.
CONCLUSIONS
AR-based applications increasing occupy an impor-
tant place in branding/marketing, tourism, educa-
tion, and many other parts of life. AR has brought the
virtual and the physical world closer and made them
highly interconnected and interdependent through
location awareness, enhanced data overlays, and
user-focused content. It also finds its applications in
364 | eCAADe 32 - CAAD Education - Volume 1
a diverse range of disciplines and ...
A number of the AR applications discussed here
exemplify an idea of "learning anytime, anywhere,"
which builds on Weiser's proposition for the role of
computation in the 21st century (Wesser, 1991). This
new role synergizes key characteristics of AR environ-
ments that include location awareness of data sets,
always-connected networks, and the ability to super-
impose images of the physical world with interactive
digital graphics. It allows for passive as well as ac-
tive interaction with information and virtual content.
Users are able not only to visually experience static
information but also to interact with data in more
dynamic and speculative ways by posing "what if..."
questions. These speculative investigations create an
environment of increased user engagement with the
benefits of experiential learning.
REFERENCES
KIM, K, SEO, B, HAN, J and PARK, J 2009 'Augmented Real-
ity Tour System for Immersive Experience of Cultural
Heritage', In Proceedings of VRCAI, Yokohama, Japan
NIEDMERMAIR, S and FERSCHIN, P 2011a 'An Augmented
Reality Framework for On-Site Visualization of Ar-
chaeological Data.', In Proceedings of the 16th Interna-
tional Conference on Cultural Heritage and New Tech-
nologies, http://www.stadtarchaeologie.at/?page_-
id=5665, pp. 636-647
NIEDMERMAIR, S and FERSCHIN, P 2011b 'An Aug-
mented Reality Framework for Architectural Appli-
cations', In Proceedings of the 8th International Sym-
posium on Location-Based Services, ed. Georg Gartner
and Felix Ortag,, pp. 192-205
WEISER, M 1991, 'The Computer for the Twenty-First Cen-
tury', Scientific American, 1991, pp. 94-100
[1] www.metaio.com/customers/case-studies/lego/
[2] the-tuts.org
[3] www.layar.com/layers/pdnar1/
[4] comeoutandplay.org/2008_comfortofstrangers.php
[5] www.layar.com/layers/highlinefinal/
Framing Parametric and Generative Structures
A Novel Framework for Analysis and Education

Ivo Vrouwe1 , Burak Pak2

1,2
KU Leuven, Faculty of Architecture
1
www.ivovrouwe.net
1,2
{ivo.vrouwe|burak.pak}@kuleuven.be

In this paper we aimed at the development of a novel tool to facilitate the

structured analysis of architectural construction principles, materials and
production methods in digital design and fabrication practices. In order to assist
the understanding and teaching of these subjects, we employed a taxonomy of
spatial design construction (Vrouwe 2013). By using the taxonomy, we analysed
and categorised 34 parametric structures published in the IJAC Journal
(2002-2014). Informed by this study, we aligned the initial taxonomy using
various framing strategies. As a result we developed a new framework for spatial
design construction specifically customised for the design and fabrication of
parametric structures which can potentially serve as a constructive tool to create
a novel design learning environment and integrated teaching strategies.

Keywords: Digital Fabrication, Parametric Design, Education, Framing,

Pedagogy

INTRODUCTION niques, the end products possess a high level of com-

As a result of the evolution of parametric and gen-
erative design strategies, over the last decade, com-
mon practices in digital fabrication have signifi-
cantly transformed. With a constructive approach,
computer controlled machines have been used and
tested to challenge the physical and digital bound-
aries to produce new tectonics and to rethink the
techniques for application in architecture and design.
As a result, parametric design and digital fabrication
have become of great significance to our spatial de-
sign education today (Kolarevic 2008).
Due to the rapid emerging traditions in digital
fabrication, combined with the shortage of fitting
frameworks to contextualise the projects and tech-
plexity in terms of their structure, geometry and ma-
teriality. In practice, these projects seem inaccessible
for well-intentioned design students to learn from on
the one hand, on the other hand, they are difficult to
teach and implement through workshops.
Therefore it is necessary to define a learning en-
vironment in which constructive education can take
place with a specific focus on the understanding and
teaching of digital fabrication practices. For this pur-
pose, the development of a novel tool is required to
facilitate a structured evaluation of architectural con-
struction principles, materials and production meth-
ods, as used in digital fabrication. Furthermore, there
is a need for an inclusive vocabulary for categorisa-

CAAD Education - Volume 1 - eCAADe 32 | 365

 tion and instruction to make techniques and projects
more accessible and teaching them more efficiently.
Reflecting on the above, this study aims at the
development and testing of a tool for the structured
analysis and teaching of digital fabrication in spa-
tial design. This tool will be derived from a master-
framework: the Supertypes-Subtypes Taxonomy for
Spatial Design Construction as developed by the
first author (Vrouwe 2013) based on the research
of prominent research by Martin (1996), Bucquoye
(2002), Ashby (2007), Kula (2009) and Engel (2007),
refer to Figure 1.
In this context, we will start our paper with a brief
introduction on the use of framing strategies in re-
search and education and present the master frame-
work (Section 'Frame Alignment of the Master Frame-
work'). Afterwards, we will illustrate the usability of
the master-framework through a survey of a wide
range of cases and present a comparison of subtypes.
Subsequently, we will present a reframed framework
for digital fabrication (Section 'Framing Parametric
and Generative Structures'). In conclusion, we will
conclude by an elaboration on the implications of our
findings on architectural design education and rele-
vant practices.
FRAME ALIGNMENT OF THE MASTER
FRAMEWORK
This research is a part of the PhD Thesis of the first au-
thor and is a continuation on the work presented in
the eCAADe 2013 Conference. In this study, a taxon-
omy for spatial design construction was introduced.
In this paper we will situate this framework in the
context of digital design and fabrication and derive
a framework particularly for this purpose. The frame-
work will be based on a common terminology to in-
tegrate data into visual frames as a basis for compu-
tation.
Framing and Frame Amplification
We used various framing strategies to construct a
framework for digital fabrication. From a wide array
of conceptions on frames and framing as a theory, in
366 | eCAADe 32 - CAAD Education - Volume 1
this research, we limited ourselves to two of them.
In the first viewpoint, frames were used in a se-
mantic environment as understood in social sciences.
In this context, framing as a concept originates from
the work of Erving Goffmann (1974) and is further
developed by Entman (1993) and Benford and Snow
(2000). In Frame Analysis, Goffmann (1974) discusses
the relevance of a condition in which a concept is un-
derstood. When something is understood within a
"world" or "reality", selective attention organises ex-
periences and generates meaning within a certain
event. In an unstable context, meaning, movements
and events change and adapt. In order to partic-
ipate in these changes, Benford and Snow (2000)
introduced four alignment strategies; frame bridg-
ing, frame amplification, frame extension and frame
transformation.
In the second viewpoint, frames were used in an
ontological environment as understood in informa-
tion sciences. In this strategy, frames are applied to
structure large chunks of information into tractable
entities (Wilensky 1987). These framing strategies
have a great tradition in object centred systems and
computer architectures. In these cases, frames are
used to structure data for representation purposes
in a stereotyped or conventional situation (Minsky
1975).
Both viewpoints bare various qualities in differ-
ent applications. In this research we used the onto-
logical strategies to construct the master-frame for
spatial design construction. In order to test the frame
for robustness, completeness and thoroughness in
its given context, semantic strategies like frame align-
ment were applied.
FRAMING PARAMETRIC AND GENERATIVE
STRUCTURES
During the development and customisation of the
tool introduced in the previous section, we analysed
a wide range of parametric structures published in
the IJAC Journal (2003-2014). We made a systematic
survey of these structures based on the following cri-
teria:
Figure 1
Master Framework
for Spatial Design
Construction
(Vrouwe, 2013).

CAAD Education - Volume 1 - eCAADe 32 | 367


• Digitally designed and fabricated
• Physically produced in 1:1 scale
• Included the word "parametric" in the full text
of the relevant article
When multiple structures were presented in a pa-
per they were treated as unique cases. As a result
of our analysis, 34 structures satisfied the introduced
criteria. Each of these cases was encoded using the
subtypes in our master frame.
Figure 2 illustrates an example of this process
with images of the structure followed by the rele-
vant icons representing our analysis. After applying
the encoding method presented above we were able
to categorise the structures regarding the processes,
orientation and structural principles as well as ma-
terials, products and processes employed (Figure 3).
According to our findings, cutting natural polymers
(wood etc.) in sheet form, chemical polymers in sheet
form and shaping natural polymers (wood etc.) in
sheet form were the most common materials, prod-
ucts and processes combinations (Figure 3 on the
bottom left).
368 | eCAADe 32 - CAAD Education - Volume 1
Figure 2
Vector-active
construction of
chemical polymer
wire, processed by
printing, connected
substantively in a
multiaxial
geometry (Example
structure from the
IJAC paper of Hack
et al. (2013))
Moreover, the most frequently used structures
were observed as biaxial/section active produced
through cutting, biaxial/surface active produced
through shaping and uniaxial/vector-active pro-
duced through machining (Figure 3 on the bottom
right).
Reflection on framing exercise
Our survey indicated a clear need for the alignment
of the master frame. In order to adjust the main con-
struction taxonomy with the digital fabrication envi-
ronment, three frame alignment strategies were em-
ployed. As the result of the use of a particular set of el-
ements in the subtypes "processes" and "structures",
frame amplification was the first strategy to be ap-
plied. With regard to processes, CNC-machinery was
chosen predominantly over hand-held tools to min-
imise room for error on the one hand and optimise
the workflow on the other. In this sense, the digi-
tal process supertype aimed at subtypes with CNC
capabilities. Because most of the students and de-
signers discussed have an education in architecture,
the new framework addressed the basic background
in structural analysis based on certain construction
 stereotypes. In order to support this theoretical back-
ground, the subtypes of parametric structures were
aligned with the stereotypes discussed.
Secondly, in a great part of the projects surveyed
above, behaviour of the model played a significant
role. Therefore, we applied the frame extension strat-
egy to incorporate these qualities into our frame-
work. In these subtypes behaviour was considered
as static when no movement or response is sensi-
ble in the model or the behaviour of the model is
very slow and unintended; smart when the object
responds in real-time to environmental stimulus; in-
telligent when the object aims at optimising a sys-
tem within a collection of parameters or predictive

Figure 3
The frequency of
materials,
processes,
products,
structures,
orientations and
connections among
34 structures
extracted from the
IJAC Journal
2003-2014 and their
combinations.

CAAD Education - Volume 1 - eCAADe 32 | 369

 Figure 4
Supertypes for the
derived framework
(Moussavi 2009;
Moussavi and Kubo
2006; Di Mari and
Yoo 2013; Vyzoviti
2011; Weinstock
2008).

370 | eCAADe 32 - CAAD Education - Volume 1

models (Negrete-Martínez 2008); interactive when
human input is required to initiate response; respon-
sive when both intelligent and interactive behaviour
is included in the system (Velikov and Thün 2012; Ox-
man 2010).
Finally, we observed that digital designers priori-
tised digital geometry of surface and solid manipula-
tion over material orientation. In order to assist this
working strategy, frame transformation was used to
manipulate the supertype "orientation" into the su-
pertypes "parametric solid" and "parametric surface".
CONCLUSION AND FUTURE SUGGESTIONS
The presented master-framework enabled the struc-
tured analysis of digital manufacturing projects. Us-
ing various alignment strategies, the introduced sub-
types were introduced into a new context-specific
framework. Building a shared vocabulary provided
a better understanding of digital fabrication tech-
niques and made it easier to communicate the con-
tent (Figure 4). It also allowed us to identify recur-
ring patterns and practices (Figure 3) which helped
us to develop a novel and aligned framework for dig-
ital fabrication.
In the future the presented framework and fram-
ing strategies can serve as a constructive tool to
create a novel learning environment and integrated
teaching strategies. This environment can be an in-
teractive mobile application which assists the stu-
dents to identify stereotypes and basic elements of
the taxonomy. Furthermore, the taxonomy can serve
as a basis for the development of a parametric design
and modelling interface which prioritises construc-
tion techniques as a starting point.
REFERENCES
Ashby, M., Shercliff, H. and Cebon, D. 2007, Materials:
Engineering, Science, Processing and Design, Elsevier,
Oxford, UK
Benford, R.D. and Snow, D. 2000, 'Framing Processes and
Social Movements: An Overview and Assessment',
Annual Review Sociology, 26, pp. 611-39
Bucquoye, M. 2002, From Bakelite to composite: design
in new materials, Oostkamp : Stichtingkunstboek,
Ghent, Belgium
Engel, H. 2007, Structure Systems, Hantje Cantz, Ost-
fildern, Germany
Entman, R.M. 1993, 'Framing: Towards Clarification of
a Fractured Paradigm', Journal of Communication,
43(4), pp. 51-58
Goffman, E. 1974, Frame Analysis: An Essay on the Organi-
zation of Experiences, Harper and Row, London, UK
Hack, N, Norman, WL, Langenberg, S, Gramazio, F and
Kohler, M 2013, 'Overcoming Repetition: Robotic
fabrication processes at a large scale', International
Journal of Architectural Computing, 3(11), pp. 285-
300
Kolarevic, B. 2008, Manufacturing Material Effects; Re-
thinking Design and Making in Architecture, Rout-
ledge (Taylor & Francis), London, UK, New York, USA
Kula, D. and Ternaux, E. 2009, Materiology: Materialen en
Technologieën: De Gids voor Creatieven, Frame Pub-
lishers, Amsterdam, the Netherlands
di Mari, A. and Yoo, N. 2013, Operative Design; A Cata-
log of Spatial Verbs, Bis Publishers, Amsterdam, the
Netherlands
Martin, J. 2006, Materials for Engineering, Woodhead
Publishing Limited, Cambridge, UK
Moussavi, F. and Kubo, M. 2006, The Function of Orna-
ment, Actar Publishers, Cambridge, USA
Negrete-Martínez, J. 2008, 'Paradigms behind a discus-
sion on artificial intelligent/Smart systems', in May-
orga, R. and Perlovsky, L. (eds) 2008, Toward Artificial
Sapience Berlin, Springer Verlag, Berlin, Germany,
pp. 201-205
Oxman, R. 2010, 'The New Structuralism: Conceptual
Mapping of Emerging Key Concepts in Theory and
Praxis', International Journal of Architectural Comput-
ing, 8(4), pp. 419-438
Velikov, K. and Thün, G. 2012, 'Responsive Building En-
velopes; Characteristics and Evolving Paradigms', in
Trubiano, F. (eds) 2012, Design and Construction of
High-Performance Homes, Routledge
Vrouwe, I. 2014, 'Ideograms as a Tool for Constructive
Sensemaking', Message, 1(1), pp. 34-43
Vyzoviti, S. 2011, Soft Shells: Porous and Deployable Ar-
chitectural Screens, Bis Publishers, Amsterdam, the
Netherlands
Weinstock, M. 2008, 'Metabolism and Morphology', in
Hensel, M. and Menges, A. (eds) 2008, Architectural
Design March/ April, Wiley Publishers
Wilensky, R. (eds) 1987, Some problems and proposals
for knowledge representation, University of Califor-
nia, Berkeley, USA
CAAD Education - Volume 1 - eCAADe 32 | 371
372 | eCAADe 32 - CAAD Education - Volume 1
Reprogramming Architecture
Learning via Practical Methodologies

Elif Erdine1 , Alexandros Kallegias2

1,2
Architectural Association (AA) School of Architecture
1,2
{elif.erdine|alexandros.kallegias}@aaschool.ac.uk

This paper aims to address innovative approaches in the pedagogical aspects of

architecture by describing the work of AA Summer DLAB and Athens | Istanbul
(AI) Visiting Schools of the Architectural Association (AA) School of Architecture
in London. The presented work is part of a research which enables a more
seamless transition from design to fabrication and from academia to profession.
The paper formulates the pedagogical and methodological approach towards the
integration of generative design thinking, large-scale prototyping,
kinetic/interactive design, and participatory design. As such, a discussion on the
methods of overcoming the fragmented nature of architectural education via the
elaboration of the methodology, computational setup, fabrication strategies, and
interaction / kinetic modes of the selected programmes is aspired.

Keywords: Computational design research and teaching, Biomimetics,

Generative Design, Kinetic / Interactive design, Participatory Design

INTRODUCTION The digital revolution in architecture, marked by

Natural systems demonstrate interrelated levels of
complexity by recycling their materials, allowing for
change and adaptation, and utilizing energy (Frazer,
1995). The complexity that is observed in natural
systems has provided the inspiration for a new ap-
proach to design and construction, becoming an es-
tablished part of the architectural discourse in the re-
cent decades. This approach is equally applicable to
the educational aspects of the practice, thus enabling
us to rethink in full the norms of the design discipline.
While architecture is defined as the style and method
of design and fabrication of physical structures, it re-
flects the technological and socio-economic circum-
stances of its' own time.
the integration of CAD/CAM tools into all design-
related practices, has substantially reconfigured the
interaction between architecture and natural sys-
tems. With the vast range of digital tools, the archi-
tect is now able to explore the correlation between
the multiple subsidiary systems operating across a
range of scales in all design related fields.
Situating itself within the complexity paradigm
and its design-oriented implications, the research
posited in this paper aims to investigate the con-
temporary applications of interaction which animate
the architectural piece according to user feedback,
thereby creating the potential for dynamic spatial ex-
periences which enhance the way by which we per-
ceive, learn and practice architecture. Concepts of

CAAD Education - Volume 1 - eCAADe 32 | 373

 real-time feedback loops and temporality in architec-
ture can play a key role in shifting the paradigm both
in practice and in theory. These concepts and their
applications are demonstrated with examples from a
series of ongoing international programmes, where
students are organized into design teams, designing
proposals in the creative atmosphere of collabora-
tive, unit-based learning environment. As such, the
paper formulates the pedagogical and methodolog-
ical approach towards the integration of generative
design thinking, large-scale prototyping, kinetic/in-
teractive design, and participatory design.
AA VISITING SCHOOLS
Moving away from the conventional educational
paradigms, the Architectural Association Visiting
School introduced in 2008, reforms the way architec-
ture is being taught world-wide. While deductive rea-
soning and the knowledge of the classics remain as
part of the teaching module, the main objective is to
purge the invisible wall between students and tutors.
The AA Visiting Schools are highly flexible educa-
tional modules which run for any length of time from
three days to a whole term depending on the agenda
and design goals. Regardless of geography, what all
the programmes share are exciting and radical levels
of invention and experimentation, all of which are to
help in developing skills in different modes of ana-
logue and digital, 2D and 3D production using the
myriad of mediums at one's current disposal. Many
of these 'laboratories', 'building programs', 'nomadic
studios' or 'schools' are formed with collaborating
partners - academic, industrial, commercial and/or
creative - while others venture out independently to
forge their own paths (AA Visiting School Prospectus
2013-2014, p.20-21, 2013).
The aim of AA Visiting Schools is to re-organize
the architectural education in response to today's
challenges. As part of this aim, the pedagogical for-
mat of these programmes is set to recognize students
for their ability to shape the future; students and tu-
tors interact on the same level, combining knowl-
edge with new ideas, expertise with eagerness for
374 | eCAADe 32 - CAAD Education - Volume 1
novelty. Learning here occurs in a diverse, interna-
tional group of students and professionals. In some
cases there is a very small cohort of no more than
15 or 20 students, which facilitates specialist input,
while in others tutors work with as many as 100 stu-
dents on the construction of 1:1 prototypes of large-
scale ingenuity. A number of programmes are based
in the world's largest global cities, others in some of
the world's remotest and harshest locations.
AA Summer DLAB, commenced in 2006, takes
place every summer at AA's London home, Bedford
Square, and its Hooke Park facilities in Dorset. Athens
| Istanbul (AI) Visiting Schools, which have been ini-
tiated in 2011, take place every spring in Athens and
Istanbul as a series of consecutive programmes. Each
series of Visiting Schools lasts for two weeks, and stu-
dent numbers range from 20 up to 40 participants.
Both Visiting Schools are organized around the con-
cept of experimentation, which involves the testing
of non-precedent approaches, novel techniques and
design thinking by analysing existing design prob-
lems thoroughly, evaluating the outcomes of the
analyses, and offering original interpretations for po-
tential outcomes. The outcomes do not present
themselves as ultimate answers to the design prob-
lems, but as part of an ongoing process of design
experimentation. More specifically, the system fol-
lowed in these programmes is a set of methods and
principles which interact with each other in multiple
ways. It is a system of complexity. This is applied
both in the design generation processes as well as
with regards to theoretical and physical parameters
of a project.
METHODOLOGY
Computers handle vast amount of data through al-
gorithms incorporating a set of design parameters
which the architect can then use and manipulate. In
AA Summer DLAB and Athens | Istanbul (AI) Visiting
Schools, the mere parametric mutation of design ob-
jects with the purpose of generating variations is ex-
ceeded by the ability to produce an architectural pro-
posal through simulation. The emphasis in design ex-
plorations becomes the process rather than the end-
result. Formation transcends form. The work of Sum-
mer DLAB and AI Visiting Schools is characteristic of
the benefits that derive from a generative design pro-
cess, particularly one which involves interaction.
The qualities of the complexity paradigm, char-
acterized by the spontaneous emergence of differ-
entiated interdependent subsystems that can adapt
to various external stimuli, are becoming the guid-
ing principles in how designers think and produce
architectural systems (Gruber, 2011). The underly-
ing condition of self-organized systems posits a sig-
nificant diversion from parametric towards genera-
tive design processes in the realm of architecture. It
can be argued that parametric design methodolo-
gies set up rules and conditions for the creation of
an anticipated outcome/set of outcomes, since there
is a linear relationship between the input parame-
ters and the output. On the other hand, genera-
tive design methodologies are articulated through
the abstraction of biological principles, such as self-
organization or evolution, thereby giving empha-
sis on the bottom-up growth of lower-level entities,
namely agents (Reynolds, 1999). This phenomenon,
which is realized with object-oriented design in the
computational paradigm, creates the potential to
correlate the complexity observed in natural systems
with the complexity in architectural systems.
Initially, students are introduced to the princi-
ples of selected complex systems in nature, serving
as the basis for the extraction and abstraction of spe-
cific rules which direct the local interactions of the
agents with each other and their environment. These
interactions lead to the emergence of complex sys-
tems, demonstrating intelligent behavior on a global
level. By fusing the qualities of natural processes with
the notion of spatial reconfiguration, the objective is
to generate design systems where the architect does
not get involved with creating an end product ac-
cording to rules/parameters, but instead initiates a
system whose effects will be continuously adapted
on a multitude of interdependent levels. As such,
while the potential of object-oriented architectures is
investigated, the current role of the architect in con-
temporary architectural discourses is also speculated
upon. Furthermore, the fact that the chosen compu-
tational method is object-oriented, in contrast with a
linear setup, creates the opportunity to focus on the
process of design generation, whereby the concept
of formation replaces form, which would act as the
sole end result in conventional practices.
The computational platform used in design ex-
plorations, the open-source environment Processing,
operates as an object-oriented programming lan-
guage (OOP). The association between the digital en-
vironment and the conceptual framework described
above facilitates the generation of seamless digital
design tools in Processing. In this respect, the com-
putational platform exceeds being a mere digital tool
and functions as an experimentation ground which
students understand and test the various phenom-
ena of complexity paradigm, such as branching, net-
work formation, and reaction-diffusion systems (Fig-
ure 1).
More specifically, the use of algorithmic design
via Processing enables participants to run various
scenarios according to different conditions. Param-
eters which are set in the digital simulation are not
ignorant of the architectural context but have direct
connection to certain possibilities of each given sit-
uation. On each consecutive version of the Sum-
mer DLAB and AI Visiting School programmes, the
design problems are explored with the application
of explicit codes which originate design outcomes
based on formulas emulating the formation process
of natural constructs. The concept of agent-based
systems found in bird formations or schools of fish,
the theory of L-systems in the shaping of plants and
trees, the recursive character of miniscule viruses or
enormous natural formations are just a few examples
which are referenced in the research of these pro-
grammes. Such natural conditions are translated in
algorithms and are set to be adjusted according to
desired design outcomes. The participant/architect
interacts with the formation throughout the entire
simulation process. Through this type of investiga-
CAAD Education - Volume 1 - eCAADe 32 | 375

tion, participants receive a direct know-how on mat-
ters which, until recently, were theoretical.
FABRICATION
In the contemporary post-industrialized world, while
the architectural proposals evolve and follow more
sophisticated forms and tackle aspects of economy
and sustainability, design teams assigned to plan the
physical proceeding for the realization are often in-
teracting with other disciplines. Currently, design,
construction, finance, and legal aspects all overlap
and interrelate even more strongly than they have in
the past.
One of the major objectives of the presented
pedagogical approach is to demonstrate the conti-
nuity of the workflow between computational soft-
ware and digital assembly procedures leading to
physical fabrication. In this respect, during the de-
velopment of each design unit's proposals, students
fabricate physical models in various scales (Figure 2).
This first step towards the realization of physical pro-
posals demonstrates how the use of generative tools
in combination with digital fabrication techniques al-
low for a coherent and smooth transition from the
"drawing board" to the building site. At this stage,
design teams also begin to experiment with the con-
cepts of interaction and kinetics in architecture with
the aim of transforming static built models into ani-
376 | eCAADe 32 - CAAD Education - Volume 1
Figure 1
Generative
form-finding
processes exploring
various biomimetic
models.
mated kinetic prototypes while considering a hierar-
chical design arrangement. With the incorporation
of the physical computation environment Arduino
in the design generation process, a continuous in-
formation loop between Arduino and Processing is
formed. In this way, students are able to manipulate
their design simulations in Processing according to
real-world physical constraints, such as lighting lev-
els, movement, and distance.
The notion of actively participating in the fabri-
cation of models is integrated in the teaching mod-
ules of the AA Summer DLAB and AI Visiting Schools.
Every participant deals with both the solution of the
given architectural problematic and the issue of real-
izing his or her proposal. More specifically, the pro-
cess of trial-and-error which occurs through the col-
laboration of different people within a team as well
as the collaboration among different teams is advan-
tageous in the understanding of current practices in
architecture. As such, the pedagogical setup acts
as an exemplary model of the processes being uti-
lized in the professional world. Distinct groups work
together with the responsibility of the delivery of
specific aspects of the project's design and building
phases. The design must not only be covering the ar-
chitectural brief but must be structurally sound and
appropriate for the use and location of the given de-
sign task. Inevitably, the success of every project
Figure 2
The initial
transformation
from the digital to
the physical
paradigm through
scaled models
fabricated with
digital prototyping
techniques.
is correlated with the effective management of the
time given. This has proven to influence the team's
design proposals as they become more aware of the
constraints of the physical world. Their proposals sur-
pass the superficial use of digital images. The in-
vestigation at this stage involves the extensive un-
derstanding of the digital three-dimensional form-
making and its' application through digital fabrica-
tion tools.
Serving as a pattern of the practice in the pro-
fessional world, the fabrication processes are multi-
scalar. Small scale physical models are produced
and occasionally disposed as their role is essential to
provide the team with an initial perception of their
concept's physical limitations. Similar to the digi-
tal form-finding processes, the fabrication comprises
of a series of steps through which the participants
gain practical, hands-on skills while interacting with
each other on different levels and stages. In the fi-
nal phase of the programmes, students are asked to
build a one-to-one scale working prototype of an in-
teractive installation. Working on an installation scale
has proven to be highly beneficial, as the fabrication
process of the installations by digital fabrication tech-
nologies and their assembly procedures by the stu-
dents allow them to understand the physical behav-
ior of the materials they work with, such as wood,
acrylic, and aluminium (Figure 3).
INTERACTION MODES
In a world exposed to rapid technological advances,
the question being addressed in this paper is not
what architecture is but what architecture can do.
The work completed in the AA Summer DLAB and AI
Visiting Schools is one that ventures into unclaimed
territory, opening up new areas of thought for re-
search and development. At this phase, Interactive
Architecture takes places as two modes of communi-
cation; one that is calibrated as User to User commu-
nication and one that is calibrated as User to Model
communication.
Interaction between user and model explores
the aspect of perception which users experience
when they interact with the built environment and
when the built environment interacts with them. The
interactivity happens through use but it is also "felt"
by the observation of the user and the interactive
model. In the context of user to user communica-
tion, the interactive/kinetic structure acts as a ves-
sel of conveying information from one user to an-
other in certain ways affecting their behavior. Here,
interactivity is formulated on anthropomorphic defi-
nitions. Complex physical interactions are made fea-
sible by the use of low-tech computational hardware,
Arduino, which can detect and react to human be-
havior. The creative fusion of algorithmic design, dig-
ital fabrication, and embedded computational intel-
CAAD Education - Volume 1 - eCAADe 32 | 377

ligence in the field of architecture is therefore discov-
ered. The research focuses not only on the benefits of
adaptation, which makes it possible to anticipate and
accommodate the response mechanisms of the built
structure, but also on the environmental and psycho-
logical ramifications of such structures.
Transformation via human interaction can take
place in a variety of scales and modes, ranging from
kinetic morphological properties and acoustic per-
formance to ambient lighting effects which bear
the potential of altering the perception and the ac-
tual configuration of space. One of such interac-
tion methods explored so far include video mapping,
which has been realized in Summer DLAB 2012 via
"Fallen Star", a large scale installation which is 4.5 me-
ters long, 3 meters wide, and 1.8 meters high. For
this installation, each design team has been asked to
challenge the perception of the architectural instal-
lation through the projection of a three-dimensional
natural growth algorithm. The specific parameters of
the algorithms can be manipulated with an IPAD in-
terface, which is also designed by student teams, en-
abling a direct interaction between the user and the
installation. The segregation between the architect
and the end-result diminishes even more with the
progression of the user interface as an architectural
tool, pointing to the extensive possibilities of partici-
patory design in architecture. The installation reacts
378 | eCAADe 32 - CAAD Education - Volume 1
Figure 3
Students working
on large-scale
prototypes at AA
Hooke Park facilities
in Dorset (image
credit: Valerie
Bennett).
to the user's input on sound, choice of algorithm to
run, and the manipulation of the algorithm itself by
changing its parameters from the interface (Figure 4).
A different method of incorporating interaction
with architecture has been the creation of a kinetic
light-diffusing architectural installation during Sum-
mer DLAB 2013. The installation, "Light Forest", is a
5 meter long, 3.5 meter wide, 1.5 meter tall (at its
maximum activated state) which can activate itself
through motion according to the distance data re-
ceived from users around it. For this design task, stu-
dents have integrated light with movement by de-
signing, fabricating, and assembling a complex body
of kinetic parts with embedded lighting pieces which
react to human movement. As the sensors receive
data on the distances and amount of visitors in space,
the tessellations of the kinetic surface react with sub-
tle movements, generating various light formations
in space. This interaction mode involves the direct
communication between user and the physical struc-
ture, but also indirectly includes the communica-
tion of users among themselves through the physical
structure. As such, the notion of permanence gives
way to the constant flux of building formations (Fig-
ure 5).
In AA Athens Visiting School 2014, the interactive
aspect of the design has been realized through the
construction of a 1-to-1 kinetic pathway. Through
Figure 4
The final working
prototype of
Summer DLAB
2012, “Fallen Star”,
exploring the
notion of
interaction with
video mapping
which is
manipulated with
an IPAD interface.

Figure 5
The final working
prototype of
Summer DLAB
2013, “Light Forest”,
which activates
itself according to
distance data
received from users.
The image on the
right shows the
activation
mechanism.

a series of proposals, the final design for a 5 meter
long, 2.5 meter high and 1.3 meter wide pathway has
been reached. This interactive/kinetic pathway con-
sists of two sets of triangulated panels. Each set has
its own purpose towards the effect of interaction for
the two modes, user-to-user and user-to-model. One
set has a pattern of LEDs which gets activated accord-
ing to the presence of humans within the passage.
The second set is made out of carefully woven elas-
tic strands. The triangles attached to these strands
rotate and twist in different manners creating open-
ings and obstructions. The path reveals itself once
two users are found in the right location in front of
each side. There is the element of time that is embed-
ded into the structure; once the users are in the pas-
sage, they have a certain amount of time before they
can exit. The longer they remain within, the more agi-
tated the structure becomes, resulting in the nervous
movement of the strands. This project is eventually
an experiment on the way humans interact with their
environment and with each other through it (Figure
6).
CONCLUSION
The application of biomimetic principles with object-
oriented architectures, participatory design, and ki-
netic transformation which are explored rigorously
throughout these programmes bear the potential for

CAAD Education - Volume 1 - eCAADe 32 | 379


creating architectures of adaptation and reconfigura-
tion. As it has been demonstrated with these installa-
tions, the concept of having a structure that is related
to motion and real-time reaction to external stimuli is
no longer an idea of the past but can be realized by
low-tech materials and technologies used in every-
day life. The pedagogical approach engaged in these
series of programmes aim to demonstrate how to
actively interweave the conceptual and practical as-
pects of innovative design paradigms, enabling stu-
dents to fully understand the theoretical, computa-
tional, and physical advantages and constraints of
such methodologies. Active engagement and expo-
sure to the procedural levels of creation enables ar-
chitecture to evolve as a supple communication net-
work, characterized as an authentic collaborative and
purposeful convening medium.
After several years of applying this specific
methodology in architectural education, various
evaluations can be comprised. Among valuations on
aspects like international collaborations, research on
cutting-edge technology, low tutors-to-students ra-
tio, and exposure to hands-on building techniques,
one outcome is believed to be of particular signifi-
cance. It is the aspect of time. This aspect has a con-
troversial character as it is equally beneficial and dis-
advantageous for the goals of these programmes. On
the one hand, it might be argued that the duration
is not long enough while testing innovative design
380 | eCAADe 32 - CAAD Education - Volume 1
Figure 6
AA Athens Visiting
School 2014
installation;
interaction with a
kinetic pathway.
concepts together with interactive/kinetic concepts.
Nevertheless, it has also revealed that students tend
to be more engaged in an intensive learning process,
whereby the limited amount of time in relation to
the limitless learning resources provided renders par-
ticipants to get more absorbed. Through the pro-
grammes' intensity, participants are introduced to a
way of critical design thinking and performing that
lingers much longer than the duration of these pro-
grammes. As such, AA Summer DLAB and the Athens
| Istanbul Visiting Schools do not act as mere tutorial
workshops; they build audiences for new ideas. They
are designed to evolve as an unpredictable crowd,
not a rational individual; re-organizing architectural
education in response to today's challenges.
REFERENCES
Frazer, J 1995, An Evolutionary Architecture, Architectural
Association Publications, London
Gruber, P 2011, Biomimetics in Architecture: Architecture
of Life and Buildings, Springer-Verlag, Wien
Reynolds, C 1999 'Steering Behaviors For Autonomous
Characters', Proceedings of Game Developers Confer-
ence, San Jose, California, pp. 763-782
[1] http://www.aaschool.ac.uk/downloads/prospectus/2013-
14/AA_VISITING_SCHOOL_PROSPECTUS_13-14_A.pdf
Learning to be a Vault
Implementing learning strategies for design exploration in inter-scalar
systems

David Stasiuk1 , Mette Ramsgaard Thomsen2

1,2
The Royal Danish Academy of Fine Arts, Schools of Architecture, Design and
Conservation (KADK) Centre for Information Technology and Architecture (CITA)
1,2
http://cita.karch.dk/
1,2
{david.stasiuk|Mette.Thomsen}@kadk.dk

Parametric design models enable the production of dynamic form, responsive

material assemblies, and numerically and geometrically analytical feedback. The
value potential for design produced through the procedural transformation of
input parameters (or features) through algorithmic models has been repeatedly
demonstrated and epistemically refined. However, despite their capacity to
improve productivity and iteration, parametric models are nonetheless prone to
inflexibility and reduction, both of which obscure processes of invention and
discovery that are central to an effective design practice. This paper presents an
experimental approach for the application of multiple, parallel computational
design modelling strategies which are tested in the production of an inter-scalar
model array that synthesises design intent, the simulation of material behaviours,
performance-driven adaptation, and open-ended processes of discovery and
categorical description. It is particularly focused on the computational potentials
embedded in interdependent applications of simulation and machine learning
algorithms as generative and descriptive drivers of form, performance, and
architectural quality. It ultimately speculates towards an architectural design
modelling method that privileges open model topologies and emergent feature
production as critical operators in the generation of flexible and adaptive design
solutions.

Keywords: parametric design, computational modelling, machine learning,

multi-objective optimisation, k-means clustering

CAAD Education - Volume 1 - eCAADe 32 | 381

 INTRODUCTION
Although parametric modelling allows designers to
dynamically produce variable geometries and ex-
ecute sophisticated analyses of performance at-
tributes, certain problems of both rigidity and reduc-
tion endemic to such design systems are well doc-
umented. Parametric models are limited because
they necessitate a predetermination of both model
topology and the feature domains that constitute the
model's parameter space. As such, the designer ei-
ther must explicitly define all parameters and rela-
tionships between model elements at the start of
the design project or risk breaking the model dur-
ing any ensuing reconfiguration (Davis 2013). This
rapid calcification of the design space is antithetic
to the experience of design as a process of inven-
tion and discovery. A second problem is associated
with risks in the oversimplification of descriptive pa-
rameters such that the defining design criteria ig-
nore or suppress potentially useful information em-
bedded in the model output. While this implicit re-
duction preserves design control and makes the cre-
ation of the model tractable (Davis 2013), the result-
ing limitations act as an impediment to a more ex-
ploratory design practice. These problems are only
compounded by an increasing sophistication in de-
sign models' ability to represent in both the produc-
tion and analysis of a number of useful characteris-
tics, such as material assembly behaviours, occupa-
tion patterns, or energy-related performances. An in-
creased facility in designing for and with these con-
siderations is central to continued innovation in de-
sign modelling (Tamke et al. 2011). However, this in-
creased capacity also offers the opportunity to explic-
itly describe the complexity of the design problem at
hand, and in response develop open-ended design
systems that have the capacity to address them (Car-
iani 2008). For this reason there remains a significant
need for the ongoing development of tools through
which we can flexibly capture and understand com-
plex interstitial dependencies across model elements
for directed performances and as a means to enhance
the pursuit of invention and discovery in the devel-
382 | eCAADe 32 - CAAD Education - Volume 1
opment of design models.
This paper traces the exploration of new
methodologies in this pursuit, specifically addressing
the related desires to maintain open feature domains
for both design and evaluation, in order to alleviate
requirements for the pre-configuration of both value
and element connectivity. It also describes the re-
lated pursuit of simultaneously embedding material
behaviours in morphogenesis and producing multi-
ple targets for performance optimisation. The aim
for the project is to establish methods for design-
ing with open topologies in which the dependen-
cies between parameters are both emergent and
changeable during the design process. To this end,
an experimental approach is implemented for multi-
ple computational strategies, applied in parallel and
sequential operation in the digital environment.
Background
This project takes its point of departure from an in-
tensive workshop with the Digital Matter Master's
studio, led by Areti Markopoulou at the Institute
for Advanced Architecture Catalonia (IAAC) in early
2014. This workshop introduced multi-objective evo-
lutionary solvers operating in a simulated material
context. Through a series of physical models us-
ing simple rattan splines and connector ties, seven
teams focused on the development of generative de-
sign algorithms that synthesise material behaviours,
the topological transformation of connectivity be-
tween constituent elements, and quantitative, multi-
objective optimisation design goals. Through the ex-
ploration of these networks as morphogenetic rule-
driven systems for incremental formation, a series
of variables available for deployment in a multi-
objective evolutionary model were then developed.
The set of simple goals that emerged from this pro-
cess of rapid physical prototyping were related to
material usage, the generation of space and struc-
tural performance and capacities. Measures related
to connectivity and material deflection define those
goals related to structural performance, and the cou-
pling of the conflicting goals of minimizing material
use while maximizing envelope size provided simple
spatial performance goals. What emerged through
these iterations was a series of rigorous pseudo-
code algorithms for both generation and evaluation.
The distinct approaches of each team established a
framework for developing a more general design sys-
tem capable of generating multiple, highly varied
configurations, that nonetheless retain formal and
organisational legibility.
Learning to be a Vault
"Learning to be a Vault" is an experiment consisting of
a significant digital exploration and consequent rep-
resentation of the design space, along with the pro-
duction of 25 1:25 models and a 1:1 demonstrator ap-
proximately 5 x 5 x 2.5 m. It has been installed as part
of the "What does it mean to make an Experiment"
exhibition at the Royal Academy of Fine Arts School
of Architecture during the Spring of 2014. The IAAC
workshop operates as a baseline for the formulation
of this experiment, which focuses on the synthesis
of multiple computational strategies implemented in
an inter-scalar modelling environment. Of particular
interest is the deployment of machine learning algo-
rithms. One possible strategy for advancing the flex-
ibility of parametric models as performative instru-
ments is tied to rethinking how parameters (or fea-
tures) are treated in the modelling space. Typically,
feature domains are established as inputs for the ge-
ometrical outputs of an architectural design model
(Davis 2013). However, there are opportunities to
recast certain features in a parametric model as ei-
ther dynamic or descriptive entities, enabling the de-
signer to engage in alternative means to search the
design space. For this experiment, the focus is tied
to using machine learning algorithms to classify large
collections of model outputs generated through a
multi-objective evolutionary optimisation solver into
legible groups - or species - of response.
In "Learning to be a Vault", there is a tight inte-
gration of three distinct modelling systems. The first
is a generative model, constructed with the consid-
erations developed through the workshop described
above. It recursively builds up networks of actively-
bent splines that have the capacity to achieve great
variety in form. Intrinsic to this model development
is not only the topological connectivity established
through the generative algorithm, but also a spring-
based form-finding simulation of the design material,
which is rattan, a soft but highly flexible wood-like
plant. The second model is a multi-objective evolu-
tionary optimisation solver, which leverages the gen-
erative model input parameters as its genotypes and
its outputs as its phenotypes. These phenotypes are
subsequently analysed for five distinct performance
measures. Finally, the outputs from this model are
analysed using k-means clustering, an unsupervised
learning algorithm that identifies intrinsic relation-
ships between the data elements of its input data
points. This analysis takes on the form of classifica-
tion, which allows for an intuitive understanding of
many outputs as belonging to legible and distinct
groups. The following section will briefly discuss ma-
chine learning in general, and in the context of com-
putational design models, will then offer focus more
on the relevant machine learning strategies used in
this particular experiment.
MACHINE LEARNING IN COMPUTATIONAL
DESIGN MODELS
Machine learning is a field of research and practice re-
lated to developing computer programs that are con-
figured to improve their performance at a given task
through experience (the acquisition and processing
of incremental data) (Mitchell 1997). It is focused
on three primary categories of interest: task-oriented
studies, cognitive simulation, and theoretical analy-
sis (Carbonell, Michalski & Mitchell, 1983). For task-
oriented studies, machine learning is largely inter-
changeable with the field of statistics, as many pre-
dictive algorithms (such as linear or logistic regres-
sions) are used extensively in both. Task-oriented ma-
chine learning is primarily concerned with the classi-
fication of data points as function values within some
descriptive domain. These domains are most often
discrete (e.g. Boolean, integer, or categorical val-
CAAD Education - Volume 1 - eCAADe 32 | 383
 ues) but can also be continuous (e.g. real numbers).
Learning models are divided into two main types: su-
pervised and unsupervised.
Supervised learning models rely on training data
where the predicted outcomes are known, so that
the model develop input (or feature) transforma-
tions and weights that will allow it in the future to
more effectively predict the outcomes for unknown
data. As each new evaluated data point used (or
instance) for training has its actual outcome deter-
mined, the learning model gains new intelligence
about its performance and recalibrates itself accord-
ing to its driver algorithm. Unsupervised models op-
erate differently. They allow for the features of data
sets to self-organize. These are primarily used for the
discretisation of data into descriptive or functional
categories, without the need for training data. In
effect, when the outcomes are unknown, unsuper-
vised learning algorithms are designed to allow for
the internal relationships within the features of a data
set to create emergent descriptions of each instance.
Both supervised and unsupervised learning models
are often used discretely to resolve subcomponents
of larger problem spaces. In effect, the outcome of
any learning model is simply a new feature for appli-
cation in a new decision space.
Both supervised and unsupervised learning ap-
proaches have precedent in architectural design
modelling. Their applications have been theorised
for nearly as long as CAD systems have been in
use, but in the last twenty years - and increasing in
step with advances to the processing power of per-
sonal computers - their implementation has become
tractable and in many cases standardised. The most
commonly used learning algorithms in design mod-
elling are genetic or evolutionary solvers, whose epis-
temological and functional maturity in the design
sciences far outreaches that of other modes. How-
ever, several of those other, lesser-used approaches
- which include neural networks of multiple types
and k-means clustering algorithms - have nonethe-
less been successfully implemented in the past and
present excellent opportunities for the future exten-
384 | eCAADe 32 - CAAD Education - Volume 1
sion of learning algorithms in design modelling. For
the purposes of this experiment, we will focus on
two of these: multi-objective evolutionary solvers
and the k-means clustering algorithm for classifica-
tion. Additionally, spring-based particle simulation
systems will be examined according to the criteria
used for establishing that an algorithm constitutes a
learning system.
Multi-Objective Optimisation
Evolutionary algorithms have been applied in the
field of architecture for over twenty years, most pop-
ularly with the pioneering work of John and Julia
Frazer in Department 11 at the Architectural Associ-
ation, starting in the late 1980's (Frazer 1995). In a
rough summary of their operation, they rely on a gen-
erative system for the parsing of inputs (genotypes)
into outputs (phenotypes), which are then subjected
to a performance analysis - most often numerically
represented. They begin with a pool of randomly as-
signed genotypes, test the performance of resulting
phenotypes, and then "breed" the most successful
offspring through the crossover (and potential) mu-
tation of genotypes. Over time, this process allows
for successful genotypes to thrive and pass on their
parameters to their children, and is repeated through
a number of generations until satisfactory objective
values have been achieved by the resulting pheno-
types. Because evolutionary solvers rely on setting
explicit targets for performance measurement, they
can be understood as supervised learning models,
with the optimisation objectives used for training.
In an architectural design context, there are key
considerations for making this approach tractable.
There must be a balance struck between the model's
ability to produce an artefact that is recognisable as
a built object, but also a capacity for the model not
overly limit the range of possible outcomes (Janssen
et al. 2000). This can be understood as a simultane-
ous desire for legibility and variety in the outputs of
such models.
As the computational tools available for execut-
ing evolutionary algorithms have become more ac-
 cessible, so has the capacity to introduce many par-
allel objectives into a single model. The concept of
n-dimensional objectives in such design models is
keyed around the Pareto front, which operates as a
means to understand the outcomes of the design
space (Figure 1). The Pareto front is defined by a
boundary of phenotypes whose optimisation values
define a convex hypervolume in the n th dimension,
with n being the total number of optimisation ob-
jectives. For example, single-objective search will re-
sult in a point, with a single, optimum phenotype. A
two-objective search will result in a convex polyline,
and a three-objective search will result in a convex
triangulated hull. Each phenotype that constitutes
the Pareto front can then be said to be "optimised"
in some capacity. The front communicates the trade-
offs that exist between phenotypes: as a phenotype
demonstrates better optimisation for one objective,
it loses value for another. The Pareto front is conve-
nient to visualise these performance trade-offs in up
to three dimensions (Caldas 2003), but becomes sig-
nificantly more difficult to understand once it defines
a hypervolume in four or more dimensions. Com-
pounding this difficulty in navigating performance
objectives is the fact that as the front's dimensionality
increases, it is comprised of a great many more con-
stituent phenotypes, potentially resulting in thou-
sands of Pareto-optimised candidates (Winslow, et al.
2010).
Figure 1
Example of a Pareto
front for a
two-dimensional
objective
optimisation
It is, however, this volume of phenotypic output
that must be addressed if the continued advances in
analytical computational methods are to be synthe-
sised in multi-objective architectural design models.
This experiment then deliberately identifies an objec-
tive space that produces such a volume of output,
and engages in the use of a second learning algo-
rithm to allow for the designer to engage in a form
of search that encourages a process of discovery.
K-Means Clustering
K-means clustering - also known as Lloyd's algorithm
- is a distance-based, unsupervised learning algo-
rithm that uses intrinsic relationships between data
points in large sets to discretise them into clusters
containing the most similar instances. K-means clus-
tering achieves this by using the Euclidean distance
between data points in n-dimensional space (where n
is the number of features used to drive the clustering
algorithm) to find solutions for similarity for a user-
specified k number of clusters. It does so through an
iterative process of computing the "nearest-distance
centroid rule", which can be visualised as a voronoi di-
agram. The algorithm begins by assigning k-number
of random points in the data as centroids for the di-
agram. It then evaluates all of the other points in
the set, and assigns them to a "cluster" based on the
voronoi boundary containment. Then, the average
value of all contained data points is used to recalcu-
late the centroid of the voronoi diagram. This pro-
cess is repeated until the cells achieve a stable state
such that data points no longer move between clus-
ters, and thus centroids no longer require recalcula-
tion.
As in the case of the Pareto front, this process is
easiest to visualise in lower-dimensional space. Fig-
ure 2 illustrates a k-means clustering algorithm run
on a two-dimensional data set as a collection of XY
points. The iterative readjustment of the voronoi
bounding space according to subsequent contain-
ment tests for data points finally results in cells that
describe boundary conditions easily identified by the
eye as being optimally discretised.
CAAD Education - Volume 1 - eCAADe 32 | 385

K-means clustering is used extensively to find
unknown relationships between data points in large
data sets. Even though it is difficult to understand
euclidean distances within data sets of high dimen-
sionality in the abstract, clusters produced by such
analyses - so long as the data being used to struc-
ture them possesses useful descriptive capacities -
generally makes intuitive sense when examining out-
comes. For example, k-means clustering is regularly
used in internet search engines for the purpose of
clustering similar search elements together, where
based on such data elements as key words and the
geography of a story's origin, related content can be
grouped together that can readily be grasped by an
observer.
K-means clustering has been implemented in ar-
chitectural models as drivers for fabrication rational-
isation. A prominent example is the facade paneli-
sation process developed by Gehry Technologies for
the Soumaya Museum in Mexico City. This freeform
structure's facade is comprised of more than 16,000
panels, each of which, in the original design geom-
etry, is unique. However, by defining a maximum
acceptable tolerance for each panel, the team was
able to deploy the k-means algorithm to reduce the
386 | eCAADe 32 - CAAD Education - Volume 1
number of moulds required for fabrication to 49. Figure 2
More generally, the architects and engineers for Evo- K-means algorithm
lute software make use of k-means clustering in their on a two-
plug-ins for facade panel rationalisation. dimensional data
For this experiment, k-means is not used to ra- set with k=9: a) data
tionalise elements for fabrication, but in an entirely array, b) boundaries
different capacity: to effectively understand, search, for k random
and discover the complex and varied design outputs sample points, c)
that high-dimensional multi-objective optimisation test for inclusion
algorithms produce. and move
centroids, d)
Simulation Systems redefine
Spring-based particle simulations are frequently boundaires, e)
used in computational models for form finding and repeat c & d until
the integration of material behaviours into the de- stability, f ) stable
sign process. They use the same principles as agent- solution
based models, with individual particles acting as
agents, influencing each other according to defini-
tions of connectivity and a rules that define the con-
sequence of interaction. Interestingly, they meet the
explicit conditions that Mitchell details in his defini-
tion of a machine learning program: through the ex-
perience of their interaction with one another over
time, simulation particles improve their performance
at the task of describing material or behavioural con-
sequences for their constituent elements.
Extensive research has been made regarding
swarm intelligence and the effectiveness of encod-
ing material capacities into simulation-based design
models (Thomsen et al. 2010), but it is worth extend-
ing this frame of reference to consider the parallels
that exist between such design systems and machine
learning algorithms. This experiment relies on the
use of material simulation for both form-finding and
the production of evaluation objectives for the evo-
lutionary algorithm.
METHOD
"Learning to be a Vault" (Figure 3) is set up to
synthesise the modelling systems and considera-
tions described above, including 1) a morphogenetic
script capable of producing a wide variety of forms
grounded in a tendency toward legibility, 2) a spring-
Figure 3
Process diagram for
"Learning to be a
Vault"
based particle simulation for form-finding and load-
ing behaviour of the actively-bent rattan member,
3) a multi-objective evolutionary solver for optimis-
ing the solution set and producing results on the
Pareto front, and 4) a k-means clustering algorithm
for searching the design space. The evaluation cri-
teria for the experiment lies in the effectiveness of
this integration, and the legibility, searchability and
meaning inscribed by the final clusters.
Generation and Optimisation
The morphogenetic script implements a recursive ar-
ray of elements. There are two types of elements
considered: primary members that attach both ends
to the floor in an elastic arch, and secondary mem-
bers that span between two primary members. Sec-
ondary members connect to primary members in
two consecutive locations such that active-bending
forces are interdependently shared between them.
A number of transformations are available to pri-
mary and secondary elements through each recur-
sion such that many forms of varying symmetry and
spatial consequence are producible (Figure 5). Once
formed, each collection is passed through a spring-
based simulation engine (the Kangaroo plug-in for
Grasshopper) to execute both form finding, and de-
flection under self-loading. Next, using a multi-
objective optimisation solver (the Octopus plug-in),
each collection is evaluated for suitability for con-
tinued breeding according to the following objec-
tives: 1) minimal deflection, 2) restriction of tight
radii, 3) targeted area covered, 4) targeted connec-
tivity, and 5) targeted height for secondary members.
Subsequent generations are then recursively passed
through the same generative and evaluative process
until suitable results have been produced.
Searching the Design Space
Based on this collection of objectives, over 80 gen-
erations more than 2000 unique Pareto optimised
phenotypes are produced (Figure 4, 5). These phe-
notypes are then analysed according to series of
descriptive paremeters developed by the designer.
In this case, the descriptive parameters applied are
a collection of numerical transformations of phe-
notype geometry. These are then passed through
the software Octave (a numerical solver capable of
rapidly executing complex machine learning algo-
rithms) for computing the k-means algorithm and
back into Grasshopper for visualisation and consid-
eration of results.
Through an iterative implementation of this pro-
cess, the designer uses both heuristics and intuition
to identify further possible transformations to the
data that might result in improved clustering differ-
entiation. In this case, 18 distinct numerically de-
scriptive variables were developed, including: total
materials used, average variance in height of sec-
ondary members, the radius of a circle inscribed in
each primary member's start point, and the same
for end points, and the distance variance between
primary member base points. On their own, such
parameters appear banal. However, when parsed
through the clustering algorithm, they combine with
CAAD Education - Volume 1 - eCAADe 32 | 387
 Figure 4
Partial section of
ten clusters (out of
80, representing
over 2000 total
phenotypes)

one another to discretise a seemingly fragmented
and intractably varied set of solutions into a collec-
tion of legible clusters.
It is here a new process of discovery emerges.
Through the invention and application of these de-
scriptive variables, one is able to steer the intrinsic re-
lationships between relatively simple aspects of the
geometry's underlying data structures toward a co-
herent and easily searched design space. This pro-
cess of iteration is critical and effectively operates as
a second stage for design execution. The clustering
of suitably large collections of phenotypes - particu-
larly those that have been optimised for a variety of
objectives - may ultimately result in a form of speci-
ation, wherein the designer is able to identify highly
distinct solutions that may be suitable for diverse ar-
chitectural applications.
The development of effective data visualisation
tools is essential for managing the information about
clusters that is produced. One measure for determin-

Figure 5
selected 1:25
models

388 | eCAADe 32 - CAAD Education - Volume 1

 ing the efficacy of a clustering analysis is the aver-
age variance for all variables of each constituent data
point relative to its cluster centroid. Here, a smaller
variance indicates clusters that better describe their
constituent data points. As such, a zero variance
requires that there be exactly one cluster per data
point. This would offer no simplication, and there-
fore the purpose of the process is for the designer to
establish a threshold that effectively finds the least
(and most searchable) number of clusters that retain
a high level of descriptive capacity. For this experi-
ment, a dashboard is developed that allows for the
designer to rapidly understand the dynamics of dif-
ferent clustering passes, with each incremental pass
adding another descriptive cluster. The dashboard
indicates how adding numbers of clusters from one
pass to the next decreases variance for the entire so-
lution, but does so at a decreasing rate, and at the ex-
pense of searchability. To facilitate finding an effec-
tive threshold, it allows for the designer to select in-
dividual clusters, see the individual phenotype most
representative of the cluster centroid, read the rela-
tive average values of the descriptors used for exe-
cuting the algorithm, and see which clusters are sim-
ilar to it both within the same clustering pass and in
those both preceding and subsequent to it (Figure 6).
Figure 6
dashboard
examining a single
cluster, and its
orientation within a
solution set
containing several
passes of increasing
cluster counts
DISCUSSION
Increased computing power and a proliferation of
tools make more advanced multi-objective optimisa-
tion design models more tractable and create a need
for new methods for managing phenotype outputs
emerges. Attendant to having more complex collec-
tions of objectives is a higher volume of outputs. So
not only does this make it difficult to visualise these
outcomes on the Pareto front, but the sheer number
of phenotypes to be processed makes a case-by-case
review difficult. This experiment represents a first
look at applying unsupervised learning algorithms
for the purpose of organising such high-dimensional
data output into coherent segments and as a re-
sult enabling the designer to engage in a secondary
operation of discovery through the development of
descriptive parameters that steer this segmentation
process.
There are a number of opportunities to further
refine and enhance the utility of the approach pre-
sented here. First of all, applying the clustering al-
gorithm between generations in the evolutionary
solver would enable the designer to implement a
form of directed breeding, with non-desirable clus-
ters removed from the pool of available solutions.
This would more tightly couple the different con-
CAAD Education - Volume 1 - eCAADe 32 | 389
 tributing model elements and lend a higher degree
of control to the designer. Also, there are multiple ap-
proaches for numerically processing descriptive fea-
tures to heighten their ability to produce better dis-
tinguished clusters, as well as for identifying thresh-
olds for ideal number of clusters for a given data set.
These approaches can be further explored and imple-
mented.
Conclusions
Machine learning algorithms present a number of
opportunities for the enhancement of computational
design practice. There are varying degrees of ma-
turity between different methods currently applied.
For example, while the use of evolutionary algo-
rithms in the design sciences has an established his-
tory and advanced epistemology, algorithms such as
k-means clustering have had more limited applica-
tion. Neural networks of multiple type have been de-
ployed. "Growing Neural Gas" networks have been
implemented for the development of use and oc-
cupation analysis in urban environments, and have
been deployed for developing way-finding analy-
ses for space plans at multiple scales (Langley et al.
2007). Backward propagating neural networks have
been used to drive robotic instruction for resilient
structural systems (Mehanna 2013). Computational
designers have long applied spring-based particle
and agent-based simulation systems in their models,
which can be seen to exhibit functional equivalence
to machine learning models, but for problems partic-
ularly well-suited to the design sciences.
Machine learning models designed to handle the
most difficult problems are often conceived of and
constructed as a complex of multiple distinct ap-
proaches, operating in some combination of paral-
lel and sequential synthesis (Flach 2012). In addition
to the operational benefit of using machine learning
algorithms directly to enhance the design capacities
of their models, such an outcome should be com-
pounded through the conception of their models as
inter-scalar ecologies of such knowledge-acquiring
algorithms.
390 | eCAADe 32 - CAAD Education - Volume 1
REFERENCES
Caldas, L.G. 2003 'Shape Generation Using Pareto Ge-
netic Algorithms Integrating Conflicting Design Ob-
jectives in Low-Energy Architecture', Proceedings of
CAADRIA 2003
Carbonell, J.G., Michalski, R.S. and Mitchell, T.M. 1983,
'Machine learning: a historical and methodological
analysis', AI Magazine, 4.3, pp. 69-79
Cariani, P. 2008, 'Design Strategies for Open-Ended Evo-
lution', Artificial Life, XI, pp. 94-101
Davis, D. 2013, Modelled on Software Engineering: Flexi-
ble Parametric Models in the Practice of Architecture,
Ph.D. Thesis, RMIT
Flach, P. 2012, Machine Learning: the Art and Science of
Algorithms that Make Sense of Data, Cambridge Uni-
versity Press
Frazer, J. 1995, An Evolutionary Architecture, Architectural
Association Publications
Janssen, P., Frazer, J. and Tang, M.X. 2000 'Evolutionary
design systems: A conceptual framework for the cre-
ation of generative processes', Proceedings of Design
Decision Support Systems in Architecture and Urban
Planning 2000
Langley, P., Derix, C. and Coates, P.S. 2007 'Meta-
Cognitive Mappings: Growing Neural Networks for
Generative Urbanism', Proceedings of th 10th Gener-
ative Art Conference 2007
de Leon, A.P. 2012 'Two Case-Studies of Freeform-Facade
Rationalization', Proceedings of eCAADe 2012
Mehanna, R. 2013 'Resilient Structures through Machine
Learning and Evolution', Proceedings of ACADIA 2013
Mitchell, T.M. 1997, Machine Learning, McGraw-Hill Sci-
ence/Engineering/Math
Tamke, M., Burry, M., Ayres, P. and Thomsen, M.R. 2011
'Design Environments for Material Performance',
Proceedings of Design Modelling Symposium Berlin
2011
Thomsen, M.R., Tamke, M. and Riiber, J. 2010, 'Sea Unsea -
Lamella Flock', in Leach, N. and Snooks, R. (eds) 2010,
Swarm Intelligence: Architectures of multi-agent Sys-
tems, LATP
Winslow, P., Pellegrino, S. and Sharma, S.B. 2010, 'Multi-
objective optimization of free-form grid structures',
Structural and Multidisciplinary Optimization, 40, pp.
257-269
PREMISE:
A discussion on open source, networks, and digital property in architecture.

Wendy W Fok
Harvard University, Graduate School of Design
http://we-designs.com/team/wendy-w-fok
wfok@gsd.harvard.edu

This research paper questions the relationship between the legal, ethical, and
economic challenges that are presented by new ways of utilizing trademark,
intellectual property, and copyright law within design production and creative
purposes of architectural design. While it does not attain a conclusive
resolution-as this is a developmental body of applied research, and includes a
great deal of exploration of reform within the legal system-the paper addresses
the systemic propositions, issues of inspiration and precedents, and case-studies
in architectural design production that are implicit in new forms of legal
transformation which could potentially address the issues within the worlds of
design, economics, and law. It is hereby proposed that by understanding these
relationships with respect to the prevalent global economic model, research in
understanding architecture and law may be able to identify ways of restructuring
and offer alternative (or, evolutionary) modes of ethical protection and equitable
reward to all members contributing to the production system.

Keywords: Digital Property, Open Source, Networks, Intellectual Property,

Extralegal Norms

This research paper questions the relationship be-
tween the legal, ethical, and economic challenges
that are presented by new ways of utilizing trade-
mark, intellectual property, and copyright law within
design production and creative purposes of archi-
tectural design. While it does not attain a conclu-
sive resolution-as this is a developmental body of ap-
plied research, and includes a great deal of explo-
ration of reform within the legal system-the paper
addresses the systemic propositions, issues of inspi-
ration and precedents, and case-studies in architec-
tural design production that are implicit in new forms
of legal transformation which could potentially ad-
dress the issues within the worlds of design, eco-
nomics, and law. It is hereby proposed that by un-
derstanding these relationships with respect to the
prevalent global economic model, research in under-
standing architecture and law may be able to identify
ways of restructuring and offer alternative (or, evo-
lutionary) modes of ethical protection and equitable
reward to all members contributing to the produc-
tion system.

CAAD Education - Volume 1 - eCAADe 32 | 391

 "Nothing is original. Steal from anywhere that res-
onates with inspiration or fuels your imagination... Se-
lect only things to steal from that speak directly to your
soul. If you do this, your work (and theft) will be au-
thentic. Authenticity is invaluable; originality is non-
existent. And don't bother concealing your thievery -
celebrate it if you feel like it. In any case, always remem-
ber what Jean-Luc Godard said: It's not where you take
things from - it's where you take them to."
• Jim Jarmusch (American independent film di-
rector, screenwriter, actor, producer, editor
and composer)
ARCHITECTURAL DESIGN AND ITS PROP-
ERTY
Architectural design has always relied on the de-
sign and creative basis of innumerable factors to de-
termine the authenticity of the product. The iter-
ative process within the development of a design
model, and processes embedded within its inspira-
tions includes the programming, socio-economic un-
derstanding, site analysis, client suggestions, mate-
rial aspects, dimensionality, renderings, computer
aided design (CAD) drawings, hand sketches, and
physical and digital models. All of these elements are
among the list of considerations that allow the abil-
ity of designers to understand the basis of a project
(Lynn, Zardini, Eiseman 2013) . Without recognizing
the motivations that contribute to the affects of a
design, which include a multi-faceted chronology of
factors and influences, the difference between an ar-
chitect (designer) and a contractor (builder) becomes
negligible, and the incentive to use an architect is
diminished. After all, most clients recognize that a
building could be constructed without an architect,
simply with a contractor and engineer; yet, arguably,
it would be significant to note that without an archi-
tect the building would only be a building, and with-
out the design, the building would not be architec-
ture.
Cohen (2007) articulates the current cultural the-
ory behind copyright laws for creative fields as a
state whereby the mainstream of copyright scholar-
392 | eCAADe 32 - CAAD Education - Volume 1
ship has grounded its grand theory in either the the-
ory of rights, or in a theory of economic analysis (Du
Mont and Janis, 2007) . According to Cohen (2007),
"rights theorists seek to derive the basis for copy-
right from the philosophy of property rights; while
others prefer a vision of copyright grounded in prin-
ciples of expressive liberty and deliberative democ-
racy." While the methodologies of creatives are un-
determinable, even among the various lineages of ar-
chitectural styles and movements, there is a common
understanding about the relationship between inspi-
ration, creative discovery, and design process, where
each simultaneously influences the other (see Figure
1).
Figure 1
Traditional to
contemporary
research,
development, and
design process
Innovations versus influences, on the other hand, are
disparate entities of conversation within the field of
design and architecture. Whereby, repeatedly the
subject of influence, or inspiration, often becomes an
issue of copying and appropriation, frequently due
to the matter of the ego, and the term: originality.
In some creative fields, such as choreography, copy-
ing is kept in check through informal (extralegal) in-
dustry norms enforced by private sanctions. In oth-
ers, the freedom to copy actually promotes creativity.
Haute couture gave rise to the term knockoff, copy-
cat, fakes, among others, yet the freedom to imitate
great designs only makes the fashion cycle run faster-
and forces the fashion industry to be even more cre-
ative (Raustiala and Sprigman. 2012). Traceability of
the lineage of influences and precedents, hence be- tion, based on an alternative compensation system.
comes an issue of a priori versus a posteri innova- Should the legal protection measures within the law
tion of a creative idea, and brings attention to the of copyright prove incapable of ensuring the secure
creator versus the innovator. Yet, the question arises distribution of creations over the internet, or as a dig-
about whether an idea could arrive out of itself with- ital format, the reform of the legal intellectual prop-
out the influences of another, and whether the legal erty and copyright system should then promote the
reform of copyright, intellectual property, and trade- ability of traceable metadata within pieces of work to
mark acts regulate these conversations within the prove as an ethical and equitable method of reward-
creative and design fields. ing creators and owners of copyrighted materials.
Cukier and Mayer-Schoenberger 2013 publsihed
CREATIVITY WITHIN NETWORKED SYS- an article in Foreign Affairs Journal which describes
the three ways to understand The Rise of Big Data.
TEMS / COMMONS
The general attack method is to 1) first understand,
The Internet has reshaped the ability of commu-
collect, and use a lot of data rather than settle for
nication by humans, but more importantly it has
small amounts or samples; 2) to shed preference for
also transformed how society gathers data and pro-
highly curated and pristine data, and instead accept
cesses information (Larsen 2014, Guattari 1989, La-
messiness, and tolerate the inaccuracy that benefits
tour 2005). This incredible amount of information
the vast amount of data which outweighs the cost of
sharing has had a tremendous impact on creativ-
using smaller amounts; 3) to stop the quest to dis-
ity, collaboration, and the question of provenance
cover the cause of things, in return for accepting cor-
of ideas. Therefore, to move the law of copyright
relations.
into the next digital millennia, the Copyright Office
Whether the mentioned approaches are the ab-
should consider the ethical and equitable measures
solute approach, the greater developmental under-
through the use of advanced technology as a means
standing between gathered data and useable data is
of protection, in relation to the appropriation of vi-
still in the stage of beta interpretation. And, these
sual, design, and sound arts.
opinions truly become the underlying issues pertain-
If the goal of copyright is to advance the creative
ing to open sourced software, as much to multi-user
process within the sciences and the arts, then the fu-
designed and crowd-sourced projects.
ture of copyright should be to identify the source of
Yonchai Benkler reveals that free software
the creation and design, along with its designer. It
projects do not rely on markets or on managerial hi-
would be beneficial if imbedded data would be in-
erarchies to organize production, while Suber (2013)
cluded within the metadata that is distributed, simi-
persists that the development of Open Access would
larly to information one would read on the informa-
allow consumers to better adapt and share perfect
tion label at a museum. Meanwhile, the alternative
copies of creations at virtually no cost through of-
is to only include enough information to identify and
fering varieties. Programmers do not generally par-
allow endorsement of the original creator ( Zimmer-
ticipate in a project because someone who is their
man, 2011). Much of these examples are close to cur-
boss told them to do so, which alleviates the discus-
rent tagging of digital songs and movies within the
sions on market-based, firm-based, and/or hybrid
entertainment industry (Fisher 2004), such as Digital
models of the economic model. However, the plus
Rights Management (DRM)-protected music ( Gasser
point on open-source participatory, collaborative,
and Palfrey, 2007).
freeware is the basic and radical challenge that it
Fisher (2004) suggests that there should exist
proposes. Therefore, it posits the rational question:
digital fingerprinting in combination with a royalty
if the person is not acting on a profit motive, then
or fee-based system through encryption of informa-

CAAD Education - Volume 1 - eCAADe 32 | 393

 what sort of progressive motive should there be? Al-
truism could potentially be a simplified conclusive
approach; however, the development of a commu-
nal trust, and development, is often a devised state-
ment of cause (Schweik, and English 2012). It has
been said, though, in behavioural economics, that
humans are conditioned to see causes even where
none exist (Benkler 2006).
The contemporary understanding of the net-
work environment looks at a generative state of rad-
ically decentralised, non-proprietary development,
based on sharing resources and outputs among
widely distributed, loosely connected individuals.
These individuals cooperate with each other with-
out relying on either market signals or managerial
commands - the basis of Benkler's (2006) commons-
based peer production.
The radical new modality of organisational pro-
duction also suggests asymmetrical purposes of "the
owner" and the relationship to the commons, includ-
ing the relationship between the use of the com-
mons, and what one could do within the commons.
This opens further the discussion on the necessary
parameters of a) whether the system is open to any-
one or only to a defined group, and b) whether the
commons system is regulated or unregulated (Ben-
kler 2006).
OWNERSHIP IN RELEVANCE TO COLLEC-
TIVE WORKS
Ownership, according to the US law¬ 17 U.S. Code §
201 (ownership of copyright), is divided into four ac-
counts (The Copyright Act of 1976) . A notable aspect
of the law states that the creator of a copyrighted
work does not always own the copyright. In some
cases, other persons or entities own the copyright.
There are also rules governing copyright ownership
when two or more people create the work. Finally,
copyright owners can assign rights to the copyright
to others, particularly for the purpose of marketing
the protected work.
The types of ownership in copyright law include,
a) initial ownership , b) Works made for hire, c) Contri-
394 | eCAADe 32 - CAAD Education - Volume 1
bution to Collective Works , d) Transfer of Ownership
, and e) involuntary transfer . To focus on the current
topic, the discussion on the ownership rights of "Con-
tribution to Collective Works" will be of significance.
The current standing of Copyright Law of the
United States (hereafter, the "Copyright Act") is in-
tended to encourage the creation of art and culture
by rewarding authors and artists with a set of exclu-
sive rights . While the Federal Copyright law grants
authors and artists exclusive right to make and sell
copies of their works, the right to creative deriva-
tive works, and the right to perform or display their
works publicly, using the original. According to the
U.S. Supreme Court, a plaintiff suing for copyright
infringement has to show "(1) ownership of a valid
copyright, and (2) copying of constituent elements of
the work that are original."
With the advent of innovation in the interna-
tional artistic arena, especially within the nature of
works attributed to appropriation art that relies so
heavily on the software and open-source advance-
ments and interchangeability of digital media and
technology (including CAD and 3D printing, among
other forms of digital tools), it brings to question the
relevance, purpose, and effective circumvention of
copyright, moving into the emerging state of the arts.
Copyright law, instead, should be looked upon
as a means to further the ability of technological
advancement, or be complemented by, technologi-
cal measures administered by the Copyright Office.
These dealings should include a better understand-
ing and assistance of technological growth and in-
novation for protecting the appropriation arts and
open-source software (Schweik, and English 2012).
Ultimately, the regulated or unregulated
incentive-based systems (Hughes 1988) each con-
tain socio-political elements that contribute to a
hierarchy of system management, which controls
the access between certain layers of information.
Both incentive-based systems bring to question the
ability- to control the ownership and user interface
between these different layers of information- that is
sold to individuals that 'buy' into a project made up of
multi-user contributions. Another concern between
unregulated system participants, and joint prod-
uct appropriation single party system participants,
within a traceable or licensed system are the ability
to retain copyrights in their contributions within the
multi-user model. This would then allow designers
to license the operation of the multi-user design to
be distributed, reused, or perhaps further explored.
A DIGITAL (ARCHITECTURAL) REFORM
Libeskind (2004) once said, "All architects are prosti-
tutes - that's what Philip John said; they'll do what-
ever it takes for the chance to build." With the recent
rise of architectural infringement discussions from
the likes of Zaha Hadid Architects, for their Wangjing
SOHO Galaxy project in Beijing set to be completed
by 2014, to Daniel Libeskind versus LAB Architecture
studio (Donald Bates), of the Federation Square in
Melbourne of 2002, the vague understanding of ar-
chitectural copyright (Quirk 2013) , or the intellectual
property rights of a designer, raises ethical questions
about how far architects would go to get a chance to
build, or even publish.
Although there exists similarities of architec-
tural representation in both Zaha Hadid Architects
Wangjing SOHO Galaxy vs Chongqing Meiquan 22nd
Century, and Studio Daniel Libeskind Jewish Museum in
Berlin vs LAB Architecture Studio Federation Square,
the approach in understanding the issues of intellec-
tual property varies between the two cases.
Whether it would be too ambitious to claim that
formal derivatives of design in architecture are driven
by scripted procedures and that code-based designs
have led to the formal similarities and exploration in
the technical means of design, application of open-
source and collaborative free-ware in architectural
software has become a common ground for design
exploration within the contemporary field of design
(Reas, McWilliams, Barendse 2010) . Should the prac-
tice of architecture continue on this natural progres-
sion within the field of code-based generation, then
similarly to open source software (OSS) development
comes the issues of code repositories and versioning
systems that comes hand-in-hand with the owner-
ship of code (Schweik, and English 2012).
There has been little to no serious architectural
dialogue on the implications of open source proce-
dures in a constructive manner that closely relates
to the ethical or equitable dealings in collaborative
work, though many theorists and practitioners have
danced around the conversation. Antonelli (2011)
explored "thinkering" thoughts about the novelties
of DIY and open source, stemming from Ratti et al.'s
(2011) experimental op-ed on the Open Source Arc
project, and Usman Haque and Matthew Fuller in-
vestigated the "Urban Versioning System 1.0" that at-
tempts to deliver a deliberated form of "open-source
urbanism that could radically change the conven-
tionalized form of city design".
Mario Carpo has extensively questioned the split
agency between the roles of participatory use of
digital technologies towards the liberal collaborative
ventures that use digital tooling within their prac-
tices, and questions the line between the author and
audience, or the authorship within this isotropic plat-
form, developed within Web 2.0 (Lessig, and Lessig
2006). Open-source software and systems could be
one form of interpretation of the networked model
within computer science, but has become a "social"
norm within the developmental status of digital de-
sign in architecture. The current generation of de-
signer's use of architectural software and structure of
practice is by every means more "social" and "collabo-
rative" than the previous generation, where the CAD-
CAM of the nineties was mostly based on controlled,
proprietary networked environments (Carpo 2011).
The current intellectual property regime for ar-
chitectural and design law is reminiscent to a state-
ment Libeskind once described about his process in
design, "(w)hen you're designing a building, the ex-
perience is kinetic. You cannot always put something
into words; otherwise you will simply produce a ver-
bal diagram. You must feel your way towards your
finished design. It's not until much later that you're
fully aware of what you're doing Cowley 2003) ." With
the ongoing and beta-stage of understanding the
CAAD Education - Volume 1 - eCAADe 32 | 395
 hurdles in the contemporary field of design within
the present legal regime, the main resolution and
suggestion is to incorporate a better understanding
of the creative process, including collaborative prac-
tices within the field of law and architecture which
incorporates designers, architects, and other inno-
vative creative fields. By understanding the existing
absences of cultural aspects within the current prac-
tice of law, the hybridized knowledge of both fields
of architecture and law could potentially offer an en-
hanced social and cultural reform.
NOTES
17 U.S. Code § 201 (a) Initial Ownership - Copy-
right in a work protected under this title vests
initially in the author or authors of the work.
The authors of a joint work are co-owners of
copyright in the work. Accessed May 1, 2014:
http://www.law.cornell.edu/uscode/text/17/201
17 U.S. Code § 201 (b) Works Made for Hire - In the
case of a work made for hire, the employer or other
person for whom the work was prepared is consid-
ered the author for purposes of this title, and, unless
the parties have expressly agreed otherwise in a writ-
ten instrument signed by them, owns all of the rights
comprised in the copyright. Accessed May 1, 2014:
http://www.law.cornell.edu/uscode/text/17/201
17 U.S. Code § 201 (c) Contributions to Collec-
tive Works - Copyright in each separate contribu-
tion to a collective work is distinct from copyright
in the collective work as a whole, and vests initially
in the author of the contribution. In the absence of
an express transfer of the copyright or of any rights
under it, the owner of copyright in the collective
work is presumed to have acquired only the privi-
lege of reproducing and distributing the contribu-
tion as part of that particular collective work, any re-
vision of that collective work, and any later collec-
tive work in the same series. Accessed May 1, 2014:
http://www.law.cornell.edu/uscode/text/17/201
17 U.S. Code § 201 (d) Transfer of Ownership -
• The ownership of a copyright may be trans-
ferred in whole or in part by any means of con-
396 | eCAADe 32 - CAAD Education - Volume 1
veyance or by operation of law, and may be
bequeathed by will or pass as personal prop-
erty by the applicable laws of intestate succes-
sion.
• Any of the exclusive rights comprised in a
copyright, including any subdivision of any
of the rights specified by section 106, may
be transferred as provided by clause (1) and
owned separately. The owner of any par-
ticular exclusive right is entitled, to the ex-
tent of that right, to all of the protection
and remedies accorded to the copyright
owner by this title. Accessed May 1, 2014:
http://www.law.cornell.edu/uscode/text/17/
201
17 U.S. Code § 201 (e) Involuntary Transfer - When
an individual author's ownership of a copyright,
or of any of the exclusive rights under a copy-
right, has not previously been transferred volun-
tarily by that individual author, no action by any
governmental body or other official or organization
purporting to seize, expropriate, transfer, or exer-
cise rights of ownership with respect to the copy-
right, or any of the exclusive rights under a copy-
right, shall be given effect under this title, except
as provided under title 11. Accessed May 1, 2014:
http://www.law.cornell.edu/uscode/text/17/201
17 U.S.C. ß 106 (2002) ("The owner of copyright
... has the exclusive rights ... : (1) to reproduce the
copyrighted work in copies or phonorecords; [and]
(2) to prepare derivative works based upon the copy-
righted work.").
REFERENCES
no authors given 1976, Copyright Act of 1976, USA, USA
Antonelli, P 2011, 'States of Design 03: Thinkering', Do-
mus, June
Carpo, M 2011, The alphabet and the algorithm (Chapter
The Fall, p. 81 -120), MIT Press
Cohen, J E 2007, 'Creativity and Culture in Copyright
Theory', UC Davis Law Review, 40(Georgetown Pub-
lic Law Research Paper No. 929527), pp. 1151-1205
Cowley, J 2003, 'Daniel Libeskind - interview', Prospect
 (http://www.jasoncowley.net/interviews/I200302_- duction transforms markets and freedom (Chapter
P.html), 83 3)', in no editors given 2006, New Haven, Yale Univer-
Cukier, K and Mayer-Schoenberger, V 2013, 'The Rise of sity Press, pp. 59 - 90
Big-Data: How it's changing the way we think about Zimmerman, DL 2011, 'Copyrights as Incentives: Did We
the World', Foreign Affairs Journal, May-June, pp. 28 Just Imagine That?', Theoretical Inquiries in Law, 2, p.
- 40 29–58
Guattari, Felix., F 1989, The Three Ecologies
Fisher, WW 2004, 'An Alternative Compensation System',
in no editors given 2004, Promises to keep: Technol-
ogy, law, and the future of entertainment, Stanford
Law and Politics , Stanford Calififornia, pp. 199-258
Gasser, U and Palfrey, J 2007, 'Case-Study: DRM-
Protected Music Interoperability and e-innovation',
in no editors given 2007, no title given, Berkman Pub-
lication Series, Harvard University
Hughes, J 1988, 'The Philosophy of Intellectual Property',
Georgetown Law Journal, 77, pp. 296-314
Larsen, LB 2014, Networks, MIT Press
Latour, B 2005, 'Network: A Concept, Not a Thing out
there', no title given, pp. 68-73
Lessig, L 2006, Code and other laws of cyberspace (Code
V2) 2nd Ed., Basic Books, New York
Libeskind, D 2004, Breaking Ground: Adventures in Life
and Architecture (A sense of place, P.19 -tribute to
World Trade Center), Penguin Books - Riverhead
Books, New York
Lynn, G, Zardini, M and Eisenman, P 2013, Archae-
ology of the digital : Peter Eisenman, Frank Gehry,
Chuck Hoberman, Shoei Yoh, Canadian Centre for
Architecture-Sternberg Press, , Montréal :Berlin
Du Mont, JJ and Janis, MD 2012, 'Functionality in Design
Protection Systems', Journal of Intellectual Property
Law, 19, p. 261
Paola, A, Adam, B, Lucas, D, Joseph, G, Dan, H, John, H,
Alex, H, John, M, Nicholas, N, Hans, UO, Carlo, R,
Casey, R, Marco, S, Mark, S, Chiara, S and Bruce, S
2011, 'Open Source Architecture', Domus, 948
Quirk, V 2013, 'The 10 Things You Must Know About Ar-
chitectural Copyrights', ArchDaily, 6 Feb 2013
Raustiala, K and Christopher, S 2012, The knockoff econ-
omy: How imitation sparks innovation., Oxford Uni-
versity Press
Reas, C, McWilliams, C and Barendse, J 2010, Form+Code
in Design, Art, and Architecture, Princeton Architec-
tural Press, New York.
Schweik, CM and English, RC 2012, Internet Success: A
Study of Open-Source Software Commons (Chapter 4
Technological and Community Attributes), MIT Press
Suber, P 2013, Open Access, MIT Press Essential Knowl-
edge Series
Yochai, B 2006, 'The wealth of networks: How social pro-

CAAD Education - Volume 1 - eCAADe 32 | 397

398 | eCAADe 32 - CAAD Education - Volume 1
Fabrication
Design By Making
Enhanced Human-Computer Interaction for Digital Conception and
Manufacturing in Architectural Education

Serdar Aşut
Istanbul Technical University
serdarasut@gmail.com

This paper presents an ongoing project which aims to develop an HCI

application. The purpose of the application is to introduce tactile experience into
digital design cycle and to have transparent links between different phases of the
design object. Thus, it will provide a hybrid design environment in which design
conception and manufacturing are integrated and enhanced learning
opportunities towards an architectural education where making is fundamental.

Keywords: Human-Computer Interaction, Tactile Experience in Design,

Architectural Education

INTRODUCTION • To enhance the communication opportunities

This ongoing project aims at developing a new
Human-Computer Interaction (HCI) application to
enable an enhanced and direct relation between the
designer and the digital design environment. The ap-
plication involves toolsets and methods which will
enable the user have a more organic relation with the
design object during digital conception and manu-
facturing processes. Its functionalities are thought
to be promising particularly for architectural educa-
tion by enhancing the modes of communication in
the design studio. It specifically considers the first
year studio, where the essential approach is not to
include complex architectural problems, but to in-
troduce design as a broader concept while focusing
on the fundamental aspects of architectural design,
such as form, scale, space and materiality.
The intentions of the project are:
• To allow the designer directly interact with the
design object through his/her body,
in the design studio and to allow and encour-
age its stakeholders collaborate on the design
object,
• To extend the limits of the design studio as an
environment of learning.
PROJECT BACKGROUND
The idea of this project initially stems from ques-
tioning the relation between the object of design
and its representation. Schön emphasizes the signif-
icance of the representation tools in design and de-
fines architectural design as an activity of producing
the representations of the things to be built (Schön
1984). According to him, design is a reflection-in-
action with talking backs which emerge as both the
maker's spontaneous reflections and the reflections
of the material back to the maker. What is missing
in this very precious definition is that it lacks of the

Fabrication - Volume 1 - eCAADe 32 | 401

 possibility to include what is real instead of mere its
representations into the design process. This project
aims at extending this concept of "talking back" by
including materiality into this definition of "reflective
conversation". It aims at questioning how it could be
possible to manifest a practice of design which di-
rectly deals with what is material or real, instead of
a mere focus on representation.
Chard seeks for a transparency between the
thought and its projection and claims that the level
of transparency is directly related with the capabili-
ties of the tools used (Chard 2005). The concept of
transparency is one of the core issues of this project
and is aimed to be achieved between design think-
ing and different phases of the design object, such as
the physical or digital models as well as the on-site
product.
Therefore the idea of Design by Making refers to
a process of design, where conception and manu-
facturing are merged together and the conversation
is performed continuously and reciprocally between
the physical and digital phases of the design object.
This requires the reintroduction of the notions of craft
and materiality into design process.
In architecture, the reintroduction of crafting as a
mode in design development refers to relocating the
architect back into the core of production, in contrast
with industrialization which had pushed his/her po-
sition to a presenter of a practice which was held by
others, a financial actor, or a rhetoric producer. This
implication re-celebrates the value of crafts which
was excluded from the practice of the architect by
modernity and leaves more room free for what craft
could bring into design.
The way we think is the property of a hybrid as-
semblage of brains, bodies and things (Malafouris
2013). And as Sennett mentions, what we can say in
words may be more limited than what we can do with
things (Sennett 2009). Including things and material
particularities into design conversation will enable a
more transparent media on which a more efficient
design communication is possible. Dearden intro-
duces the concept of material utterance which is a sit-
402 | eCAADe 32 - Fabrication - Volume 1
uated communication act that depends on the par-
ticularities of speaker, audience, material and genre
(Dearden 2006). He claims that all materials have
properties that moderate the ease or difficulty of cre-
ating a material utterance, like for different materials
have different density, ductility, plasticity and mal-
leability. It is important for the designer to experi-
ence these varying particularities which are intricate
to each different material. This can only be achieved
by tactile experience as all the senses including vision
are extensions of the sense of touch and the senses
are specializations of the skin, and all sensory experi-
ences are related to tactility (Pallasmaa 1996). There-
fore it is necessary to include this notion of material
utterance with all its particularities into digital design
cycle.
These notions are actually not exempt from dig-
ital design practices. The new techniques and meth-
ods of digitally enabled making are reaffirming the
long forgotten notions of craft, resulting from a de-
sire to extract intrinsic qualities of material and de-
ploy them for particular effects (Kolarevic and Klinger
2008). Digital environment enables and encourages
the integration of conception and manufacturing.
Moreover, the common practice of representation is
no longer valid, and the digital is now the real itself.
What is seen on the screen is not the representation
of the design object, but the objectified design idea
itself. Its presence is yet in a unique phase, which still
nestles all its material features and is always capable
for phase transition. The geometric model seen on
the screen, the parametric definition of the model,
the prototype produced by the 3D printer or the digi-
tized model produced by the 3D scanner are all differ-
ent phases of the same objectified idea and are able
to transform into each other instantly.
Design, in this manner, performs an exciting sim-
ilarity with vernacular forms of making. Here, just like
in vernacular, making and design are integrated prac-
tices and the maker's direct relationship with the ob-
ject is essential. However, the difference is that a di-
rect bodily relation is yet not fully sufficient.
On the other hand, by the introduction of in-
formation and communication technologies (ICT), a
strong paradigm shift has occurred in architectural
education too. This paradigm shift has implied a sig-
nificant awareness in academia and altered the ex-
pectations from education. The school is no longer
an institution of didactics, but an enhanced environ-
ment of research and experience. The rising influ-
ence of novel concepts, such as Maker Culture, Hack-
erspace or Fab Labs, forces architectural education
to have a stronger emphasis on learning rather than
teaching. Even though these phenomena are not en-
tirely new, the developments in ICT and the spread
of digital culture make them the de facto bases of
learning and production for the future practition-
ers. Hence, the school needs to be an environment
where the necessary facilities and encounters which
are necessary for enhanced learning are provided.
This refers to enhancing the possibilities of communi-
cation and encouraging making in the design studio.
The application which is presented in the next
chapters is being developed towards these princi-
pals in order to enhance the already existing possi-
bilities provided by the digital media, enabling the
designer have a more direct and organic relation with
the design object, integrating design conception and
manufacturing, including the notions of materiality
and crafting, and utilizing this new experience for
enhanced learning in architectural education. The
application is a proof of concept, and currently be-
ing developed. It is still necessary to include more
features, test its functionalities with architecture stu-
dents, and conduct an in-depth theoretical discus-
sion in relation with the project principles.
APPLICATION ENVIRONMENT
3D Motion Detection
The application runs with Leap Motion Controller,
a three dimensional motion detection device. It
tracks the three dimensional movements of hands,
fingers and certain tools without a physical touch
and uses them as input to digital applications. The
reason for choosing Leap Motion Controller for this
project is that it is more precise in detecting the hand
movements, it provides a more user friendly Soft-
ware Development Kit (SDK) which supports several
programming languages including JavaScript, it is
cheaper, smaller, lighter and is easier to operate with,
comparing with similar devices on the market.
There already exist several applications which
use the controller for HCI in different fields such as
gaming, healthcare, education and data visualiza-
tion. In this project, motion detection is thought
to be a promising opportunity to include tactile ex-
perience into design process.There could have been
three different approaches of application develop-
ment to include motion detection into digital design
applications:
1. Developing a standalone application.
2. Developing a plug-in which operates with the
existing CAD software.
3. Developing an application which allows cus-
tomized desktop control.
A standalone application would offer a more flexible
and independent solution. However this is not the
aim of this project because resource and time wise
it is not seen efficient to develop CAD software from
scratch instead of utilizing an existing one. Therefore
the aim is to develop a customized desktop control
application in the first phase, then to develop a plug-
in for existing CAD software in the second phase.
In this sense, the project utilizes the existing tech-
nology; and its innovative aspect is to build up new
methods for using the technology and new relations
between the tools.
CAD Software
The project is developed on SketchUp Make 3D mod-
elling software. The reason for choosing SketchUp
Make is that it is for free, is easy to learn and use and
it supports Ruby application programming interface
(API) which is fairly more user friendly when com-
pared to other available interfaces. Moreover, mod-
elling in SketchUp Make is mostly intuitive, which is
a feature that makes it a proper environment in rela-
tion with the project aims.
Fabrication - Volume 1 - eCAADe 32 | 403

APPLICATION FEATURES
Several features are gradually being included in the
application for the integration of conception and
manufacturing in design. Eventually it aims at pro-
viding a toolset to be utilized in all phases of digital
design cycle. The features are as follows:
• Imaginary Tactile: Provides 3D digital mod-
elling by using hands.
• Simultaneous Physical Modelling and Digi-
tizing: Provides physical modelling and simul-
taneous digitizing by using hands and certain
tools.
• Simultaneous Design and Making: Provides
link between the digitized model and robotic
manufacturing.
• Tool Variety in Motion Scanning: Provides a va-
riety of tools to be used with the controller.
In this project it is aimed to develop only the first two
features. The other features are going to be devel-
oped as separate projects. The outcomes of the first
feature and the context and aims of the second fea-
ture are presented in this paper.
404 | eCAADe 32 - Fabrication - Volume 1
Imaginary Tactile
This feature allows the user to build a digital 3D Figure 1
model and modify it by using his/her hands, where Imaginary Tactile
both hands can operate simultaneously (Figure 1).
There already exist a couple of applications which
integrate the controller with CAD software, such as
AutoCAD and Leap Motion Integration, Ossewa Solid-
Works Plug-in, Leap Motion for Grasshopper and Leap
Motion Plug-in for Autodesk Maya. These applications
utilize the controller analogues to a mouse. Also,
there are a couple of applications which experiment
utilizing the controller on SketchUp. They utilize
a third party application such as Touchless, AirInput
or GameWAVE to interact with SketchUp by provid-
ing an alternative to a mouse. These applications
are for controlling the desktop therefore can inter-
act with any software with varying degrees of func-
tionalities. Likewise, the Imaginary Tactile feature al-
lows the user interact with SketchUp through a third
party desktop control application which gets input
from the controller (Figure 2).
The most important challenge for this feature is
to customize the hand gestures and define the rela-
tions between the hand gestures and the software
functions in the best possible way in order to provide
a seamless workflow. In other words, the existing
ways of desktop interaction with the controller need
to be tailored for 3D modelling in SketchUp. For this
purpose, the GameWAVE application is chosen to be
modified as it allows easier and more efficient config-
uration. Eventually, two different configurations with
different modes of interaction are developed:
• Imaginary Tactile with 2 Modes
• Imaginary Tactile with 3 Modes
Imaginary tactile with 2 modes. This configuration
mediates the mouse movements and clicks, as well
as the Pan, Orbit, Zoom, Undo and Select functions of
the software by using the controller. The idea is to
let the user reach any function of the software in a
mouse-like fashion, while letting him/her run certain
Camera and the most used Tools and Edit functions as
Figure 2
Desktop Control
Application

Table 1
Gestures and
functions for the
first two modes of
the Imaginary
Tactile
configuration.

Table 2
Gestures and
functions for the
third mode of the
Imaginary Tactile
configuration.

Fabrication - Volume 1 - eCAADe 32 | 405

 well as the Escape and Switch Mode functions by hand
gestures. For this purpose, the assigned functions are
categorized as stable (Trigger Mode) or active (Steer-
ing Mode) functions and they are associated with cer-
tain gestures of the right or left hand, where the right
hand mediates the active functions and the left hand
mediates the stable functions (Table 1). It is impor-
tant that the selected gestures should be intuitively
sympathetic with the assigned function.
This configuration allows a more organic inter-
action with the digital environment comparing with
mouse use, because the use of the hands in 3D space
provides a more efficient perception of the form.
However, the transparency between the designer
and the object is not yet in the desired level. Because,
the designerly logic behind the held operations is not
much different than what the common practices in-
clude. Therefore the only progress achieved with this
configuration is the gained ability of using the hands
in a 3D fashion, especially for model viewing func-
tions, and the fun and excitement brought with this
new experience.
Imaginary tactile with 3 modes. This configuration
is achieved by enriching the previously used two
modes with the left hand trigger mode. The idea be-
hind providing one more mode is to allow the user
to access certain functions of the software by using
hand gestures instead of pointing them on the tool-
bar. This new mode provides access to the Line, Rect-
angle, Circle and Arc functions of the Draw menu, and
the Push/Pull, Move and Rotate functions of the Tools
menu by using hand gestures. Similar to the inter-
action with 2 modes, the categorization of the soft-
ware functions and their association with the hand
gestures through an intuitive manner are essential
within the configuration (Table 2).
Interaction with 3 modes provides a more or-
ganic relation between design thinking and the de-
sign object comparing with the use with 2 modes.
Because the user does not need to point the tools on
the toolbar for the most commonly used functions of
the software, but he/she activates them by perform-
ing hand gestures. Therefore this configuration pro-
406 | eCAADe 32 - Fabrication - Volume 1
vides an experience in which the designer feels like
he/she uses not only his/her hands in the 3D space,
but also certain tools even though the feeling is imag-
inary.
On the other hand, the functions that are avail-
able through hand gestures are limited because the
GameWAVE offers a limited number of gestures to be
customized. So that the user still needs to point the
tools on the toolbars or in menus to reach many func-
tions of the software. Also, the hand gestures pro-
vided for customization by GameWAVE are all pre-
defined. Therefore a perfect intuitive relation be-
tween the sought function and the hand gesture can-
not always be obtained. Because of these reasons,
the Imaginary Tactile feature can prove the concept
of Design by Making only up to a certain limit. As a
concept, it is able to prove that; a more organic re-
lationship between the designer and the digital de-
sign environment can be achieved with the existing
technology, the use of hands in 3D space provides
a more transparent media between design thinking
and the design object, the bodily actions provide a
more direct interaction with the object during the de-
sign process, and such an enhanced interaction with
the digital design environment is fun and exciting.
The next phases of the project focus on the
missing aspects which are necessary to develop a
stronger proof of the concept of Design by Making.
Figure 3
Simultaneous
Physical Modelling
and Digitizing
Simultaneous Physical Modelling and Digi-
tizing
The purpose of this feature is to eliminate the dis-
tance between the physical design object and the
digital one (Figure 3). The common practices of
digital design and manufacturing involve sequential
steps such as the generation of the digital model
first, then its physical manufacturing. In the cases
which require further digital processing of the physi-
cal model, a digitizing process, which usually is oper-
ated by 3D object scanning, is executed (Figure 4).
This feature is developed in order to achieve
a process in which the production of the physical
model and its digitizing is simultaneous. Instead of
scanning the completed physical object, its produc-
tion process is scanned by tracking the operations of
the working hands and tools. Thus, the physical and
digital phases of the design object are produced si-
multaneously, and the concept of "reverse engineer-
ing" becomes an internal component of the design
process (Figure 5).
This feature requires the tracking of the tools
as well as hands used in the physical production as
hands are not always enough to work on a physical
object. Currently, the controller can recognize and
track long, thin and straight objects such as a pen-
cil. In order to be able to involve the several types of
objects used in physical manufacturing, more devel-
oped motion detection devices are necessary. This
is going to be the focus of the later phases of this
project, whereas the current aim is to utilize the ex-
isting capabilities of the device for this purpose.
The idea of using a 3D printing pen or a glue gun
is thought to be applicable for the current phase, as
it is possible to build 3D physical objects with these
tools and to track them using the controller. 3Doo-
dler Pen is evaluated as the best tool to be tested, as
it is proved to be stable and precise enough to build
physical objects and its form is similar to a pen, which
makes it easier to be recognized by the controller.
3Doodler Pen is a 3D printing pen that melts and
then cools plastic thread while the pen is moved in
3D space. It uses ABS or PLA plastics like the common
3D printers. One can see it as a freehand and manual
3D printer, as it provides additive manufacturing and
uses similar types of materials like 3D printers. How-
ever, unlike a 3D printer, the production of the object
with this method is not in a layer by layer fashion, but
more visual and depictive. There is no digital data
used for physical production. On the contrary, with
this feature of the project, this production method is
used as a mode in which the intuitive and depictive
behaviours of the designer are transformed into dig-
ital data in order to produce the digital phase of the
design object.
It is possible to provide the feature of simultane-
ous physical modelling and digitizing by two differ-
ent approaches. The first approach would be using a
desktop control application like in the previous fea-
ture. The GameWAVE application does not track ob-
jects; therefore it is necessary to develop a new ap-
plication for desktop control which can track hands
and the 3D printing pen simultaneously. The second
approach would be providing the same functionality
by a plug-in which specifically works with SketchUp.
Currently the application performs as a separate cus-
tomized desktop control application.
The most important aspect of this feature is that
the controller tracks the movements of the pen while
building the physical model and this input is used
to produce the digital model simultaneously. There-
fore the procedure of the digital production should
be convenient with the way the pen is used. In other
words, the logic behind the production of the geom-
etry should be identical for both the physical and the
digital modes. Considering the use of the pen, this
convenience can be achieved by the Line and Free-
hand functions of the software. Both geometric el-
ements are defined in the same way, by tracing of
the pen tip for the physical model, and by tracing
of the cursor position for the digital model. There-
fore the controller tracks the movements of the pen
in the physical 3D space and accordingly moves the
cursor position in the digital 3D space. However, dur-
ing physical modelling, the pen is moved not only to
build the geometry but also to reposition. Therefore
Fabrication - Volume 1 - eCAADe 32 | 407
 Figure 4
The workflow in the
common practices
of digital design

Figure 5
The complete
digital design cycle

408 | eCAADe 32 - Fabrication - Volume 1

it is necessary to activate and deactivate the func-
tions during the pen movements in order to avoid
building excessive geometric elements in the digital
model.
The application is developed in JavaScript using
the Leap Motion SDK. Currently, only two modes are
developed in order to keep it as simple as possible
until the tracking of the pen reaches a robust level.
The modes currently available are;
• Pen mode: controls mouse movements to trace
the geometry for the digital model,
• Hand mode; controls Line, Freehand and
mouse left-click by using 3 different gestures.
This feature is not yet robust enough because of a
couple of problems which are gradually being solved.
The main problem is that the size and shape of the
3D printing pen is slightly out of the limits of what
the controller can track. Until the controller is de-
veloped to be able to track the 3D pen or another
3D printing pen which has a size and shape within
the limits is possible to be used, the application algo-
rithm performs the task by tracking a certain point on
the hand which holds the pen instead of the pen it-
self. The other problem is that the controller's sensors
are directed upwards when it is in its standard op-
erating position and it can most effectively track the
hands and tools within a limited field of view which
is above the device. In order to keep the controller in
its standard position and operate the physical mod-
elling within the field of view, a glass surface is put
above the controller and used as the base for the
physical model.
The firsts tests executed using this application
provides good enough results to see the direct rela-
tion between the physical model and the digital one.
However, currently models are not entirely identical.
This is mainly because of the sensitivity difference be-
tween the digital and physical environments. This re-
sults excessive geometric elements to appear in the
digital model and will be solved by developing the
application algorithm. Also, certain movements of
the hand are tracked on wrong axes because of the
axis recognition algorithm of the software. The best
solution to avoid this problem is to develop the ap-
plication as an embedded plug-in in further steps. So
that it can communicate with the software algorithm
in the best possible way.
Yet, the application is currently good enough to
prove that the idea of Design by Making is applica-
ble and such a direct relation between physical mod-
elling and digital modelling is promising and can be
developed further.
CONCLUSIONS
The application which is being developed in this
project provides a new design experience where con-
ception and manufacturing are integrated and recip-
rocal feedbacks between different phases of the de-
sign object are performed continuously. It reintro-
duces the notion of craft into design process. It al-
lows the designer have a sense of tactile experience
where design development is performed on the de-
sign object itself instead of its representations. It pro-
vides a transparent link between different phases of
the design object and different types of media. It can
also be seen as a hybrid facility for sketching, where
the outcomes of sketching are both physical, digital
and three dimensional, as it is a natural way for the
designer to explain and understand complex ideas
and to perform visual and spatial reasoning as de-
fined by Do (Do 2002).
One can also argue that these features will pro-
vide an enhanced learning environment in architec-
tural education. In the common practices of digital
design, design development is performed distinctly
in digital environment which is treated as a represen-
tational media. This results the lack of fundamen-
tal aspects of architectural design such as material-
ity and scale in design thinking, particularly from the
students' point of view. This application proves that
it is possible to recall these aspects by literally putting
the hands and tools into the digital environment.
Moreover, such a practice of design will en-
hance the possibilities of communication between
Fabrication - Volume 1 - eCAADe 32 | 409
 Figure 6
Distant
collaboration
through online
shared digital
model by Design by
Making

the stakeholders of the design studio. The com- Kolarevic, B and Klinger, K (eds) 2008, Manufacturing Ma-
mon practice is based on a process where the stu- terial Effects: Rethinking Design and Making in Archi-
dent develops a solution to the given design prob- tecture, Routledge
Malafouris, L 2013, How Things Shape the Mind: A Theory
lem, presents it through representation tools, and
of Material Engagement, MIT Press
receives verbal comments and evaluations upon it. Pallasmaa, J 1996, The Eyes of the Skin: Architecture and
Learning is by the student's own practice, observing the Senses, John Wiley & Sons
the other students' practices, and by receiving verbal Schön, DA 1984, 'The Architectural Studio as an Exem-
critics from others. With the help of this application, plar of Education for Reflection-in-Action', Journal of
it will be possible to work collaboratively on the de- Architectural Education, 38 (1), pp. 2-9
Sennett, R 2009, The Craftsman, Yale University Press
sign object itself. As the application can track more
than two hands and tools simultaneously, not only
more students, but also teachers can actively partic-
ipate in the design development. The collaboration
can even be distant by using an online shared digital
model (Figure 6). Therefore an efficient environment
of making can be achieved which will profoundly in-
troduce the concept of Maker Culture into architec-
tural education.

REFERENCES
Chard, N 2005, 'Drawing Instruments', Architectural De-
sign, 75 (4), pp. 22-29
Dearden, A 2006, 'Design As Conversation With Digital
Materials', Design Studies, 27 (3), pp. 399-421
Do, EY-L 2002, 'Drawing Marks, Acts, and Reacts, toward
a Computational Sketching Interface for Architec-
tural Design', Artificial Intelligence for Engineering De-
sign, Analysis and Manufacturing, 16, pp. 149-171

410 | eCAADe 32 - Fabrication - Volume 1

Physical input-driven offline robotic simulation through a
feedback loop process
Odysseas Kontovourkis1 , George Tryfonos2
1,2
Department of Architecture, Faculty of Engineering, University of Cyprus, Cyprus
1
http://ucy.ac.cy/dir/en/component/comprofiler/userprofile/odykont
1,2
{kontovourkis.odysseas|at07tg2}@ucy.ac.cy

This ongoing research describes a feedback loop procedure where physical inputs
are used as the medium for offline robotic simulation. The purpose is to
investigate the ability of industrial robots that are currently used in
manufacturing processes to work in a flexible and productive manner whilst
providing a continuous feedback loop between physical inputs and fabrication
artifacts. In order to achieve this, a methodology is developed that involves the
use of data acquisition devices to enable the transference of information from the
physical to the digital environment and then to use this data as real-time
parameters to control the robot's behaviour during fabrication. The aim is to
achieve active involvement of robots in the manufacturing process to address
complex construction issues and to ensure accuracy, a reduction in
manufacturing defects and flexibility in the materials used. This investigation is
accompanied by relevant experiments to exemplify the potential of control
mechanisms to be used in prototyping case studies.

Keywords: Physical input, Robotic simulation, Feedback loop, Manufacturing

process, Material control

INTRODUCTION loop process is currently coming to the fore.

The continued development of strategies in digi-
tal fabrication that are used in the construction of
complex morphologies and the need for accuracy
and precision of the physical results obtained intro-
duces the potential for new directions in manufac-
turing processes that move beyond the simple linear
workflow between design and construction. Hence
the need for effective control of physical prototypes
during the construction process by computational
mechanisms that allows the continuous evaluation of
results according to physical changes via a feedback
As described in Cybernetics (Wiener 1965), the
term feedback refers to a process that occurs inter-
nally within a system where a part of the output is fed
back to the input. Similarly, in architecture character-
istics can occur during the construction process that
could be fed back to the digital design to allow con-
tinuous control of the physical end product. Apart
from the way morphological objectives are gained
through robotic manufacturing techniques, physical
characteristics, scale and materials play an important
and influential role during this process and are of-

Fabrication - Volume 1 - eCAADe 32 | 411

 ten combined with other simulation strategies (Brau-
mann and Brell-Cokcan 2012a).
In an initial interpretation of the above pro-
cess, feedback logic can be calculated in relation to
the physical characteristics and simulation of ma-
terials (Gramazio and Kohler 2008) resulting in a
three-dimensional digital design, tool-path defini-
tion and execution by the robotic machine. Using
this method, the construction of the ICD/ITKE Re-
search Pavilion 2010 (Fleischmann et al 2012) ex-
emplifies the development of a construction model
where the behaviour of timber strips are simulated
and then the robot is used to cut and fabricate the
strips and their joints. Iterative feedback loops be-
tween design and manufacture can be created by de-
veloping parametric software and plug-ins that are
capable of simulating and controlling the behaviour
of robotic machines and offer real-time interaction
with the physical manufacturing process (Braumann
and Brell-Cokcan 2012b). This is strengthened by
their ability to integrate design and simulation pro-
cesses combined with additional input data from the
physical world into a single digital environment. An-
alytically, physical input from sensors is incorporated
into the program, processed and then robotic mo-
tion is simulated. In parallel with the digital simula-
tion of the physical form, the actual robotic machine
has the ability to perform the manufacturing process
without any additional changes to the algorithm. The
example of the Aggregate Structure (Dierichs et al
2012) follows similar logic. In this case, a modular sys-
tem that is continuously controlled by a robotic ma-
chine is developed which involves the encoding of
the physical information derived from the emptying
412 | eCAADe 32 - Fabrication - Volume 1
of aggregate units and the controlling of the place-
ment of the aggregate in a physical and a digital en-
vironment.
Within this frame, digital simulators can control
the feedback behaviour of robotic machines and,
therefore, can allow experimental investigation into
the evolution of design proposals during construc-
tion development. Such robotic control could offer a
number of additional advantages to the current man-
ufacturing process including real time, open-ended
construction that adapts to the designer's needs and
to any unforeseen changes that may occur in the
physical environment. Future applications could be
designed for developing structures with the ability of
providing feedback to allow refinement and readjust-
ment of their morphology to suit any changes in the
structure's surrounding environment. Another ap-
plication could be for structures containing flexible
material to continually report on its behaviour. The
following chapter describes our proposed methodol-
ogy.
INVESTIGATED METHODOLOGY
The logic of feedback loops in robotically-controlled
construction can be interpreted in different ways,
including feedback focusing on the design process
and materials or construction based on the use of a
robotic machine. This paper suggests the develop-
ment of a cyclically iterative feedback loop process
between four key aspects. These are: physical in-
puts, computer control, robotic machines and digi-
tal prototypes attempting to redefine the morphol-
ogy of physical structures according to specific en-
vironmental influences during the construction pro-
Figure 1
Flowchart of the
feedback loop
process
 cess. Consequently, readjustment is achieved in ac-
cordance with the changes of the physical proto-
type and the designer's intervention. The efficient
use of articulated robotic machines depends on re-
peated readjustment of material for optimising com-
plex physical prototypes (Figure 1).
The behaviour of the physical structure is
recorded by sensors and this information is then
translated into data that can be input to the pro-
gram. This data then activates an appropriate reac-
tion of the robotic machine. Through this sequence
of steps, the robot attempts to optimise the con-
struction process and produce prototypes that are
influenced by any changes that may have occurred
in the physical environment resulting in reconfigura-
tion of their initial appearance.
Analytically, using Microsoft Kinect (a depth map
camera) or webcam devices, the initial physical form
is recorded to a digital environment and, with the use
of Firefly components [1] (a plug-in for Grasshopper
[2]) (Payne 2013), it is converted into a list of points
(xyz). Using the initial list and incorporating any de-
cisions that may be made by the designer, the digital
geometry is redefined as a new list of points. The new
list is compared with the original list and new tool-
paths for controlling the articulated robotic machine
Figure 2
Flowchart of the
proposed control
methodology
are created.
The changes that occur in the physical proto-
type are continuously recorded to capture the phys-
ical form's deformation and then to create a new list
of points. Subsequently, the two lists are compared
in relation to their distance and position. The pro-
posed robotic control methodology is responsible
for redefining the position of points and the robotic
tool-paths in the digital environment. The process of
readjustment and control of the robotic machine is
achieved by the following basic rules (Figure 2):
1. If the number of points in the new list (List B)
exceeds the number of points in the initial list
(List A) then the extra points are removed via
generated tool-paths.
2. If the number of points in the new list (List B)
is less than the number of points in the initial
list (List A) then the difference in the number
of points is added to the system via generated
tool-paths.
3. If the number of points in both lists (Lists A
and B) is equal, tool-paths are created be-
tween the points in the new list (List B) and
points in the initial list (List A) to redefine the
physical geometry.
Fabrication - Volume 1 - eCAADe 32 | 413
 4. If the distance between the points of the new
list (List B) is larger than expected then the ex-
tra points are removed.
As the robotic machine is responsible for the removal,
addition or reconfiguration of points via generated
tool-paths, the continuous flow of information from
the physical to the digital environment results in the
dynamic redefinition of geometry in space.
EXPERIMENTAL CASE STUDY A:FEEDBACK
CONTROL OF THE POSITIONS OF STRUC-
TURAL UNITS
In order to investigate the manufacturing process
based on the idea of a feedback loop between
the physical input, the digital environment and the
robotic machine control, experiments based on the
proposed methodology are developed using offline
robotic simulation.
Since this study is concentrated on the use of
physical inputs as the driving force to accelerate
robotic behaviour for material control, the initial in-
vestigation focuses on the mechanisms for physi-
cal information recording and computer processing.
This is achieved by simulating the behavior of the
robotic machine and, hence, the readjustment of the
digital form. The experimental results aim to evalu-
ate the proposed control methodology of a feedback
loop based on the four basic rules mentioned above.
Physical data and digital processing
The first study focuses on the identification and trans-
lation of physical information into digital input data
as well as on the control and reaction of the robotic
machine according to any changes that have oc-
curred in the physical environment. By using a Kinect
camera, the movement of black acrylic rectangular
units on a white background is observed. Their two-
dimensional centre position is recorded into a digi-
tal environment by using Grasshopper/Rhino [3] soft-
ware and the 'edge detection' component in Firefly.
Then for the motion control of the robotic machine,
tool-paths in the form of a 'pick and place' procedure
414 | eCAADe 32 - Fabrication - Volume 1
are created, which aim to restore the initial state of
the physical environment (Figure 3).
Figure 3
Robotic response:
physical
interference,
generated
tool-paths and
robotic behaviour
simulation
Analytically, the initial physical geometry is
changed due to external factors that cause the in-
creasing, decreasing and repositioning of the acrylic
units. In a continuous procedure, the new positions
of the units' centre points are recorded and com-
pared with the initial list (List A). Thus, by using the
first three rules (1st, 2nd and 3rd rules) described
in the previous chapter, the methodology is respon-
sible through simulation to restore the new list of
points, repositioning the units in their correct posi-
tion.
Digital control of prototype through robotic
simulation
The first study described in the previous paragraphs
recognises the ability of algorithms to respond in real
time to any changes that may have occurred in the
physical environment to continuously restore the po-
sition of the units. This second study aims to clar-
ify the relationship between digital control of the
geometry and the construction of a prototype us-
ing a robotic machine's behaviour simulation. Thus,
in order to simulate the construction of a three-
dimensional geometry, the robotic simulation pro-
gram HAL [4] (a plug-in for Grasshopper) is used to
generate a real time robotic motion (Schwartz 2012).
By using a parallel gripper as the end-effector, the
Figure 4
Process of
prototype
construction and
units repositioning

robotic machine assembles and adjusts acrylic struc-
tural units (dimensions 4cm x 4cm x 1cm) within the
overall structure and these are influenced by external
changes that have occurred in the physical environ-
ment.
Subsequently, the desired geometry is defined
by a curve system that generates a surface. Through
the separation of the existing surface and the intro-
duction of digital points in Rhino by the designer and
taking into account any static factors, the existing
surface is defined. This results in the creation of initial
units handled by the robot.
During the construction process of the units,
the user can redesign the surface by deforming the
curves or by influencing the manufacturing process
via the addition of new units. Thus, the proposed
control methodology uses the 1st, 2nd and 3rd rules
to adjust the position of the units according to the
desired geometry. Because the changes of curves di-
rectly effect the number and position of the units, the
robotic control behaviour includes local movement
of existing units within a desired position, the addi-
tion of new units or the removal of unnecessary ones
(Figure 4).
In the first study, the ability of the robotic ma-
chine to react in real time to control the geometry
of design based on decisions effecting the manu-
facturing process is observed. The relationship that
is developed between the designer and the robotic
procedure can be considered as dynamic when the
robotic machine's actions are simultaneously based
on the decisions taken by the designer. While in pre-
vious cases the complexity, precision and the need
for structural stability justifies the use of a robotic
machine, the relationship that is developed between
sensor devices (Kinect or webcam) and materials ap-
peared to be supportive of the construction process.
This research seeks for a dynamic relationship
between digital results, manufacturing processes
and physical prototypes within a continuous feed-
back loop procedure. In this case, each modification
of a prototype's physical condition made by the de-
signer or the robotic machine requires physical data
recording.
EXPERIMENTAL CASE STUDY B: FEEDBACK
CONTROL OF ELASTIC MATERIAL SYS-
TEMS
Based on the results derived from previous experi-
ments and by using similar control methodology, this
third study aims to clarify the relationship between
the material's composition and the construction ca-
pacity of complex forms.
Taking into consideration the four key aspects
mentioned in the 'Investigated Methodology' chap-
ter and in order to avoid any failures of the robotic
machine during the fabrication process, this paper
continues by introducing a comprehensive study
dealing with the offline robotic simulation process
(Biggs and MacDonald 2003) of the proposed control
methodology.
Specifically, this experimental investigation uses

Fabrication - Volume 1 - eCAADe 32 | 415

 silicone elastic threads as the material. The current
study intends to define a manufacturing process that
can be considered as a form-finding procedure ap-
plied in the physical environment for creating tensile
mesh structures in real time. Through the additive
manufacturing process, the designer is able to con-
trol the increasing smoothness (levels of smoothness
are explained in each study) of the overall structure
(Pottmann et al 2007). In order to achieve this, the
control methodology is concentrated on the 2nd, 3rd
and 4th rules (see 'Investigated Methodology').
This experiment introduces an end-effector that
controls the local addition of silicone elastic threads
into the tensile mesh system. In order to achieve
this, the investigation is concentrated on the devel-
opment of a multifunctional end-effector tool (Brau-
mann and Brell-Cokcan 2012c) with material han-
dling and local welding abilities. All welds that are
created are recognised and continuously recorded
by the image recording device. The positions of the
welds are then translated into nodes within the over-
all tensile structure and are converted into input data
for the proposed algorithm (a detailed description
of end-effector functionality will be given in a future
work).
The increasing smoothness of the prototype is
416 | eCAADe 32 - Fabrication - Volume 1
examined through two main mesh typologies, i.e.
triangular mesh and square mesh (Pottmann et al
2007). The first typology aims for the overall mesh
structure development through triangulation and
the second typology through polygons. The gener-
ated mesh structures consist of nodes derived from
the physical world (in this case points are used as in-
put data) and handled in the Rhino digital environ-
ment. To record the refinement of the existing mesh
structure, the designer is able to define different lev-
els of smoothness that specify the repeatability of ad-
ditional material (Figure 5) with the use of Grasshop-
per (a plug-in for Rhino).
Triangle mesh prototype
The study of triangle mesh typology begins by in-
troducing three points that are increasing sequen-
tially with at least one common edge that creates
a number of initial triangulations in the Rhino envi-
ronment. Then for each triangulation a mesh con-
figuration is predefined, initially by dividing individ-
ual edges into eight points that are connected to
the opposite edges, creating lines that are sorted ac-
cording to their length. In this way, a list of seven
lines for each pair of edges is created that defines the
robotic tool-paths. In turn, the generated lines are
Figure 5
Comprehensive
flowchart of the
proposed
smoothness control
methodology
 divided into between two and eight points, depend-
ing on their position in the sorted list, creating points
of intersection. Those points are defined as the de-
Figure 6
sired nodes for the robotic welding process. The de-
Evolutionary
signer then specifies the level of smoothness and the
development of
robotic system can begin the fabrication procedure.
triangular mesh
After the initial mesh configuration is specified,
smoothness
the real time robotic control that involves material
procedure
behaviour simulation is then developed. Using Kan-
garoo (a plug-in for Grasshopper) (Piker 2013) and
applying the 'spring behaviour' (Fleischmann and
Ahlquist 2009; Kontovourkis et al 2013), the simula-
tion of silicone elastic threads forming the mesh ty-
pology is achieved by determining, in parallel, mate-
rial deformation and robotic behaviour.
Analytically, the length of the thread is controlled
during the construction process by first calculating
the distance between predetermined nodes under
'spring behaviour' control. This is reached using the
simple equation: Spring Length = 0.7 x Start Length.
Then, to add elastic threads that connect the nodes,
their length in the overall system under tension is
checked. If their simulated length is less than the
desired digital length then the specific thread con-
nections are excluded from the construction process
(usually this is observed in connections close to the
naked edges). This results in the formation of four-
sided polygonal patterns near the naked edges.
By taking into consideration the results derived
from the physics-based simulation, the robotic tool-
paths are defined as follows: sorted line 5 for all
edges resulting in smoothness level 1, sorted line 5
and 6 for all edges resulting in smoothness level 2
and sorted line 5, 6, and 7 for all edges resulting in
smoothness level 3.
In the case of online fabrication, the generated
physical nodes are recorded through the camera sen-
sor and their position in space is compared with the
results derived from the digital environment. In or-
der to find consistency between triangular units, the
distance between nodes in the physical output is cor-
related with the expected distance of the nodes. By
following the rules of a control algorithm, the fabrica-
tion procedure continues until the desired smooth-
ness of the mesh structure is achieved (Figure 6).
The process of smoothness is characterised as
continually additive, where the designer is able to
add new points to define the overall mesh system.
Through the camera sensor, these points can be
taken into account and used in other triangulated
mesh systems. The new triangulations effect the ex-
isting system, triggering a new smoothness proce-
dure that is continuously repeated until the required
smoothness is achieved (Figure 7).
The outcome of the smoothness procedure can
be characterised as the result of a feedback loop be-
tween the material, the manufacturing process and
design decision making. In every step of the cycli-
cal iteration process, the robotic machine is effected
by the target configuration and desired smoothness.
By adding nodes and material, the smoothness of the
overall system is increased and, in parallel, the initial
geometry is redefined, optimising the dynamic be-
haviour of the elastic material.
Polygon mesh prototype
The study of quad mesh typology is based on the
evolutional development of the triangle mesh using
a similar process to find correlations and discuss re-
sults.
The quad mesh is defined by four points in the
Rhino digital environment. For the sequential devel-
opment of basic quadrilaterals, the designer defines
two additional points that are connected to the two
Fabrication - Volume 1 - eCAADe 32 | 417

points of the nearest naked edge of the existing sys-
tem. Each edge of the quadrangle is divided into
three points (start, midpoint and end). The middle of
each edge is successively joined with each other for
the creation of a new quadrilateral. The new edges
define the tool-path that directs the action of the
end-effector. Joint marks are defined at the start and
end point of each tool-path.
In this case, the simulation of the robotic fabri-
cation procedure is directly combined with Kanga-
roo simulation observing results in real time. Elas-
tic material simulation is controlled by tension be-
haviour using 'spring force' with the length calcu-
lated by using the simple equation: Spring Length
= 0.7 x Start Length. The system's level of smooth-
ness is determined by the designer by repeating the
418 | eCAADe 32 - Fabrication - Volume 1
Figure 7
Results of the
triangular mesh
typology
process with each addition of new material. Each it-
erated step is defined as the action of the algorithm
to measure and react to changes in the existing sys-
tem and to create new quadrilaterals to update the
existing ones. In this case, the maximum repetition
degree to be controlled in the Grasshopper environ-
ment is limited by the scale of the prototype (the ini-
tial distance between nodes) compared to the scale
of the end-effector tool (this also influences the size
Figure 8
of any added nodes). The iterative results of simula-
Evolutionary
tion are shown in Figure 8.
development of
The following observations are made using the gen-
polygonal mesh
erated results:
smoothness
• The triangles within the quadrilaterals are the procedure using
result of the intervention of the algorithm. four edges
This is due to the addition of the material
that allows deformation of triangulations into
quadrilaterals. Figure 9
Alternative
• The fabrication procedure creates large poly- polygonal mesh
gons in the middle of the geometry that could smoothness
cause instability of the three-dimensional sys- procedure with
tem. minimum four
edges
As a result, the quadrilateral forming procedure is al-
tered in a process for a multifaceted formation. The
algorithmic control can be applied to create more
than four nodes. For the new polygon formation, the
robotic machine adds the new material to the sys-
tem and defines the new nodes. The deformation of
the system is caused by the new material influenc-
ing the division of subordinate threads at their mid-
point position. Using the camera sensor, the new
nodes and the repositioning of the existing nodes are
Figure 10
Results of the
polygonal mesh
typology

Figure 11 recorded. By transferring the information into the

Installation of digital environment, the geometry is redefined and
robotic machine the algorithm recognises the new polygons. Then,
and camera devices from the midpoint of the edge of each polygon a new
for the fabrication polygon is formed internally and its points are recog-
process nised as new nodes. The lines of the polygons are
converted into robotic tool-paths, which are used by
the robotic machine to add the elastic material. The
process between the physical manufacturing and the
digital feedback loop is repeated until the required
level of smoothness is achieved (Figure 9).
The feedback loop procedure between the phys-
ical and the digital information and the design-
The results obtained from the production of the
er/robotic machine interaction leads to an evolution-
triangle mesh are adapted to the characteristics of
ary process of adding new material to an existing
the selected material. Depending on the basic tri-
mesh system. The main purpose of this procedure is
angulations, the new mesh system can generate tri-
to investigate and adjust the levels of smoothness of
angular or square patterns. The calculation of the
structural systems. By using an iterative robotically
added material is predetermined by the initial nodes
driven procedure, the existing physical prototype is
and is redefined in the physical environment. The
strengthened by the addition of new material and
equilibrium of the system based on the tension forces
adapting the intervention of the robotic machine.
applied due to the elasticity of the elements causes
The redefinition of the nodes leads to a dynamic de-
the continuous deformation of the structure until it
formation of form and these results are influenced by
reaches the desired shape.
the relationship between designer, robotic machine,
In the case of a polygon mesh, the addition of
digital control and material behaviour (Figure 10).
material to the basic quadrilateral shapes happens
Discussion on triangle and polygon mesh gradually in a bidirectional manner combined with
prototype typologies the physical environment that transforms the exist-
The experimental simulation of the two mesh typolo- ing prototype into a new multifaceted product. In
gies based on the methodology and algorithmic pro- this case, the recording procedure of the camera sen-
cedure described below leads to the versatility of ma- sor is an indispensable part of the evolutionary de-
terial and the generation of different outcomes (Fig- velopment and smoothness of the prototype. The
ure 11). procedure of adding material to the mesh combined

Fabrication - Volume 1 - eCAADe 32 | 419


with the application of tensile forces causes the re-
definition of nodes in the digital environment. The
reformation of digital geometry is the result of the
behaviour of the physical prototype in every level of
smoothness (Figure 12).
CONCLUSION
In this ongoing experimental study, an attempt to es-
tablish a cyclically iterative feedback loop process be-
tween design and robotic machine control driven by
physical input data is demonstrated. Through the ap-
plication of data acquisition devices and parametric
plug-ins, the assembly and redefinition of structural
elements and the addition of elastic material that cre-
ates tensile mesh structures in digital space is investi-
gated. By simulating the real-time dynamic relation-
ship between proposed designs, manufacturing pro-
cesses and the physical behaviour of added materi-
als, the capability of a robotic machine to adjust dig-
ital geometries in space precisely and to react to any
changes that may occur in the physical environment
is examined.
This paper argues that the application of feed-
back loop logic involving physical data collection and
robotic machine control might open up new direc-
tions in digital fabrication research. This technolog-
ical shift towards an adaptive construction process
that is influenced by design decisions and flexible
material behaviour in real time can allow new investi-
420 | eCAADe 32 - Fabrication - Volume 1
Figure 12
Generated elastic
mesh prototypes
with eight initial
points using the
proposed control
methodology: A.
Possible triangular
mesh result, and B.
Possible polygonal
mesh result
gations to respond to new manufacturing demands.
Potential applications of such technology could be
found in the fabrication procedures of complex struc-
tures that continuously readjust, redefine and refine
their morphology in the physical environment and al-
low structural accuracy and a reduction of manufac-
turing defects via the cyclical iterated feedback loop
logic.
Future development in regard to the proposed
control methodology aims to examine and investi-
gate further a number of case studies in order to re-
duce the amount of defects that occur during de-
sign and fabrication based on digital and physical
mechanisms. In addition, different aspects of the
proposed process including the functionality of the
end-effector, control of the robotic machine and the
elastic behavior of material will be investigated and
developed further. Finally, the application and test-
ing of proposed methodology in physical conditions
using an actual robotic machine is important for val-
idating the current hypothesis.
REFERENCES
Biggs, G and MacDonald, B 2003 'A survey of robot pro-
gramming systems', Proceedings of the Australasian
conference on robotics and automation, Brisbane
Braumann, J and Brell-Cokcan, S 2012a, 'Digital and
Physical Tools for Industrial Robots in Architecture:
Robotic Interaction and Interfacesrell-Cokcan', Inter-
national Journal of Architectural Computing, 10, pp.
541-554
Braumann, J and Brell-Cokcan, S 2012b 'Real-Time Robot
Simulation and Control for Architectural Design',
Digital Physicality - Proceedings of the 30th eCAADe
Conference, Czech Republic, pp. 479-486
Braumann, J and Brell-Cokcan, S 2012c 'Digital and Phys-
ical Computing for Industrial Robots in Architec-
ture: Interfacing Arduino with industrial robots', Be-
yond Codes and Pixels: Proceedings of the 17th Inter-
national Conference on Computer-Aided Architectural
Design Research in Asia, Nanjing, pp. 317-326
Dierichs, K, Schwinn, T and Menges, A 2012, 'Robotic
Pouring of Aggregate Structures', in Brell-Cokcan,
S and Braumann, J (eds) 2012, Robotic Fabrication
in Architecture, Art and Design, Springer Wien New
York, Austria, pp. 196-205
Fleischmann, M and Ahlquist, S 2011 'Cylindrical Mesh
Morphologies: Study of Computational Meshes
based on Parameters of Force, Material, and Space
for the Design of Tension-Active Structures', Compu-
tation: The New Realm of Architectural Design [27th
eCAADe Conference Proceedings], Ljubljana, pp. 39-
46
Fleischmann, M, Knippers, J, Lienhard, J, Menges, A and
Schleicher, S 2012, 'Material Behaviour: Embedding
Physical Properties in Computational Design Pro-
cesses', Architectural Design, 82, pp. 44-51
Gramazio, F and Kohler, M 2008, Digital Materiality in Ar-
chitecture, Lars Müller Publishers, Baden
Kontovourkis, O, Phocas, MC and Tryfonos, G 2013, 'Pro-
totyping of an Adaptive Structure based on Physi-
cal Conditions', International Journal of Architectural
Computing, 11, pp. 205-226
Payne, AO and Johnson, JK 2013, 'Firefly: Interactive Pro-
totypes for Architectural Design', Architectural De-
sign, 83, pp. 144-147
Piker, D 2013, 'Kangaroo: Form Finding with Computa-
tional Physics', Architectural Design, 83, pp. 136-137
Pottmann, H, Asperl, A, Hofer, M and Kilian, A 2007, Ar-
chitectural Geometry, Bentley Institute Press, Exton,
Pennsylvania USA
Schwartz, T 2012, 'HAL', in Brell-Cokcan, S and Braumann,
J (eds) 2012, Robotic Fabrication in Architecture, Art
and Design, Springer Wien New York, Austria, pp. 92-
101
Wiener, N 1965, Cybernetics, Second Edition: or the Control
and Communication in the Animal and the Machine,
The MIT Press
[1] http://fireflyexperiments.com/
[2] http://www.grasshopper3d.com
[3] http://www.rhino3d.com/
[4] http://hal.thibaultschwartz.com/

Fabrication - Volume 1 - eCAADe 32 | 421

422 | eCAADe 32 - Fabrication - Volume 1
Bio-inspired and fabrication-informed design strategies
for modular fibrous structures in architecture
Leyla Yunis1 , Ondřej Kyjánek2 , Moritz Dörstelmann3 , Marshall Prado4 , Tobias
Schwinn5 , Achim Menges6
1,2,3,4,5,6
Institute for Computational Design, University of Stuttgart, Germany
1,2,3,4,5,6
http://icd.uni-stuttgart.de
1,2
{leylis2007|ondrej.kyjanek}@gmail.com 3,4,5,6 {moritz.doerstellman|
marshall.prado|tobias.schwinn|achim.menges}@icd.uni-stuttgart.de

Research pavilions can serve as architectural scale demonstrations for the

materialization of experimental forms and structures. Pavilions seek to prove and
change methods of design and construction mechanisms in order to achieve
desires such as material efficiency, novel spatial qualities and performative
needs. The case of the ICD/ITKE Research Pavilion 2013-14 highlights the use of
fiber composites in order to achieve a core-less filament winding modular system
from bio-inspired lightweight structures through robotic fabrication. This paper
describes the multi-disciplinary design and construction process of this pavilion
that created a structure of out 36 unique components.

Keywords: bio-inspired, fiber composites, multi-disciplinary design, robotic

fabrication, modular system construction

INTRODUCTION from a multidisciplinary study on the elastic defor-

Biology-inspired projects often involve a team of re-
searchers and students who work through a collab-
orative design methodology. For example, the Silk
Pavilion by the Mediated Matter group at the MIT Me-
dia Lab involved a team of researchers and students
who worked on a holistic and sustainable design of
"non-woven fiber-based constructions". By studying
the material and fabrication methods of Bombyx-
mori Silkworm and the use of simulation tools, the
team constructed a large scale pavilion through a
species that naturally prints in 3-D and a digitally fab-
ricated framework (Oxman, 2011). Furthermore, the
ITKE at the University of Stuttgart also explored and
implemented the use of a biomimetic façade shading
system, Flectofin, whose characteristics were derived
mation of the Strelitzia reginae flower (Knippers and
Speck, 2012). Such projects stimulate the integration
of fields of study such as biology, architecture and en-
gineering in order to implement innovative designs
and novel construction processes.
Utilizing a collaborative design methodology to
facilitate a biologically informed product allows for
greater exactitude of all stages of the project from
design to product. The joint research studio be-
tween the University of Stuttgart and the University
of Tübingen which consisted of both researchers and
students from an architectural and biological back-
ground, working together to combine knowledge
and implement it into a full-scale demonstrator. A
parametric design tool was used to fuse the research,

Fabrication - Volume 1 - eCAADe 32 | 423

 fabrication, and progression of the work from various
fields of study.
By combining topics such as architecture, biol-
ogy and robotic fabrication, new levels of creativity
are developed with a scientific basis while adopting
a reliable robotic fabrication method. Due to the set
schedule of the project, the prototypes evolved from
hand-made models to a single-robot setup, and fi-
nally matured into a robust two-robot method to pro-
duce core-less filament modules. The abstraction of
natural morphologic principles allows for the oppor-
tunity to use robust solutions, which when adapted
to robotic fabrication strategies can create a precise
and performative system. This research leads to a 1:1
pavilion as a demonstrator for testing the tolerances,
material behavior and form of this system. (Figure 1)
424 | eCAADe 32 - Fabrication - Volume 1
CONTEXT
Biomimetic design strategies in architec-
ture
The term biomimetics was formulated in the 1960s
by American engineer Otto H. Schmitt by combin-
ing the Greek words bios (life) and mimesis (imitate).
It refers to a method of abstracting principles found
in biology to solve complex problems in technology
(Schmitt, 1969). With the increasing sophistication of
tools and methods we can now not only study pat-
terns and techniques in nature that were previously
out of scope but also garner further understanding
of the guiding natural principles, in order to abstract,
transfer, and successfully apply them to the built en-
vironment.
Biomimetics has different methodologies: "(a)
Bottom-up process of biomimetics (biology push).
Figure 1
Full-scale prototype
at the University of
Stüttgart (ICD/ITKE)
 (b) Top-down process of biomimetic research (tech-
nology pull)" (Knippers and Speck, 2012). This pa-
per describes a bottom-up process from which a bio-
logical role model investigation led to an innovative
use of robotic tools for filament winding. The use of
robotics in architecture is implemented by adopting
industrial machinery through custom software inter-
faces and personalizing end-effectors (Brell-Crokcan
and Braumann, 2012). This design process took
advantage of CAM (computer-aided manufacturing)
methods using pre-existing 6-axis robotic arms, and
profiting from the ease of use and variability that is
pertinent in both iterative and flexible CAD (com-
puter aided designs).
Fiber composites are an active area of research in
industrial or engineering arenas. They allow products
Figure 2 to become structurally efficient and light and are tra-
Hand-made models ditionally manufactured by applying malleable ma-
of the construction terial over molds. In natural systems, fibrous struc-
system tures can achieve a wide range of performative qual-
ities such as flexibility, a greater diversity of mate-
rial density through anisotropic material organiza-
tion, as well as differentiated structural reinforce-
ment without the need for molds. In the ICD/ITKE
Research Pavilion 2012, a core-less filament winding
system was developed where a robotic arm placed
the fiber around a framework that was removed af-
ter the prototype was cured and complete (Weigele,
2012). The ICD/ITKE Research Pavilion 2013-14 fur-
ther developed a bottom-up methodology utilizing
a reconfigurable and removable framework to create
core-less winding modules with a two-robotic setup
(Prado and Dörstelmann, 2014).
METHODS
Biomimetic investigation
The process of abstracting biomimetic principles
makes it possible to design architectural pavilions
whose role is to optimize material usage and weight,
provide new architectural qualities and structural
systems, and therefore understand natural systems
on a deeper functional and methodological level
(Knippers and Speck, 2012). The biomimetic inves-
tigation was searching for new performative mor-
phologies, as well as innovative fiber composites ma-
terial organization principles which could be devel-
oped into robotic fabrication strategies.
Four topics were sought as a basis for preliminary
biomimetic investigations: 1) fiber layouts, 2) fabri-
cation methods, 3) joint-systems, and 4) functional
morphology, all through conceptual biological role
models. Arthropod cuticles, known for their high-
performance fibrous structures, were researched by
biologist for potentials to adapt and perform in an ar-
chitectural realm. Simultaneously, there was a con-
tinuous effort to investigate fiber models that pur-
sued an understand of fiber behavior and their con-
struction framework. (Figure 2)
The elytra are protective hind-wings for beetles
can withstand bending and compression forces and
demonstrate high aerodynamic performance. (Van
de Kamp, 2010). This role model first became a focus
of investigation due to its lightweight capacity and
the structural height that serves as the protective
characteristic for beetles' flying wings. After further
investigation by the biologists, the elytra also pre-
sented several species morphological variations that
influenced the performative characteristics of the
pavilion design. Hence, influencing the concept of
the pavilion both in terms of its micro-fiber arrange-
ment and also regarding the geometric layout of the
fibrous structures on a global scale.
Fabrication - Volume 1 - eCAADe 32 | 425
 The biological investigation provided a notion
of the construction of the system of the elytra. The
fibers are embedded in a matrix of proteins, and
piled into a series of helicoidal fiber layers in order
to resist stresses in all directions (Lomakin, 2004).
These are hollow fiber bundles that connect upper
and lower surfaces act as a "bridge pier" in between
the wings (Figure 3). The trabeculae, or hollow fiber
bundle columns, are arranged throughout the ely-
tra in a non-uniform manner, while the remaining ar-
eas constitute blood cavities. In ground beetles, the
blood cavities are reduced and there is a higher den-
sity of fibers. Flying beetles, which tend to be more
lightweight, have a reduced number of fibers and
larger cavities (Van de Kamp, 2010). Therefore, the
investigation focused on eight different flying bee-
tles that showed density patterns on the trabeculae,
proportionate to the thickness of the elytra, and their
mechanical characteristics vis-a-vis the beetle's body.
Part of the project's investigation was to de-
velop a modularized system of fabrication. This
would allow for reuse of construction parts of vari-
able elements, and would create a pavilion that could
be easily transported. Furthermore, the fabrication
constraints would not permit a creation of an en-
tire beetle elytra structure, but rather abstracted
individual trabeculae, as components for a global
system. These components resembled hyperbolic
columns that were fabricable as coreless filament-
wound modules. Although biologists had presented
426 | eCAADe 32 - Fabrication - Volume 1
more interest in the micro-scale fiber arrangements,
the need for architects to analyze the elytra as a
whole presented an opportunity to study the ar-
rangement of the hyperbolic columns along the
wing in order to abstract the pattern in which the
fiber columns are located. The analysis of particu-
lar species also showed the cantilevering capacity of
the elytra as well as the multiple connection of hyper-
boloid columns. (photo) The matrix found in the fiber
layers translated into the resin epoxy-based mixture
used in composite systems.
Development of the fabrication logic
The building components for the double-layered sys-
tem required two set of frames made up of rim-
shaped polygons. Each side of the polygons was di-
vided with a regular spaced set of hooking points, or
control points, where the fiber was wrapped around,
creating a hyperboloid in an iterative way. This con-
stituted the fiber syntax. The frames could change
in size, symmetry, angle and position shift, as well as
in their winding logic. This provided the opportunity
to optimize the material usage by reinforcing areas
where more fiber was required to ensure fiber inter-
action. The components, or hyperboloids, each had
seven layers of fibers that included a glass fiber base
followed by a number of carbon fiber structural vec-
tors. (Figure 4) These were differentiated codes that
ensured the structural capacity of each element by
itself and transferred a cross-bracing system of rein-
Figure 3
Sections of the
elytra: A) Cassidia
Beetle , B)
Chrysomela
vigintipunctata
Beetle
Figure 4
Fiber layer system
of the hyperbolic
column and the
robtic setup
forcements along the whole global system when the
elements were connected.
Handmade models were created to test con-
struction parameters such as polygon shape, binding
tension, and fiber manipulation, in order to transfer
proper winding patterns and fiber interactions onto
a computational domain. The winding logic of each
component was transferred as CAD information by
creating a winding syntax polyline that was able to
easily store the order of the hooking points and at
the same time gave a rough visual control of the fiber
mandrel. This translated into a two-robot system
setup that guaranteed the precise and automated
fiber placement. It also meant simplifying the pro-
cess since the robots ensured the code-based wind-
ing logic and also provided the opportunity to iterate
or change the logics to create a stable hyperboloid
component.
Synthesis of biomimetic research and devel-
opment of suitable fabrication strategies
The production was based on a non-woven concept,
meaning that the layering of winding logics and their
binding tension was necessary. The density of the
fiber of each component varied according to its po-
sition in the global design, aiding in force transfer
among the various components. However, the fiber
syntax also integrated restrictions on a fabrication
platform, so all elements needed to be conceivable
for material manipulation and on a robotic domain.
The exploration of the different beetle species,
furnished enough information to transfer to the
global design. According to the abstracted prin-
ciples, the double layered system could have the
following characteristics: 1) hyperbolic column ar-
rangements with a higher density on the outside
rim, 2) a bifurcation of the components leading to
Fabrication - Volume 1 - eCAADe 32 | 427
 a 3) cantilever, 4) proportions of the thickness of
the components from a thicker base to a thinner
apex and 5) differentiated winding logics that would
transfer forces parallel on each individual surface
and through helicoidally wound elements connect-
ing bottom and top surfaces.
The fabrication was led by these preliminary
characteristics, but at the same time had its own re-
strictions: 1) winding logics that left enough space
for the frame to be removed, 2) size manipulation to
avoid collisions of the end-effector and the robotic
arm, 3) sufficient thickness for the elements to be
wound inside and 4) polygon shape and angle ranges
for the robots to construct their own frames and
wind comfortably. Winding codes were developed
through algorithms that made them applicable for all
the elements comprised in the system.
INTEGRATIVE DESIGN AND CONSTRUC-
TION PROCESS
Collaborative studio experience
The project was established as a vertical integration
in a design-to-build oriented studio. An interdisci-
plinary team of students and researchers provided
the opportunity to test and investigate on many as-
pects thoroughly, producing large-scale prototypes
that entailed a robust production with real structural
results. The studio workflow divided students into
eight different groups of research: 1) biology, 2) com-
ponent, 3) component system, 4) material system, 5)
structural system, 6) joints, 7) effector, 8) robotics.
The students took the conceptual design and prelimi-
nary investigation of the beetle elytra, and expand on
the topics assigned to them. Through this method-
ology, the hyperboloid component required enough
information to work structurally as a component on
its own, connect itself to another component, and be
part of a component system global design. More-
over, the material system team was in charge of ex-
amining the fiber spool setup and epoxy resin base
properties, leading to a robotic setup which required
a personalized end-tool or effector in order to wind
the differentiated building elements.
428 | eCAADe 32 - Fabrication - Volume 1
Group work was separated in order to grasp all
the factors regarding the elytra and plausible abstrac-
tions, both individually and as a loop-feedback sys-
tem for the pavilion construction. The research con-
sisted of generic test models to inspect the software
or tools used by each particular group, which would
mature into descriptive characteristics able to feed
a production loop. Furthermore, the limitations of
time and availability of tools made some steps more
dynamic or static, leading to a faster progression of
some characteristics that started to create a number
of rulesets. For example, before the arrival of the two
robots for the setup, the fiber tests and code genera-
tion would be first based on a single robotic system.
The wooden frame used as an effector showed struc-
tural weakness and therefore led to an empirical ap-
preciation of the capacity that the effector required in
order to diminish the deviation of the wound proto-
type to its minimum. The effector group then estab-
lished a highly structural yet flexible system, whose
frame assembly had geometrical restrictions but en-
sured reconfigurability for all elements in the produc-
tion. (Figure 5)
Figure 5
Project group
division and
collaboration
The collaborative group work was organized in
order to represent, test through prototypes, and set
constraints that would guide the other aspects of the
project. Working on the same virtual platform facili-
tated the possibility to make any adaptations to the
design and fabrication strategies. Every aspect was
of equal importance, meaning that the construction
 system evolved was perfected in every detail towards new design studio semester. It was a good oppor-
the culmination of the project. tunity for students who developed the process to
pass on the knowledge to the new students and help
Fabrication setup and organization them familiarize with the material and techniques as
The fabrication process consisted of a 12-axis robot well as with the robotic fabrication itself.
setup. This was composed of two synchronized 6-
Figure 6
axis industrial arms KUKA KR210 R3100. Each robot
Time-flow diagram
was equipped with custom made steel space frames,
of the fabrication
on which two outline rims were mounted made out
loop explaining
of reusable steel bars that defined the shape of the
processes
component as explained above. The bars were con-
overlaping in the
nected with 5 axis cut plywood guides that were
production of the
unique for each component. The fiber source, where
components.
the fiber was dipped into an epoxy resin before it was
hooked around control points, was attached to a con-
crete slab between the two robots.
The construction of each component was di-
vided into four stages: 1) end-effector prefabrication,
2) end-effector assembly, 3) core-less filament wind-
ing and 4) component survey. Every stage in the pro- CAM process development
cess depended on the previous one, creating a closed Robots can be usually programmed in two ways: 1)
production loop. In line with the concept of proto- an online/automatic method, in which the robot is re-
typing digital fabrication processes contrast to proto- ceiving the information indirectly and the "program-
typing products, the fabrication method could be im- mer" has only limted control over the program itself
proved with every new component produced, based (e.g. programing by demonstration or inductive pro-
on the feedback from the previous one. The compo- gramming), 2) an offline/manual methods, in which
nent survey also allowed the opportunity to measure the behavior information for the robot is generated
the real result of the prototype, making the research offline and the robot is provided with precise com-
complete by having concrete data of what was be- mands in specific programming language (e.g. the
ing produced. The measurability was an aspect that KUKA Robot Language - KRL). (Biggs and MacDonald,
arose due to the facilitation of working with robots, 2003). In this project an offline method was applied
emphasizing the importance of combining robotics due to the parametric nature of the design and thus
and architecture during a learning architecture expe- the possibility to take advantage of a CAM solution.
rience. For each step of the production a new custom-
The time schedule of the production loop is ex- made algorithm was developed. However, since the
plained in the critical path diagram (Figure 6). The beginning, the main focus was placed on develop-
main critical path was the time when the robots were ing the winding path generation that was based on
working on assembly, winding, and curing. There- the syntax polyline, shapes of the component's out-
fore, it was possible to easily fast track the whole pro- line polygons and the orientation of their parts (fig.
cess and shorten the path average up to two days 00). The algorithm itself had three main steps: 1) sort-
per component, even to one day fabrication loop for ing - breaking down the syntax polyline and trans-
the smallest ones. The whole production phase was forming it into frames based on the orientation of
organized as student-to-student tutoring part of the the sides of the outline polygon, 2) generating -
defining the safe path for the fiber source by hitting

Fabrication - Volume 1 - eCAADe 32 | 429

 checkpoints around its way from point- to-point and
adding a hooking movement in the end and 3) trans-
lating - reorienting control points frames to end effec-
tor frames, transferring the position data into a KRL
structure and saving it in a text file. (Figure7)
Each line in the robot code represents one spe-
cific position of the end effector within the working
envelope. The movement is therefore accomplished
by aligning the frame of the end-effector with the tar-
get frame by repositioning the arm as little as pos-
sible. Although the robotic setup was comprised of
two robots, only the master robot was running the
winding code. The slave robot locked a specific off-
set position of its flange in the coordinate system of
the master's flange and kept it unchanged during the
entire winding and curing phase.
After a few test runs with the largest compo-
nents, it was also evident that the size of the compo-
nents made it necessary to optimize the workspace.
The most viable solution, without readjusting the
physical position of the fiber source, was to rotate
the virtual frame of the base, which was set as the
tip of the fiber feeding tube. To be able to do that,
all positions outside the working envelope were lo-
cated. Moreover, an evolutionary algorithm was cre-
ated and tested before each winding sequence to
minimally rotate the base to fit in all these crucial po-
sitions. Subsequently, an effector frame optimization
algorithm was added to position the space frames of
both robots as parallel as possible to each other as
well as simple collision detection.
RESULTS AND CONCLUSION
The ICD/ITKE Research Pavilion 2013-14, based on
the beetle elytra as a fiber reinforced double-layered
lightweight system, developed a two-robot robust
setup that was able to create differentiated core-less
filament elements. The design allowed the structure
to have a parameter based on material efficiency and
fibrous geometric performances.
430 | eCAADe 32 - Fabrication - Volume 1
Figure 7
Based on the syntax
polyline (black) is
generated path of
the fiber source
(red), that is further
translated into the
end effector path
(green).
The research on biological role models, the trans-
fer of performative principles of structural morphol-
ogy as well as the further development of a core-
less robotic filament winding technique resulted in
a proposal of a double layered modular structure
able to adapt geometrically according to the set of
defined parameters. Fiber layout, component ge-
ometry, and global component arrangement were
principles abstracted from the microscopic fiber lay-
outs and global distributions of the fiber bundles of
the beetle's elytra. The ability to construct many
unique components using a minimal amount of ma-
terial resources led to a modular-based system de-
sign. This paper presents the fusion of computational
data from biological analysis, computational design
and robotic manufacturing, resulting in the proto-
typical structure composed out of 36 unique compo-
nents. (Figure 8)
Figure 8
Two unique
components that
demonstrate the
scalability of the
system that was
created
Figure 9
Interior of the
Pavilion showing
the scale of the
components
(ICD/ITKE)
The research process provided the opportunity
to create a modular construction system, that was
not only tested on a full-scale scenario, but also pro-
vided to measure and analyze the result of its con-
struction. The pavilion's elements each have an in-
dividual fiber layout which results in a material effi-
cient load-bearing system. The diameter of the com-
ponents range from 0.536 to 2.5 meters (Figure 9).
The biggest element weighs 24.1 kg. The research
pavilion covers a total area of 50 m² and a volume
of 122 m³ with a total weight of 593 kg. The struc-
ture reaches a height of 3.4 m. The multidisciplinary
design and fabrication strategies from a bottom-up
biomimetic implementation resulted in the construc-
tion of the pavilion as well as a robust fabrication sys-
tem for the construction of core-less filament wound
hyperbolic components.
ACKNOWLEDGEMENTS
The biomimetic research was conducted in collabo-
ration with the Department of Evolutionary Biology
of Invertebrates directed by Prof. Oliver Betz and
the Department of Palaeontology of Invertebrates di-
rected by Prof. James Nebelsick, both University of
Tübingen, as well as Dr. Thomas van de Kamp from
the ANKA / Institute for Photon Science and Syn-
chrotron Radiation at the Karlsruhe Institute of Tech-
nology.
The construction of the full-scale prototype was
supported by the Competence Network Biomimetics,
KUKA Roboter GmbH, the SGL Group, and Sika AG.
Additional advice and support was received from Dr.-
Ing. Markus Milwich from the ITV Denkendorf and
Dr.-Ing. Thomas Stehle, Rolf Bauer, Michael Reichers-
dörfer from the Institute for Machine Tools from the
University of Stuttgart.
The authors would like to particularly thank the
team who worked on this project:
ICD Institute for Computational Design - Prof.
Achim Menges
ITKE Institute of Building Structures and Struc-
tural Design - Prof. Jan Knippers
Scientific Development
Moritz Dörstelmann, Vassilios Kirtzakis, Stefana
Parascho, Marshall Prado, Tobias Schwinn
System Development, Fabrication & Construc-
tion
WiSe 2012 - SoSe2013:Desislava Angelova, Hans-
Christian Bäcker, Maximilian Fichter, Eugen Grass,
Michael Herrick, Nam Hoang, Alejandro Jaramillo,
Norbert Jundt, Taichi Kuma, Ondrej Kyjánek, Sophia
Leistner, Luca Menghini, Claire Milnes, Martin
Nautrup, Gergana Rusenova, Petar Trassiev , Sascha
Vallon, Shiyu Wie, and Leyla Yunis
WiSe 2013:Hassan Abbasi, Yassmin Al-
Khasawneh, Yuliya Baranovskaya, Marta Besalu,
Giulio Brugnaro, Elena Chiridnik, Eva Espuny,
Matthias Helmreich, Julian Höll, Shim Karmin, Georgi
Fabrication - Volume 1 - eCAADe 32 | 431
 Kazlachev, Sebastian Kröner, Vangel Kukov, David
Leon, Stephen Maher, Amanda Moore, Paul Poinet,
Roland Sandoval, Emily Scoones, Djordje Stano-
jevic, Andrei Stoiculescu, Kenryo Takahashi, Maria
Yablonina with the help of Michael Preisack

REFERENCES
Brell-Çokcan, S and Braumann, J 2013, 'Introduction',
in Brell-Çokcan, S and Braumann, J (eds) 2013,
Rob|Arch: Robotic Fabrication in Architecture, Art, and
Design, Springer-Verlag/Wien
Chen, J X, Ni, Q Q and Q, Li 2005, 'Biomimetic
light-weight composite structure with honeycomb-
trabeculae', Biomacromolecules, 12(2), pp. 321-322
Van de Kamp, T and Greven, H 2010, 'On the Architecture
of Beetle Elytra', Entomologie Heute, 22, pp. 191-204
Knippers, J and Speck, T 2012, 'Design and construction
principles in nature and architecture', Bioinspiration
& Biomimetics, 7, pp. 1-10
Lomakin, J 2011, 'Mechanical properties of the Beetle
Elytron, a biological composite material', Biomacro-
molecules, 12(2), pp. 321-335
Neville, AC 1975, The Biology of the Arthropod Cuticle,
Springer-Verlag, New York
Schmitt, OH 1969 'Some interesting and useful
biomimetic transforms', In Proceeding Third Inter-
national Biophysics Congress, Boston, p. 297
Weigele, J, Scholz, M, Schwinn, T, Reichert, S, LaMagna,
R, Waimer, F, Knippers, J and Menges, A 2013 'Fi-
brous Morphologies: Integrative design and fabrica-
tion of fibre-reinforced structures in architecture us-
ing robotic filament winding', Proceedings of eCAADe
Conference 2013, Delft, pp. 549-558

432 | eCAADe 32 - Fabrication - Volume 1

Isoprototyping
Rapid Robotic Aided Fabrication for Double Curvature Surfaces

Pavlos Fereos1 , Marios Tsiliakos2

1
UIBK - UCL - formDEcode 2 UIBK - formDEcode
1,2
www.formdecode.com
1,2
{Pavlos.Fereos|Marios.Tsiliakos}@uibk.ac.at

IsoPrototyping is a research initiative, undertaken at the Institute fur

experimentelle architektur.hochbau, within the context of the Vertiefung Hochbau
and Sonderkapitel des Hochbaus courses, which specialize on building
construction. Through the case study of an iso-surface spatial configuration, this
research targets the exploration of innovative digital prototyping methodologies,
that would allow rapid and cost-efficient fabrication, capable of manufacturing
any given double curved surface. The ABB industrial robots of REX-Lab
programmed in combination with custom designed, recalibrated dry-mold,
surface-producing apparatus, formed the framework for a proficient, yet flexible,
process describing and fabricating implicit non-linear systems.

Keywords: industrial robots, pin-board, rapid-prototyping, dry-recalibrated

mould, digital-fabrication

INTRODUCTION The research question raised, therefore, was

The Vertiefung Hochbau and Sonderkapitel des
Hochbaus course's incentive derives from the obser-
vation that despite the increasing popularity of these
implicit morphologies within the architect's agenda,
the fabrication outcome has been mostly confined to
3d-printing. While the provided product is accept-
able in terms of representation, is not however con-
tributing to any advancement towards larger scale
fabrication. In large scale projects that include panel-
ing or cladding with constant, yet relatively smooth,
change of curvature without any standardization or
breakdown into repeatable components, the mate-
rial waste for one-time-usage moulds can exceed by
far in volume, the actual product.
whether a non-conventional fabrication method
could be developed having significant advantages
over other known rapid customization procedures,
yet maintaining an adequate level of accuracy. The
hypothesis was formed around the argument of
whether via a flexible and re-usable tool, capable of
being re-calibrated rapidly and accurately via robotic
control, a transformable dry mold could be devel-
oped to the point where the whole fabrication sys-
tem could be superior in terms of cost-efficiency, ma-
terial waste and production pace.
A generalization of this process is intended up
to the level of any given double curved surface, for
a large scale façade project, while addressing param-
eters such as choice of material, the healing process,

Fabrication - Volume 1 - eCAADe 32 | 433

 the complexity of the model and the post-fabrication
assemblage. In this context, the selection of an iso-
surface structure (figure 1) proved a demanding task,
due to its constantly changing curvature over a vast
range of scales and sizes within a single model.
BACKGROUND RESEARCH
Similar approaches exploring double curved surface
fabrication methodologies provided a hands-on ex-
perience on the challenging character of this in-
vestigation. Notable are the 3d puzzle prototypes
by Axel Kilian (Kilian 2003) and the flat aluminum
stripes articulation of self-supported implicit surfaces
from Marc Fornes (Fornes 2011) , forming structures
resembling minimal and iso-surface configurations.
While these two focused on a more linearly tessel-
lated assembly, Iso-Prototyping instead of creating a
repetitive or flexible component that by assemblage
generates multiple varieties of double-curved sur-
faces, looked into creating a flexible re-usable me-
chanic system that generates infinite number of non-
identical components.
On the robotic fabrication side, a similar and syn-
chronous approach of heat forming plastic plates was
developed by RobotsInArchitecture (Braumann and
Brell-Cokcan 2013) , where the heating and forming
process were confined on a local topological level,
rather than on the overall module.
434 | eCAADe 32 - Fabrication - Volume 1
ISO-SURFACES
Reasons for choosing iso-surfaces as a case
study
The iso-surfaces case study was selected due to their
intrinsic nature deriving from the two-dimensional
isolines (Greek ίσος=equal). Iso-surfaces in a simi- Figure 1
lar fashion represent 3D-topology of the same value The initial
within a certain spatial volume, being primarily rep- Iso-surface case
resented by the marching cubes algorithm (Lorensen study
and Klein 1987) , which extracts the polygonal mesh
of a three-dimensional scalar field or voxel-space.
Their emerging implementation in contemporary ar-
chitectural research, origins from their inherent ca-
pacity to generate a formal language of an otherwise
un-achievable spatial intricacy.
Digital experimentation with iso-surfaces com-
menced via the Grasshopper3d platform , utilizing a
series of add-ons to render designable their recipro-
cal properties. In particular Nudibranch [3] to gen-
erate the value-fields, Millipede [2] for polygonizing
the scalar field (Bourke 2012) and Weaverbird [4] for
smoothing the resulting mesh. The design output
depended greatly on user-defined constraints, with
only global design restriction being that, the iso-
surfaces had to be in pairs with a constant distance
between them. Hence a substantial amount of com-
plexity emerged, with necessity of being realized into
physical models via several methods of fabrication.
Different fabrication techniques
A series of different fabrication techniques where ex-
plored, prior to the final decision to push forward the
recalibrated iso-machine, in order to assess different
material behaviours, the surface quality of the result
and the required time, cost and material waste gen-
erated per piece. These were:
• Waffled structures (cross linked profiles made
out of laser-cut mdf ) to which a thin metal
mesh was attached on both sides as a base for
plaster being applied at the end (figure 2).
• CNC milled foam moulds for casting plaster
(figure 2).
Figure 2
Testing different
fabrication
techniques

• Pressure application on heated polystyrene the resulted 2,5D computer-controlled tangible sur-
plates (figure 2). face is similar to our proposal. Yet with the introduc-
tion of fluid power for actuation, sensors and micro-
THE ISOMACHINE controllers aiming into constructing a "massive hy-
draulic actuator", the decision to have all the com-
A Transformable dry Mould
ponents integrated into the "Digital Clay" can not
As a solution on how to achieve a continuous,
only increase the cost but also increase malfunctions
smooth and double-curved surface, a choice was
when such a delicate mechanic system is employed
made to work upon the concept of the decorative
to apply deforming-pressure to hot polystyrene sur-
toy 'Pin Art' (figure 3), where an array of metal pins
faces.
is being distributed between three boards that when
Similarly the FEELEX (Iwata et al. 2001) de-
pressure is applied from one side, the pins move cre-
veloped a haptic surface by employing DC motors
ating a pixelated representation the object that is
with a minimum of 6x6cm array grid and resolution.
pushing them. The same logic was implemented in
This limitation due to the size of the smallest motor
the examined machine prototypes, with the differ-
available (4cm diameter ) resulted into the FEELEX-2
ence being that no shape applies pressure on the
which incorporated a piston-crank mechanism to al-
pins. Rather they get positioned by robots controlled
low one servo motor to calibrate more than one rod,
by an algorithm. As a result the outcome of this po-
Figure 3 sitioning is then used as the mould for forming the
The Pin-Board final surface.
Game The 'Pin Art' concept provides an infinite number
of possible surface describing configurations, capa-
ble of operating as dry moulds manipulating heated
polystyrene boards bought right off the local market,
overcoming the dependence on liquid-form materi-
als with time-consuming setting requirements.

Similar Approaches
In terms of similar approaches to a recalibrated mold,
the "Digital Clay" project (Zhu and Book 2006) cre-
ated a haptic surface described as a "3D Monitor
whose pixels can move perpendicularly to the screen to
form a morphing surface". Although aimed for an en-
tirely different and bi-directional interactive purpose,
in this case 25 - a grid of 5x5. Yet this had as a limita-
tion the small height displacement of the rods, linked
to the rotation of the servo motor and the linkage.
The presented in this paper approach towards all
the above projects was to overcome the limitations
arising from incorporating moving mechanisms (ei-

Fabrication - Volume 1 - eCAADe 32 | 435

 ther DC motors or fluid actuators) into the machine
such as low rod resolution, vibrations, small height
movement, multiple servo motors or delicate equip-
ment non-resistant to pressure by employing ABB in-
dustrial robots. Our hypothesis was based on the ar-
gument that the 1mm robotic precision could accu-
rately relocate the rods into a large number of po-
sitions, without the actual rods needed to bear any
mechanism other than a break to hold them into po-
sition.
The IsoMachine concept
By investigating this system's articulation through
various physical models it was evident that certain
design and fabrication limitations had to be resolved,
with the requirement for a break system to lock the
rods into position while under pressure, to be the
most important one, along with the optimum res-
olution of the grid and the requirement to control
efficiently the amount of pressure applied on the
polystyrene boards.
In the case of the interaction between the mate-
rials used, in order to achieve a smooth surface prod-
uct, several layers of other materials had to be ex-
amined between the wooden rods and the plastic
heated frame. Furthermore, one single array of rods
proved insufficient to provide the demanded pres-
sure on plastic when curvature was complex, thus
436 | eCAADe 32 - Fabrication - Volume 1
two mirrored and consecutive arrays of rods had to
be employed.
In addition, the robotic control of the proposed
system had to be successfully utilized. The REX-Lab
of the University of Innsbruck consists of three ABB
robots; hence the double array of rods was not an
issue. In fact both sides of the isosurface would be
calibrated to their positions simultaneously by the
two rear robots while the third one, with a blow-torch
heating device as an extension of its arm, could heat
the polystyrene board and then place it into posi-
tion. Despite the relatively low heating temperature
required for the polystyrene (120C), this last option
was aborted due to the absence of an ABB gripper
and the size of the heated plates, making it impossi-
ble to heat locally.
ISOMACHINE V1.0
The apparatus
Two plates holding 400 wooden rods on a grid of
20x20, each one consisted of three acrylic laser-cut
frames, have been positioned on a wooden frame
which enabled them to move towards each other
by sliding on aluminium corner sectioned elements.
This movement has been articulated through a series
of steel wire connections, which enabled the whole
system to move just by pushing or pulling a single
lever. As a break system, a rubber material was in-
Figure 4
The Iso-Machine
v1.0, exploded
model and rods
Figure 5
Mirrored arrays of
rods with cotton
spheres at the tip,
forming continuous
double curved
surface

Figure 6
Procedural diagram
of the fabrication
stages

troduced in-between the three acrylic frames, which
expanded towards the rods, when the frames were
fastened together, increasing friction thus prevent-
ing them from moving under pressure (figure 4).
Finally, at the tip of each wooden rod a cotton
sphere was placed, to provide a soft, flexible and
heat-absorbing edge while touching the plastic and
allowing the utilization of sticks on only every other
grid hole, reducing the calibration time by 50%. (fig-
ure 5).
Fabrication process
Fabrication was a procedural linear process, where
a set of inputs (figure 6), such as the heated plastic
and the robotic calibration, lead to a controlled trans-
formed result.
It is a rational assumption that due to the finite
dimension of the material and the machine, struc-
tures larger than this size have to be tessellated in
advance. Whilst the machine is calibrated by the
robots, in a process that lasts about five minutes (on
medium robot speed), the rigid polystyrene is be-
ing heated in an industrial oven (figure 7), for about
twelve minutes. Laser-cut wooden frames ensure the
preservation of the piece's outline, when the fixed
on a wooden frame polystyrene plate is inserted in
the machine and the lever is pushed to bring the two
calibrated frames together forming the surface. The
healing time of this process is about 45 to 60 sec-
onds, providing a total fabrication time of 12 to 15
minutes per module. This compared to the 6 hours
of CNC milling required for building foam-moulds
added to the 16-20 hours required for the plaster to
rest, proved efficient.

Fabrication - Volume 1 - eCAADe 32 | 437

 Figure 7
Heating the rigid
polystyrene

Figure 8
Snapshot of the
dynamic
grasshopper
procedural script

Figure 9
The first case study
model and material
result

Figure 10
On the left the 2
frames (v2.0 and
v2.5)exploded
model, on the right
the breakdown of
components of
v2.5.

438 | eCAADe 32 - Fabrication - Volume 1

Robotic Calibration Method v1.0
In terms of software control, a custom associative
code was developed, where from one single input
item, a double curved surface, the entire system au-
tomatically was recalculated, readjusting to the new
parameters and then transferred as a list of com-
mands into the robot controller. HAL [1] Grasshop-
per plug-in for Rhino was used, along with several
other surface analysis and a few custom written com-
ponents (figure 8).
The inputs are the tessellated structure's differ-
ent pieces, reinforcing the initial hypothesis of a rapid
prototyping approach that from one single appara-
tus a vast number of different configurations can be
harvested. In addition a custom designed tool at the
tip of the robot arm was used comprising of a straight
wooden stick with a carved cavity at its edge to en-
sure proper fit on the rods.
First case study and results
As a case study to perform the experiment for as-
sessment, a double iso-surface model was selected.
Output data from this system can be comprised to
recorded observations, the final surface outcome,
cost and time management. It is evident that there
is almost no material waste from this process, even in
malfunctions the plate can be reheated and reused.
In addition the fabrication time per piece is almost
2% of other methods. Assessing the system in terms
of cost control, time efficiency, material waste man-
agement and the global character of this process,
can be proved a fairly straightforward task. How-
ever, the comparison to the digital model and the
effective translation and re-assembly to one unified
iso-surface is rather intricate and ambiguous. Post-
fabrication control and connectivity systems have to
be carefully planned and examined in order to fully
assess a prototyping system similar to the proposed
(figure 9).
Assessment of v1.0
While the proposed system proved successful in re-
spect to certain initial expectations, such as its global
character, material waste, cost efficiency and fabrica-
tion time, it under-performed in providing an exact,
in respect to the digital model, double-curved sur-
face piece as an end product. A specific aspect in
need of optimization was the resolution and size of
the rods array. The more rods the more exact the cur-
vature representation would be, and having greater
rods in length would result to large curvature de-
viation within the same module. This requirement
brought forward a period of research for a new ma-
terial to replace the wooden sticks, lighter but also
rigid. A choice was also made to remove the frame of
the machine and reconstruct the two moving parts
and connect them via a rail. A new break system also
had to be designed capable of locking into position
around 1600 rods at the same time.
ISO-MACHINE V2.0 & V2.5
The apparatus
The new layout consisted of two opposing frames.
The first was stationed and made out of wood while
the second made out of steal with a rolling system
attached to its base. Each one is holding each 1600
rods made out of plastic sticks (balloon sticks bought
off the market) which passed through three acrylic
frames perforated on a 40x40 grid (figure 10).
In-between these three frames two wooden
frames were interpolated, which moved in oppos-
ing direction via side-screws locking the rods into
position, acting effectively as breaks. The new plas-
tic rods required a rubber cup at their top to allow
a smooth contact with the heated plates and avoid
the melting of the plastic rods (figure 11). This grid
of 40x40 rods with a maximum length of 38cm pro-
vided us with a high resolution three-dimensional
dry-mould which through initial tests provided us
with extremely smooth and substantially accurate re-
sults of polystyrene plates (figure 12). Yet the large
amount of rods (1600 from 400 before) and their ex-
tended length (38cm from 20cm) generated a very
long calibration time per surface which had to be ad-
dressed.
Fabrication - Volume 1 - eCAADe 32 | 439

Robotic Calibration Method v2.0
The calibration method of the v1.0 machine was a
simple 3-point based loop movement:
• starting point - robot head on zero position
• middle point -robot head on point-on-surface
position
• ending point - robot head return to starting
position (to avoid hitting sideways the other
rods)
The above steps were repeated as many times as the
number of rods. Although this worked sufficiently
on v1.0 for a total amount of 600 rods (3x200 us-
ing odd and even slots alternating), the amount of
point coordinates required for v2.5 was increased by
800% to 4800 (3x1600=4800). This combined with
almost doubled rod length pushed the calibration
time up by as much as 1500% (depending on the
surface shape), which rendered the time-efficiency of
our method obsolete.
To address this issue we first designed a new
cross-shaped calibration-head to attach to the robot.
It had four different 3D-printed edges, three being
square-shaped and the fourth a simple pin-head on
440 | eCAADe 32 - Fabrication - Volume 1
Figure 11
Elements of
IsoMachine V2.5
Figure 12
Results from
Isomachine V2.5
the size of the rods (figure 13). This allowed the robot
to shift edge by rotating it's sixth-axis and also to
push into average position up to 100 rods with one
robot movement. Also the precision of the machine
built enabled the calibration of only one frame which,
if positioned face-to-face with the second, it allowed
the opposing rods also to be pushed into the mir-
rored position.
Figure 13
The new calibration
method with the
cross-shaped robot
edge tool.
Furthermore, to effectively reduce the time
of calibration via clustering, two custom routines
were written within the Grasshopper3d environment
based on two different basic algorithms:
• K-Means algorithm for clustering the surface
points into groups based on Y-axis displace-
ment.
• Square-packing algorithm to best fit the four
edges of the cross-head to each cluster of
points.
Figure 14
On the left the
clustering of surface
points by K-Means
algorithm, bottom
right the square
packing of one
cluster and on top
right the output of
the script with the
line-paths ready for
calibration by the
robot.
The K-Means algorithm partitioned the 1600 surface
points into a set range of clusters (min 5, max 10) and
return the solution with the smallest overall distance
the robot had to travel to reach each cluster's mean
(center point) and back. This allowed the calibration
of each cluster separately, reducing substantially the
path the robot had to travel to avoid side-hitting the
rods (figure 14).
The square packing script had as input the clus-
ters of points returned from the K-Means algorithm.
It then fit separately into each (flattened and with
irregular boundary) cluster the four square sizes of
the new calibration head, starting from larger to
smaller. Using exhaustive search, the algorithm pro-
duced four different solutions starting from the four
extreme edge points based on XZ coordinates. To
speed things up, the algorithm first tested whether
each point was lying in the center of a possible square
by checking only whether four points existed on the
corner coordinates of the square. If that was true,
then the algorithm tested the whole square perime-
ter's coordinates for point-match. To ensure that
each square's center point was also a point on the
grid, it was drawn using a step value from the center,
thus each square had a side length of an odd num-
ber. Therefore if the number of point-matches satis-
fied the following rule:
(2 · st + 1)2 + (2 · st − 1)2 = m (1)
where 'st' is the step used to build the square and 'm'
the number of matching points hits returned by the
script, then the algorithm would add a square drawn
from the center of the point under evaluation. From
that moment on every other possible square would
have to not intersect, share edge or corner points
with the ones already placed.
The fact that there were always a number of re-
mainder single points not grouped meant that the
solution with the fewest squares packed wasn't al-
ways the optimum. Therefore the algorithm returned
not the solution with the fewest squares, but the so-
lution with the smallest sum of squares plus remain-
ing points. The four new groups of squares (and
their associated points) were then assigned a new
common Y coordinate, which was the smallest of all
points in that group, and moved to that temporary lo-
cation via their respectful square-sized surfaces and
using their centers as starting points. When the robot
finished the first iteration of this process, it then re-
peated it for three more times, excluding each time
the largest surface and starting with the next in line,
until all rods were eventually positioned into final co-
ordinates by the single-pin edge.
The above calibration method reduced the time
needed down to a range of 8%-15% of the original
required, depending on the surface under analysis. A
more sophisticated algorithm, implementing better
method than brute-force, with perhaps more start-
Fabrication - Volume 1 - eCAADe 32 | 441
 ing points for square packing than just four corners
would have reduced the time by an extra margin of
2-5%, something considered, for the scale of this re-
search, of minimal overall effect. For that reason the
whole process, via the employment of this algorithm,
was considered as sufficiently optimized.
CONCLUSIONS
The level of smoothness of the produced surfaces
and their curvature precision was satisfying, consid-
ering the complexity of the digital model initially set
to physically build. Also the latest machine version
proved very resilient to faults and breakdowns, cre-
ating the expectation that with a substantial budget
invested, a larger and industrial-level machine could
be produced with greater capabilities and increased
precision. Yet after a long process of development
a conclusion was made that as long as there's no
liquid casting involved in the fabrication of double-
curved surfaces, then the material properties will al-
ways define the limitations of the system. Pressuring
a material to reshape it eventually increases its over-
all surface area while maintaining its original volume.
This stretches the material to its limits, while always
producing an approximation of the intended shape,
never an exact replica. Furthermore, even though
the increased resolution of the rods improved the
smoothness of the surface curvature, the edge con-
dition remained throughout the process problem-
atic. And since the precise connection of one surface
component to another relied on the accuracy of their
edges, as the deviation from the digital model accu-
mulated along the build-up, it eventually made the
overall assemblage a challenge.
However, if the employment of this machinic sys-
tem is shifted from the task of producing a large
spectrum of accurate non-identical components, to
that of producing a range of similar geometries, or
minor variations of the same component, it could
perform much better. And the proven low mate-
rial waste, cost efficiency and fast production pace
of this method could help in finding applications
to small scale components construction, benefitting
442 | eCAADe 32 - Fabrication - Volume 1
from more accurate and efficient fabrication than
vacuum forming used today in the industry.
In terms of the ABB Robots used, an automa-
tion of the whole process in a way that no human in-
put is necessary could perhaps also find applications
in the industrial market. This could be achieved by
constructing a large stationed grid-wall of rods on
one side and by incorporating one machine frame
on the robot on the other side, making it effectively
an extension of its robotic arm. This way no other
moving parts will be required, the robot will directly
apply pressure against the grid-wall, forcing the in-
terpolated polystyrene plates into shape, while the
whole system will utilize a second robot only for cal-
ibration. This would certainly advance the fabrica-
tion process and provide a more consistent use of the
robotic aided rapid IsoPrototyping methodology.
AKNOWLEDGEMENTS
The authors would like to thank Dr. Marjan Col-
letti, head professor at Hochbau institute for his con-
stant support, Thibault Schwartz for providing an ed-
ucational HAL license for the project, Allison Weiler
for her support in using the robots, and all the stu-
dents who participated so productively in the course
: Avila Gomez Jaime, Aschberger Anna, Bernhard
Markus, Biel Isabel, Bierwirth Christoph, Brandauer
Ulrike, Brandtner Markus, De Soto Ana Isabel, Fleck-
enstein Julia, Fraolini Julia, Garcia Javier, Gavrilovic
Pedja, Grasser Alexander, Gruber Maximilian, Hartl
Rainer, Honke Sebastian, Huff Beatrice, Kinzele Tun,
Klein Thomas, Leitgeb Spoerk Pascal, Maric Marko,
Pernstich Daniel, Rieder Benjamin, Riepl Luisa Maria,
Rust Philipp, Schöpf Jan, Schuster Daniel, Simon Bob,
Stanzel Jörg, Stefanova Melina, Steinkogler Anne,
Stieler David, Tholen Lena, Uitz Theresa, Unterfrauner
Maximilian, Walzer Alexander, Weittenhiller Chris-
tian, Zeiler Thomas and Zluga Philipp.
REFERENCES
Bourke, P 2012 'Polygonising a scalar field',
www.paulbourke.net/geometry/polygonise, Ac-
cessed 1 February 2014
Braumann, J and Brell-Cokcan, S 2013 'Robotic Forming',
RobotsinArchitecture, http://www.robotsinarchitecture.org/
Accessed 22 June 2013
Fornes, M 2011 'nonLin/Lin Pavilion -Theverymany',
http://theverymany.com/constructs/10-frac-centre/,
Accessed 3 February 2013
Iwata, H, Yano, H, Nakaizumi, F and Kawamuri, R 2001
'Project FEELEX: Adding Haptic Surface to Graphics,',
Proceedings of SIGGRAPH 2001
Kilian, A 2003 'Fabrication of Partially Double-Curved
Surfaces out of Flat Sheet Material Through a 3D
Puzzle Approach', Connecting-Crossroads of Digital
Discourse[Proceedings of the 2003 Annual Conference
of the Association for Computer Aided Design In Archi-
tecture], pp. 75-83
Lorensen, W and Klein, H 1987, 'Marching Cubes: A
High Resolution 3D Surface Construction Algo-
rithm', Computer Graphics, 21(4), pp. 163-169
Zhu, H and Book, W 2006 'Digital Clay Next Generation
Man-machine Interface', International Conference on
Robotics and Automation, Orlando
[1] http://hal.thibaultschwartz.com/
[2] http://www.sawapan.eu/
[3] http://digitalsubstance.wordpress.com
[4] http://www.giuliopiacentino.com/weaverbird/

Fabrication - Volume 1 - eCAADe 32 | 443

444 | eCAADe 32 - Fabrication - Volume 1
Design to fabrication integration and material
craftsmanship
A performance driven stone architecture design system based on material,
structural and fabrication constraints and criteria

Sina Mostafavi1 , Matthew Tanti2

1
PhD Candidate, TU Delft 2 MSc in Architecture, TU Delft
1
www.hyperbody.nl 2 www.rapidstudio.nl
1
s.mostafavi@tudelft.nl 2 m.tanti@student.tudelft.nl

This paper presents a computational design methodology through describing of a

case study on stone building system. In addition to establishing a performance
driven form-finding methodology, the objective is to redefine local craftsmanship
methods as industrial fabrication techniques in order to introduce the
constructability of the design solutions as one of the main performance criteria.
Therefore, the focus of the methodology is to facilitate architectural design
processes through developing of customized computational design tools and
workflows for data integration and concurrent performance evaluation. The
research starts with the hypothesis that the technological advancements in digital
design and fabrication can lead to re-exploration and improvement of traditional
building techniques with local materials. The paper explains different stages of
the methodology and the way the chained design to fabrication processes would
lead to constructible, structurally possible and optimal design solutions of small
scale and simple symmetric design solutions to complex topologies at the scale of
larger complex buildings.

Keywords: digital materiality , design information exchange, compression-only

stone strcuture, Computer Aided Craftsmanship , robotic fabrication

INTRODUCTION goals is to extend the performance of stone archi-

The case study and the described design process in
this paper will challenge the performative limitations
of working with stone as a purely mass produced
material. Through the development of a methodol-
ogy for data integration and information exchange
between different computational subroutines, the
tecture through mass customization of the individual
stone elements. With these objectives in mind, the
research investigates into the vanishing discipline
and building technique of stereotomy, a building sys-
tem which has been in decline since the industrial
revolution and its knowledge is in danger of getting

Fabrication - Volume 1 - eCAADe 32 | 445

 lost due to its extensive demand for high level craft-
menship and geometrical knowledge. Through the
use of digital design and fabrication these require-
ments could be met and also exceeded, enabling
new opportunities for re-interpretation and improve-
ments in the performance of this material system.
Considering the aforementioned goals and chal-
lenges at fundamental and theoretical level this re-
search tries to answer first this question of what kind
of architecture could emerge if we shift from the
modernist paradigm of mass production and impor-
tation of foreign materials; and focus on local mate-
rials, high-level craftsmanship and mass customiza-
tion through robotic fabrication? Secondly, it ques-
tions how can computer aided craftsmanship help us
re-explore the vernacular in a contemporary way and
extend the performance of a local material system?
Furthermore, more specifically at technical level the
focus of the research is on developing reusable and
generic computational design to fabrication work-
flows for design and construction of buildings using
stone building systems.
The study does not see digital fabrication as a
unidirectional process, but a bidirectional commu-
nicative process were the material properties are also
digitized and simulated. This allows for both material
and fabrication characteristics to be applied in an in-
tegrated manner for the same architectural goal. At
theoretical level this idea can be traced in the works
of contemporary philosophers such as DeLanda that
brought the idea of material morphogenesis in which
by 'teasing' a particular material a form emerges out
from its properties rather than being imposed on it
(DeLanda 1998). Therefore the first step adopted
in heading towards an integrated design approach
in stone masonry structures was the study of the
material itself. Stone's physical properties is depen-
dent to the region and formation process. From the
three main stone types, (igneous, sedimentary, meta-
morphic) the article focuses on sedimentary stone.
The material's inability to resist tension stresses was
the leading factor why a compression only structural
system was adopted throughout the whole process.
446 | eCAADe 32 - Fabrication - Volume 1
Such a system eliminates flexural stress on the ma-
terial by taking full advantage of form. This geomet-
ric requirement introduces us to the concept of the
'thrust line', a theory well documented by Jacques
Heyman (Heyman, 1995), which is illustrated in the
Figure 1.
Figure 1
Thrust line in
dotted red. Hinges
develop when
thrust line is not in
held the section of
The study is based on a design case where the reader the material
is guided through the four basic steps applied within
this integrated design system. In this chained design
to fabrication system the several common inputs are
defined and the outputs of each stages will be the
input set of the other stages. The challenges and
the developed methods for design information ex-
change and translation of the output data sets will be
described. Moreover, through the processes the ar-
ticle gives a brief review on the computational tools
based on graphic statics and spring-based particle
systems to introduce the state of the art in this field
of research. Furthermore, a brief review on the hard-
ware and software used to robotically cut the stone is
also given. This review gives a solid reasoning for the
choice of technologies and applications chosen. Dur-
ing the elaboration of the process, the study is con-
cerned about the following questions:
• How a dynamic design information exchange
can facilitate the design processes to have
more flexibility both in term of generating
semi optimal solutions and selecting the de-
sirable alternatives by the designer?
• How to interconnect a multitude of different
factors for different performance criteria in or-
der to achieve an integrated design to fabrica-
tion procedure?
• How to integrate various processes of perfor-
mance mapping and simulation? In this case,
 how to integrate structural analysis in the as-
sembly process in order to introduce local dif-
ferentiation in thickness?
The final section of the paper, shows how this
methodology can be used in a more complex archi-
tectural project where the design intent, context and
program have an important impact on the final topol-
ogy and morphology.
THE DESIGN CASE AND THE METHODOL-
OGY
The study deals with the challenges involved in work-
ing with topologies where two (or more) vaults in-
terconnect; a problem which has not been tackled
deeply yet. The reason for considering this type of
topology is to give more flexibility to the design sys-
tem for form finding of more complex geometries
at larger building scales. These kind of topologies
pose a challenge to already well established methods
of simulating equilibrium in compression only struc-
tures with existing tools that proves to be accurate
and a robust for single shells, but their fundamental
computational logic and process does not support
interconnected shells. Therefore the chosen topol-
ogy required an alternative method of form finding.
In following sections the descriptions of the four in-
terconnected stages of the process clarify how this
challenge is tackled and how the structural and fab-
rication constraints and criteria are considered and
embedded in the design system.
STAGE 1: FROM MATERIAL PROPERTIES TO
GEOMETRIC REQUIREMENT
In this stage, having a clear idea of the topology is the
starting point. The prototypical topology used for
this design case is kept simple for clarity and ease of
understanding the method developed and applied.
In an architectural design context, this topology is
solely influenced by the design concept for every
specific project. The topology in question can be
seen as two vaults at top each other as seen in Fig-
ure 2 . The inner supports of the top vault are lo-
cated on the apex of the bottom vault, meaning that
forces from both vaults have to interact and behave
as a whole for the structure to be statically equili-
brated. To run these hanging chain simulations, a
spring-based particle system was used in connection
to the physics engine Kangaroo in Rhino Grasshop-
per. To do this, it is important to have a sealed digital
model where no naked edges exist between the dif-
ferent parts.
In the early stages, topological experimenta-
tions were done by manual modeling, converting the
NURBS (Non-uniform rational B-spline) surfaces into
a mesh, and further transforming the mesh edges
into springs for relaxation. For further refinement of
the relaxed result, a series of iterations were needed.
This lead to create a parametric chain geometry
which was able to adapt quickly depending on the
architectural needs (Figure 2). Apart form the alter-
ation of the chain geometry itself; the anchor points,
the springs' stiffness and the loads imposed also give
rise to other variations. In this case, rather than im-
posing a specific geometry on a material, the geom-
etry of the chain is given by the designer and then,
altered accordingly by the dynamic relaxation pro-
cess to satisfy the geometric requirements for the
structure to work in compression only. In this way
the system became an interactive tool where a direct
communication between top down architectural de-
cisions and bottom up structural geometric require-
ments come together to find an informed balance
between the two; relating back to the philosophical
concepts of material morphogenesis. In this particu-
lar case, the only hard criteria involved in choosing
a semi-optimal option was based on structural val-
idation through FEA software and making sure the
compression stresses does not reach the compres-
sion strength of the material (Figure 3). Other softer
criteria involve the height, quality of the space en-
closed and overall visual proportions of the structure.
Since the relaxation process gives a mesh geometry
as an output, the resultant geometry can be directly
used as input for a different number of performance
analysis criteria. For future studies solar radiation and
thermal mass calculations could also affect the over-
Fabrication - Volume 1 - eCAADe 32 | 447

all form and more specifically in this case differentia-
tion in the thickness and the challenge is to see how
such conflicting performance criteria could be con-
sidered simultaneously.
STAGE 2: FROM GEOMETRIC REQUIRE-
MENTS TO THICKENED SHELL
In this stage, the process focuses on introducing dif-
ferentiation in the thickness of the shells to con-
trol material distribution and structural performance
of the design. Rather than giving the shell one
global thickness the process explores the idea of us-
ing structural analyses to inform the design system to
differentiate the local thicknesses. Again in this stage
differentiation in the thickness of the shell can be in-
fluenced by other simultaneous simulations like solar
radiation and thermal mass calculation.
The concept of differential local thickness in this
study shifts towards the assembling process, and
how this can be beneficial. The main principle of a
compression only structures is that the thrust net-
work needs to be between the inner (intrados) and
448 | eCAADe 32 - Fabrication - Volume 1
Figure 2
Top left:
topologic study,
Top middle:
parametric chain
topology,
Top right:
mesh to springs,
Bottom:
iterations
outer face (extrados) of the vault. There might be the
case when the section of material in the shell is not
thick enough to keep the thrust network within it;
this scenario creates tension forces on the shell which
could lead to hinges and also collapse. A thicker
shell is able to contain a bigger number of catenar-
ies for different load cases, therefore more it's sta-
ble (Heyman 1995). When designing compression
only structures, it is important to check for stabil-
ity also while it is being constructed. In this design
exercise, the thrust network shown in the Figure 3
only apply when the structure is fully constructed.
This means that the structure would be unstable until
the top shell is assembled. Figure 4 (Left) shows the
thrust network of the bottom shell supporting itself
(in red) imposed on the thrust network of the struc-
ture as a whole. (blue). Where the two thrust net-
works deviate from each other, it reveals the need for
more thickness for the shell to be stable throughout
the whole construction process as seen in Figure 4
(Right).
Figure 3
Left:
Chosen dynamic
relaxation result.
Right:
Structural
validation using
FEA software.
Figure 4
Studies on
thickness variation
For the shell to be discretized into individual blocks
at a later stage, the geometry needs to be translated
back into a NURBS geometry. This gives the flexibility
to adjust accordingly and not being bounded by the
irregular mesh edges and vertices. The thickness has
been created by generating two faces (intrados and
extrados) on each side of the thrust network. For op-
timal structural performance, it is important that the
thrust network is in the center between the two new
surfaces. The same logic for the generation of the
chain NURBS geometry in stage one was used to cre-
ate the intrados and the extrados. When compared to
the high problematic issues involved in converting a
mesh geometry to a NURBS curve, this step is rather
simple and if care is taken in inputting the right pa-
rameters, a high level of precision can be achieved.
Due to the fluidity of the digital model, the designer
is given full flexibility to adjust thickness were he/she
deems fit. In this case the minimum thickness is of
(20cm) and gradually get thicker (55cm) towards the
oculus or top skylight opening were the two shells
meet.
STAGE 3: FROM THICKENED SHELL TO PER-
FORMATIVE VOUSSOIRS
After determining the local thicknesses of the shell,
the structure needs to be subdivided into the indi-
vidual stone voussoirs or blocks. This stage calls for
the full integration of structural constraints, material
considerations, aesthetics and craftsmanship. One of
the main parameters in stone architecture is the di-
rection and flow of the voussoirs in relation to the
thrust network (Block & Rippmann 2011). It is imper-
ative that the shared faces between the stone vous-
soirs are as perpendicular to the thrust network as
possible, with an acceptable deviance of less than 30
degrees. Further deviance would result in slippage
of the stone elements and instability of the struc-
ture (Huerta 2001). When dealing with sedimentary
stone, the orientation of the bedding lines should
also be perpendicular to the main principle stress
(Cachia 1985). Furthermore it is of outmost impor-
tance for the ashlars to have a staggered bond, this
creates friction forces along the sides which increases
resistance against slipping. Different types of tessel-
lations can be explored as long as the top and bot-
tom faces touch more than one stone voussoir (Block
& Riippmann 2011). To achieve such complex dis-
cretization of the surface, a relatively simple method
was applied. Rather than focusing on intensive com-
putation; topology was used to the design's advan-
tage. The NURBS surfaces created in the previous
stage, needed to go through another redrawing se-
Fabrication - Volume 1 - eCAADe 32 | 449

quence before tessellation. In this phase, the geom-
etry is translated form a trimmed surface (Figure 6,
top-left) into an untrimmed surface by creating a net-
work surface. By using the surface boundaries as the
main curves (red curves in figure 6 top-middle) for
the network surface, it guarantees that the flow of
the surface iso-curves approximate the direction of
the thrust network. The new surface was later split
along the lateral direction (green curves in figure 6
top-middle), thus creating continuous longitudinal
strips along the surface (along the red curves). These
longitudinal strips were further split in a way that the
seams between the voussoirs are always perpendic-
ular to the longitudinal direction. By allocating a dif-
ferent number of divisions per strip, and moving the
seams of the edges, it was made sure that a stagger-
ing bond would always be generated. A safety fea-
ture was included in the logic so as to mark any seams
which deviate more than a 20 degree angle from the
principle stress flows (Figure 6, bottom-left).
For the voussoirs to be robotically fabricable with
line-based end effector or head, all the surfaces in the
stone voussoirs had to be of a ruled nature. The indi-
vidual stone elements had to be further articulated
to achieve this feasibility. This was done by choos-
ing one specific direction along which the curvature
would need to be discretized into straight lines. In
this example the curvature in the voussoir was left
along the continuous longitudinal direction of the
surface. This creates a continuous aesthetic of con-
450 | eCAADe 32 - Fabrication - Volume 1
Figure 5
Thrust network
perpendicular to
shared voussoir
faces and natural
stone bedding
lines.
centric strips along the entire structure. (Figure 6,
top-right & bottom-middle) A feature which became
useful in the process is the ability to parametrically
control the number and differentiate the thickness
of the longitudinal strips, this enabled the strips to
be thinner and more condensed where the curva-
ture of the vault was higher. This tweaking allowed
for the surface to be discretized where really needed
thus adhering to the original curvature as much as
possible in the lateral direction (along green curves
in figure 6 top-middle) The top and bottom longitu-
dinal curves were taken as guiding curves whereas
the lateral direction followed the straight geometry
of the straight line of the robot end effector (Figure 6,
bottom-middle). By keeping these geometrical con-
straints in mind, further articulation of the voussoirs
can be done which would result in a higher degree on
performance embedded in the stone element. Figure
6 (bottom-right) shows how the stone block could be
further crafted in a way to create thermal breaks and
water tightness through removal of the material in
the middle layer.
STAGE 4: FROM VOUSSOIRS TO FABRICA-
TION INFORMATION
The final stage focuses on the fabrication of the ac-
tual blocks from the extracted stone megalith. To
reduce as much waste as possible, the stone ash-
lars were grouped together depending on their thick-
Figure 6
Re-generating
NURBS untrimmed
topology and
tessellation
procedure.

ness. The individual stone component is nested in a vised before going for the actual cut. The speed of
larger stone slab which had been split from the me- the robotic arm and rotation speed of the abrasive
galith. Apart from minimizing waste as much as pos- wire have to be tested and calibrated depending on
sible, the nesting algorithm also aligns the stone ash- the type of stone and finishing required.
lars according to the bedding lines in relation to the
Figure 7
principle stress flows.
Rapid Code
When all the geometry in a particular slab is
generation through
nested, it is used as input information for the robotic
Grasshopper
arm simulation. This makes sure that the geomet-
custom
ric information in all the stone blocks is feasible be-
components
fore moving from the digital realm to the physi-
cal. In this case study the simulation has been done
with open source 'Mussel' software, which allows for
the transliteration of geometry into ABB industrial
robotic arm movements via RAPID code. This simula-
tion sheds light on any overlooked fabrication issues
which might cause a break in the production process. ARCHITECTURAL APPLICATION AND DIS-
The most common breaks usually arise from the lim- CUSSION
its of the robotic arm's reach and joint rotation. Fur- The following images (Figures 8,9,10 &11) shows the
thermore, issues such as collisions of the tool with implementation of such a methodology into a com-
other objects (including the robotic arm itself ) may plex architectural project. In this case, the project
also arise. After the simulation, a dry run is also ad- is situated in the Maltese islands, a place where glo-

Fabrication - Volume 1 - eCAADe 32 | 451

 Figure 8
left: Design case
topology. Right:
Topology of
implementation
project

Figure 9
Left: Iterations of
Dynamic relaxation
procedure. Right:
Analysis and
validation of the
result.

Figure 10
Left: Converting
mesh geometry
into untrimmed
NURBS
surface.Right:
Voussoir
tessellation.

452 | eCAADe 32 - Fabrication - Volume 1

Figure 11
figure 5.7-
Rendering of final
morphology in the
surrounding
environment.

Figure 12
Physical printed
prototype of the
design case

bigerina limestone is the only resource material avail-
able for construction on the island. This system is
suited for such a context as this renewed building
system could lead to a new type of regionalist archi-
tecture where the cultural landscape built in stone
throughout its history can keep on flourishing in a
new interpretation.
The main distinctive difference between the de-
sign case and this project, is the topology. The im-
ages below show how the same logic can be ap-
plied for a more complex topology. The issue was
solved by first defining a central point for every sup-
port. These points were used as centre points, where
the voronoi principle was used to divide the shell
into manageable part. The applied method also
made sure that the shell was split perpendicularly
and mid way from the supports. This logic was fur-
ther tweaked in order to support additional architec-
tural requirements like building program and circu-
lation. Figure 8 (left) shows the logic applied in the
paper's design case and figure 8 (right) shows how
the same logic was applied to define the topology of
the project mentioned.
The techniques of using stone in a freeform ma-
sonry architecture brings a lot more challenges and
opportunities which can be investigated in future re-

Fabrication - Volume 1 - eCAADe 32 | 453

 searches and case studies. The authors are interested
and encourage further research in how solar radia-
tion and thermal mass calculations could also affect
the local thickness of the stone elements or the air
gaps between two layers of stone. As future research,
we are interested in delving further into the design
opportunities found in the heterogeneity of stone
and explore how informed material extraction/cut-
ting (through CT Scanning) and design could both
influence each other by allowing a bidirectional rela-
tionship. Although the research did consider the pro-
cess of assembly, there is a lot more to be explored
and detailed which would result in more efficiency
on site through the reduction of falsework needed
to prop the structure before achieving structural in-
tegrity. One other objective is also to find ways how
the developed system could be adapted for other
fabrication techniques such as the 3D printing of an
inert natural material. The following Images illus-
trates some of the initial physical prototyping with
Stereo lithography type of printing with artificial or
plastic based material (Figure 12). The hypothesis for
the next step can be further formulated based on this
presumption that like the way the orientation of bed-
ding lines in sedimentary type of stones is important
in structural performance of the stone blocks, it could
be important to explore how it is possible to control
the direction of the layers of depositing material in
printing for better performance.
REFERENCES
Bechtold, M 2009 'New stone shells: design and robotic
fabrication', , Proceedings of International Associa-
tion for Shell and Spatial Structures (IASS) Symposium
2009, Valencia
Block, P and Rippmann, M 2012 'New Design and Fab-
rication Methods for Freeform Stone Vaults Based
on Ruled Surface', Computational Design Modelling
2012, pp. 181-189 Springer
Cachia, J 1985, The Mechanical and Physical Properties of
The Globigerina Limestone as used in Local Masonry
Construction, University of Malta
Clifford, B 2012, Volume: Bringing Surface into Question,
The Skidmore, Owings & Merril Foundations, Mat-
terdesign,, Boston
454 | eCAADe 32 - Fabrication - Volume 1
DeLanda, M 1998 'Deleuze and the Open ended Becom-
ing of the World', Chaos/Control: Complexity Confer-
ence,, University of Bielefeld, Germany
Dernie, D 2003, New Stone Architecture, Laurence King
Publishing, London
Evans, R 1995, The Projective Cast: Architecture and Its
Three Geometries, MIT Press, Cambridge
Fallacara, G 2009 'Toward a Stereotomic Design: Ex-
perimental Constructions and Didactic Experiences',
Proceedings of the Third International Congress on
Construction History, Cottbus, p. 553
Feringa, J and Søndergaard, A 2013 'Fabricating Ar-
chitectural Volume Stereotomic investigations in
robotic craft', Fabricate, 2013
Gramazio, F, Kohler, M and Oesterle, S 2010, 'Encoding
Material, The New Structuralism: Design, Engineer-
ing and Architectural Technologies', Architecture De-
sign July/August 2010, Wiley, Volume 80, Issue 4, pp.
108-115
Harding, J and Shepherd, P 2011 'Structural Form Find-
ing using Zero-Length Springs with Dynamic Mass',
IASS Annual Symposium: IABSE�IASS 2011: Taller,
Longer, Lighter, London
Heyman, J 1995, The Stone Skeleton. Structural Engineer-
ing of Masonry Architecture, Cambridge University
Press, Cambridge
Kilian, A 2004 'Linking hanging Chain Models to Fabrica-
tion', ACADIA 2004
Leach, N, Turnbull, D and Williams, C 2004, Digital Tecton-
ics, John Wiley & Sons
Menges, A 2012, 'Material Computation: Higher In-
tegration in Morphogenetic Design', Architectural
Design,March/April 2012, Wiley Publications,London,
Volume 82, Issue 2,Special Issue: Material Compu-
tation:, pp. 14-21
Mostafavi, S, Morales Beltran, M and M.Biloria, N 2013
'Performance Driven Design and Design Informa-
tion Exchange', Proceedings of the 31st eCAADe Con-
ference – Volume 2, Delft University of Technology,
Delft, The Netherlands, pp. 117-126
Oxman, R 2008, 'Performance-based design: current
practices and research issues', International journal
of architectural computing (IJAC), 6, pp. 1-17
Pilgram, D, McGee, W and Kaczynaki, M 2011 'Robotically
Fabricated Thin-shell Vaulting: a method for the in-
tegration of multi-axis fabrication processes with al-
gorithmic form-finding techniques', ACADIA 2011
Siegesmund, S and Snethlage, R (eds) 2011, Stone in Ar-
chitecture: Properties, Durability, Springer
From rapid prototyping to automated manufacturing
Aant van der Zee1 , Bauke de Vries2 , Theo Salet3
1,2,3
Eindhoven University of Technology
1,2,3
{a.v.d.zee|b.d.vries|t.a.m.salet}@tue.nl

In this paper we present an outline of a newly started project to develop a tool

which connects BIM to a manufacturing technique like 3D printing. First we will
look some promising manufacturing techniques. We will design a small dwelling
and export it into a BIM, from which we will extract our data to generate the path
the nozzle has to follow. The chosen path is constrained by the material
properties, the design and speed of the nozzle. To validate the system we develop
a small VR tool in which we mimic a manufacturing tool.

Keywords: Rapid prototyping, rapid manufacturing, robotics, automation,

Building Information Model (BIM)

INTRODUCTION RELATED RESEARCH

It is good practice in architecture to make prototypes
of the design at hand. The reason is often for own
usage or presentation to stakeholders. Mostly card-
board or wood is used to make these models. In
the 1980s early examples of 3D printing occurred,
though the printers then were large, expensive and
highly limited in what they could produce. These in-
dustrial 3D printers have been used extensively for
rapid prototyping and research purposes. In the
1990's the term "3D printing" was coined at MIT when
then graduate students Jim Bredt and Tim Anderson
modified an inkjet printer to extrude a binding solu-
tion onto a bed of powder, rather than ink onto pa-
per.
These rapid prototyping techniques did start re-
search to see if these techniques can be used as
a manufacturing technique, in order to automated
the building industry. But in order to be useable in
the building industry the manufacturing techniques
must connect to a BIM.
Rapid prototyping is a group of techniques used to
quickly fabricate a scale model of a physical part or
assembly using three-dimensional computer aided
design (CAD) data. The first techniques for rapid pro-
totyping became available in the late 1980s and were
used to produce models and prototype parts. To-
day, they are used for a much wider range of applica-
tions and are even used to manufacture production-
quality parts in relatively small numbers. These rapid
prototyping techniques gave a new impulse in the re-
search to automate the building industry. For men-
tioned rapid prototyping processes can be divided
into two mainstream categories each with a number
of techniques:
1. Additive manufacturing process:
• Fused depositing modeling (=FDM)
• Stereo lithography (=SLA)
• Selective laser sintering (=SLS)

• Laminated object manufacturing (=LOM)

Fabrication - Volume 1 - eCAADe 32 | 455

 • Electron beam melting (=EBM) Stereo lithography
SLA is a process, which employs a vat of liquid ultra-
2. Subtractive manufacturing process: violet curable photopolymer "resin", and an ultravio-
let laser to build parts' layers one at a time. For each
• CNC milling: layer, the laser beam traces a cross-section of the part
pattern on the surface of the liquid resin. Exposure to
• Laser cutter the ultraviolet laser light cures and solidifies the pat-
tern traced on the resin and joins it to the layer below.
• Wire cutting Stereo lithography requires the use of support-
ing structures, which serve to attach the part to the
We will discuss briefly the above mentioned rapid
elevator platform, prevent deflection due to gravity
prototyping processes. It is just an explanation of the
and hold the cross sections in place so that they resist
process, in order to make a classification of these pro-
lateral pressure from the re-coater blade. Supports
cesses possible.
are generated automatically during the preparation
of 3D Computer Aided Design models for use on the
Fused depositing modeling
stereo lithography machine, although they may be
FDM works by laying down material in layers. A plas-
manipulated manually. Supports must be removed
tic filament or metal wire is unwound from a coil and
from the finished product manually.
supplies material to an extrusion nozzle, which can
turn the flow on and off. The nozzle is heated to
Selective laser sintering
melt the material and can be moved in both horizon-
SLS is a technique that uses a high power laser to
tal and vertical directions by a numerically controlled
fuse small particles of plastic, metal, ceramic, or
mechanism, directly controlled by a computer-aided
glass powders into a mass that has a desired three-
manufacturing (CAM) software package. The model
dimensional shape. The laser selectively fuses pow-
or part is produced by extruding small beads of ther-
dered material by scanning cross-sections generated
moplastic material to form layers as the material
from a 3-D digital description of the part (for exam-
hardens immediately after extrusion from the noz-
ple from a CAD file or scan data) on the surface of
zle. Stepper motors or servo motors are typically em-
a powder bed. After each cross-section is scanned,
ployed to move the extrusion head.
the powder bed is lowered by one layer thickness,
FDM begins with a software process, developed
a new layer of material is applied on top, and the
by Stratasys, which processes an STL file (stereo
process is repeated until the part is completed. Un-
lithography file format) in minutes, mathematically
like some other additive manufacturing processes,
slicing and orienting the model for the build process.
such as stereo lithography (SLA) and fused deposi-
If required, support structures are automatically gen-
tion modeling (FDM), SLS does not require support
erated. The machine dispenses two materials - one
structures due to the fact that the part being con-
for the model and one for a disposable support struc-
structed is surrounded by un-sintered powder at all
ture.
times.
The thermoplastics are liquefied and deposited
by an extrusion head, which follows a tool-path de-
Laminated object manufacturing
fined by the CAD file. The materials are deposited in
LOM is a process where layers of adhesive-coated pa-
layers as fine as 0.04 mm thick, and the part is built
per, plastic, or metal laminates are successively glued
from the bottom up - one layer at a time.
together and cut to shape with a knife or CNC or laser
cutter.

456 | eCAADe 32 - Fabrication - Volume 1

Electron beam melting
EBM is a type of additive manufacturing for metal
parts. It is often classified as a rapid manufactur-
ing method. The technology manufactures parts by
melting metal powder layer by layer with an electron
beam in a high vacuum. Unlike some metal sintering
techniques, the parts are fully dense, void-free, and
extremely strong.
This solid freeform fabrication method produces
fully dense metal parts directly from metal powder
with characteristics of the target material. The EBM
machine reads data from a 3D CAD model and lays
down successive layers of powdered material. These
layers are melted together utilizing a computer con-
trolled electron beam. In this way it builds up the
parts. The process takes place under vacuum, which
makes it suited to manufacture parts in reactive ma-
terials with a high affinity for oxygen, e.g. titanium.
The melted material is from a pure alloy in pow-
der form of the final material to be fabricated (no
filler). For that reason the electron beam technology
doesn't require additional thermal treatment to ob-
tain the full mechanical properties of the parts. The
EBM process operates at an elevated temperature,
typically between 700 and 1000 °C, producing parts
that are virtually free from residual stress, and elimi-
nating the need for heat treatment after the build.
Rapid manufacturing
Advances in the mentioned rapid prototyping tech-
niques have brought about the ability to use materi-
als that are appropriate for final manufacture. These
advances in material make it possible to directly man-
ufacturing finished components. Additive manufac-
turing is defined by ASTM ( = American Society for
Testing and Materials) as the "process of joining ma-
terials to make objects from 3D model data, usually
layer upon layer, as opposed to subtractive manu-
facturing methodologies such as traditional machin-
ing." Additive manufacturing techniques can be used
anywhere throughout the production cycle from pre-
production (= rapid prototyping) to full scale pro-
duction (= rapid manufacturing). The advantages of
rapid manufacturing lie in the relatively inexpensive
production of small numbers of parts, even if they are
so-called freeform. This ability to manufacture inex-
pensive full scale freeform components triggered off
research in the building industry to see if they could
adapt these rapid manufacturing techniques.
It is our opinion that these additive processes can
be subdivided into two main categories:
1. Where material is placed onto a layer of the
same material (layered);
2. Where layer-wise material is 'sintered' (sin-
tered).
Both of these categories and especially the 'fused
depositing modelling' en 'selective laser sintering'
methods are used in by the building industry for
rapid manufacturing processes. There are a few
promising real life techniques based on the above
mentioned categories, which are still under develop-
ment.
1. Layered.
• Contour Crafting (CC) is a additive layered
fabrication technology developed by Dr.
Behrokh Khoshnevis of the University of
Southern California. Contour Crafting is lay-
ered manufacturing system which uses poly-
mer, ceramic slurry, cement to build large
scale objects. The extrusion process forms the
smooth surface of the object by constraining
the extruded flow in the vertical and horizon-
tal direction to trowel surfaces. Attaching the
system to a gantry system (see figure 1) the
system is capable of "printing" houses.
• Kamermaker: DUS-architects have developed
a 3D printer which is able to print rooms with
a maximum size of 3 x 3 x 2.5 meter. Using
the same technique they are experimenting
to print a so-called "grachtenpand" in Amster-
dam (see figure 2). The Kamermaker is an up
scaled version of the Ultimaker, the 3D desk-
top printer. In contrast to the filament of the
Fabrication - Volume 1 - eCAADe 32 | 457
 Ultimaker, the Kamermaker uses a bio plastic Figure 1
granulate which enters an extruder. In the ex- Contour crafting
truder the granulate is heated and pressed to- with gantry
gether to a homogeneous liquid. This liquid
is brought to the printer head which extrudes
the melted material along a generated path.

• Concrete Printing Process is developed at

Loughborough University in the UK The con-
crete printing process comprises a printing
head which is digitally controlled by a CNC
machine to move in X, Y and Z directions
via three chain-driven tubular steel beams.
A material container is mounted on top of
the printing head and connected to a pump Figure 2
to convey the material to the printing noz- Grachtenpand
zle which is activated by the CNC machine.
The deposition material, concrete, has been
designed to be extruded through a nozzle
to build layer-by-layer structural components
(see figure 3).

2. Sintered.

• D-SHAPE a 3D printing technique, based on

stereo lithography and selective sintering, de-
veloped by the Italian Enirco Dini. D-shape
uses a a powder deposition process, which
is selectively hardened using a binder. Each
layer of build material is laid to the de-
sired thickness, compacted and then the noz- Figure 3
zles mounted on a gantry frame deposit the Concrete printing
binder where the part is to ne solid. Once a
part is complete it is then dug out of the loose
powder bed. Dini used the technique to build
the so-called "Radiolaria" a 1.6 m high archi-
tectural art work (see figure 4).

Next to the research of developing rapid manufactur-

ing techniques, there is still ongoing research on the
implementation of robotics in the building industry,
some research is still in its infancy state:

458 | eCAADe 32 - Fabrication - Volume 1

 • Assembly robotics
• Drones:
• Exo skeletons
For on-site use of rapid manufacturing techniques
systems based on selective sintering have a big dis-
advantage compare the layered based systems. Se-
lective sintering uses a powder bed, and the finished
object has to dug out.
Figure 4
Radiolaria
Each manufacturing technique is capable of de-
liver building components of significant size. But this
means only a step forward in the indoor manufac-
turing process to personalize prefabricated compo-
nents.
We can conclude that the research of layered
techniques is focused on monolithic constructions,
be it adobe or concrete. If we continue in this di-
rection, and developed an on-site 3D manufactur-
ing system which, delivers monolithic buildings, this
will raise the question whether clients will accepted
those building, or do they prefer houses with a tra-
ditional brick or wood cladding. If the clients prefer a
traditional look, the advantage of fast manufacturing
on site is counter acted by the slow process of giv-
ing the building a traditional look. Research had to
be done in 3D manufacturing systems with at least
2 materials (nozzles) or a combination of 3D printing
techniques and assembly robotics
RESEARCH QUESTION
Compared with the car-industry, the level of automa-
tion in the building industry is low. This low level of
automation is due to the fact that in the building in-
dustry each object is a unique object and each build-
ing site is a unique location.
The main reason of this high level of automation
in the car industry is a modern manufacturing con-
cept: Computer Integrated Manufacturing (CIM). The
CIM systems permit to balance the flexibility in the
product with the manufacturing productivity. This
relationship is one of the key factors of the success.
The so called platform concept is one of the
newest advances of the CIM system. It is based on
the use of a number of elements in various models.
The same platform design, engine, electronics, etc.
are used not only in different models of cars of the
same company but also in the cars of other compa-
nies. This concept reduces a vehicle cost and makes
the automobile companies more competitive (Bala-
guer, 2008).
It is our opinion that in building industry, now re-
search, development and use of a Building Informa-
tion Model (BIM) is growing; automation of the build-
ing process will become the next main research area.
With BIM technology, one or more accurate vir-
tual models of a building are constructed digitally.
They support design through its phases, allowing
better analysis and control than manual processes.
When completed, these computer generated models
Fabrication - Volume 1 - eCAADe 32 | 459
 contain precise geometry and data needed to sup-
port the construction, fabrication and procurement
activities through which the building is realized (East-
man 2011).
BIM has a big impact on the design side of the
building process and the way stakeholder's commu-
nicate with each other. It is our opinion that BIM also
will influence the other side of the process, the man-
ufacturing side, be it prefabrication or use of robotics
(in the broadest sense of the word). The main diffi-
culty on the manufacturing side is related to the na-
ture of the 'working' environment, which is highly
unstructured. Working in this environment involves
handling heavy objects, elements made with big tol-
erances, low-level standardization, and medium level
of industrialization and pre-fabrication. The develop-
ment of a systematized approach to construction us-
ing largely dry, prefabricated components delivered
just in time has advanced the degree of automation.
As a result there is heavy traffic between construction
plant and construction site.
Because of the global financial crisis the yearly
demand for housing is reduced. De Ridder states
"Large quantity is replaced by high quality, small
and personalized production units and a focus on
added value for the client" (de Ridder 2012). This pro-
cess will transform the building industry, especially
the construction techniques. As stated earlier the
automation in the building industry focuses nowa-
days on prefabrication in closed environments. Few
initiatives are realized in the personalisation of the
dwelling, see WoonConnect1. With the introduction
of 3D printing techniques in the building industry the
door is opened to personalized production units be
it made in indoor or outdoor environments. It is our
opinion that the use of 3D printing on-site is more in-
teresting.
In our research we focus on the Netherlands. The
climate in the Netherlands has a big impact on the
used materials in the building industry.
The process of on-site 3D manufacturing will
have big impact on the design and used materials of
dwellings. If we look at the traditional dwelling typol-
460 | eCAADe 32 - Fabrication - Volume 1
ogy the houses on a row ("rijtjes huis"), we see that
now nowadays the main load-bearing walls are per-
pendicular to the street. As a result the depth of the
dwelling is bigger than the width. With a on-site 3D
manufacturing system, where the nozzle is attached
to a gantry crane the main loadbearing walls will be
parallel to the street and as a result the width will be
bigger than the depth of the dwelling. This layout will
have a big impact on the layout of the new to build
neighborhood.
Goal of this research is to connect BIM to auto-
mated construction techniques. The research ques-
tion we like to answer: is it possible to use BIM to au-
tomate the construction process on-site and or in a
closed environment (pre-fabrication). This research
question can be subdivided in to two sub questions:
1. Which type of manufacturing is the most
promising in the building industry?
2. Is it possible to use BIM to navigate the "robot",
or put it differently is it possible to develop an
interface between BIM and the control unit of
an automated construction system (= ACS);
RESEARCH APPROACH
It is our opinion that real life 3D printing based on lay-
ering is the most feasible on-site 3D printing as man-
ufacturing technique. As explained earlier the nozzle
"draws" the walls layer by layer.
In order to build a interface we need to extract
the geometry out of the BIM. More precise we use the
IFC code as base for the transformation from the BIM
to an 2D layered pathway which the nozzle has to fol-
low.
From the BIM we need to extract the geometry
of the walls and the locations of the windows and
doors. This information will be used to calculate the
x,y,z- movement of the nozzle. The BIM will also be
used to generated the required support construction
which may or may-not have influence on the path the
nozzle has to traverse. But besides to retrieve the or-
thogonal movement, we have to calculate the short-
est, fastest and economic path the nozzle needs to
 travel. The deposition material constrains the path.
3D printing requires the concrete to have no slump
and harden fast. Therefore, the print path and mate-
rial properties are strongly linked together. If ∆t is the
time it takes to go from point a on layer i to point b
on layer i+1 and point b lies exact on top of point a,
we can conclude:
• If ∆t is too short the wall will collapses the wall
is too wet to bear its own load;
• if ∆t is too long the concrete of the previous
layer will no longer be wet and the structural
properties will be anisotropic.
So ∆t is related to the length of the path the noz-
zle has to go and the speed of the nozzle. The path
is related to the design of the building (placing of
the walls, windows and door). We can conclude that
there is a strong relationship between the shape of
the object (building or building component), the de-
mand (physical requirements), the deposition ma-
terial properties and both manufacturing technique
and print path (figure 5).
Figure 5
Relationship
between design,
material, demands
and printer
Research will start to see if the idea to print
with multi materials is feasible. With two materials,
(lightweight)concrete and an insulation material, ef-
ficient sandwich construction can be realised. The
second material can also be used to print the re-
quired support construction.
CONCLUSION
The research of BIM is nowadays mainly focused on
the office-side of the automation in the building in-
dustry. For a fully automated building industry we
need to connect BIM also to manufacturing side. This
research will show that it is possible to map the com-
plex 3D data in a BIM onto an orthogonal movement,
which can direct a nozzle in an economic way.
We will design a small dwelling and export it into
a BIM, from which we will extract our data to direct
the nozzle. To validate the system we develop a small
VR tool in which we mimic a manufacturing tool. The
VR tools must have the same degree of freedom as
the real-life manufacturing tool. Our tool will direct
the VR nozzle to draw the dwelling. With this simula-
tion we can conclude:
• If it is possible to use a BIM as control unit for
manufacturing;
• Will the nozzle travel the best feasible path.
REFERENCES
Balaguer, C and Abderrahim, M (eds) 2008, Trends in
Robotics and Automation in Construction, InTech,
http://www.intechopen.com/books/robotics_-
and_automation_in_construction/trends_in_-
robotics_and_automation_in_construction
Eastman, C and Teicholz, P 2011, BIM Handbook, a guide
to building information modeling, John Wiley & Sons,
inc
Khoshnevis, B 2004, 'Automated Construction by Con-
tour Crafting-Related Robotics and Information
Technologies', Automation in Construction, 13(1), pp.
5-19
Lim, S 2012, 'Developments in construction-scale ad-
ditive manufacturing process', Automation in Con-
struction, 21, pp. 262-268
de Ridder, H 2012, Legolisering van der bouw, Industrieel
maatwerk in een snel veranderende wereld, Maurits-
groen.mgmc, Amersfoort
Fabrication - Volume 1 - eCAADe 32 | 461
462 | eCAADe 32 - Fabrication - Volume 1
Material Swarm Articulations
New View Reciprocal Frame Canopy

Evangelos Pantazis1 , David Gerber2

1,2
University of Southern California
1,2
dilabwebsite.wix.com/dilab
1,2
{epantazi|dgerber}@usc.edu

Material Swarm Articulations, is an experiment in developing a multi-objective

optimization system that incorporates bottom up approaches for informing
architectural design. The paper presents an initial built project that demonstrates
the combination of a structural form finding method, with an agent based design
system through the digital fabrication processes. The objective of this research is
to develop a workflow combined with material and construction constraints that
has the potential to increase performance objectives while enabling geometric
complexity and design driven articulation of a traditional tectonic system. The
emphasis of the research at this stage is to take advantage of material properties
and assembly methods applied to a digital design and simulation workflow that
enables emergent patterns to influence the performance of the space.The paper
illustrates the research through a prototype of a self standing canopy structure in
1:1 scale. It presents results of the form finding, generative patterning, digital
fabrication affordances and sets and agenda for next steps in the use of
multi-agent systems for design purposes.

Keywords: Computational design, agent-based system, digital fabrication,

parametric design, reciprocal frames, form finding, multi-objective optimization,
multi-agent systems for design

INTRODUCTION AND RESEARCH OBJEC-
TIVES
In contemporary design discourse and practice the
rapid development and availability of computational
design tools are amplifying both conceptual and
technical capacity for manipulating complex geo-
metrical configurations and introducing informed
and articulated pioneering design possibilities. (Ger-
ber and Lin, 2013, Tsiliakos, 2012). Moreover com-
puterized fabrication technologies have enabled the
generation of unique mass customized parameter-
ized parts for almost the same cost of standard-
ized production issuing in the post fordist paradigm
(Scheurer et al., 2005). As a result, a great number
of architectural projects worldwide are being real-
ized partly following a "file-to-factory" pattern, as the
progress and availability of fabrication technologies
has allowed the manufacturing of complex geome-

Fabrication - Volume 1 - eCAADe 32 | 463

 tries to a large extent. As these new technologies
in conjunction with contemporary social phenom-
ena suggest a more participatory design model that
supports customization and the sharing of knowl-
edge and tools, the notion of the vernacular in ar-
chitecture, with local common rules and cultural val-
ues, is ripe for re-evaluation as to its design possibili-
ties (Morel, 2006). The implication of the post fordist
paradigm in this context is to enable a more diverse
set of design opportunities, styles and palettes while
furthering the projects towards meeting global chal-
lenges for performance criteria inclusive of cost, en-
ergy, comfort, as well as social suitability and pur-
pose.
While the introduction of digital fabrication in
building industry is rapidly narrowing the tolerances
between represented and realized form (Scheurer,
2005), recent investigations are looking into perfor-
mative approaches of form generation that for in-
stance seek to achieve a continuous digital chain
from design exploration to fabrication which not only
rationalizes structure but also evaluates the materi-
alization process that leads to fabrication. (Hensel
et al., 2010) Driven by increasing computational
capacity and availability of data and simulation,
some contemporary research is oriented towards in-
tegrating rationalization and optimization methods
during the first steps of computational form find-
ing.(Gerber, 2007) In terms of our use of computation,
our project's form can be defined as an interaction
between internal components and external forces
(Kwinter and Davidson, 2008). These internal compo-
nents are described as separate design drivers, which
in turn, are synthesized into an integrated computa-
tional design tool, which take into account external
forces, i.e. load cases and environmental parameters.
These integrated design and performance drivers are
described as agents with different motivations (i.e.
material, social) that interact with each other within
an environment, and exchange information that ul-
timately increase the complexity of the system as a
whole (Baharlou and Menges, 2013). In this con-
text, this work introduces a methodological frame-
464 | eCAADe 32 - Fabrication - Volume 1
work that explores the applicability of form finding
techniques and multi agent systems as a generative
bottom up strategy for material and locally respon-
sive design exploration coupled with traditional top-
down design strategies and analysis tools. A primary
objective at this stage of the research is to investigate
how the integration of generative systems with con-
straints relating to structural systems, environmental
criteria, and material and tectonic properties can lead
us to enriched and complex design outcomes, a com-
plexly curved canopy comprised of a novel reciprocal
curvilinear frame. A second objective of the research
is to reconsider architectural design strategies, by re-
vising a vernacular building system that of the recip-
rocal frame, and combining it with an agent based
model for informing the design process. A final ob-
jective is to test and validate the design methodol-
ogy that bridges the digital to the physical while ad-
dressing real world architectural constraints of time,
cost, and construction. Advances in building tech-
nology and structural systems are traditionally an in-
dicator of technological progress for a particular lo-
cality and culture.(Lin and Gerber, 2014) In the con-
text of the technological leap and mass customiza-
tion paradigm driven by access to computation we
also consider it essential that the re-examination of
traditional design, tectonic and structural method-
ologies be addressed. While the focus of the work
is first to prove out the methodological approach an
underlying agenda is also to ensure the approach is
supportive of cultural design diversity.
BACKGROUND AND CONTEXT OF RECIP-
ROCAL FRAMES.
The principle of reciprocity in structural design and
construction i.e. the use of load bearing elements
to compose a spatial configuration wherein they are
mutually supported one another has been known
since the antiquity (Baverel, 2000). Etymologically
reciprocity derives from the 'Latin' recirpocus, which
is composed by the two parts 'recus' meaning back-
wards and 'procus' translated as forwards. The word,
reciprocity, implies the practice of exchanging things
with others for mutual benefits. Such a definition
emphasizes the obliged stressed return of a certain
action. The development of reciprocal frames has
not had a linear history and the evidences of its
knowledge and application around the world seems
to be unrelated to one another. However a com-
mon point in the use of this system is the use of
timber as constructive material in both Occidental
and Oriental culture. Thus it is worth noting that it
was more of a practical and construction issue in Eu-
rope, for development of planar spanning configura-
tions, while in Asia, for more ceremonial realization of
three-dimensional structures.
The first reciprocal frame structures are traced
back to Chinese and Japanese religious architecture
in the 12th century, seen in the wood constructed
roof support systems of the mandala roof. In Eu-
rope, the concept of spanning distances longer than
the length of the available timber beams was the
main reason for the use and development of the
reciprocity principle (Larsen, 2008). During the 13th
century Villard de Honnecourt conceptualized in his
sketches roof support structures that were based on
this principle. Later in the 16th century Leonardo Da
vinci, who laid the foundation for a scientific study of
reciprocal structures, explored at least five different
spatial configurations based on the principle of reci-
procity, experimenting upon regular and non-regular
2D and 3D geometrical configurations. Sebastiano
Serlio addressed the problem of planar roof construc-
tion with short beams in his book on architecture
dated to 1556 . A comparable structure system made
of reciprocally supporting bar-shaped elements is
the Zollinger system which is mainly used in tim-
ber roof construction, where Friedrich Zollinger ob-
tained a patent for it in 1923 (Kohlhammer and Kot-
nik, 2011). It has been through the development of
sophisticated timber products-such as glulam trusses
and plywood-that produce long spanning structural
elements through adhesive technology that have led
to the replacement and lack of further development
of reciprocal frames and similar structure systems.
Currently, the principle of reciprocity continues
to stimulate the interest of designers and researchers
and it has again become a topic for academic re-
search. (Thomas Kohlhammer, 2010). Architectural
applications include the Mill Creek Public Housing by
Louis Kahn (1952-53), the Bunraku Puppet Theater by
Seiwa, Kazuhiro Ishii (1994) and the Pompidou Metz
museum by Shigeru Ban (2008) to name a few can be
found around the world. Moreover, a set of experi-
mental works related to structural, geometrical and
constructive issues of 'reciprocal' structures are ap-
pearing, such as the Forest Park structure by Shigeru
Bahn and ARUP AGU, the Serpentine pavilion by Cecil
Balmond and Alvaro Siza (2005), the H-edge pavilion
by Cecil Balmond and students from Penn University
(Pugnale et al., 2011) and the research pavilions by at
EPFL Lausanne (Nabaei and Weinand, 2011). In those
projects the reciprocal principle has been explored
by using different materials, element sections, joints
and planar or 3-dimensional configurations provid-
ing the fundamental evidence that adaptations of
this typology should still be further investigated in
for a diversity of architectural styles, patterns, per-
formance characteristics and finally local sensibilities.
Based on a specific type of reciprocal frame, we re-
examine the applicability of such a vernacular struc-
tural system by analyzing its functional and material
behavior. The research models a design system that
fosters design exploration by incorporating issues of
recyclability and material efficiency coupled with the
design and spatial comfort performance objectives.
This is partially achieved by implementing digital fab-
rication techniques but also through the incorpora-
tion of agent based design technologies enabling
emergent and intrinsic performance.
RECIPROCAL FRAMES DEFINED
A reciprocal frame is a structural system, formed by a
number of short bars that are connected using fric-
tion only. Most importantly the reciprocal frame can
span many times the length of the individual bars.
Reciprocal refers to the fact that such structures are
composed of a number elements (referred to also as
"short beams") that structurally interact through sim-
Fabrication - Volume 1 - eCAADe 32 | 465
 ple support binding in order to create more complex
structures of dimensions much greater than the sin-
gle elements from which they are composed. The ap-
plication of the reciprocity principle requires: a) the
presence of at least two elements allowing the gen-
eration of forced interactions; b) that each element of
the assembly must support and be supported by an-
other one; and c) that every supported element must
meet its support along the span and never at the ver-
tices in order to avoid the generation of a space grid
with pin joints. The space structures that conform
with the above requirements are called reciprocal
and are constituted of at least two interlaced linear
elements, where the final form is relative to a basic
component type, its material as well as the connec-
tion technology. The components can be identical or
non-identical but should follow a specific global tes-
sellation pattern. The joining of the components at
the node points can generally be carried out without
mechanical connections, solely by pressure and fric-
tion (Figure 1). To support the frictional forces sim-
ple connection techniques such as tying together or
notching of the elements can be used. In fact, in the
context of wood processing techniques it's recog-
nized that the complexity of the connection technol-
ogy, becomes an important feature in order to distin-
guish different structural propositions both from fi-
nancial and structural aspects (Nabaei and Weinand,
2011). From a structural point of view, each individ-
ual element in the system works as a single beam.
This beam lies at each of its edges either on another
component or if it forms the edge of the structure,
on the supports of the entire system. Each element
bares the supporting force of one of the neighboring
elements and optional dead loads or live loads. The
interest of such structural system lays in the fact that
the "global" form is determined by the "local" condi-
tion of the building elements.
Structural Form finding
The term, structural form finding is commonly used
to recall a morphogenetic process used to find the
relaxed form of grid-shells under a certain bound-
466 | eCAADe 32 - Fabrication - Volume 1
ary condition and load conditions. Many techniques
have been developed both in the academia as well
as in practice in order to determine geometries that
work in pure tension or compression under their
self-weight. What is common between various form
finding techniques is the fundamental property that
the final form is the direct result of the equilib-
rium and is influenced intensively by the material-
ity and the boundary condition applied (Nabaei and
Weinand, 2011). Many designers historically have ex-
perimented with hanging models and other physi-
cal methods for finding efficient structural from act-
ing in pure tension or compression. Antoni Gaudi
employed hanging models to solve structural issues
for projects like Casa Mila Pedrera, while Heinz Isler
and Frei Otto have done extensive work developing
highly accurate physical experiments for exploring
structural forms (Kilian and Ochsendorf, 2005). Al-
ternatively such forms are called funicular, which et-
ymologically derives from the Latin word funiculus,
meaning thin cord. This refers to the shape taken by
a thin cord acting in pure tension under a given set
of loads. A new generation of researchers within the
field of design and computing are developing com-
putational tools implementing the same principles
in digital environments (Van Mele et al., 2013). They
provide tools that often employ particle spring sys-
tems and aim to educate designers as to the effects
of forces on the form of structure as well as provide an
interactive form finding environment that was previ-
ously restricted to physical models.
Figure 1
Images from
models, built
examples, and
studies
implementing
different types of
reciprocal frames
(right to left)
 Particle-spring systems are based on lumped
masses, called particles, which are connected by lin-
ear elastic springs. Each particle in the system has
a position, a velocity and variable mass, as well as a
summarized vector for all forces acting on it (gravity,
loads etc). These forces can be calculated using mass
less connectors between particles, called springs (Kil-
ian and Ochsendorf, 2005). The magnitude of these
forces is based on the spring's offset from its rest
length, and thus supports can be added by restrain-
ing the displacement of specific particles. The sys-
tem initiates in a non-equilibrium state, and particles
move until they reach their equilibrium positions for
a given set of parameters. In this research we investi-
gate the possibilities of such a system for the intro-
duction of structural evaluation environments into
the design process from the early stage of our de-
sign workflow. (Figure 2). In particular we explore a
self-standing canopy shell structure by approximat-
ing a hanging model of networked catenary arches.
For that we developed a custom definition and al-
gorithm using a particle spring system through the
interface of an open source plugin (Kangaroo) that
operates within a commercial 3d software package
(Rhinoceros-Grasshopper).
Figure 2
Form finding
process with given
constraints (left)
and analysis of
generated surfaces
(right)
Agent Based Design Systems
An agent-based design system consists of large num-
ber of agents that follows simple local rules and in-
teracts within an environment (Gilbert, 2008). Criti-
cally, the research addresses the fact that functioning
in a similar fashion that goes from local relations to
emergent global phenomena the reciprocal principle
asks for the implementation of a bottom up design
approach. Such an approach maintains a generative
computational protocol to generate varying design
possibilities, while maintaining specific material con-
straints or local limitations. While most designers en-
gage with the information from simulations as a form
of validation or as an aid in the decision making pro-
cess, there are numerous precedents that have used
the information from simulation as a platform and a
driver in the generative process (Miranda and Coates,
2000, Ireland, 2009). As a first step towards such a de-
sign paradigm we developed an agent-based system
for introducing a perforation pattern that is informed
by an environmental analysis. Specifically, a flocking
algorithm is combined with data resulting from a sun
radiation analysis in order to investigate the perme-
ability of the selected structural system. As a first step
basic motion behavior mechanisms are introduced to
agents that spawn at specified points on the surface
of the components. Further behaviors (attraction, re-
pulsion) are triggered by values provided from en-
vironmental analysis, specifically sun radiation mea-
sured as perceived energy (kWh) / area (sqm). The
basic motion behavior mechanisms are introduced
according to the definition of Reynolds who cate-
gorizes them in three layers; action, selection, steer-
ing and locomotion (Reynolds, 1987). At issue for
our research are the mechanisms of these generative
agent-based systems how they are bound to material
properties, fabrication and construction constraints.
Material Properties and Joint Conditions
Timber and specifically plywood has been selected
as the preferred study material for this research for a
number of reasons. Firstly, timber is sustainable as a
construction material as it can be grown again and
is relatively inexpensive compared to other build-
ing materials. Secondly, although historically per-
ceived as a liability for a construction material, tim-
ber's fibrous nature can be considered an advan-
tage in terms of both the material's functional and
aesthetic characteristics. Additionally, in reciprocal
structures components may vary in size and plywood
Fabrication - Volume 1 - eCAADe 32 | 467
 offers the flexibility to be processed into different
shapes and lengths through our digital fabrication
workflows. The design and detailing of the compo-
nents' connections are intrinsic to each frame mem-
ber whereby the whole system and consequently the
mechanical behavior of joints, in terms of degrees
of freedom, friction, deformability and displacement
capability directly contribute to define the global be-
havior of the structure (Balfroid et al., 2011).The con-
nection points become an integral part of the build-
ing component, and their complex and or unique
geometry can be milled or directly fabricated into
the volume (See Figure 3). The wood joinery tasks,
historically manual and tedious, are now facilitated
by means of advanced CNC facilities. Of interest is
our ability to produce performance driven complex
forms that can emerge from the aggregation of a ba-
sic module type by introducing notches with variable
size and angle within the topology of the reciprocal
frame.
468 | eCAADe 32 - Fabrication - Volume 1
RESEARCH METHOD
The methodology of the research has been to work in
an incremental fashion with the overall objective of
being able to measure improvements in design out-
comes in formal terms, in design performance and
optimization terms, and finally in terms of a file to
factory process affordances, all the while working
with real world site, material, assembly, and cost con-
straints. For proving such an approach, a case study
of a self-standing shell structure at 1:1 scale is realized
and is sited on the rooftop of a cultural center in the
center of Athens, Greece. (See Figure 4) As stated pre-
viously one aim is to develop computational frame-
work for designing deployable structures that are in-
formed by environmental data and that adapt to eco-
nomical and material restrictions that also allow for
local, vernacular stylistic propensities to be discussed
in future research. A design to production workflow
was developed that proceeds through the following
steps: 1), form-finding of a canopy shell surface of
limited area with two support conditions through the Figure 3
use of a mesh relaxation algorithm; 2) discretization Diagram showing
of the generated surface through iso-curves that are the sliding
populated with interlocking building components, connection of two
where each component follows the principle of reci- modules and the
procity and is modeled with an associative paramet- assembly of a
ric geometry modeler; 3), informing the geometry whole arch
of the basic element based on the selected material
(thermo formed plywood) to optimize its transver-
sal section and render it resistant to the forces un-
der which the structure would be placed; 4), perform
environment analysis specifically solar radiation, and Figure 4
use this data for informing the probabilistic of a multi- Design Process
agent "swarm" system (MAS) controlling the perme- Framework
ability of the structure by introducing perforations
on the building components; 5), examine the funda-
mental mechanical properties of a single arch using
Finite Element Method (FEM) while considering the
non-linear contact boundary condition, this is per-
formed with the static behavior of the structure un-
der the self-weight load case; and 6), prototype at 1:1
scale using a new type of thermoform-able panels of
plywood, where we optimize material usage by using
the minimum allowable thickness and by grouping
the components to be fabricated in panels with non
standard dimensions that correspond to available ve-
neer width rather than fixed plywood panel dimen-
sions.
As a first step the design of the experiment(s) be-
gins with setting the boundary and support condi-
tions of the structure. By implementing a mesh relax-
ation algorithm a series of surfaces is generated with
the same support conditions. Based on a curvature
analysis, a selection of these initial designed surfaces
are parametrically discretized, their iso-curves are ex-
tracted and populated with a basic component type
in an automated and parametric fashion. The third
step is to inform the component design based on the
selected material and the condition of the neighbor-
ing elements. Furthermore each component is ana-
lyzed for the sun radiation it receives for a given pe-
riod and a value (kwh/sqm) is given for each mesh
point. The weighted values are passed on to the
multi-agent design system that controls the perme-
ability of the structure, by drawing on the surface
the reaction of the agent swarm as paths for areas
to protect or remove material. Agents are created at
the mesh points of each component and the associ-
ated value at each points influence the agent's flock-
ing behavior. The path of the agents is exported and
translated in a perforation pattern via dashed slits
for an improved solar radiation protection and com-
fort beneath the structure. The last step includes the
prototyping at 1:1 scale of the whole structure out
of curved thermo-formable plywood panels that are
CNC milled flat and thermoformed subsequently.
COMPONENT TO PAVILION FORM RELA-
TIONSHIP
Our research develops a basic module of a two el-
ement reciprocal frame structure that is parametri-
cally defined and investigated. Instead of using tradi-
tional linear or planar elements, mutually supported
curved panels are instead used, for reasons of both
formal design aspiration but as well as for creating
a thickened pillow shell providing for a more com-
plex shading condition. The self-standing reciprocal
frame structure presented in this study shows an ex-
ample for the design practice when the final form
is driven by the connection technology, where a re-
laxed modular global form is discretized by means
of mutually supported panels. The proposed slide
connection scheme inspires a new family of recipro-
cal frames, where instead of linear members (beam
or bar), folded or formed members are mutually sup-
ported. The connection between the building com-
ponents is integrated as a notch with specified an-
gle within the geometry of members, unlike the tra-
ditional reciprocal frame system where the connec-
tor members are regular. A V-form module of given
angle is fabricated through thermoforming of ply-
wood panels and is then spatially multiplied using
consecutive rotations and translations that follow the
tangent vectors of a discretized catenary arch. The
structure can be decomposed into three principal
module types each consisting of curved panels (con-
vex and concave) with locally specific angles in re-
lationship to their generatively form found neigh-
bors. These modules are then interlocked sequen-
tially along their uniquely milled U shape cuts, to
form an arch. The inter-panel stability is provided by
rigidity of slide connection and axial contact of re-
ciprocal panels. The structural performance of the
whole structure improves when more than 2 arches
are connected together. This is due to the fact that
they act like a truss, with only axial compressive and
tensile forces. Bending moments and shear forces
minimize when in network arches (Tveit, 1987).
A MULTI-AGENT SYSTEM FOR MATERIAL
FABRICATION
The workflow has been built on a series of open
source simulation environments, an associative para-
metric design and visual programming environment
Grasshopper/Rhinoceros and a series of plug-ins for
integrating performance simulations and generating
data that can then be read by a custom agent based
swarm algorithm written in the Processing 2.0 envi-
ronment. The specific Grasshopper definition incor-
Fabrication - Volume 1 - eCAADe 32 | 469
 porates Kangaroo, a physics simulation solver used duced. The design and programming of the structure
for running a mesh relaxation algorithm on a shell was at first optimized for discretizing an irregular sur-
structure with given boundary conditions. The gen- face in curved panels using a single pressing mold.
erated surfaces are parametrically discretized in iso- The size of the structure was parameterized based
curves, which are further divided in linear segments. on a maximum amount (area) of material, that of 54
Interlocking components are placed sequentially in sqm. The whole production workflow included; 1),
pairs on the division points of each iso-curve and material processing; 2), manufacturing of the press-
are oriented parallel to the tangent vectors. The ing mold; 3), CNC cutting of the components; 4), ther-
panels are further analysed with Ladybug, another moforming and post processing of the components;
Grasshopper plugin that runs a Radiance based sim- and 5), final assembly and erection on site (Figure 6,
ulation. All components are numbered and unrolled 7).
in flat panels with all the perforation and notch lines
projected. Thus material calculations can be done re- Figure 5
lating the covered area of the shell surface. Diagram illustrating
The simulation data is exported as a text file that the integration of
is passed to the custom flocking algorithm in the Pro- the agent based
cessing environment in 3D where the agents read the model with the
vertices of the component surfaces and coupled data enviromental
values. The agents are spawned and programmed analysis for the
to make movement and trajectory decisions based generation of
on the local information including the intensity val- perforations
ues from the sun simulation, proximity to neighbors,
and trails left by other agents. Each agent has the Figure 6
capacity to read the data from the simulation, which Diagram showing
is paired with its corresponding point in a mesh ob- the whole
ject as well as data related to its neighbors' and con- fabrication process
straints as trajectories in time. Thus, the agent's en- in steps
vironment is a collection of points to which it is con-
strained, and each point is assigned an intensity score
based on the data from the simulation. The agent's
trajectories then become a generative geometry for
material organization and material re-organization
which happens in a collective recursion. The agent-
based trails are exported again as a text file into the Figure 7
Rhinoceros / Grasshopper scripts to be incorporate as Top view with all
another layer of information that controls the perme- the modules
ability of the panels through a process of CNC driven arrayed and
material erosion through milling. (See Figure 5) perspective view of
the assembled
structure
FABRICATION PROCESS AND 1:1 PROTO-
TYPING
In order to test the structural feasibility of such a de-
sign approach a prototype at 1 to 1 scale was pro-

470 | eCAADe 32 - Fabrication - Volume 1

 Of great significance for material efficiency was custom panels as the film can be cut according to
the creation of plywood panels in sizes that corre- the available veneers for material. It was empiri-
sponded to the available veneer sizes and not the cally discovered that the film adhesive and thermo-
standard dimensions. This saved material as plies forming process produced panels with more elas-
were used "as is" and reduced material processing tic behavior than the traditional gluing technique
time as multiple plies needed not to be stiched to- which uses phenol-formaldehyde based wood glue
gether in larger panels before being glued to form and high voltage electricity for curing the glue while
the final plywood panel that was CNC milled. More- the panel is being pressed. This facilitated the assem-
over the module's geometry was optimized in order bly process as it allowed the panels to slide in place
to be fabricated by a single mold for reducing pro- more easily. Given the common cost and material
duction costs.. The panels were thermo-glued at first constraints locally but arguably globally for projects
flat and were milled afterwards using a 5-axis CNC like these, the project sought to be highly resourceful
machine. The cut pieces were then reheated and with material usage aided by computational design
pressed consecutively in the mold to take the final tools and CNC manufacturing. The thin self-standing
shape. A master parametric 3d model was developed shell structure with a footprint of 24 sqm required 54
in Grasshoper-Rhinoceros that generated all the cut- sqm (0.4 cubic meters) of thermo-formed birch ply-
ting files for the 148 components in a file-to-factory wood. The total scrap wood did not exceed 5 sqm
process where custom routines were developed to and was used to create the seating below the struc-
give each component their precisely calculated slots ture.
for the sliding joints, all in gradually shifting posi-
Figure 8 tions and variable angles in order for the pavilion to
Photo from the achieve its irregular funicular form. All components
finished structure were uniquely labeled and numbered in order to fa-
on site cilitate assembling and dismantling of the structure
without the need of experts or detailed drawings.
The plywood components were manufactured in
Greece at a factory outside Athens and transported
to the site. On site assembly of the pavilion was com-
pleted in two days, by a group of 5 non expert work-
ers.A selection of photos illustrating the project is
Figure 9 shown in (See Figure 8,9)
Detail photos from
the finished Material behavior and performance
structure on site Through our approach to the reciprocal frame we
investigated the potential of using curved plywood
components instead of planar ones. For that reason
we selected an innovative material called UPM Grada,
which is specifically designed for the manufacturing
of form pressed plywood panels. UPM Grada uses the
application of an adhesive film which allows the ply-
wood to be thermo-formed after production by ap-
plying pressure through a custom made mold. The
UPM Grada technology allows for the fabrication of

Fabrication - Volume 1 - eCAADe 32 | 471

 In further comparison to the precedent work
(Weinand, 2011) we have attempted to define bench-
marks and measures for the affordances of our ap-
proach. By adapting the principle of reciprocity and
a specific family of reciprocal frames with mutually
supported curved panels we benchmarked the con-
struction of a maximum 600 cm span out of 45 cm
long, 7.5mm thick interlocking panels that require
no additional joinery. Previous structure, following
the same principle, realized by prof. Y. Weinand,
S. Nabaei and students in Lausanne had achieved a
span of 740cm with 21 mm thick panels (Nabaei and
Weinand, 2011). While the project is clearly indica-
tive and statistically descriptive of improved span to
material depth we expect to continue to investigate
the project for further efficiencies of span, material
usage, and applicability to complex curvature and
other design vernaculars. What is also a major thrust
of our future investigation is that of the impact of the
swarm generated pattering and its influence on the
structural as well as highly coupled comfort dynam-
ics for complexly shaped canopy shell structures.
CONCLUSION AND DISCUSSION
The paper presents our very first built experiment
which combines research into the development of
a workflow that combines the possibilities for form
finding through both bottom up strategies through
the use of materially, structurally, and environemn-
tally informed swarm agents with that of more com-
mon structural form finding. The canopy also, from
inception through to its use, incorporated intuited
and them empirical real world material, assembly,
cost, and human constraints that informed the de-
velopment of the system. While we have yet to mea-
sure the results of all of our research objectives, those
of material optimization, multi-agent based design
objectives, structural performance and fabrication
times and efficiencies are the most concrete so far.
As stated we will continue to investigate these fur-
ther through more empiricism but as well will further
develop the use of the multi-agent system for its ar-
ticulation possibilities and for their ability to collec-
472 | eCAADe 32 - Fabrication - Volume 1
tively negotiate even more optimal material and en-
vironmental conditions for the reciprocal frame sys-
tems. Reconfigurable structures and resource saving,
mutually supported elements via reciprocal frames
will be further investigated in concert with digital
form finding and articulation. Clearly we can ob-
serve if the internal forces are known, each element in
such structure can be adjusted to its local stress and
therefore an optimized material consumption can be
achieved. Computation and digital Fabrication pro-
cesses make such approaches easily realizable and
economically viable.
The paper presents research into Material Swarm
Articulations, a multi-objective design systems that
incorporate a bottom up MAS approach for architec-
ture through the combination of form finding tech-
niques with agent based design system and digital
fabrication processes. The aim of this work was to
develop a workflow combined with a fabrication pro-
tocol that has the potential to bridge design inten-
tions with a structural system and realistic parame-
ters (budget) with material constraints. The focus
here was to take advantage of intrinsic features of
the material properties and assembly methods and
apply simple local rules in a manner that enables
emergent phenomena to arise. In such an approach
the design intention is codified by these local and
global constraints and their relations rather than by
a prevailing "architectural gesture." The design opti-
mization becomes an iterative process, where both
solution and starting condition are constantly oscil-
lating towards an equilibrium defined by ,multiple
performance criteria in response to the given topog-
raphy and the user's design intentions and adjust-
ments. This study presents an initial analysis of the
generative and automation possibilities using a dig-
ital fabrication workflow driven by a simulated and
recursively informed material swarm aggregation for
a novel reciprocal frame design.
ACKNOWLEDGEMENTS
We would like to thank and acknowledge Xenakis
Curved Plywood S.A, Romantso Cultural Center, Ikea
Stiftung in Switzerland and Onassis Foundation for
their support of the project and of the authors; and
to the team of helpers Iason Pantazis, Constanti-
nos Schoinas, Anastasios Spyridwnos, Dimitris Char-
itatos, Thanasis Demiris, Iraklis Kassimis, Dionysis
Dikefalos, Viktoras Gogas and Rodrigo Shiordia for
their tireless support and contributions. This material
is in part based upon work supported by the National
Science Foundation under Grant No. 1231001. Any
opinions, findings, and conclusions or recommenda-
tions expressed in this material are those of the au-
thor(s) and do not necessarily reflect the views of the
National Science Foundation.
REFERENCES
BAHARLOU, E. and MENGES,, A. 2013 'Generative Agent-
Based Design Computation', eCAADe 2013
Balfroid, N, Kirkegaard, N and Henning, P 2011 'Robust-
ness of Long Span Reciprocal Timber Structures', The
35th Annual Symposium of the IABSE 2011, the 52nd
Annual Symposium of the IASS 2011 and incorporating
the 6th International Conference on Space Structures
BAVEREL, O. S. 2000, NEXORADES: A FAMILY OF INTERWO-
VEN SPACE STRUCTURES, Ph.D. Thesis, UNIVERSITY
OF SURREY
Cabrinha, M 2008 'Gridshell Tectonics: Material Values
Digital Parameters', Proceedings of the 28th Annual
Conference of the Association for Computer Aided De-
sign in Architecture (ACADIA)
Gilbert, N 2008, Agent-based models, Sage
Ireland, T 2009 'Emergent space diagrams: the appli-
cation of swarm intelligence to the problem of au-
tomatic plan generation', Joining Languages, Cul-
tures and Visions: Proceedings of the 13th Interna-
tional CAAD Futures Conference
Kilian, A and Ochsendorf, J 2005, 'Particle-spring systems
for structural form finding', INTERNATIONAL ASSOCI-
ATION FOR SHELL AND SPATIAL STRUCTURES, 148, p.
77
Kohlhammer, T and Kotnik, T 2011, 'Systemic behaviour
of plane reciprocal frame structures', Structural Engi-
neering International, 21, pp. 80-86
Kwinter, S and Davidson, C 2008, Far from equilibrium: es-
says on technology and design culture, ACTA Press
Larsen, O 2008, Reciprocal frame architecture, Routledge
Van Mele, T, De Laet, L, Veenendaal, D, Mollaert, M and
Block, P 2013, 'Shaping Tension Structures with Ac-
tively Bent Linear Elements', International Journal of
Space Structures, 28, pp. 127-136
Miranda, P and Coates, P 2000 'Swarm modelling. the
use of Swarm Intelligence to generate architectural
form', Proceedings of the 3rd Generative Art Confer-
ence
Nabaei, S. S. and Weinand, Y 2011, 'Geometrical de-
scription and structural analysis of a modular timber
structure', International Journal of Space Structures,
26, pp. 321-330
Pugnale, A, Parigi, D, Kirkegaard, P H and Sassone, M
2011 'The principle of structural reciprocity: history,
properties and design issues', The 35th Annual Sym-
posium of the IABSE 2011, the 52nd Annual Sympo-
sium of the IASS 2011 and incorporating the 6th Inter-
national Conference on Space Structures
Reynolds, C. W. 1987 'Flocks, herds and schools: A dis-
tributed behavioral model', ACM SIGGRAPH Com-
puter Graphics, pp. 25-34
Scheurer, F 2005, 'Turning the design process downside-
up', in Martens, B and Brown, A (eds) 2005, Computer
Aided Architectural Design Futures, Springer
Scheurer, F, Schindler, C and Braach, M 2005 'From de-
sign to production: Three complex structures mate-
rialised in wood', 6th International Conference Gener-
ative Art
Sistaninia, M, Hudert, M, Humbert, L and Weinand, Y
2013, 'Experimental and numerical study on struc-
tural behavior of a single timber Textile Module', En-
gineering Structures, 46, pp. 557-568
Tsiliakos, M 2012 'Swarm Materiality: A multi-agent ap-
proach to stress driven material organization', Digi-
tal Physicality: Proceedings of the 30th eCAADe Con-
ference, pp. 301-309
Tveit, P 1987, 'Considerations for Design of Network
Arches', Journal of Structural Engineering, 113, pp.
2189-2207
Weinand, Y 2011 'Innovative Timber Constructions',
IABSE-IASS 2011
Fabrication - Volume 1 - eCAADe 32 | 473
474 | eCAADe 32 - Fabrication - Volume 1
Digital Fabrication Technology in Concrete Architecture
Pedro Filipe Martins1 , José Pedro Sousa2
1,2
Faculdade de Arquitectura da Universidade do Porto
1
www.faup.up.pt
1,2
{pcarvalho|jsousa}@arq.up.pt

Technological innovation has been an important driving force in architecture,

enabling and inspiring architects and engineers by giving them new tools for
solving existing problems. In the last two decades, the exploration of digital
design and fabrication technologies has stimulated the development of a variety
of interests and strategies to materialize increasingly complex and customized
solutions in architecture, with traditional building materials.Reinforced concrete
is the most widely used material in the building industry today and throughout its
history has been the subject of vast research into its performance as a
construction material and its tectonic potential in architecture. As such, the
introduction of digital fabrication processes in concrete construction represents
the biggest prospect for renovation of our built environment and at the same time,
presents particular difficulties and opportunities, which are now being addressed.
In an effort to investigate the alternative design and material possibilities in
concrete emerging from the use of digital fabrication technologies in architecture,
this paper proposes a focused view of digital fabrication applied to concrete
construction with two areas of research. By framing the research in the context of
reference works in concrete architecture of the 20th century, this paper describes
and illustrates taxonomy of existing and possible types of integration of digital
fabrication technologies in concrete architecture in the realms of Practice and
Research.This characterization allows the authors to frame the relation between
material, technology and architecture in different environments regarding the
same material, extracting a clear image of existing processes, their potential and
shortcomings, as well as expectations for future developments.

Keywords: Digital Fabrication, Concrete, CAM, Robotics, Sustainability

INTRODUCTION new architectural possibilities.

Since the rediscovery of roman concrete and the in- Reinforced concrete is a compound material
vention of reinforced concrete in the nineteenth cen- composed of two elements: the cementitious matrix
tury, concrete has been the focus of continuous tech- and the reinforcement. It can thus be considered a
nological advances, driving with it an expectation for hybrid material, which combines the properties of its

Fabrication - Volume 1 - eCAADe 32 | 475

 constituents to achieve a different set of character-
istics. This hybrid nature, which is further character-
ized by a duality of states (i.e. evolving from liquid
to solid) through a mould, has been at the core of
contemporary discussion about the intrinsic plastic-
ity of this material in architecture. For instance, Forty
(2006) suggests the idea of understanding concrete
as a process rather than solely as a material. Indeed,
its rich and diverse history of technological innova-
tions through research and practice has prevented
concrete to have a fixed underlying traditional and
accepted aesthetics.
This multi-faceted background of a material has
provided a fertile ground for exploration for new
materialities. This has become increasingly relevant
in the last decades, with the dissemination of the
emerging digital design and manufacturing tech-
nologies (CAD/CAM). Together, they have inspired a
growing interest in rethinking concrete, while help-
ing and builders to bridge the gap between design
and construction. In this context, the subject of this
paper is to investigate and outline the emergent op-
portunities for concrete architecture that have been
supported by the use of digital fabrication processes.
HISTORICAL PERSPECTIVE
As a starting point it is important to contextualize this
subject by examining some innovative examples in
the history of concrete architecture. Past architects
and engineers have produced relevant works in con-
crete, which have been a source of inspiration for
those who are trying to innovate the application of
this material in architecture.
Thus, the following themes were selected to set
a framework to analyse the reference works of con-
crete architecture:
1. Complex Forms
2. Surface Textures
3. Material Composition
While form and surface are the basis of the majority of
the examples surveyed, changes in the material com-
476 | eCAADe 32 - Fabrication - Volume 1
position of concrete have been used to improve its
functional properties and to renew its aesthetic qual-
ities.
In this context, this research contextualizes these
traditional recurring themes to use them as a basis to
analyse and characterize the recent 'digital' works.
Complex Forms. The first reinforced concrete proto-
types, the concrete boat and conical vases of Lam-
bot and Monier in 1848-50, exposed at an early age
the opportunity of materializing complex forms in
this material. Decades later, several geometrically ap-
pealing images of American silos and industrial build-
ings in concrete, were published by Walter Gropius in
1913 and re-used in Le Corbusier's manifesto of 1925,
illustrating this early link between concrete and ex-
pressive forms in architecture.
Reinforced concrete made its way into architec-
ture through the development of innovative con-
struction systems (e.g. Hennebique system), which
opened new spatial possibilities. However, it was its
plastic properties of conforming to any shape that
made it a prime material for exciting formal explo-
rations, which were subsequently adopted by mod-
ernist architects as the medium for a new architec-
ture.
In the first decades of the 20th century, archi-
tects and engineers seek to control the process of
materializing concrete, testing the limits of its ma-
terial, formal and aesthetic possibilities. Among the
different types of buildings, the exploration of con-
crete forms was particularly relevant in the thin shell
constructions. Pioneered in early 30s by architects
and engineers like Eduardo Torroja, Felix Candela and
Miguel Fisac, this type of constructions showcased
the plasticity of concrete in assuming self-supporting
and continuous curved surfaces of reduced thick-
ness. The Zarzuela Hippodrome (1935), the Palmira
Chapel (1959) or the Jorba Laboratories (1965) are
some clear examples of that. In most cases, the com-
plex geometries exhibited in these works were based
on exploring and combining ruled surface geome-
tries. One of the most notable cases is Felix Can-
dela who spent most of his career working with hy-
perbolic paraboloid surfaces, which were carefully
combined and adapted to a large set of architectural
programmes (Garlock 2008). One particular aspect
of these constructions was the fact that their form-
work was built with rigid wooden planks following
the same geometrical creation rules. This defined a
direct relation between material behaviour, form and
construction technology, which has become increas-
ingly relevant in the contemporary concept of digital
materiality (Gramazio and Kohler 2008).
Surface Texture. Although the previous works had
implicit qualities regarding the expression of their
skin, they were mainly concerned with the formal ex-
pression of concrete. In the second half of the 20th
century, architects started to insist more with the sur-
face finishing of their concrete constructions. Le Cor-
busier's "béton brut" experiments, in which concrete
cast with rough wood boards was left bare, prompted
an exploration into the possible aesthetic properties
of exposed concrete. American architects, like Louis
Khan Paul Rudolph or I.M. Pei, developed through-
out the 60s numerous works where the exposed con-
crete skin was treated to achieve expressive quali-
ties. In works such as the Yale Art Gallery (1953), the
Art and Architecture Building (1963), the Everson Mu-
seum (1968) and the Jonas Salk Institute (1965) sur-
faces textures ranged from rough to smooth features.
In these cases, texture resulted from the formwork
design or by mechanical or manual post processing
techniques (Legault 2005), achieving varying visual
and tactile sensations with the same material by ma-
nipulating the properties of its finished surface. An-
other relevant example consisted in the experiments
of Miguel Fisac with flexible formwork, where the in-
tegration of new technologies within the process of
concrete creation introduced deep changes in the
textural and visual properties.
From this scenario, one can extract the relevance
of controlling the production process of concrete as
a medium for achieving expressive surface textures
and infer that changes to this process can impart new
material qualities.
Composition. Although, when referring to the
works discussed before, traditional literature focuses
mostly on shape and surface texture, much of those
defining features were dependant on a careful atten-
tion to its mixture composition. Going back to the
notion of reinforced concrete as a hybrid material,
where steel was incorporated into a matrix of cement,
sand and gravel, it is possible to understand the op-
portunity and potential of changing the composition
of concrete. For instance, most of the works of Pier
Luigi Nervi, like the Sports Palace in Rome (1958),
featured a structural and formal complexity that was
made possible by his research into the composition
(i.e. reinforcement) of concrete and the subsequent
development of ferroconcrete (Iori and Poretti 2005).
Moving from the structural to the aesthetic proper-
ties, the composition of concrete has also been ex-
plored to achieve different visual expressions. Forty
(2012) mentions Perret's careful selection of aggre-
gates by their colour to feature in exposed concrete,
while Legault (2006) mentions the case of Pei's use of
selected mineral additives to achieve white or earth
coloured surfaces in his projects.
The works described in this survey, suggest rel-
evant clues to understand the possibilities that con-
crete construction can follow through digital fabri-
cation. Nevertheless, they also reveal some of the
greatest problems of materializing expressive works
in concrete. As it was already discussed, formwork
played a major role in the characteristics of the ma-
terial and in the building process. Formwork con-
struction conveyed a large part of expenditures in
materials, labour and assembly time. Moreover, the
particular types of formwork traditionally available
(timber and, later, steel) introduce significant geo-
metrical constraints on possible forms and undiffer-
entiated repetition in most concrete constructions,
which can only be partly overcome with additional
building time and costs. Also, relevant efforts were
put into achieving desired surface finishes through
the careful control of the formwork materials and
building processes, as well as by means of controlling
the mixture of its components. However, these in-
Fabrication - Volume 1 - eCAADe 32 | 477
 terventions were still tied to the standardization con-
straints of the industry.
In this context, the recent use of digital fabri-
cation technologies have started to tackle some of
these constraints, proposing new ways of materializ-
ing concrete and revisiting and further exploring tra-
ditional concepts of form, texture and composition in
concrete architecture.
THE INTEGRATION OF DIGITAL FABRICA-
TION IN CONCRETE ARCHITECTURE
To examine the integration of digital fabrication tech-
nologies in the production of concrete architecture
this paper considers the realms of practice and re-
search. While the first side testifies the real impact
in the built environment of using such technologies,
the second one indicates the tendencies that can
shape the next innovative applications of concrete in
architecture.
Analysis methodology. The first step of the analy-
sis consisted in surveying existing literature on rele-
vant built works in which digital fabrication played a
key role in the materialization of concrete elements.
A set of key examples was then selected to illustrate
the spectrum of similar and diverse approaches that
could sustain the proposal of a taxonomy of the dig-
ital fabrication strategies.
To construct this analysis, a structure based on
three consecutive categories was defined to organize
the studied works:
1. Intervention mode
2. Digital Fabrication Technology
3. Theme
478 | eCAADe 32 - Fabrication - Volume 1
Facing the existing landscape of digital fabrica-
tion technologies and the dual state of the material
(i.e. liquid and solid), the (1) Intervention Mode topic
comprises two possibilities - direct and indirect. This
division separates the cases where digital fabrication
technologies are directly employed over the building
components in concrete, from those where they are
applied to other elements, which will then, indirectly,
affect the production of concrete elements. In the lat-
ter case, a further distinction is considered by consid-
ering the possibility of intervening in the formwork or
in the reinforcement. Furthermore, because the pro-
duction of formwork is the focus of the majority of the
surveyed works, this topic had to be subdivided tak-
ing into account different strategies based on their
materiality: rigid, flexible or dynamic.
The (2) Digital Fabrication topic was divided into
subtractive (Table 1), additive and formative cat-
egories, as proposed by Branko Kolarevic (2003).
These topics were then subdivided into the specific
fabrication technologies in order to illustrate the dif-
ferences within the same family of fabrication pro-
cesses.
Finally, specific (3) Themes were identified to
each example according to three key concepts in the
materiality of concrete architecture: Form, Surface
and Composition. Given that these three goals, or in-
terests, can be found in the relevant historical exam-
ples, critical comparisons between past and current
works can be developed.
Examining the developments in Practice
The use of digital fabrication technologies became,
in the mid 1990's, a indirect solution to the prob-
lem of building complex geometry in concrete, by
enabling the production of custom moulds. The ca-
Table 1
Digital Fabrication
in Concrete
Architecture –
Developments in
Practice
pabilities of CNC (Computer Numerical Control) sub-
tractive processes were employed in the production
of complex formwork in EPS (Expanded Polystyrene
Foam). This material proved to be capable of with-
standing the forces of the casting process and at the
same time, was sufficiently soft to be quickly and ac-
curately milled into a complex surface, reducing fab-
rication costs. The Table 1 summarizes the analysis
about the use of digital fabrication technologies in
the production of concrete elements in architectural
practice.
Early examples of this production strategy can
be found in the Neue Zolhoff project in Dussel-
dorf, Germany (Gehry Partners, 2000) and the Big
Belt House in Montana, USA (William Massie, 2000).
In both cases, digital fabrication technologies were
used in the production of precast structural compo-
nents (beams and walls) with geometrical features
difficult to achieve with traditional formwork sys-
tems. This condition allowed the architects to further
explore the formal plasticity of concrete.
As seen in the Table, CNC milling of homoge-
neous materials dominates the production in prac-
tice of complex geometries in concrete. Recently,
these processes were developed and scaled up to
the construction site, overcoming the need for pre-
fabrication. The Science Center in Wolfsburg, Ger-
many (Zaha Hadid Architects, 2005) and The Spencer
Dock Bridge in Dublin, Ireland (Future Systems /
Amanda Levete, 2008) are exemplar in showing the
intrinsic relation between the design geometry and
the fabrication possibilities. In the Phaeno Science
Center, a careful analysis and subdivision of the de-
signed form into single-curved and doubly-curved
parts was developed in order to be, respectively, con-
structed with standard wooden formwork and with
milled EPS formwork (Kara 2008). This feature had
an immediate impact on the surface finishing qual-
ities, given that one can clearly identify the wooden
textures for the ruled sections of the cones and the
smooth surface in the doubly curved transition areas.
The Spencer Dock Bridge goes further with the ma-
terialization of doubly curved surfaces. Using milled
formwork in high-density EPS, the building was built
in a continuous way by pouring concrete onsite.
Another application of CNC milling of EPS has
been for the production of sculptural voids in con-
crete surfaces, as was the case of the O-14 Tower
(Reiser+Umemoto 2007). At the same time, other
recent built projects employed the same technolo-
gies for the production of non-standard formwork
in wood. The Mercedes Benz Museum (UNStudio,
2006) and the Rolex Learning Center (SANAA, 2010),
both featured curved surfaces in concrete that were
achieved through the precise CNC cutting of pla-
nar plywood panels to compose the desired double-
curved geometries.
Although in the mentioned works it is clear that
the application of digital fabrication in the produc-
tion of formwork has enabled exciting formal explo-
rations in concrete, it is also revealing that new prob-
lems emerge from this condition.
One issue regards the surface smoothness.
When comparing these processes to traditional
wood or steel formwork in standard construction,
smooth surfaces are much harder to achieve through
the materials commonly used in milling than with
the traditional standard sheet materials. Moreover,
in terms of formal freedom, the CNC milling process
has its own geometrical constraints, depending on
the number of available movement axis and the size
of formwork stock material.
Problems with sustainability are also inherent to
subtractive processes. The gradual removal of ma-
terial layers produces large amounts of waste mate-
rial that are not easily reusable. For instance, in the
specific case of milled EPS, the need to apply non-
recyclable coating agents to achieve satisfactory sur-
face finishes makes the recyclable process necessary
quite unviable. Facing the scale and complexity of
architectural construction, research in digital fabri-
cation technologies and materials engineering is re-
quired to overcome these problems (Table 2).
Fabrication - Volume 1 - eCAADe 32 | 479
 Table 2
Digital Fabrication
in Concrete
Architecture –
Developments in
Research

Examining the Developments in Research Figure 1

The academia has been increasingly active in investi- TailorCrete Project -
gating the use of digital fabrication technologies in Wax Formwork
the production of concrete elements. On the one
hand, research groups have tried to address solutions
to some of the problems identified in the practice. On
the other hand, they have tried to suggest new av-
enues that eventually can inspire new developments
in practice. The Table 2 resumes the current ten-
dencies in the realm of Research, which are then de-
scribed and analysed.
Indirect intervention / Rigid Formwork / Subtrac-
tion. The fabrication of formwork noticeably contin- Figure 2
ues to be a core issue in research, as it is reflected by Pre:Vault Pavilion
most of the examples considered. However, there is
an interest in exploring different strategies for its pro-
duction with the aid of digital fabrication. To surpass
the limitations of using standard EPS milled form-
work, the TailorCrete project (Gramazio and Kohler
2011) [1] recently developed a wax-based formwork
for complex geometries with the aid of robotic pin
actuators (Figure 1). This strategy showed compara-
ble efficiency to EPS systems but performed better
in terms of the economy and ecology of the process,
and of the surface finishing qualities of the fabricated
pieces. At the same time, a greater level of geometri-
cal freedom and texture expressions were made pos-
Despite the traditional relation of concrete with
sible with EPS formwork by the adoption of 6 and 7-
continuous surfaces, the integration of digital fabri-
axis robotic arms instead of the standard 3-axis CNC
cation technologies in the production of rigid form-
routers commonly used for milling [4].
work has been studied for the prefabrication of cus-
tomized discrete components as illustrated by the

480 | eCAADe 32 - Fabrication - Volume 1

Figure 3 Pre:Vault research pavilion (Figure 2) (Larsen et al.
Robotic Slipform 2012). In this case, laser cutting enabled the mass-
research prototype customized production of digitally scripted sets of
unrolled geometric surfaces that were subsequently
folded into moulds for unique building components.
This digital process points towards the emerging de-
sign possibilities for building complex and highly
customized structural components in concrete. A
similar approach was featured in the research project
"Fisac Variations" for the reinterpretation of Miguel
Fisac's bones system, using a robotic hotwire to pro-
duce EPS moulds for customized concrete elements
(Raspall 2013).
Indirect intervention / Flexible Formwork / Sub-
tractive. Non-rigid and non-static formwork systems
have been less used in the history of concrete pro-
duction; However, they are also starting to integrate
digital fabrication technologies in their specific pro-
cesses. Although various researchers are currently
exploring the tectonic, formal and textural potential
of flexible textile formwork, built prototypes and lit-
erature suggest the use of craft-based approaches for
building components such as columns, beams and
walls (West 2009) [2] (Manelius 2012). Digitally driven
processes, such as the case of the Fatty Shell exercise
Figure 4
(Mcgee and Pigram 2012) [3] show the relevance of
MeshMould
integrating CNC technologies to cut unfolded com-
research prototype
plex patterns in fabric formwork solutions.
Indirect intervention / Dynamic Formwork / Ad-
ditive. Dynamic formwork systems consist in those
strategies that consider the displacement in space
(e.g. translation) of rigid moulds to produce forms.
Approaches that fall in this topic join some of the
benefits of rigid formwork (e.g. the reduced diffi-
culty of achieving smooth surfaces) with a higher de-
gree of formal freedom through spatial movement.
In the case of the Robotic SlipForm (Figure 3) process Indirect intervention / reinforcement / formative
developed at the ETH in Zurich, the integration of a and additive. To overcome the complex and manual
multi-axis robotic arm to control the vertical and rota- labour process of creating steel reinforcement cages,
tional movement of the formwork in space, calibrat- the "Tailorcrete" research project has explored the au-
ing the deposition speed of the wet concrete, greatly tomation of this process through the formative pro-
increases the formal capabilities of an otherwise ge- cess of robotic bending[4].
ometrically limited strategy. (Kristensen 2013)

Fabrication - Volume 1 - eCAADe 32 | 481


Establishing some similarities to the ferrocon-
crete production system used extensively by Nervi,
the problem of reinforcement and, simultaneously,
of the formwork production is being addressed in
the on-going research by Fabio Gramazio & Matthias
Kohler "Mesh-mould" (Figure 4). This work investi-
482 | eCAADe 32 - Fabrication - Volume 1
gates the use of a robot to 3D print the 3D mesh com- Figure 5
ponent that serves as reinforcement and guides the Contour Crafting
creation of the concrete element. With this fabrica- wall prototype
tion system, the design of the 3D mesh can be cus-
tomized to address variation, for instance, in size, ge-
ometry or density. (Hack 2013)
Direct intervention / Additive fabrication. A more
recent trend in the field consists in using additive fab-
rication processes to directly produce concrete ele-
ments. 3D printing technologies, in which materi-
als are sequentially solidified layer by layer to pro-
duce customized objects, have been translated to the
construction scale and used concrete as a base ma-
terial for printing building components. The use of
these processes introduces a higher degree of geo-
metric freedom in the production while avoiding the
need of formwork. Currently, two of the most con-
solidated examples of this strategy are Contour Craft-
ing (Figure 5) (Koshnevis 2006) and "Freeform Con-
struction" (Figure 6) (Lim et al. 2012), which already
proved the feasibility of 3D printing functional con-
crete elements. In the first case, mainly wall seg-
ments were produced, optimized to speed-up the
production and obtain smooth surface finishes. In
the second case, the built prototypes proposed a Figure 6
larger emphasis on formal freedom. (Lim et al. 2012) Free Form
Facing the inherent promises, this technologi- Construction panel
cal trend still reveals some limitations. For instance, prototype
attaining formal freedom is mostly dependent on
two key issues: printing resolution, which is mainly
defined by the height of concrete layers and the
diameter of the printing head, and the need for
secondary structures to support the hanging parts.
These constraints contribute to design lower curva-
tures in height, when compared with concrete ele-
ments made with milled formwork, and feature addi-
tional difficulties in creating voids in planes orthogo-
nal to the base printing plane. Also, in terms of sur-
face finishing, both technologies still leave distinctive
marks in produced elements, which are not fully de-
sired. This feature must be improved towards its im-
plementation in practice.
An alternative approach using the same tech-
nology was presented by Neri Oxman (2011), en-
titled "Variable-Density Graded Fabrication of Con-
crete". Instead of the design geometry of a product,
this research is focused instead in its material compo-
sition. Using additive fabrication in a robotic printing
platform, it looks for controlling the material proper-
ties and cellular structure of concrete, following the
inspiration from the gradient structures presented in
natural systems. This approach enables decoupling
form from expected structural behaviour in concrete
structural elements, as well as allowing the modula-
tion of some properties (e.g. colour, texture) in con-
crete surfaces.
Direct intervention / Subtractive fabrication. All
previous applications of digital fabrication technolo-
gies occur in different moments prior to or during the
casting of concrete. However, a few were found that
directly transform the produced concrete element in
other forms or textures.
For instance, and considering the use of a con-
crete composite, the [c]shape Pavillion built at the
Architecture Association in 2008 employed over 850
glassfibre-reinforced concrete panels, which were
cut with CNC technology. Also, there are some exper-
iments of cutting prefabricated concrete blocks using
a diamond wire attached to a robotic arm in the con-
struction of self-supporting assemblies (McGee et al.
2012). Both situations point out to the possibility
of CNC post-processing operations directly over the
concrete elements for the materialization of geomet-
rically complex assemblies.
CONCLUSION
Recognizing the current dissemination of using dig-
ital fabrication technologies in architecture this pa-
per investigate the state of the art of their application
in the production of building elements in concrete.
From the presented survey, its seems that the use of
these processes has served more to fulfil the desire
of materialising expressive forms rather than that of
achieving customized surface effects.
In pursuing an increasing level of formal com-
plexity in the built works, this study reveals the nat-
ural predominance of recurring to the use of sub-
tractive processes to produce customized formwork.
This general tendency in practice, raises an impor-
tant concern with sustainability, which makes the
research efforts on additive processes a promising
technological alternative.
Although the surveyed works in practice and re-
search are concerned with surface smoothness, few
examples have been found trying to exhibit new
texture effects from digital fabrication means. In
most cases, the surface texture of exposed concrete
is mainly regarded as technical concern to be ad-
dressed rather than as a topic of design exploration.
Nonetheless, the existing technologies and examples
shown, in direct and indirect subtractive fabrication,
suggest the possibility for the expansion of textural
effects. One possible path lies in acknowledging the
inherent properties of specific technologies and in-
tegrating them into the design stage, as was found
in some of the historical examples mentioned in the
second chapter of this paper.
Finally, although it has been a recurrent research
topic in contemporary architecture, the exploration
of the material composition is still underdeveloped
in relation with other digital fabrication trends. The
concept of its controlled variability is therefore an ex-
citing new theme that can yield interesting aesthetic,
formal and structural results.
The wide set of technologies mentioned
throughout the analysis establishes different rela-
tions with the materiality of concrete, depending on
the specific role each plays. Whether working directly
on concrete elements or, indirectly, in the formwork
production, it is interesting to notice how technology
options can be intrinsically linked with the design
languages. The works presented in this paper thus
show the exploration of a variety of geometries such
as free-forms, ruled surfaces, single-curved shapes or
extruded forms.
Because no identifiable process can be used to
achieve all desirable outcomes in construction, it is
foreseeable that several research works will make
their way into the architectural practice. It is our un-
Fabrication - Volume 1 - eCAADe 32 | 483
 derstanding that the these particular languages as
well as other material imprints emerging from spe-
cific technological processes can become relevant
sources for new effects and material expressions in
concrete and should be further developed in the fu-
ture.
ACKNOWLEDGMENTS
This work is part of a PhD research developed with
the support of the FCT - Foundation for Science and
Technology (ref. SFRH / BD / 79227 / 2011). It is
also integrated in a Research Project that is finan-
cially supported by FEDER funds through the Op-
erational Competitiveness Program - COMPETE and
by National Funds through FCT - Foundation for Sci-
ence and Technology, in the scope of the project
"PTDC/ATP-AQI/5124/2012".
The authors would like to thank Professors Fabio
Gramazio & Matthias Kohler, Dr. Berkoh koshnevis,
Dr. Richard Buswel and Dr. Ole Egholm Perdersen for
their help in illustrating some of the works surveyed
in this paper.
REFERENCES
Forty, A 2012, Concrete and Culture: A Material History,
Reaktion Books
Forty, A 2006, 'A Material Without a History', in Cohen,
JL and Moeller, GM (eds) 2006, Liquid Stone. New Ar-
chitecture in Concrete., Birkhauser Architecture, pp.
34-45
Garlock, MEM and Billington, DP 2008, Félix Candela: En-
gineer, Builder, Structural Artist, Yale University Press
Gramazio, F and Kohler, M 2008, Digital Materiality in Ar-
chitecture, Lars Muller
Hack, N, Lauer, W, Langenberg, S, Gramazio, F and
Kohler, M 2013, 'Overcoming Repetition: Robotic
fabrication processes at a large scale', International
Journal of Architectural Computing, 11, pp. 285-299
Iori, T and Poretti, S 2005 'Pier Luigi Nervi’s Works for the
1960 Rome Olympics', Actas del Cuarto Congreso Na-
cional de Historia de la Construcción, Cádiz
Kara, H (eds) 2008, Design Engineering: Adams Kara Tay-
lor, Actar
Khoshnevis, B, Hwang, DP, Yao, K and Yeh, Z 2006,
'Megascale fabrication by contour crafting', Interna-
tional journal of Industrial and System Engineering, 1,
484 | eCAADe 32 - Fabrication - Volume 1
pp. 301-320
Kolarevic, B 2003, Architecture in the Digital Age: Design
and Manufacturing, Taylor & Francis
Kristensen, EL, Gramazio, F and Kohler, M 2013 'COM-
PLEX CONCRETE CONSTRUCTIONS: Merging exist-
ing casting techniques with digital fabrication', Pro-
ceedings of the 18th International Conference on
Computer-Aided Architectural Design Research in Asia
(CAADRIA 2013), pp. 613-622
Larsen, NM, Pedersen, OE and Pigram, D 2012 'Realisa-
tion of complex precast concrete structures through
the integration of algorithmic design and novel fab-
rication techniques', Advances in Architectural Geom-
etry 2012
Legault, R 2006, 'The Semantics of Exposed Concrete', in
Cohen, JL and Moeller, GM (eds) 2006, Liquid Stone.
New Architecture in Concrete., Birkhäuser Architec-
ture, pp. 46-56
Lim, S, Buswell, RA, Le, TT, Austin, SA, Gibb, AGF and
Thorpe, A 2012, 'Development in construction-scale
additive manufacturing processes', Automation In
Construction, 21, pp. 262-268
Manelius, AM 2012, FABRIC FORMWORK: Investigations
into Formwork Tectonics and Stereogeneity in Archi-
tectural Constructions, Ph.D. Thesis, Royal Danish
Academy of Fine Arts Schools of Architecture, De-
sign and Conservation, School of Architecture
McGee, W, Feringa, J and Søndergaard, A 2012, 'Pro-
cesses for an Architecture of Volume: Robotic Wire
Cutting', in Brell-Çokcan, S and Braumann, J (eds)
2012, Rob | Arch 2012: Robotic Fabrication in Architec-
ture, Art, and Design, Springer, pp. 62-71
Oxman, N, Keating, S and Tsai, E 2011 'Functionally
Graded Rapid Prototyping', Proceedings of VRAP: Ad-
vanced Research in Virtual and Rapid Prototyping
Raspall, F, Imbern, M and Choi, W 2013 'Fisac varia-
tions: an integrated design and fabrication strat-
egy for adaptable building systems', Open Systems:
Proceedings of the 18th International Conference on
Computer-Aided Architectural Design Research in Asia
(CAADRIA 2013), pp. 55-64
[1] http://www.dfab.arch.ethz.ch/web/e/forschung/164.html
[2] http://www.umanitoba.ca/cast_building/assets/-
downloads/PDFS/Fabric_Formwork/Thin-Shell_Con-
crete_From_Fabric_Forms_SCREEN.pdf
[3] http://cargocollective.com/fabroboticsnet/Fatty-
Shell-Flexible-Formwork
[4] www.tailorcrete.com
Shape, Form and Geometry 1
Acoustically Efficient Origami Based Partitions for Open
Plan Spaces
Developing a Design Tool

Sophia Vyzoviti1 , Nicolas Remy2

1
Assistant Professor, Department of Architecture, University of Thessaly 2 Researcher,
Laboratoire CRESSON, Ecole Nationale Supérieure d’Architecture de Grenoble
1
http://www.arch.uth.gr/en/ 2 http://www.cresson.archi.fr/EQ/EQnr.htm
1,2
{vyzoviti.sophia|nicola.remy}@gmail.com

The paper investigates the management of acoustic and privacy problems in

open-plan spaces through the implementation of lightweight architectural
partitions developed by origami tessellations. Integrating knowledge from
parametric modelling, acoustics and design for user needs, a design tool for
acoustically efficient, flexible, interior partition systems is developed. The paper
elaborates on three components of the design tool: form generation, acoustic
performance and spatial performance. The form generation component employs
parametric models of origami tessellations to generate the partition system. The
acoustic performance component employs acoustic simulation and prediction to
regulate the containing volume as well as the system's surface materials. The
spatial performance component evaluates form and material through qualitative
criteria for privacy and flexibility according to user needs.

Keywords: Parametric origami, acoustic design, interior partition systems,

design tool development

INTRODUCTION has been considered as an innovative solution, effi-

Since Le Corbusier's definition of plan libre as one of
the five points for a new architecture (Curtis 1986) the
open-plan has evolved into an influential design prin-
ciple in Modern and contemporary Architecture. Suc-
ceeding the iconic projects of the first half of 20th
century, open-plan layouts have been widely applied
in the design of working and learning spaces, as
spatial solutions to the reformative programs in the
structure of labor and education after the 70's. Par-
ticularly in the case of office layout, the open-plan
cient both in terms of spatial allocation since it allows
increased net usable area, higher occupancy density
and ease of reconfiguration (Duffy 1992 in Kim and
de Dear 2013) and beneficial to productivity by facil-
itating communication and interaction between co-
workers (Brand and Smith 2005 in Kim and de Dear
2013). Nevertheless post occupancy evaluation stud-
ies of office spaces have proved that, open-plan lay-
outs are widely acknowledged to be more disruptive
due to uncontrollable noise and loss of privacy (Kim

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 487

 and de Dear 2013). Similarly, excessive noise lev-
els and lack of acoustic privacy have been frequently
identified as the most undesirable aspects of open
plan schools (Shield, Greenland and Dockrell 2010).
The standard technical solution to the acous-
tic privacy problems of open-plan layouts would be
a uniform ceiling with acoustic properties. How-
ever with standard acoustic ceiling solutions it is
difficult to facilitate several activities in the same
large room and therefore a spatial-acoustic device
is needed need to separate functions into smaller
acoustic places. In this framework the research in-
vestigates the management of acoustic and privacy
problems in open-plan spaces through the imple-
mentation of lightweight partitions.
At the basis of the research hypothesis we con-
sider that in comparison with flat paneling, pleated
surfaces manifest enhanced acoustic performance.
Folding augments the available surface in a given
area enabling increase in the quantity of sound ab-
sorbent material. Furthermore due to their tortuous
morphology pleated surfaces facilitate sound diffu-
sion and reflection. Therefore the form generating
presupposition for the definition of the partition sys-
tem is a set of parametrically defined folded surfaces
and origami deriving tessellations specifically.
The acoustic advantages of pleated surfaces are
evident in the design of concert halls, best exem-
plified in the recent Lelystadt Agora Theatre (UnStu-
dio 2007) or the Festspielhaus at Tirol (MHM Archi-
tects 2012). However investigations in the reciprocity
between acoustic performance and pleated surfaces
with applications in open plan spaces are limited in
the state of the art. Distortion II (Peters et all 2011) is a
small scale architectural partition, designed as an ex-
perimental installation that creates visual and acous-
tic affects within an open-plan space, employing a
parametrically designed single surface that responds
to two acoustic extremes: a sound-amplified zone,
and a sound-dampened zone. In this project a trihe-
dral folded plate unit is adopted leading to a struc-
turally strong system with design flexibility. In the
refurbishment of Surfrider Foundation Europe head-
488 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
quarters, Gardera-D Architecture firm develop a par-
tition in the shape of a pleated 'wave' that isolates
two acoustically conflicting programs -an exhibition
space and a working space-allowing their cohabita-
tion (Gardera-D 2012). Made of a layering of com-
posite wood panels habitually used in transportation
construction the wave allows maximum treatment in
term of acoustical insulation and absorption. The
convex surface of the partition which is facing the
exhibition area is reflective while its concave surface
facing the working area is sound absorbing.
Acoustic roofs employing origami deriving
pleated surfaces are more common in the state of
the art. The resonant chamber (rvtr 2011) acoustic
roof paneling system employs the inherent flexibil-
ity of the defining geometry, which is a well-known
origami archetype invented by Ron Resch, to trans-
form the acoustic signature of a space. Composed
of reflective, absorptive, and electro-acoustic pan-
els, the pieces can dynamically adjust their shape
to expose or hide these surfaces, thus altering sonic
conditions. In a similar manner, tunable sound cloud
(fishtnk 2013) is designed as a responsive architec-
tural system that enhances one's auditory experience
of space. The prototype is an interactive, pleated,
acoustic roof is composed of triangular panels lined
with a series of Arduino micro-controllers.
Considering the state of the art, the distinct iden-
tity of the research for 'Acoustically Efficient Origami
Based Partitions for Open Plan Spaces', lies in the in-
tegration of acoustically regulating formal and mate-
rial characteristics of the partition system, with a set
of user defined performance criteria. The formal pre-
supposition for an origami deriving shape grammar
intends to develop a self-supporting, lightweight
structure whose kinetic ability can enhance flexibil-
ity and adjustability in addition to its acoustic perfor-
mance.
 A DESIGN TOOL FOR ACOUSTICALLY EFFI-
CIENT ORIGAMI BASED PARTITIONS
Methodology
The research intends to develop a Design Tool for
structurally efficient, flexible, interior partition walls
where evaluation of their morphology according to
user needs and acoustic performance is integrated
in the design process. The Design Tool integrates
computational form-generation, acoustic simulation
and prediction and user defined performance crite-
ria. The design assumption for integrating the dis-
crete sets of data is a chaining mechanism between
performance, operations and form, considering that
certain morphological characteristics of the system
facilitate spatial events that satisfy particular user
needs.
Figure 1
Conceptual
structure of the
Design Tool.
Table 1
Conditioning
acoustic privacy:
spatial boundaries, The three discrete components of the design
articulations and tool, spatial performance according to user needs,
enclosures. form generation and acoustic evaluation are elabo-
rated further.
Spatial performance
Definition of spatial performance criteria according
to user needs draws knowledge from Post Occu-
pancy Evaluation studies concerning users' satisfac-
tion and comfort in working and learning spaces. Dis-
traction by noise and loss of privacy were identified
as the major causes of workspace dissatisfaction in
open-plan office layouts (Kim and de Dear 2013). Par-
titioning has been evaluated as a physical measure
that can supports visual and acoustic privacy without
obstructing interaction (Shield, Greenland and Dock-
rell 2010).
The intention behind design of the interior parti-
tioning system is the creation of acoustic sub-spaces
which would provide differentiated degrees of pri-
vacy to the users of open plan office, education or
cultural spaces. The partitioning system ought to be
flexible -both by design and by implementation- in
order to accommodate the user need for privacy pro-
viding boundaries, articulations and enclosures with
distinct spatial, visual and acoustic effects. When the
partitioning system is curved, the potential acoustic
conditions on the convex part of the shell are dif-
ferentiated from the concave side of the shell. The
concave side has less volume, more absorption, and
therefore less reverberation time though it would still
be strongly acoustically coupled to the main space.
Consequently, it becomes obvious that by modify-
ing the curvature of the partitioning system, from a
linear to a curvilinear and a circular arrangement dif-
ferent degrees of acoustic and spatial privacy are be-
ing achieved. In order to integrate this modification
option in the design process parametric modelling is
employed to produce curvilinear assemblages of the
pleated component.
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 489
 Form generation
One of the starting points of the research is the
development of form finding processes inspired by
Origami, the Japanese art of paper folding. Based on
simple deployment mechanisms, Origami derivative
forms and particularly those based on the Yoshimura
and Miura patterns allow a controlled transfer of
forces without the need for any secondary support
structure. These forms also demonstrate kinetic be-
havior through translation and rotation. The unique
combination of stability and kinematics evident in
origami derivative forms renders them appropriate
for flexible interior partitioning systems that are self-
supporting and deploy-able. Regular origami pat-
terns can be deployed in the Rigid Origami Simulator
developed by Tomohiro Tachi (2007-2009) comple-
menting modelling with Rhino. The shape grammar
of origami provides an archive of pleated compo-
nents that are aggregated into kinetic arrays. Com-
ponents may be designed according to structural, ki-
netic and aesthetic criteria, the analysis of which ex-
ceeds the scope of the current paper.
Based on arrays of the pleated component, the
partitioning wall is composed as a parametric system
with variable adjustment of curvature in plan and el-
evation. Dimensions of component are defined ac-
cording to ergonomic data. The number of compo-
nents is regulated according to the desired volume
of the partition. A Grasshopper script is developed
to manipulate the form generative curves of the par-
tition system, which we call power curves and dis-
490 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
tinguish between base and elevation curves. This
typology of curves has been defined in accordance
with the acoustic and privacy criteria discussed in the
previous section. By manipulating the curvature of
the array through power curves, each component is
parametrically differentiated while maintaining con-
nectivity and cohesion within the partitioning system
overall form. Base curves define the spatial configu-
ration of the partition in plan according to desirable
degrees of privacy. Elevation curves define the sys-
tem's ability for differentiation in response to users'
needs for visual contact, light or in aid of structural
efficiency. As proof of concept a series of scaled mod-
els are digitally fabricated. Evaluating the physical
models we can conclude that the kinetic potential of
the origami generated structure augments the versa-
tility of the partitioning system and enables its real
time manipulation. Comparing the fabricated pro-
totypes, we can conclude that linear arrays manifest
high kinematics; the system is able to acquire multi-
ple curvatures assuming a variety of base curve con-
figurations, while the ones differentiated through el-
evation curves are less flexible and would rather be
considered as static partitions. Figure 2
A set of origami
Acoustic performance based pleated
The main objective of the morphogenetic research components.
is the creation of parametric models for partitioning
systems that can adjust their geometry in order to
correct the acoustic performance of the containing
volume as well as their surface materials. By altering
the form generating geometry of the partition sys-
tem or the material characteristics of the component
surfaces in specific ways the acoustic comfort level of
the room can be adjusted in terms of ideal reverbera-
tion times according to the use and according to the
volume of the room. A Grasshopper script is devel-
oped in order to improve the acoustic characteristics
of the partitioning system including numerical data
describing the coating material's parameters. As it is
known in the field of acoustics, each material has its
own absorption coefficient of alpha Sabine. Provid-
ing that, the designer can calculate the acoustic im-
 pact of the object during the design -equivalent ab-
sorption area gain, reverberation time modifications,
and speech intelligibility indexes- comparing it to the
tables of audio data that are given. Thus, an evalua-
tion system is becoming part of the design process
and measurable acoustic performance criteria must
be set to correspond to the design intent. Designers
are enabled to contemplate design options through
these evaluative feedback mechanisms.
One can criticize this methodology by arguing
that these simple calculations, deriving from the Wal-
ter Clement Sabine formula, are valuable for the entire
Figure 3
Partition systems:
Variability of a
linear array through
elevation and base
curves.
volume and not precise enough. Actually, the form of
the partition is only considered in these evaluations
by the equivalent absorption area gain that the objet
creates and not by its form and position in space. For
this reason we wanted to go further and evaluate the
partition's potential to regulate the acoustic perfor-
mance of the room and also to create small different
acoustic spaces in the same room.
TEST CASE
The test case intends to verify the acoustic efficiency
of a partition developed according to the Design Tool
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 491
 methodology, employing acoustic simulation. The • S1 : the room empty (wooden floor aw=0,05,
goal of the simulation was to evaluate the acoustic roof and lateral walls are all made of Plaster
performances of an S-curved partition, formed as a rough on lath, aw=0,05)
space articulating boundary. For that, we use Pachy-
derm Acoustical Simulation (version 1.0b3 by Arthur • S2 : the room with the flat partition with the
Vanderharten, 2011 ) plug-in for Rhino that simu- same materials on both sides as described be-
lates the impulse respond of the room and calcu- low.
lates from it several well-known room acoustic cri-
teria used for theaters, operas, concert halls. These • S3 : the room with S-curved partition with a
criteria are mostly used in a context of artistic repre- global thickness of 2 cm using on the con-
sentation when we know exactly the position of the vex surface common wooden panel (aw=0,10)
sound source (singers, musicians, performers) and and on the concave surface sound absorbing
the positions of the receivers (audience). perforated wooden panels with mineral wood
In our context, sources and receivers might be on the inner side (aw=0,75).
mobile and some criteria often used have no sense
Scattering and diffusion factors wall also introduced
to be evaluated such as Strength, Clarity (C80), Defini-
as they are used in the literature and set up depend-
tion D(50). They are really valuable for music but less
ing of the materials used.
meaning full for ordinary audio-scene in offices (see
We wanted to simulate the potentials given by
the interesting discussion developed by Nilsson and
the S-curved partition. It's why we declared in the
Helström, 2010). On the other hand, it was interest-
model a sound source "far away" from the partition
ing for us to evaluate several reverberation durations
and 2 receiver points, one on the same side of the
that are more linked with the ordinary perception.
partition as the source is (receiver 1 or r1), and one on
"Classical" Reverberation times calculated on the de-
the other size of the partition (receiver 0 or r0), in the
crease of 60 dB on the impulse sound generated give
most protected place offered by the S-curved parti-
results that are often discussed because we, as listen-
tion.
ers, we never experience such a decrease in the "real"
life. We hear the reverberation of the room trough a Figure 4
loud and impulsive sound but very quickly another Acoustic simulation
sound will mask the whole decrease of the first one. of s-curved
For this reason we focus our work on the simula- partition: sound
tion on short reverberations durations described by source and
the following criteria as: Early Decay Time, T-15 (de- receivers positions.
crease of the first 15ms) and T-30. They describe the
behaviors of the first reflections of the sound source
until they reach the listener's ears. Many studies have
shown that this first reflections have already all the
sound qualities of the room and receivers interpret
them properly even they will never hear the whole
decrease of the reverberation. We calculated also the
reverberation time using the Sabine theory too.
To simulate the acoustic performance of our par-
tition we place it in an imaginary room (15 by 10 me-
ters) and run several simulations:

492 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

 Results
One first result of our simulation is regarding Sabine
reverberation time values. In the first scenario aver-
age Reverberation Time was approximately around
2,2 seconds whatever the position of the receiver
around the partition. With the S-curved partition, the
average RT decreases to 1,07 seconds. It means that
the global reverberation of the room is divided by a
factor 2. In the room, reverberation is more comfort-
able for any activities related to the ordinary life in
open-plan offices (working, speaking, ..). These re-
sults from simulation confirm also previous calcula-
tions that we've developed through the Grasshopper
script as we describe previously.
Results from simulations of the Early Decay Time
are described as follows:
Figure 5
EDT simulations: 3
scenarios for the
reception point r0
Figure 6
3rd simulation
scenario for
reception points r0
and r1
As shown on the previous chart, decrease of
the EDT from simulation 1 to simulation 3 is clearly
demonstrated. For all octaves, the EDT decreases
about half a second which means that it improves the
acoustic qualities inside the partition from a source
emitted in the same room but on the other side. It
means that, actually, people can easily go inside the
S-curved to look for more intimacy if they want to talk
with someone or have a personal chat on the phone.
Improvements are also demonstrated in comparison
with a flat partition (simulation 2) and it can be ex-
plained by the fact that the origami based form offers
more surfaces which increase the equivalent area of
absorption.
The results can be shown also if we compare the
EDT value close to the source (receiver 1, direct field)
and the receiver protected by the partition (reverber-
ated field) for the third simulation:
Even the global reverberation of the room has
decreased, while locally simulation shows how the
partition can create two sub-spaces in which acous-
tic qualities will differ. Similar results were found with
the other criteria that we ran in calculation (T-15 and
T-30).
To conclude, we can say that the S-curved parti-
tion creates locally an acoustic affordance, as James
J. Gibson (1977) described it for visual perception, an
added acoustic value that future users would have
the possibility to exploit.
DISCUSSION
The Design Tool for Acoustically Efficient Origami
Based Partitions can be useful in a case of collab-
oration between designer and user since it makes
explicit the available design options concerning the
form and material of the partition. In terms of com-
putational design, the Design Tool is not seamless at
present; it combines different software, scripts and
plug-ins. However the criteria are clearly defined
and each step in the procedure manifests clear re-
sults. The acoustic and visual privacy criteria lead
to the definition of power curves which control the
overall form of the partition system employing the
form generative script. The material selection for
the facets of the pleated component employing the
acoustic script, defines its acoustic profile. Results
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 493
 from the scripts entail design modifications and as-
sist the overall design decision making.
From the acoustic point of view, the paper
presents the possibility to develop a Design Tool that
can evaluate an architectural object as a partition in a
room. This first step is very helpful to improve acous-
tic sensibility for designers that are not aware about
acoustics in details. However, it ought to be under-
stood that the introduction of this partition must re-
alize conflicting goals from the acoustic point of view.
In this sense, the choice of an origami based fold-
ing partition system is particularly interesting in the
sense that it is freestanding, rigid and composed of
facets. These facets can then be made of sandwich
panels which can be made of an absorbent material
on one side and a reflective material on the other.
This adds a new perspective for architectural design
by integrating sound qualities and acoustic poten-
tials in the design process.
Furthermore, it is evident from the test case that
we needed to use more precise tools to evaluate the
potential of the S-curved partition. On a usual com-
puter, simulation took a lot of time to run the calcula-
tion with 2 receiver points. It means that such devel-
opment can be done only in the final phases of the
design when we want to be more efficient and pre-
cise. Even results are encouraging, it shows that, one
more time, acoustic comes at the end of the project
and it use in corrective processes: to validate or not
the choices made by the designer considering the
form, the position in space, and the materials of the
partition. We think that the next step of such stud-
ies might be in the development of plug-in or appli-
cation that can work in parallel with modelling. It
would be ideal if calculation could be done instantly,
then we could imagine designing form by visualizing
acoustic criteria in real time.
sity of Thessaly by Christina Angeli and Aikaterina
Charisiadi, under the supervision of the authors .
REFERENCES
Curtis, WJR 1986, Le Corbusier: Ideas and Forms, Rizzoli,
New York
Gibson, J 1977, 'The Theory of Affordances', in Shaw, R
and Bransford, J (eds) 1977, Perceiving, Acting, and
Knowing, NJ: Lawrence Erlbaum
Kim, J and de Dear, R 2013, 'Workspace satisfaction: The
privacy-communication trade-off in open-plan of-
fices', Journal of Environmental Psychology, 36 , pp.
18-26
Nilsson, E and Hellstrom, B 2010 'Room Acoustic Design
in Open-Plan Offices', Proceedings of 10eme Congres
Francais d'Acoustique, Lyon
Shield, B, Greenland, E and Dockrell, J 2010, 'Noise in
open plan classrooms in primary schools: A review',
Noise and Health, 12(49), pp. 225-234
[1] http://architype.org/project/surfrider-foundation/
[2] http://www.world-architects.com/en/projects/40605_-
Festival_hall_for_the_Tyrol_Erl_Festival
[3] http://www.bradypeters.com/project-distortion-
ii.html
[4] http://rvtr.com/research/resonant-chamber/
[5] http://www.fishtnk.com/fish/tunable-sound-cloud/
[6] http://www.unstudio.com/projects/theatre-agora

ACKNOWLEDGEMENTS
This paper has benefited from the results of the
course special research topic 'Acoustic/Paramet-
ric/Origamic/Design for open plan spaces' devel-
oped at the Department of Architecture, Univer-

494 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Measurability of Loos' rejection of the ornament
Using box-counting as a method for analysing facades

Wolfgang E. Lorenz
Vienna University of Technology, Institute of Architectural Sciences,
Digital Architecture and Planning
http://www.iemar.tuwien.ac.at
lorenz@iemar.tuwien.ac.at

As evidence from recent years has demonstrated, box-counting provides an

objective fractal analytical method to evaluate the visual complexity of
architecture. This paper for the first time explores the potential of box-counting
with regard to the work of the Viennese architect Adolf Loos (1870-1933). Loos is
seen as the pioneer of modern architecture, as someone who anticipated the
International Style. This impression derives from his resentments towards the
ornament, expressed especially in his texts. However, Loos did not reject
ornamentation in general. Thus, the group of smooth plastered facades provides
a narrowed view on his overall architectural concept. A more differentiated view
on Loos' oeuvre is not new; however, the author further develops the possibilities
of describing facades geometrically by using an implementation of the fractal
analytical method, especially created for facades. This paper not only focuses on
the possibility of grouping facades with similar characteristic values, but
considers other aspects of Loos' design such as space as well.

Keywords: Box-counting, Adolf Loos, Complexity, Fractal geometry

MOTIVATION of Loos' style. On closer inspection, two aspects

Benoît Mandelbrot (1981, 1982), the "father" of frac-
tals, argues that modern architecture expresses an
affinity to scale-bound objects, and hence looks
smooth. However, the author reveals that Loos' view
does not necessarily yield smooth facades. From the
very beginning Loos appears as a pioneer of mod-
ern architecture of plain, unornamented buildings,
due to both his theoretical work and the first pub-
lication of Steiner House in 1910. The smooth fa-
cade of Steiner House is, however, only one facet
are particularly noticeable about Loos' oeuvre: First,
the outer appearance of his buildings is not nec-
essarily smooth, but, on the contrary, diversified.
Throughout his whole work, smooth cubic designs al-
ternate with classic-style villas, timber constructions
and conversions. Second, Loos accomplishes a mani-
fold sophisticated whole by focusing on two strate-
gies, the difference and interaction between an in-
ner and outer order and the refined use of symme-
try. Those aspects support the author's assumption

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 495

 that the complexity of Loos' facades is both, indepen-
dently of the year of construction and of varying de-
gree. This paper has, therefore, two objectives: in or-
der to verify the diversity of Loos' facades, their com-
plexity is first analysed with the help of an objective
comparison method, and then, based upon this anal-
ysis, the division into groups of similar characteristics
is discussed. At the same time, the results are also
evaluated in relation to Loos' essential idea of space.
INTRODUCTION
Beside Peter Behrens and Frank Lloyd Wright, Loos
ranks among the most important representatives of
the first modern movement (Hitchcock 1994). His
thoughts about architecture do not only find their
theoretical expression - published in numerous arti-
cles - but manifest themselves in his buildings. How-
ever, Loos' concepts cannot be judged by his proba-
bly most frequently cited and famous pamphlet "or-
nament and crime" (Loos 1997 [1908]) only. In his
theoretical articles, he primarily criticizes the archi-
tectural practice of his time - the arbitrary use of or-
namentation without any reference to an underly-
ing concept. It was the time of Historicism when ar-
chitectural and decorative elements of the past were
used devoid of their former significance. However, he
also attacked new movements in architecture which
developed their own independent ornamentation.
The "Secession" movement, for instance, a Viennese
variety of Art Nouveau, did not meet his expectations
either (Tournikiotis 1994).
Influences
It was during his stay in the United States be-
tween 1893 and 1896 that Loos decisively shaped
his thoughts about culture and architecture. In the
purism of an American suitcase he found the defini-
tion of modern style: "functional is beautiful!" (Scheu
1909). Loos may also have been influenced by the on-
going discussion of Louis H. Sullivan on the restrained
use of ornament (Ungers 2002). Frequently used ele-
ments, such as the two-storey central living hall, the
fitted wardrobes and the fire place (niche) reveal a
496 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
clearly American-British influence.
Additional sources of influence were his journeys
to the Greek Islands and North Africa. The cubic
and terraced buildings by Loos, for instance, bear re-
semblance to the flat cubes with sharply cut win-
dows found at both destinations. Moreover, the con-
cept of Raumplan (spatial plan) shows a striking sim-
ilarity to the architecture of Greek islands (e.g. of
Euboea). Both, the central high hall with adjusted
gallery and the stairs of the outdoor area with in-
termediate places serve as inspiration (Kurrent et
al. 1998). Raumplan is a term coined by his stu-
dent H. Kulka (1931) describing the artistic three-
dimensional arrangement of spaces.
"Ornament and Crime" and Steiner House
The publication of Loos' (1997 [1908]) well-known ar-
ticle "ornament and crime" started fierce discussion.
According to Loos, contemporary artists were not
able to produce adequate ornaments, and therefore
it was better to dispense with ornamentation at all
(Loos 1997 [1908], [1924]). Moreover Loos was con-
vinced that the production of ornament wasted ma-
terial and time. He, however, pointed out himself
that he had never completely refused ornamentation
(Loos 1997 [1924]).
In 1909 the owners of Goldman & Salatsch com-
missioned Loos with his first larger - and proba-
bly most controversial - work: the Looshaus on the
Michaelerplatz in Vienna, completed in 1911. It was
called 'the Viennese house without eyebrows', since
the usual window roofings of that time were miss-
ing. The prominent location on Michaelerplatz oppo-
site the back entrance of the Imperial Palace further
enhanced public controversy. It was said that Em-
peror Franz Joseph not only severely objected to the
Looshaus across the square, but also drew the cur-
tains so that he would not have to look at it at all.
Another building created at the same time al-
lowed Loos to realize his ideas of a building reduced
to simple geometric shapes: Steiner House. In 1910,
the year of completion, the smooth plaster facade
of the front view was published in the journal "Der
 Architekt". Without additional images from inside or
floor plans, the sole publication of the front view not
only established a connection with the Looshaus, but
contributed to a preconceived opinion about Loos'
architectural concept. Pevsner described Loos' rad-
ical rejection of ornamentation as the architect's de-
cisive contribution, which he interprets as an antic-
ipation and confirmation of the International Style
(Kühn 1989, Khan 1998).
"Raumplan"
The concept of the interrelation of rooms in Loos' de-
sign is today known as Raumplan. Loos himself never
used this denomination. His student Kulka (1931) in-
troduced the term to describe differentiated spatial
thinking. Three-dimensionality of architecture and
interaction of rooms replace the two-dimensional
way of thinking. The height and the position of a
room depend on its purpose: The more private, the
lower its height and the closer to the top of the build-
ing. The central "semi-private" hall occupies up to
two storeys. The individual flights of stairs, rarely
positioned one over the other, coil upward next to
the rooms. This creates a composition of split levels
with interesting experiences of space. Müller House
shows a comprehensive realisation of Raumplan.
Figure 1
Box-Counting BACKGROUND
method Basically, box-counting, the fractal analytical method
demonstrated with used in this paper, identifies the minimum number
Steiner House (gray of boxes of a certain size completely covering the ob-
coloured boxes are ject under consideration. The grid method - as a sim-
counted) plified extension - forms the basis for a script in VBA
for AutoCAD, called "Box-counting steps", written by
the author (Lorenz 2009, 2012). In particular, this im-
plementation takes into consideration the peculiari-
ties of architectural analysis: It uses a medium of the
planning architect and due to the fact that it consid-
ers vector graphics instead of pixel graphics, neither
the thickness of lines nor the resolution of the image
influences the result. With this simplified method, a
grid with a certain mesh size is placed over the object.
For facades the object consists of all relevant lines,
which is a matter of distance and visual perception.
Relevant lines mark striking changes of direction or
of significant material differences. In the next step,
those boxes of the grid which cover the object (fig-
ure 1) are counted. As already demonstrated by Carl
Bovill (1996) the largest and smallest mesh size again
depends on the distance of the observer to the object
and on the eye angle. In turn, the distance is a func-
tion of the height of the building and three different
angles, 18°, 27° and 45° (Maertens 1884). Based on
these studies the author defines the upper and lower
bounds for box sizes (Lorenz 2013). Finally, while
reducing the box-size the amount of covered boxes
changes as well. The box-counting dimension is de-
fined by:
log (N )
DB = lim ( ) (1)
→0 log 1
If a relationship between size and scale exists - at
least for a certain range of scale before the elevation
dissolves in its one-dimensional lines -, this can be
clarified in a log-log graph. Data points then nearly
follow a straight line or more precisely they are close
to a regression line. Finally, the slope of the regres-
sion line (equal to DB) gives the ratio of irregularity.
According to Lewis Frey Richardson (1961), the
length of coastlines is hard to determine. This is due
to the fact that the length depends on the scale of
the map used. Thus, he was able to establish a re-
lationship between total length and scale. This re-
lation was given by a power law. Benoît Mandel-
brot, the "father" of fractals, related the exponent to
what he called fractal dimension. Fractal dimension
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 497
 is a characteristic value for fractals. In order to es-
timate the fractal dimension different methods ex-
ist, for instance box-counting, which Bechhoefer and
Bovill (1994) applied to architecture for the first time.
Analyses and Evaluation
According to C. Bovill (1996), box-counting calculates
the approximate visual complexity of facades. How-
ever, the result of any measurement depends on var-
ious influences (Foroutan-pour et al. 1999, Lorenz
2003, 2009). In order to minimize influences, and as a
result of previous works (Bovill 1996, Foroutan-pour
et al. 1999, Ostwald et al. 2008), the author's imple-
mentation "box-counting steps" allows adjustments
of influential parameters via an input form. As a con-
sequence, the analysis of each facade considers a set
of measurements with different adjustments. The set
is then critically assessed by statistical methods (visu-
alized in a box-plot). In order to obtain the most sig-
nificant possible result, it is important that
• every single measurement offers only little or
no deviation: a coefficient of determination R²
= 0 indicates no relation, while R² = 1 means
highest possible correlation,
• the set of measurements has a small in-
terquartile range, which is the range of the
box-plot containing 50% of all values: the
smaller this range, the smaller the fluctuation
of data-points.
For the data analysis by the author several values are
important (the median, the interquartile range, the
smallest and the average coefficient of determina-
tion and the range of scale), since they provide infor-
mation about coherence across scales.
Results
The author focuses on whether the complexity of
Loos' designs changes over the years or remains on
a similar level. It is, therefore, the measurement of vi-
sual complexity that provides new insight into Loos'
work, both in terms of his buildings and his theoreti-
cal work. The paper analyses nine residential houses
498 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
from all periods of Loos' creative work, including four
unrealised projects and two conversions (table 1 and
figure 2). The author's new drawings as vector graph-
ics ensure equal treatment of the data. The study
work of the Technical University of Munich (Kurrent et
al. 1998) served as an important source. While seven
buildings consist of four elevations each, two, as an-
nexes to a neighbouring plot, have three elevations
each. In sum, the analysis of this paper comprises
34 elevations. Each individual measurement set (for
a single elevation) consists of ten different measure-
ment settings. As they, in turn, consider different
starting points, the total number amounts to 60 sin-
gle measurements for each facade. In order to guar-
antee some consistence of all measurements, several
influences on the results have been minimized as fol-
lows:
• every elevation has been prepared in the
same manner (i.e. the author considered the
same sort of architectural elements, and used
the same scale of preparation),
• every set consists of the same number of in-
dividual measurements with exactly the same
measurement settings,
• the average coefficient of determination for
each facade amounts to at least 0.998 and
• thus, in order to compare different building
sizes, the smallest and largest mesh size is set
in relation to the height of the facade (dis-
played as a percentage value).
Table 1
List of analyzed
buildings
Figure 2
Analyzed buildings
(left to right - listed
in table 1 starting
from the top)

A sharp bend in the data curve interquartile-range of the box-plot that displays the
When analysing the data, we note that the west view difference (right image of figure 3). The following
of Steiner House displays a sharp bend in the data buildings belong to the category with a sharp bend
curve of every single measurement (left image of fig- in the data curves:
ure 3). The author has already been able to observe
a similar behaviour in case of two other representa- • Scheu House (all views)
tives of modern architecture: Villa Savoye by Le Cor-
• Steiner House (all views; west and east view
busier (Lorenz 2012) and Villa Tugendhat by Mies van
with a clearer sharp bend)
der Rohe (Lorenz 2013). The sharp bend clearly sepa-
rates two ranges of scale with different slopes: while
• Mandl House (all views; west and east view
the first section displays a steeper slope, the second
with a clearer sharp bend)
one becomes flatter. As the former comprises larger
mesh-sizes, it can be deduced that the elevation is
Figure 3 of greater complexity when viewed from a greater
Analysis of the west distance. At a certain scale the behaviour suddenly
view of Steiner changes and the elevation tends to be smoother.
House: one single Thus, the gradient angle of the regression line and,
measurement (left) subsequently, the box-counting dimension (as its
and box-plot equivalent) flatten alike. The turning point is equiv-
diagram of the set alent to the change from the dominance of the win-
of measurements dows towards the smooth plaster surface. However,
(right) when all measurements are considered, it turns out The smooth facades of Steiner House with reg-
that the data material of the second section is more ularly arranged and sharply cut windows seem to
consistent than the data of the first section. It is the have influenced Le Corbusier. However, Le Corbus-
ier in contrast aligns the window flush with the outer

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 499

 surface, which has the effect that the volume of the
building becomes even more apparent (Hitchcock
1994). In his dissertation, the author has already
pointed out the similarity of the results between the
garden (east) facade of Steiner House and the south-
east facade of Villa Savoye by Le Corbusier (Lorenz
2013). Both display a clearly sharp bend in their
data curves (Lorenz 2012). For the first section of
larger distances, the median value of the garden fa-
cade of Steiner House amounts to 1.73. The first
section of the south-east facade of Villa Savoye, in
turn, achieves 1.60, a slightly lower value. However,
for the second range the differences between both
buildings are marginal: 1.26 for the garden facade
of Steiner House and 1.28 for Villa Savoye. These re-
sults support the conclusion that both facades are of
a higher visual complexity viewed from a larger dis-
tance than from a smaller one.
The street (west) and the garden (east) facade
of Steiner House, the three facades of Scheu House
and the street (west) and the garden (east) facades
of Mandl House display a similar median. All val-
ues remain between 1.65 and 1.78. However, what
is striking about Steiner House is that for the two side
walls the point of change appears earlier - at a mesh-
size of about 2.3 meters. Moreover, their medians for
the first section are smaller - in both cases 1.46 (for
further details see section "Opposite sides with sim-
ilar results"). All views of Mandl House, represent-
ing the other extreme, display higher values in the
second section, with higher complexity of this scale
range. This is mainly due to the fact that it is a con-
version and expansion. Visually, the result is under-
lined by a higher level of detail on smaller scales: cor-
nices, wooden elements and the pattern of smaller
elements of the railing.
Similar results for all elevations of one and
the same building
Some of the analysed buildings show similar results
for all views over a large range of scale (table 2).
This suggests a similar complexity not only for all
sides but also independently of the considered scale.
500 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Scheu House clearly belongs to this group of build-
ings, as the medians of all three elevations remain
between 1.65 and 1.68. The reasons for those results
are twofold: First, size-differences between elements
are similar and, second, their number increases when
size decreases. Primarily, the three levels of size
include the whole elevation, the windows and the
smaller elements of the skylights. By contrast, sym-
metry is no decisive criterion, as only the east view of
Scheu House is symmetrical.
Table 2
Buildings with
similar results in
most of their views
The outlier of Mandl House concerns only the
second section of the north facade. The lower me-
dian is due to the large facing surface of the mansard
roof and to the large windowless area of the facade.
On the opposite side the smoother effect is reduced
by the tent-roof covered tower-like annex and the
flat-roofed side annex by Loos, but also by the pillars
for the open canopy. Thus, the south side receives
similar characteristics as both gable-views of the half
hipped roof with different architectural elements.
Khuner House, a wooden construction with bal-
conies on a natural quarry stone foundation, offers a
completely similar characteristic for all views. A sim-
ilar behaviour occurs at the project for Dr. Müller's
daughter. In this case, only the south view represents
a slightly smaller value (1.76). This is due to the outer
smoothly plastered chimney, which occupies part of
the facade.
Opposite sides with similar results
For the first section the opposite sides of Steiner
House display similar results (table 3). Parallels
and similarities can be identified between either the
street and the garden view or both side views. For all
views the design of the windows is - apart from their
formats - similar, depending on the scale of consider-
 ation. The same is true for the order of magnitude of
elements from the large to the small scale. Neverthe-
less, the measurements of the less visible side-views
result in smaller medians. While the street and the
garden view display symmetry (only slightly broken),
this characteristic changes to a free distribution at
the other two facades. However, the differences be-
tween the two pairs of facades occur due to the ratio
between empty (smooth) areas and those with sig-
nificant lines. Larger smoothly plastered areas at the
side-views correspond to smaller medians. However,
both the street and the garden view display high val-
ues above 1.7 until the sharp bend in the data curve
occurs.
Table 3
Buildings with
similar results for
opposite views
Strasser House, although a conversion, presents
a similar division. Both side views, again, appear
smoother due to larger empty areas of the facade.
In turn, higher medians for the street view and the
rear side reflect higher complexity, this time for a
larger scale range. On smaller scale, for instance,
plate connections of a quarter-round tin roof deter-
mine in both cases the entire top parts. Furthermore,
the otherwise continuous classical cornices are inter-
rupted at the side views and the ratio of the vertically
structured balcony balustrades is smaller.
The unrealized Villa Stross, a representative of
Loos' classic villa-style architecture, belongs to this
group as well. The street and the garden view are
dominated by strict symmetry and axiality. Differ-
ences occur due to the deep, high loggia and colossal
Ionic columns of the garden view and the extensions
and terraces of the street view (Risselada 1988). In
contrast, the sides views are less structured and again
consist of large empty areas.
DISCUSSION
As could be demonstrated in the previous section,
the study of Loos' work with regard to its complex-
ity reveals no correlation between the year of design
and the degree of complexity. This confirms the as-
sumption that Loos' design concept cannot be re-
duced to the rejection of ornamentation, but com-
prises different outer appearances. Moreover, most
results - apart from those with separate data-curves
- reveal astonishingly high levels of complexity for a
large range of scales. In order to perform a more de-
tailed analysis the following section utilises a descrip-
tion of Loos' buildings according to design features.
Smooth cubic and terraced buildings
Buildings of plane facades with sharply cut win-
dows, summarized in the group of cubic and terraced
buildings, repeatedly occur throughout Loos' work.
According to Loos, the object of utility (and most
pieces of architecture) should be considered sepa-
rately from the art, and thus requires no decoration
(Loos 1997 [1909], [1924]). The aesthetics of a utility
object is exclusively derived from a material-specific,
functional and resources-conserving realization.
If one considers the interrelation between inside
and outside, two subgroups emerge: (1) facades,
which are designed from inside to outside, where the
outer appearance is, however, subject to a separate
order, and (2) facades, whose openings, at first, ap-
pear disorganized and arbitrary, but correspond to
the internal order (Kurrent et al. 1998). While Steiner
House belongs to the first subgroup, Scheu House -
according to Kulka (1970) the first terrace house in
Central Europe - belongs to the second. Although all
facades of Steiner House avoid ornamental decora-
tions, it is especially the garden facade with the up-
per end of a flat roof where nothing distracts from the
clear form. Nevertheless, some characteristics of this
side of the building refer to the vocabulary of clas-
sicism: the middle and side risalits and the strictly
symmetrical arrangement. These belong to the outer
order. In contrast, the eastern view of Scheu House
clearly demonstrates the correspondence between
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 501

the size of openings and the purpose of the room be-
hind. Scheu House, although part of the smooth cu-
bic examples, displays again high values at least for a
certain scale range.
An apparently smooth facade does not necessar-
ily lead to a sharp bend in the data curve. For ex-
ample, Schröder-Schräder House by Gerrit Rietveld
shows a consistently constant slope down to very
small mesh sizes (Lorenz 2013). Bojko House, a three-
storey terraced cubic building by Loos has similar
characteristics, but remained a project. In contrast
to Steiner House and Scheu House, none of the three
views displays a sharp bend in the data curve. This
suggests that Bojko House, a later representative of
smooth cubic buildings, is more balanced with re-
gard to scale. It displays high results for a large range
of scales. Only the western facade has a slightly
higher value because of both the material of the en-
closure and of an annex and the balcony with hori-
zontal lines of the parapet ranging across the entire
width (figure 4). However, the scale ranges are the
same for all three facades.
Classic villa-style architecture
Loos' rejection of ornamentation from earlier ar-
chitectural epochs does not necessarily result in a
complete rejection of references to past styles. Al-
though the use of classical or neo-classical vocabu-
lary of forms appears mainly at unrealised buildings
(Tournikiotis 1994), it is not completely unfamiliar to
Loos' work (compare Steiner House). However, the
502 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Figure 4
Smooth cubic and
terraced buildings;
Buildings with
reference to
classical antiquity
main topic is the continuous development of the in-
ner order, which Kulka (1931) later called the Raum-
plan. While Villa Konstandt already shows room lev-
els, it is Villa Stross that indicates the principles of
Raumplan (Kurrent et al. 1998). The classical facade
with its own order does not oppose the inner or-
der. The use of symmetry demonstrates the conflict.
While the facades offer one level of reflection sym-
metry, the inside is rather a (time) sequence of sym-
metries. Time sequence means that every room has
its own symmetry whose axis is not continued to the
next room. When one crosses one room after the
other, one passes from the one symmetry to the next.
Symmetry in a room includes cases where a prop or
a chimney is juxtaposed (mirrored) by a wall closet.
When the outside mirrors the inside, they finally offer
complex interlocking between symmetries and spa-
tial axes.
The results for all facades of Villa Konstandt and
Villa Stross show certain variation (figure 4). In the
case of Villa Konstandt, differences are due to the
sloping terrain. The base area of all four views con-
sists of exposed masonry, which, in general, leads to
a higher fracture. The significantly higher-fractured
base at the north-east and the north-west view is re-
flected in the higher medians. The median finally
decreases continuously from north-east, north-west,
and south-east to south west, the latter having the
largest ratio between smooth facade and windows.
Compared with the smooth cubic buildings, it is
striking, that the results are similarly high, at least for
Figure 5
Results for timber
constructions and
conversions of
existing buildings
a certain range of scale. This leads to the conclusion
that the differences in visual complexity do not de-
pend upon the fact whether the building belongs to
the group of cubic buildings or of classic villas.
Timber construction
Among the analysed buildings timber constructions
achieve the highest values (figure 5: Khuner and
Müller). These wooden constructions already an-
nounce a tendency towards later modern move-
ments (Hitchcock 1994). A stone plinth of the Khuner
country House serves as a base for the wooden con-
struction with a flat sloping gable-roof, an appropri-
ate roof for the pre-alpine area. Constantly high re-
sults of the measurements are due to the design-
related pattern of smaller elements from the win-
dows to the exposed wood trunks (blockhouse). At
the same time, the number of components increases
when their size decreases - from the whole over larger
cuts over windows and finally to the timber structure.
Inside, following the Raumplan, the two-storey living
hall, providing access to the bedrooms via a three
sided gallery, is the centre of the house. This is dis-
played to the outside via a large window front.
The outer appearance of the House project for
Dr. Müller's daughter, also called "the last House"
(Risselada 1988), is characterized by the separation of
the individual layers of wood parts. This again leads
to a steeper data curve in the log-log graph. All re-
sults of timber constructions are of similar height (fig-
ure 5) and have a similar scale range. The only excep-
tion concerns the south facade of the house project,
due to the outer smoothly plastered chimney.
Conversion of existing buildings
The existing structure strongly influences and limits
house rebuilding, especially on the interior and ex-
tensions. While Steiner House and Scheu House only
have a flat space sequence, Villa Mandl already in-
cludes a two-storey hall. However, Strasser House
with different spatial levels is finally one of two con-
versions that featured Raumplan for the first time (Ris-
selada 1988). Reconstructing the staircase opens up
spaces from intermediate landings. Additional ele-
ments, for instance bay windows, loggia and a round
tower, dominate the completely redesigned outer fa-
cades (Risselada 1988).
CONCLUSION
The box-counting method is suitable for comparing
different elevations. It measures the complexity for
a certain range of scales. Contrary to the assump-
tion that a building by Loos is smooth from high to
low scale (due to his rejection of ornament) some
results underline high complexity in a broad range
of scales. This is true for the timber constructions,
but also for the classic villa-style architecture. As ex-
pected, the wood design results in the highest val-
ues because they appear roughest. It is interesting
that the depth of scales, the range which offers corre-
lation, is not necessarily different from the cubic ex-
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 503
 amples. While the first examples of cubic style still
have a sharp bend in the data curve similar to the
Villa Savoye by Le Corbusier, later ones, for instance
Bojko House project, seem to be more balanced. In
any case it is obvious that Loos' rejection of the use of
(contemporary) ornament does not lead to a uniform
expression of complexity. Facades of various degrees
of complexity, expressed in the data set, accompany
Loos' entire creative period. This shows that Loos'
creative power cannot be reduced to any uniform
dogma, but rather reacts to different requirements.
The recurring classical elements are rather a reinter-
pretation than an unconditional dedication to an en-
tirely new radically smooth surface design and fit into
his entire design concept.
REFERENCES
Bechhoefer, W and Bovill, C 1994 'Fractal Analysis of Tra-
ditional Housing in Amasya, Turkey', IAESTE, Tradi-
tional Dwellings and Settlements Working Paper Se-
ries, Vol. 61, Tunis, pp. 1-21
Bovill, C 1996, Fractal Geometry in Architecture and De-
sign, Birkhäuser, Boston, Mass
Foroutan-pour, K, Dutilleul, P and Smith, DL 1999, 'Ad-
vances in the implementation of the box-counting
method of fractal dimension estimation', Applied
Mathematics and Computation, 105(2-3), pp. 195-
210
Hitchcock, H-R 1994, Die Architektur des 19. und 20.
Jahrhunderts, Aries Verlag, München
Khan, H-U 1998, Taschens Weltarchitektur: International
Style – Architektur der Moderne von 1925 bis 1965,
Benedikt Taschen Verlag, Köln
Kulka, H 1931, Adolf Loos. Das Werk des Architekten, An-
ton Schroll, Wien
Kulka, H 1970, 'Bekenntnis zu Adolf Loos', Alte und mod-
erne Kunst, XV(113), p. 24
Kurrent, F, Horn, S and Weidlein, M (eds) 1998, Adolf Loos
1870-1933, 40 Houses, Anton Pustet, Salzburg
Kühn, C 1989, Das Schöne, das Wahre und das Richtige,
Birkhäuser
Loos, A 1997 [1908], 'Ornament und Verbrechen', in Opel,
A (eds) 1997 [1908], Trotzdem, Prachner, Wien, pp.
78-88
Loos, A 1997 [1909], 'Architektur', in Opel, A (eds) 1997
[1909], Trotzdem, Prachner, Wien, pp. 90-104
504 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Loos, A 1997 [1924], 'Ornament und Erziehung', in Opel,
A (eds) 1997 [1924], Trotzdem, Prachner, Wien, pp.
173-179
Lorenz, WE 2003, Fractals and Fractal Architecture, Mas-
ter's Thesis, Vienna University of Technology
Lorenz, WE 2009 'Fractal Geometry of Architecture –
Implementation of the Box-Counting Method in a
CAD-Software', Proceedings of eCAADe 2009, Istan-
bul, pp. 697-704
Lorenz, WE 2012 'Estimating the Fractal Dimension of
Architecture: Using two Measurement Methods
implemented in AutoCAD by VBA', Proceedings of
eCAADe 2012, Prague, p. 505–514
Lorenz, WE 2013, Fraktalähnliche Architektur - Einteilung
und Messbarkeit: ein Programm in VBA für AutoCAD,
Ph.D. Thesis, Vienna University of Technology
Maertens, H 1884, Der optische Maßstab: oder die Theorie
und Praxis des a�sthetischen Sehens in den bildenden
Ku�nsten, Wasmuth, Berlin
Mandelbrot, BB 1981, 'Scalebound or scaling shapes: A
useful distinction in the visual arts and in the natu-
ral sciences', Leonardo, 14, pp. 45-47
Mandelbrot, BB 1982, The fractal geometry of nature, W.H.
Freeman, San Francisco
Ostwald, MJ, Vaughan, J and Tucker, C 2008, 'Characteris-
tic Visual Complexity: Fractal Dimensions in the Ar-
chitecture of Frank Lloyd Wright and Le Corbusier',
Nexus: Architecture and Mathematics, VII, pp. 217-
231
Richardson, LF 1961, 'The problem of contiguity: an ap-
pendix of statistics of deadly quarrels', General Sys-
tems Yearbook, 6, pp. 139-187
Risselada, M 1988, Raumplan versus Plan Libre. Adolf Loos
and Le Corbusier 1919-1930, Delft University Press,
Delft
Scheu, R 1909, 'Adolf Loos', Die Fackel, 283-84, pp. 25-37
Tournikiotis, P 1994, Adolf Loos, Princeton Architectural
Press, New York
Ungers, L 2002, Über Architekten, Dumont Literatur und
Kunst Verlag, Köln
Unsupervised Symmetric Polygon Mesh Mapping
The Dualism of Mesh Representation and Its Implementation for Many
Layered Self-Organizing Map Architectures

Matthias Standfest
ETH Zürich, IA - Chair of Information Architecture, Switzerland
http://www.ia.arch.ethz.ch
standfest@arch.ethz.ch

With this paper we present a fully automated semantic shape similarity detection
based on N-rings with further potential for shape synthesis in a topological
correct feature space. Therefore a way of symmetric encoding of geometry,
optimized for the use as feature-vector in self-organizing maps, is introduced.
Furthermore we present a modified kernel for the detection of the best matching
unit in self-organizing maps especially designed for a data topology differing
from the default predecessor/successor structure. Finally we provide the results
of a conducted experiment clustering building blocks of an area in Zürich,
Switzerland.

Keywords: Unsupervised Machine Learning, Geometry Clustering,

Self-organizing Map, Mesh Synthesis, Probabilistic Modelling

BACKGROUND features of an underlying architecture (cf. Langen-

One intrinsic problem of the architectural design pro-
cess is the issue of how to compare different designs
and how to interpolate between them. As a neces-
sity for any synthesis driven design approach this
wicked problem has yet to be fully automatized, es-
pecially in consideration that existing methods fail in
practicability because of expensive manual data pre-
processing. So far in the field of architecture those
methods can be divided into two main groups, cor-
responding to the particular kinds of compared data:
either the comparison of shape determining rules in-
stead of a shape itself (cf. Stiny, Mitchell 1978; cf.
Fezer 2009; cf. Eastman 1972; cf. Rittel 1977; cf.
Alexander et al. 1977; cf. Dave et al. 1994) or the com-
parison of key figures computed as arbitrary shape
han et al. 2011; cf. Dillenburger 2010; cf. Derix et al.
2012). Clearly, for both approaches pre-processing
of data in different ways is necessary. While the first
group is focusing on the analysis of decision trees
leading to a shape, in order to allow sustaining a lim-
ited ability for synthesizing interpolated shapes, ex-
actly this ability rapidly decreases with the increasing
complexity of the solution-space; up to a point where
finally generalizable approaches like the highly ab-
stract Case Based Reasoning (cf. Dave et al. 1994)
need to fully sacrifice the ability for shape-synthesis
in order to maintain a certain kind of interpretable
and general ontology. Additionally, when compar-
ing shape determining rules a larger solution space
is mainly achieved by increasing the complexity of

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 505

 the underlying rule sets. This problem can be partly
avoided by comparing a set of designs according to
arbitrary figures (cf. Dillenburger 2010; cf. Derix et
al. 2012) or topology informed labels (cf. Langenhan
et al. 2011) rather than the decision trees describ-
ing the shapes preceding design processes. Such
highly abstract and generalized models for comput-
ing key figures provide easy applicable semantics for
a much larger solution space, because they mini-
mize the costs for a useful comparison by reducing
the amount of semantic information needed. While
these methods render only a handful basic architec-
tural elements required for being labelled during the
data pre-processing stage, the figures used for com-
parison lack the potential of being translated unam-
biguously into architectural shapes. Thus we iden-
tified that the potential for shape-synthesis on the
one hand and the generalizability of the labelling
method on the other hand tend to be mutually ex-
clusive. However, for a seamless integration of a pro-
cedure of this kind into any architectural design pro-
cess it is a necessity to provide both analysis and syn-
thesis capabilities. In order to do so and to provide
exactly those discussed processes this paper investi-
gates, inspired by actual developments in the field of
computer graphics, the potential of a method based
on a purely geometric representation of architecture.
The goal was to work with the latent semantics hid-
den underneath the topological contexts of the mesh
geometries.
METHOD
Already known from different methods for architec-
ture or urban planning (cf. Stadler et al. 2007),
most approaches in computer graphics need to pre-
process the purely geometrical information provided
for enriching it with certain semantic or topological
information first. The main focus hereby lies on the
segmentation into semantically tagged shapes (cf.
Chaudhuri et al. 2011) which is necessary to compare
and cluster shape elements according to different ge-
ometrical or topological features (cf. Funkhouser et
al. 2004) or to automatically label identical functions
506 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
(cf. van Kaick et al. 2013). Others even established a
workflow where this information is used to combine
existing shapes into meaningful new combinations
(cf. Chaudhuri et al. 2011). And thereby they provide
methods for both analysis and synthesis of polygon
meshes - the missing link for the architectural design
process. While the main problem of those techniques
is the costly supervised or even manual segmenta-
tion, a problem also leading to a growing demand for
pre-segmented data repositories (cf. van Kaick et al.
2013), this paper presents a way of substituting the
necessary semantic information by a purely contex-
tual one, analogue to unsupervised image (cf. Bengio
2009) and text processing (cf. Bellegarda 2004).
Most methods in clustering or probabilistic mod-
elling of geometry are either based on arbitrary fea-
ture descriptors for the mesh segments like angles,
diameters and much more complicated ones (cf. Sidi
et al. 2011) or topology informed semantic labeling
(cf. Kalogerakis et al. 2012). We already described
those tactisc when previously mentioning the ways
of comparing designs in the field of CAAD. The two
problems arising are yet again the need for large la-
beled (pre-segmented) data sets and, regarding the
feature descriptors, the difficulty of how to decide
which of the vast amount of possible descriptors are
relevant for clustering in general or if not a contextual
approach would be preferable to the intrinsic one (cf.
Standfest et al. 2013). Both problems are shown to
be avoidable by using unsupervised learning meth-
ods for creating high-level features from only unla-
beled data (cf. Le et al. 2011). And undoubtedly
this is also part of the observable trend towards min-
imizing the amount of semantic information needed
for state of the art data analysis. To contribute to
this development, this paper is focusing on the prob-
lem of comparing unlabelled geometry data, by us-
ing both many layered neural networks and n-gram
inspired mesh processing. We show how emergent
descriptors allow clustering of architecture compara-
ble to state of the art approaches in geometry cluster-
ing while the same method still manages to maintain
synthesizing abilities (figure 1).
Figure 1
result of the
conducted
experiment on
clustering 48
randomly chosen
building blocks
according to the
latent semantics of
the unlabelled
mesh geometry

1999). Adopted for constrained Delaunay triangula-

Figure 2 tion (cf. Chew 1987) this means we work with the
constructing a triangle relations instead of the triangles themselves.
2-ring (right) as a A shift easily accomplished by constructing the fold-
collection of the ing transformation for each pair of neighbours (fig-
outer leaves of the ure 3). We define this fold by one rotation angle a
nested 1-rings and one pair of relative uv transformations. Due to
(middle) with the Nested N-Rings the fact that certain angles appear much more often
corresponding root As an alternative to the popular goal of achieving se- than others (e.g. 180° where the mesh is planar or
1-ring (left) mantically correct sub graphs (cf. Kalogerakis et al. 90° as the most common edge) this design is espe-
2012) the segmentation of polygon meshes may also cially optimized for architectural or urban geometries
aim for same sized patches. This can be done either and supports thereby subsequent clustering efforts
via intersecting the mesh with r radius balls (cf. Mi- (as does minimizing the number of triangles by using
tra et al. 2006) or as in our case by creating patches constrained Delaunay triangulation).
of similar underlying topology. Like N-grams in lan-
guage processing (Shannon 2001 [1948]), small topo-
Figure 3
logic configurations instead of predefined subdivi-
triangle folding
sions now carry latent semantics. Therefore we chose
process, showing
to compare each triangle according to its topologic
the rotation axis as
neighbourhood, a format we further call N-ring (fig-
well as the u- and
ure 2).
v-transformation
These N-rings of each triangle are strongly re-
lated to N-grams and are used to replace, from nat-
ural language strongly influenced, standard seman-
tics with its statistical model (cf. Bellegarda 2004).
To achieve a fully contextual encoding without in-
trinsic semantics (cf. Deleuze et al. 2008), we build
upon the dualism in mesh representation already
used for polygon mesh compression (cf. Rossignac

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 507


Analogue to this contextual encoding paradigm,
N-rings with distance greater than one edge may be
composed out of smaller rings as alternative to con-
structing them from scratch, allowing the hierarchi-
cal processing of geometry (figure 4). For robust-
ness reasons we designed those nested rings to cre-
ate overlapping branches, resulting in the compos-
ing of a N-ring of size i out of four N-rings of size i − 1
or 4i−1 N-rings of size 1. This way we ensure a fixed
data topology for the whole recursive process. The
presented kind of segmentation is fully unsupervised
and works with any unlabelled triangular 2-manifold
polygon mesh without boundary. We optimize the
meshes before converting them into training vec-
tors (as basis for the obligatory Delaunay triangula-
tion planar straight-line graphs are detected first af-
ter reducing some noise like wielding vertices or flat-
tening of peaks). Nevertheless the underlying data
encoding topology differs fundamentally from well
investigated predecessor-successor patterns of data
streams (like text or audio signals) causing conse-
quences for the further below described best match-
ing unit (BMU) kernel of the used self-organizing map
(SOM).
Mesh-SOM
The presented mesh segmentation allows working
with a fixed data topology for every nested N-ring of
arbitrary size in form of a tree graph with one root
node and three leave nodes, thus a star S3 with 3
edges what we further call a claw. Every node of
the graph is again a nested claw itself (with at least
one overlapping edge for connection purposes) and
may be decomposed until the lowest order of unir-
ing representation is reached (figure 5). Because the
SOM is trained layer by layer the conceptual complex-
508 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
ity of each level increases and therefore bigger trees Figure 4
may be interpreted as more abstract features than composing a higher
smaller ones - a common method when compared level ring out of
to multi-level usage of Gabor filters (cf. Gabor 1946) four lower level
or N-grams which are essential for deep believe net- rings (red: root,
works (cf. Bengio 2009). This means that we compare dark green: leave A,
the triangles not only by its N-rings but according to cyan: leave B, light
their immediate but increasingly detailed neighbour- green: leave C)
hood. As a result overlapping edges (a third of the
branches is oriented towards the root claw) improve
the robustness of our encoding significantly.
Figure 5
nested hierarchical
data topology (red:
root, dark green:
leave A, cyan: leave
B, light green: leave
C) of level 1 (not
nested), level 2
Replacing predecessor/successor-pairs with (biring) and level 3
claws leads to a further change in the feature vec- (3-ring)
tor managing of our process. The clearly identifiable
root ring and the order of the claw's leaves deter-
mined by the normal of the initial triangle result in
only one remaining ambiguity: the index of the ini-
tial leave. When converted into a vector the values
of the root node are always the first few dimensions,
but the following topology of the three neighbour-
ing rings (the leaves) forms a 3x3 circulant matrix
C. For synthesis each nested claw (subring) needs to
be rotated for finding the matching fold of the root
claw (figure 6), because of this uncertainty. On the
other hand, for the analysis phase the BMU kernel
is adopted to find the best of three variations (the
most frequent initial claw) of a feature vector and is
thereby increasing the quality of its SOM.
Figure 6
circular data
topology leaves 3
possible index
leaves per claw
 This limitation to three possible combinations
per vector per SOM level allows to approach the un-
derlying combinatorical problem in a nested recur-
sive manner (figure 7) which proofs to be much faster
than proceeding just the direct way. Only the ini-
tial learning level has twice the amount of variations
to be compared. The simple reason is that we try
to further densify the solution space of the SOM by
the possibility of flipping the normal of each triangle
(changing the order of its neighbours). Finally com-
paring the different variations of each vector is done
by adopting the used GPU SOM BMU kernel (cf. Wit-
tek 2013) to not only find the best matching unit, but
the best matching unit possible out of n variations -
we call this kernel further BMUofN. This modification
is the very reason why we are able to produce nec-
essary findings otherwise impossible to compute in
a time sensitive manner (table 1).
Figure 7
computational
costs due to
possible tree
combinations
Table 1
relation between
recursive and direct
data processing
The rest of the process is basically following the
2-grams based web-SOM method for uncovering la-
tent semantics of texts (cf. Ritter et al. 1989): For
each N-ring level one SOM is first sorting the con-
cepts (the different kinds of triangles) according to
its contexts in the form as described. For better train-
ing results those feature vectors are also standardized
but not normalized (cf. Blayo 1992). Subsequently a
so called domain map for each polygon mesh is cre-
ated. This is done by creating the domain feature
vectors on basis of histograms over the triangle map
which again is smoothened by a Gaussian convolu-
tion kernel (cf. Kaski et al. 1998). As a result we
get SOMs arranging the compared meshes accord-
ing to lower or higher level emergent features which
establishes a process of strictly unsupervised cluster-
ing of unlabelled architectural geometries. Summing
up, we present a way of algorithmic modelling (cf.
Breiman 2001) geometrical semantics as an alterna-
tive to the intrinsic ones provided by semantically en-
riched pre-processed data.
RESULTS
For a first test of the method we compared a set of
48 building blocks (the level of detail is 1) randomly
chosen from the area of Zürich Altstetten, leading
to the comparison of 3064 triangles in total. The
resulting U-matrices of the triangle maps (figure 8)
clearly show significant clustering of similar triangle
contexts. Although this kind of semantics may dif-
fer from the kind the human mind would create. Fur-
ther research needs to be conducted in how far this
algorithmic modelled one is able to be used to sub-
stitute previously missing semantic classes (cf. Sidi
et al. 2011). In addition to the emergent clustering
we observed that our modified BMU kernel leads to
much clearer results, e.g. at the initial training level
(figure 8). Finally the rendering of the triangle map
of level 3 reveals the lack of necessary additional data
to do high quality high level abstraction. A problem
which is in accordance to the existing studies show-
ing the necessity of big data sets for deep learning
algorithms (cf. Le et al. 2011). This underdetermina-
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 509
 Figure 8
triangle maps with
toroid topology,
t.l.t.b.r.: level 1
(default kernel),
level 1 (modified
kernel), level 2
(modified kernel),
level 3 (modified
kernel). The color
scale reflects the
cluster borders,
blue cells are close
neighbours while
bright ones are
further apart.

tion of the achieved clustering as result of deficient
data is closely related to the concept of algorithmic
modeling itself (cf. Lyre 2004).
To quantify the quality of the SOMs regarding the
clustering for analysis purposes as well as the accu-
racy for synthesis purposes we used well established
figures like the topographic error and the quantiza-
tion error (cf. Uriarte et al. 2008). As a modification
of the last one mentioned we calculated the aver-
age Euclidian distance not only over the whole fea-
ture vector but instead over the single three dimen-
Table 2
sional components each ring consists of. Thus the
SOM statistics (level
component quantization error (QC -Error) is more re-
1* is calculated with
vealing in this context and is better suited for com-
default BMU kernel
paring the different SOM levels on a unified basis (ta-
method)
ble 2). The computed statistics of the test scenario
emphasize the impressions we already got from U-
matrix renderings: first of all the significant differ-
ences between the default BMU kernel and our modi-
fied one which is densifying the solution space. Then

510 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 9
domain maps of
level 1 (based on
the distribution of
unirings), 2 (based
on the collection of
birings) and 3
(based on further
nested 3-rings)

we observe the lack of data resulting in a rather great
topographic error within our most abstract SOM. And
finally we get positive results of our second level tri-
angle map regarding a convincing synthesizability.
Figure 10
sample deviation of
trained level-3
feature vector to its
original before (top)
and after (bottom)
leveling the error by
using the
underlying
redundancy
Despite the still limited dataset the three result-
ing domain maps already show reasonable clustering
and arrangements of the evaluated building blocks
(figure 9). Regardless of the significant quality differ-
ences between the underlying triangle maps of level
two and three the two corresponding domain maps
of these levels show surprisingly few differences. This
is partly caused by the Gaussian convolution filter we
applied to smoothen the histograms and partly result
of a probable limit of reasonable N-ring sizes - a limit
research in phrase-based language processing as al-
ready suggested (cf. Zollmann et al. 2008).
Finally our tests confirm the self-stabilizing ef-
fect of our robust nested encoding method (figure
10). When re-translating, the produced QC-errors of
level 3 feature vectors are between 30% and 40%
smaller than without redundancy and appear espe-
cially small in close neighborhood to the root trian-
gle.
CONCLUSION
The dominant outcome of this paper is a fully au-
tomated shape similarity detection with the poten-
tial for shape synthesis in a topological correct fea-
ture space. Therefore a way of symmetric encod-
ing of geometry, optimized for the use as feature-
vector in self-organizing maps, is introduced. The
conducted experiments further illustrate how differ-
ent unlabelled polygon meshes can be aligned ac-
cording to latent semantics. In future this could
have the potential of rendering expensive, manu-
ally edited, semantically enriched geometry reposi-
tories obsolete.The successful implementation of the
described techniques is directly connected to a vast
number of design related theories. One can interpret
the synthesized scenarios as concretization of possi-
bilities (cf. Flusser 1994) while the nodes of the SOM
itself can be viewed from an actor-network theory
perspective (cf. Wassermann 2010). The underlying
duality of the mesh encoding is discussed as double
articulation (cf. Deleuze et al. 2008) and the appli-

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 511

 cation of the method enforces the opinion that de-
sign is rather a redesign than design ex nihilo (cf. La-
tour 2009). Of course, this also has an effect on the
discussions of authorship in architecture (cf. Carpo
2011) and on the role of geometric representations in
the design process and its communication (cf. Evans
2011).
Further consequences for the design process are
at least as manifold as the theoretic implications: in
future a generalizable analysis of architectural and ur-
ban structures is a necessity not only for synthesis
driven design approaches but also for any kind of ar-
chitectural impact assessment. Especially in consid-
eration of Big Data and algorithmic modelling, this
process is able to form a geometric data stream so
that other streams can be mapped onto, so data min-
ing processes can finally re-manifest into geometric
output. As a next step in the development of this
methodology we are working on modified point set
registration (cf. Gelfand et al. 2005; Wang et al. 2008)
for re-translating the feature vectors of the domain
maps back into polygon meshes to better visualize
the output of the trained probabilistic shape space.
REFERENCES
Alexander, C, Ishikawa, S and Silverstein, M 1977, A pat-
tern Language, Oxford University Press, New York
Bellegarda, JR 2004, 'Latent Semantic Language Mod-
eling for Speech Recognition', in Johnson, M (eds)
2004, Mathematical foundations of speech and lan-
guage processing, Springer Science + Business Me-
dia, New York, pp. 73-103
Bengio, Y 2009, Learning deep architectures for AI, Now
Publishers, Hanover, Mass
Blayo, F 1992, 'Kohonen Self-Organizing Maps: Is the
Normalization Necessary?', Complex Systems, 6(6),
pp. 105-123
Breiman, L 2001, 'Statistical modeling: The two cultures',
Statistical Science, 16(3), pp. 199-231
Carpo, M 2011, The alphabet and the algorithm, MIT Press,
Cambridge, Mass
Chaudhuri, S, Kalogerakis, E, Guibas, L and Koltun, V
2011, 'Probabilistic reasoning for assembly-based
3D modeling', ACM Transactions on Graphics, 4, pp.
35-43
Chew, LP 1987 'Constrained Delaunay Triangulation',
512 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Proceedings of the third annual symposium on Com-
putational geometry, pp. 215-222
Dave, B, Schmitt, G, Faltings, B and Smith, I 1994 'Case-
based design in Architecture', Artificial intelligence in
design, pp. 145-162
Deleuze, G, Guattari, F and Massumi, B 2008, A thousand
plateaus, Continuum, London
Derix, C, Gamlesæter, Å, Miranda, P, Helme, L and Kropf,
K 2012, 'Simulation Heuristics for Urban Design', in
Müller Arisona, S, Aschwanden, G, Halatsch, J and
Wonka, P (eds) 2012, Digital urban modeling and sim-
ulation, Springer, Berlin, Heidelberg, pp. 159-180
Dillenburger, B 2010 'Space Index: A retrieval-system for
building-plots', Proceedings of eCAADe 2010, Zürich,
p. 893–899
Eastman, CM 1972, 'Preliminary report on a system for
general space planning', Commun. ACM, 15(2), pp.
76-87
Evans, R 2011, 'Durch Papier Sehen', in Voorhoeve, J
(eds) 2011, Welten schaffen, Diaphanes, Zürich, p.
157–194
Fezer, J 2009, 'A Non-Sentimental Argument', in Geth-
mann, D and Hauser, S (eds) 2009, Kulturtechnik En-
twerfen, Transcript, Bielefeld, p. 287–304
Flusser, V 1994, 'Vom Subjekt zum Projekt', in Bollmann,
S and Flusser, E (eds) 1994, Schriften, Bollmann, Ben-
sheim, pp. 9-160
Funkhouser, T, Kazhdan, M, Shilane, P, Min, P, Kiefer, W,
Tal, A and Rusinkiewicz, S 2004, 'Modeling by exam-
ple', ACM Transactions on Graphics (TOG), 23(3), pp.
652-663
Gabor, D 1946, 'Theory of communication. Part 1: The
analysis of information', Electrical Engineers-Part III:
Radio and Communication Engineering, 93(26), pp.
429-441
Gelfand, N, Mitra, NJ, Guibas, L and Pottmann, H 2005
'Robust Global Registration', Eurographics Sympo-
sium on Geometry Processing 2005
van Kaick, O, Xu, K, Zhang, H, Wang, Y, Sun, S, Shamir,
A and Cohen-Or, D 2013, 'Co-hierarchical analysis
of shape structures', ACM Transactions on Graphics,
32(4), pp. 69:1-69:10
Kalogerakis, E, Chaudhuri, S, Koller, D and Koltun, V 2012,
'A probabilistic model for component-based shape
synthesis', ACM Transactions on Graphics, 31(4), p. 55
Kaski, S, Honkela, T, Lagus, K and Kohonen, T 1998,
'WEBSOM–self-organizing maps of document col-
lections', Neurocomputing, 21(1), pp. 101-117
Langenhan, C, Haß, S, Weber, M, Petzold, F, Liwicki, M
and Dengel, A 2011 'Investigating research strate-
gies for accessing knowledge stored in semantic
 models', Proceedings of eCAADe 2011, Ljubljana, pp. SJ 2008, 'Simultaneous nonrigid registration of mul-
403-411 tiple point sets and atlas construction', IEEE Trans
Latour, B 2009 'A Cautious Prometheus ? A Few Steps To- Pattern Anal Mach Intell (IEEE transactions on pattern
ward a Philosophy of Design', Proceedings of the 2008 analysis and machine intelligence), 30(11), pp. 2011-
Annual International Conference of the Design History 2022
Society, pp. 2-10 Wassermann, K 2010 'SOMcity: Networks, Probability,
Le, QV, Ranzato, MA, Monga, R, Devin, M, Chen, K, Cor- the City, and its Context', Proceedings of eCAADe
rado, GS, Dean, J and Ng, AY 2012 'Building high- 2010, Zürich, pp. 197-205
level features using large scale unsupervised learn- Wittek, P 2013, 'Somoclu: An Efficient Distributed
ing', Proceedings of the 29th International Conference Library for Self-Organizing Maps', Preprint -
on Machine Learning, Edinburgh, Scotland, UK http://arxiv.org/pdf/1305.1422, tba, p. tba
Lyre, H 2004, Lokale Symmetrien und Wirklichkeit, Mentis, Zollmann, A, Venugopal, A, Och, F and Ponte, J 2008 'A
Paderborn systematic comparison of phrase-based, hierarchi-
Mattausch, O, Panozzo, D, Mura, C, Sorkine-Hornung, O cal and syntax-augmented statistical MT', Proceed-
and Pajarola, R 2014, 'Object Detection and Clas- ings of the 22nd International Conference on Compu-
sification from Large-Scale Cluttered Indoor Scans', tational Linguistics, Manchester, pp. 1145-1152
Computer Graphics Forum, 33(2), p. to appear
Mitra, NJ, Guibas, L, Giesen, J and Pauly, M 2006 'Prob-
abilistic fingerprints for shapes', Proceedings of Eu-
rographics Symposium on Geometry Processing, pp.
121-130
Rittel, HWJ 1977, On the Planning Crisis: Systems Analy-
sis of the" first and Second Generations.", Institut für
Grundlagen der Planung IA, Universität Stuttgart,
Stuttgart
Ritter, H and Kohonen, T 1989, 'Self-organizing semantic
maps', Biological Cybernetics, 61(4), pp. 241-254
Rossignac, J 1999, 'Edgebreaker: Connectivity compres-
sion for triangle meshes', Visualization and Computer
Graphics, 5(1), pp. 47-61
Shannon, CE 2001 [1948], 'A mathematical theory of
communication', Mob. Comput. Commun. Rev. (ACM
SIGMOBILE Mobile Computing and Communications
Review), 5(1), pp. 3-55
Sidi, O, van Kaick, O, Kleiman, Y, Zhang, H and Cohen-Or,
D 2011, 'Unsupervised co-segmentation of a set of
shapes via descriptor-space spectral clustering', ACM
Transactions on Graphics, 30(6), p. 126
Stadler, A and Kolbe, TH 2007 'Spatio-semantic coher-
ence in the integration of 3D city models', Proceed-
ings of the 5th International ISPRS Symposium on Spa-
tial Data Quality, Enschede
Standfest, M and Moosavi, S 2013 'Fuzzy Modelling with
Self-Organizing Maps', Proceedings of Design Mod-
elling Symposium 2013, Berlin, p. 243–256
Stiny, G and Mitchell, WJ 1978, 'The palladian grammar',
Environment and planning B, 5(1), pp. 5-18
Uriarte, EA and Martin, FD 2008, 'Topology Preservation
in SOM', International Journal of Applied Mathematics
and Computer Sciences, 2(9), pp. 874-877
Wang, F, Vemuri, BC, Rangarajan, A and Eisenschenk,

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 513

514 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Constructing a shape grammar. The Ducal palace façade
Constructing a shape grammar. The Ducal palace façade

Filipe Coutinho1 , José Pinto Duarte2 , Mario Kruger3

1,3
Departamento de Arquitectura da Universidade de Coimbra
2
Faculdade de Arquitectura da Universidade de Lisboa
1
filipecoutinho@darq.uc.pt 2 jduarte@fa.utl.pt 3 kruger@ci.uc.pt

This paper shows the way the Ducal palace façade shape grammar was built,
evidencing some steps involved in the process, its derivation and partial outputs
gained from the grammar evaluation.

Keywords: Alberti, shape grammar, rule base design, design automation,

transformation in design

INTRODUCTION have knowledge of the Renaissance style.

After 4 decades of shape grammar research there
is not (m)any document showing systematically the
process of shape grammars construction. This arti-
cle aims to trigger the subject in this particular field,
showing not just its relevance in terms of enlighten-
ment of a procedure (if it is possible), but also, the ne-
cessity to provide future grammarians with a synthe-
sis of others approach. To illustrate it, a brief presen-
tation of the main steps taken to achieve the Ducal
palace façade in Vila Viçosa, Alentejo, Portugal gram-
mar and its interpretation is presented.
The construction of the Ducal palace façade
in Vila Viçosa, Alentejo, Portugal took place circa
1532/1537 extending the existing palace called Paço
de D. Jaime (after 1513).
It was commissioned by D. Teodósio I, having
110m long facade with 23 bays, being one of the
longest in Iberic peninsula of its times. According
with its patron, the formal and ornamental solution
should be in romano style.
Several of the architects and builders involved in
the design, such Nicolau de Frias, Pero Vaz Pereira
and Manuel Pereira Alvenéo (Teixeira, 1997), may
Theresearch present in this paper was part of
Alberti Digital research project finished in Septem-
ber 2013. The goal of the project was decoding
Alberti's treatise De Re Aedificatoria (Alberti, 2011)
by inferring the corresponding shape grammar us-
ing the computational framework provided by de-
scription grammars (Stiny, 1981) and shape gram-
mars (Stiny and Gips, 1972) and the transformations
in Languages of designs (Knight, 1994) and the for-
malism involved in such transformations.
The main propose was to compare the gram-
mar of the treatise with the grammar of actual build-
ings to determine the extension of Alberti's influence
on Portuguese architecture in the counter-reform pe-
riod, comparing the grammar of the treatise with the
grammar of existing buildings to determine the ex-
tension of Alberti's influence on Portuguese archi-
tecture in the counter-reform period (Kruger, 2011).
The Ducal palace façade shape grammar is part of
a wider grammar named column systematization
shape grammar.

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 515

 Figure 1
Fill in the gaps
diagram.

Figure 2
Point cloud model.
Elevation, plan and
section of the Ducal
palace façade.

516 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

METHODOLOGY
The Ducal palace façade shape grammar is a par-
allel grammar that used different grammars to es-
tablish part of their rules. Two of those grammars
rules were extracted from Alberti`s treatise (the col-
umn system and the intercolumn shape grammars).
A third generic grammar was constructed to provide
rules specifically to apply those grammars.
This paper will briefly show 4 different phases of
work. Structurally they are organised as: 1) The ele-
ments used to construct the grammar; 2)The generic
rules used in to generate Ducal palace façade; 3)
Computational implementation and visualisation of
the grammar; and finally 4) Evaluation process.
NOTES ON THE SHAPE GRAMMAR CON-
STRUCTION.
The treatise can't provide all the information needed
to build a grammar able to generate the desired fa-
cade. By that fact a strategy was adopted to full fill
those blank spaces in the treatise. That is the pas-
sages with contradictory descriptions or even erased
passages or even lack of information. A strategy was
prepared called fill in the gaps (FITG). Looking closer
to FITG it is possible to see diagrammatically (see Fig-
ure 1) the steps needed to build such a grammar.
Ducal palace in Vila Viçosa was the building men-
tioned to be generated with the grammar. There
was a previous survey of the building using a phase
based laser scanner (FARO Focus 3D) and 24 different
stations were considered from which 8 point clouds
were done. A merging of all this point clouds may be
seen in Figure 1.
Constructing a grammar is all except a straight
pass. In this process it was used the LCS system, the
musical consonances and the numbers descripted in
the book IX, chapters 5 and 6. Finally, information
was gained from real buildings and models made out
of those buildings. See diagram 1.
The LCS system is a system described by Alberti
at Book VII Chapter VII. It represents the way to gen-
erate the ornament mould with letters, like: an L that
does a Fillet, a L+C does an ovolo, a L + C (reversed)
does a channel. A L+S does a gullet, and finally a L+
S (reversed) does a wave.
The numbers represented in the Book IX that
where taken in consideration are the ones from mu-
sical consonances representing a system of propor-
tions, like: 2:3 (Diapente); 3:4 (Diatesseron); 1:2 (Dia-
pason); 1:3 (Diapason-diapente); 8:9 (Sesquioitava) e
8:7 (sequiseptima). Also taken in consideration were
the Innate proportions represented by the roots √2,
√3, √8, √12 e √16 = 4.
The data used to analyse the different build-
ings from Alberti was based in photographic survey-
ing, 2D drawings and some measurements taken in
the site. The same procedure was taken at the Por-
tuguese architecture in the counter-reform period.
Two of those buildings were surveyed in detail using
a reversed modelling process. With that a cloud point
model was obtained from a Terrestrial Laser Scanning
(Figure 2).
TRANSFORMED AND GENERIC RULES
The grammar was obtained with the column sys-
tem and the intercolumn grammars and few other
rules. The goal of this research is to see the degree
of coincidence between the rules in use on the two
grammars mentioned before and ones of the Ducal
palace façade shape grammar. Some rules had to be
transformed, that is, added, subtracted, and changed
(repositioned and resized).
To generate the facade some of the rules of the
treatise had to be transformed with immediate im-
pact on their parameters. These parameters were
then compared with the ones from the treatises. . In
Figure 3 some transformations were made in order to
achieve a corinthian capital similar to the one exist-
ing in the Ducal palace façade. The corinthian cap-
ital described by Alberti has 8 leafs spread in 2 lev-
els. The one existing in the Ducal palace as 3 leafs. A
transformation was operated resulting in a new con-
figuration and new proportions, namely, its height
that changed from 1/3D (that is 2parts of the entire
height) to 1D (that is 6parts of the entire height).
As the treatise does not fully specify how to
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 517
 Figure 3
Transformations of
the rule 7 cc of the
Corinthian capital

Figure 4
Set of rules of
recognition

518 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 5
Meta-structure
rules

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 519

 generate a facade, some rules had to be produced.
These rules are generic rules and they comprehend
specifications for different generative process scales.
Those scales are large (e.g.related to site plot and/or
pré-existing structure recognition), medium (focused
on the elements of the column system), and small
(regarding the constituent parts of the elements of
the column system), leading to 5 different derivation
stages organizing the subsequent generic grammar
that as:1 - Recognition rules; 2 - Meta structure rules;
3 - Rules to collocate windows and doors; 4 - Inter-
column grammar (set of descriptions to transform
columns in pilasters); 5 - Rules to collocate column
system elements.
Recognition rules maps some generic elements
of a building such as parts of walls, floors and other
pré-existing structural elements. Figure 4 shows rule
1st that recognise the interior and exterior floors
placing a label A; rule 2st places a dashed line related
with an interior slab, rule 3st add a dashed line related
with an interior slab and a cornice, and rule 4st add a
slab at previous dashed lines. Rule 1st, 2st, 3st works
in two views (elevation and vertical section) and rule
4st in one view (elevation). Rule 5st inserts a label
A and a wall, rule 6st maintains the label A and in-
serts a wall, rule 7st introduces a wall between two
openings, and rule 8st introduces a label P related to
a door. Rule 5st, 6st and 7st works in two views (plan
and horizontal section) and rule 8st in two views (plan
and elevation)
520 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Meta-structure rules as 5 rules that may be seen
in Figure 5.Rule 1vp inserts a door in the point P from
previous rules and two labels (triangles) that show
the direction next rules are going to take. Rule 2vp
inserts column system elements labels (pd - doric pi-
laster and cd - doric capital). This rule as a set of equa-
tions such as L = 12 w´ + k1 · w + 21 wc + w`,
where w = intercolumn width, w` = ½ door width,
wc = column width; L = facade module. This rule
drives to the right side of the facade. Rule 3 vp is
similar maintaining the same equations except the
L1 = 12 w` + k2 · w + 2w` + 12 wc + r, where r =
facade remain. This rule drives to the left side of the
facade. These 2 rules as the expression
 
∑
k1+k2
 w+4w`+wc+r; ∀w`, wc, w, r R∧kN
j=1
(1)
as the sum of the two sides of the façade.
The rule 4vp adds a doric pilaster and a ionic cap-
ital labels at second floor. Finally rule 5vp adds a doric
pilaster and a corinthian capital labels related to third
floor and erases the triangle label.
Rules to collocate windows and doors may be
seen in Figure 6. Rule 1jp, 2jp, differentiates the bays
with doors from the ones without doors, adding la-
bels to insert windows or doors. Rules 3jp and 4jp
Figure 6
Rules to collocate
and detail windows
and doors
Figure 7
Rule 2ic from
Intercolumn
grammar
transforming
columns in
pilasters.
are related to doors detailing, and rules 5jp to 9jp are
related to windows detailing.
The Intercolumn grammar rule 2ic, seen in Figure
7, is applied to transform columns in pilasters and it
comprehends a set of descriptions as:
<wall > → <2colunas >;
<bd > → <S(bd) >;
<bj > → <S(bj) >;
<f> → <S(f) >;
<cd > → <S(cd) >;
<cc > → <S(cc) >;
<ccm→><S(ccm) >.
In order to fill the facade with ornament, it's evoked
the column system shape grammar applying all the
elements needed. (Coutinho, 2011).
Note that in an advanced stage of construction of
a shape grammar the rules transformed are going to
be integrated in the grammar rules showing a dialec-
tic process of reconstruction of the grammar simul-
taneous to its derivation. That is that the grammar
derivation is not the end of the process until being
obtained a satisfactory and meaningful result. Fig-
ure 8 shows the Ducal facade derivation. Drawings
are being generated and detailed in elevation, plan
and section simultaneously trough a parallel gram-
mar application.
IMPLEMENTATION AND VISUALISATION
OF THE PALACE FACADE GENERATION
The application of the parameters and descriptions of
the grammar, using the software Grasshopper (GH),
allowed the automatic implementation of the gram-
mar.
The GH program allowed achieving different for-
mal solutions through the manipulation of those pre-
defined parameters.
The implementation of shape grammars in GH
environment corresponds to a parametric system
that simulates the generative capacity (derivation) of
the grammars inferred from the treatise.
This GH program created a set of medium scale
and small scale that were applied simultaneously
generating different solutions, allowing some degree
of flexibility. See Figure 9.
This set of systems aimed to contribute to an in-
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 521
 Figure 8
Ducal palace façade
shape grammar
derivation.

522 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 9
Ducal palace façade
physical models
made with digital
fabrication technics
and virtual models
from a GH program.
terpreter, which is a system of applying a rule set
automatically generating new elements with gram-
mar, denoting a behavior with semantic and syntac-
tic value providing shape emergence (Stiny, 2011).
Several physical artefacts were fabricated. First
experiments comprehended an ionic capital milling
using a 3 axis milling machine Pronum 4020. Mod-
els were milled from an expanded polystyrene (EPS)
block.
The CNC machining strategies consisted in one
rough passage using a 10 mm milling cutter and
two diagonal finishing passages using a 3mm cutter,
thereby allowing a better definition of the physical
model.
The final Ducal palace facade physical model was
fabricated with a laser cutter.
This technique was elected due the large pro-
portions of the facade. The aim was to manufacture
a prototype with a reasonable scale without losing
the definition of the elements of the column system.
The final facade prototype has about 80cm width and
15cm height.
This technique enabled both the cut of empty
windows and doors, such as the engraving of the dif-
ferent column elements. The constructed models as
well as some stages of manufacture can be seen in
Figure 9.
There`s evidence that Alberti`s treatise influ-
enced Ducal palace facade design?
The column system shape grammar allowed
compiling the resulting treatise rules in order to gen-
erate different buildings providing them with orna-
mentation.
However, it was not always possible to directly
use those rules, so new rules were found through
transformations. This new rules were then used to
understand the impact of the treatise in the design
of Ducal palace façade.
The simple linear regression method (SLRM) was
chosen to perform the analysis of the rules applied in
the facade design and construction, measuring the
degree to which the treatise has influenced the de-
sign and construction of buildings. To implement
SLRM was used the statistical software SPSS..
For verification of the degree of influence two
variables were used. The independent variable (IV)
that is the values of the column system from the trea-
tise, and the dependent variable (VD) consisting on
the rules applied in the Ducal palace façade values.
The SLRM was chosen to understand the rela-
tionship between the two variables, and if the IV pos-
itively influenced or not the DV. With this method
is possible to describe the relationship between the
two variables. This relationship may be seen through
the line
Y i = 0 + 1 · Xi + i; (2)
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 523
 where X is the independent variable; Y is the depen-
dent variable or predict; β 0 is the constant which rep-
resents the intersection of the straight line and the
vertical axis; β 1 is a constant representing the slope
of the line; and εi is the residual factor.
The objectives of SLRM are: perform measure-
ment on how much one variable is explained by an-
other, that is how much DV is explained by IV; quan-
tify the intensity and direction of the linear relation-
ship between two variables; predict the DV from the
IV, and; infer that the model is adequate to explain
the linear relationship between two variables.
The dispersion model gives the quality measures
of the model where the correlation coefficient (R)
measures the intensity and direction of the linear re-
lationship.
R2 is the coefficient of determination which mea-
sures the proportion or percentage of variation of the
DV that is explained by IV. This varies between [0, 1].
Where R2 = 0, DV cannot be explained by IV.
If R2 = 1 means that the DV can be 100% ex-
plained by IV. The coefficient of determination then
varies between 0% and 100%. It is assumed that 50%
means that the DV can moderately be explained by
IV. In this analysis we focus on the readings R2.
To apply the SLRM is necessary to assess 5 pre-
supposition (Carvalho, 2008): Linearity of the studied
phenomena; Random variables with null value: E (ε i)
= 0; Constant Variance of residual random variables:
Var (ε i) = σ 2; Independence of residual random vari-
ables: Cov (i ε, v j) = 0 i ≠ j;- Normal Distribution of
residual random variables: ε i � N (0, σ2).
The treatise and transformed rules proportion
parameters represented by height h of the column
system element were selected to be analysed with
SLRM. These proportions are dependent on a con-
stant diameter D which represents the measure of
the column imoscape projection.
The 98 rules of the column system contains: 8
rules from shaft grammar; 7 rules from doric base;
7 rules from ionic base; 18 rules from doric capital;
18 rules from ionic capital; 10 rules of the corinthian
capital; 2 rules from composite capital; 13 rules of
524 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
the doric entablature; 17 rules ionic entablature; 1
rule from corinthian entablature and 5 rules from the
pedestal.
The rules which contain more than one param-
eter are divided into sub rules. For example the rule
of the shaft Rshaft 6 has h = 1/8D, h1 = 7D and h2 =
27/5D and was treated in sub rules like Rfuste6 h =
1/8D, Rfuste6A h1 = 7D, Rfuste6B h2 = 27/5D to facil-
itate the data base management.
The sample had 261 x 2 (different building) = 522
parameters. That is N = 522, were N is the number of
observation.
The result of the Constant Variance of residual
gave an abnormal result, being the large majority of
the data divergent with the line (seen in Figure 10)
not revealing that the DV can be explained by IV. This
result suggests that the rules used in the design of
Ducal palace facade are not the from Alberti`s trea-
tise.
Figure 10
Figure 10.
Regression
Standardized
Residual
CLOSURE
The grammar presented lead us to conclude that
there is no evidence of Alberti`s De Re Aedificatória
descriptions use on the design of the Ducal palace
façade.
The experiment showed in this paper is part of a
wider research that analyzed 3 other building, 2 from
Alberti (Rucellai palace in Florence and Sant`Andrea
church in Mantua) and another Portuguese build-
ing, São Vicente de Fora in Lisboa using the same
methodology evidencing different results.
The shape grammar proved to be efficient in pro-
viding and organizing the data to be evaluated with
SLRM.
This paper also had the goal to expose, as much
as possible, the process of construction of this spe-
cific grammar, contributing to a better understand-
ing of its specificities and extension of such genera-
tive tool.
Kruger, M, Duarte, J and Coutinho, F 2011, 'Decoding
De Re Aedificatoria: using grammars to trace Al-
berti’s influence on Portuguese classical architec-
ture', Nexus Network Journal, 13(1), pp. 171-182
Stiny, G 1980, 'A note on the description of designs', .
Environment and Planning B: Planning and Design, 8,
pp. 257 - 267
Stiny, G 2011, 'What Rule(s) Should I Use?', Nexus Network
Journal, 13(1), pp. 15 - 47
Teixeira, J 1997, 'O Paço Passo a Passo. A estratégia ar-
quitectónica ducal (Sec. XVII-XVIII)', Monumentos, 6,
pp. 10-13

ACKNOWLEDGEMENTS
This work was part of the Digital Alberti project
funded by Fundação para a Ciência e Tecnologia
(FCT), Portugal, and hosted by CES at the Univer-
sity of Coimbra (PTDC*/ AUR*/64384*/2006). The
project was coordinated by Mário Krüger. Fil-
ipe Coutinho was funded by FCT with PhD grant
SFRH*/BD*/81260*/2011. Special thanks to Prof. Luis
Mateus and to Faculdade de Arquitectura da Univer-
sidade Técnica de Lisboa for the laser scanner FARO
Focus 3D use, and Luis Cabral for Mathematical revi-
sion.

REFERENCES
Albert,, L, B 2011, Da Arte Edificatória, Fundação Calouste
Gulbenkian, Lisboa
Bryman, A and Cramer, D (eds) 1992, Análise de dados
em ciências sociais: introdução às técnicas utilizando
o SPSS, Celta, Oeiras
Carvalho, H 2008, Análise Multivariada de Dados Qualita-
tivos Utilização da Análise de Correspondências Múlti-
plas com o SPSS, Silabo
Coutinho, F, Castro e Costa, E, Duarte, J.P and Kruger,
M 2011 'Interpreting De Re Aedificatoria – a shape
grammars of the orders system', Proceedings of Inter-
national Conference Eccade,2011, Ljubliana, Eslove-
nia, pp. 788-798
Knight, T 1994, Transformations in Design: A Formal Ap-
proach to Stylistic Change and Innovation in the Visual
Arts, Cambridge University Press

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 525

526 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Local-reconfigurable Freeform surface with plywood
From the perspective of Japanese Tsugite-Shiguchi

Koki Akiyoshi1 , Hiroya Tanaka2

1,2
Keio University, Japan
1
kokiakiyoshi.com 2 http://fab.sfc.keio.ac.jp/
1
apoanian@gmail.com 2 htanaka@mva.biglobe.ne.jp

This research exhibits a novel construction method for Freeform surfaces with
plywood, without using metal joints and bending. By introducing the perspective
of Japanese Tsugite-Shiguchi, the research aims for a drastic change from
node-oriented thinking to module-oriented thinking. This paper focuses on the
investigation of how to simplify fabrication processes, how to realize the
environmental capabilities of Freeform wood structures, and how to provide
redundancy and stability to the whole architectural system. In order to challenge
these problems, we examined three discretion methods. As a result, we have been
successful to produce a double-layered surface, filled with triangular mesh,
implemented only by cutting one sheet of plywood. Moreover, the system has also
acquired a new nature: local-reconfigurability, wherein it can react and adapt to
fit local parameters and requirements.

Keywords: Digital fabrication, Freeform timber, Without metal and bending,

Discrete surface, Minimal components for mega-assembly

INTRODUCTION Of course, there are simplified fabrication pro-

Two-thirds of the land in Japan is covered by moun-
tains and forests. By using this rich natural resource
efficiently and effectively, Japanese people have built
their own wooden culture. Nowadays, interest in tra-
ditional wood construction methods is increasing in
the world because of the nature efficient character-
istics of wood. It is naturally renewable, fully recy-
clable, energy efficient, and reduces CO2 emissions.
On the other hand, recently open-source housing
projects, such as WIKI-HOUSE and MIT Larry Sass's re-
search, have become well known in the maker move-
ment.
cesses for building houses. However, in areas with
harsh climates such as hot and humid Japan, these
systems cannot sufficiently respond to climates be-
cause they can only express identical forms. There-
fore, the problem is that the existing wooden con-
struction methods have difficulty in realizing the
Freeform surface, which is able to be represented us-
ing computer graphics.
To see the previous research to realize the
Freeform structure by wood, there are three major
shortcomings.

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 527

 • First, the steam bending and 5-axis machining
processes, which are almost entirely related
research avenues that have been employed as
fabrication methods, are far too complex to
be available to the general public. The com-
plexity of fabrication hinders the generaliza-
tion of the wooden Freeform structure. Com-
plexity means a lot of labor, which demands
immense time, cost, and skilled professionals.
• Moreover, in order to be actively used, the
method needs to have an architectural perfor-
mance. In previous studies and methods, al-
most all of them were used for a facade or a
temporary pavilion. Think of the human body.
Our body is not only made of bones but also
consists of skin and flesh. The skin responses
to the outer environment, while flesh regu-
lates it, and the bones support them. In order
to perform better, the structure should also
have the ability to put skin and pack flesh in
addition to building the bones.
528 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
• In addition, the most serious problem is that
once you assemble, it is hard to deconstruct.
This is most glaringly apparent in the Frei
Otto's Mannheim Pavilion. If a problem oc-
curs to a part of the building, you will have to
replace the whole wall or structure. In other
words, the system does not have redundancy.
The purpose of this paper is to simplify fabri-
cation processes, realize the greater performance of
Freeform structure, and provide redundancy and sta-
bility to the whole architectural system. The advan-
tage of using wood essentially lies in the reconfig-
urability it has, and Japanese wooden architecture
have been metabolized by grafting the damaged
member and replacing it by relaying the new mem-
ber. In order to use this nature of wood, the sys-
tem should consist of minimal components, which
are reconfigurable. In the following chapters we will
try to solve these problems by looking from the per-
spective of Japanese Tsugite-Shiguchi and will also
demonstrate three approaches through prototyping
-"Wooden Fabric" (Figure1).
Figure 1
WOODEN FABRIC
by KOKI AKIYOSHI
Figure 2
Shiguchi.
Collaboration with
additive
manufacturing and
subtractive
manufacturing.

METHOD Shiguchi, a method to combine three beams in a two-

Tsugite-Shiguchi
This research is based on the perspective of Japanese
Tsugite-Shiguchi, which is a variety of Japanese tra-
ditional joining techniques without using nails and
glue. Tsugite is a technique to connect materials
to augment the lack of length of materials, while
Shiguchi is a technique to connect materials at a spe-
cific angle. In this respect, we employ this method as
a key concept to realize the system, which can easily
be reconfigured and reconstructed.
It is said that there are around 200 types of
Tsugite and Shiguchi, however, we invented new
types of Tsugite-Shiguchi for the system. This is
because we decided to assemble three beams as
a node, with Shiguchi. In the existing pattern of
dimensional plane did not exist. With the new type
of Shiguchi proposed in this paper, one out of three
beams can be divided into two parts in the middle
(Figure 2).
Instead, the new type of Tsugite had to be de-
signed so that the divided beams can be connected
next to normal beams. So, we designed interlocking
details to fit each beam together well (Figure 3).
Discrete Geometry
To realize feasible segmentation, the target geome-
try has to be resolved in a discrete way. In this pa-
per, we used triangle meshes to do this. The primary
reason to use a triangle for this research is that we in-
tend to acquire strength due to the goal of creating

Figure 3
Tsugite. It is a
further evolution of
the present
Daimochi-Tsugi.

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 529

 Figure 4
Discrete geometry.
For example, the
minimum surface is
imported.

a Freeform surface that can be used as a structural Discrete Fabrication

framework. The second reason is that we intend to We think that the most rational construction method
add surface covering, or "skin". is the integration of additive manufacturing and sub-
Of course, CNC milling machines are able to cut tractive manufacturing. Of course, the 3D-printer can
curves as numeric curves, however, the region de- create complex objects out of nothing. However,
fined by curved materials is the Freeform surface. But, when you want planar shapes, it is obvious that the
by approximating each curve to a poly-line, the re- easiest way to produce them is to cut and derive from
gion can be defined by only using planar surfaces. In mass-produced materials. This becomes even truer
this method, you can see the module as three beams when the scale of the structure becomes larger such
meeting at a node and not as six beams meeting at as architecture.
the node (Figure 4). Figure 5
Discrete Materials.
Discrete Materials Taking dowel holes
Beams we used in this research were not square tim-
for guidance,
ber such as two-by-fours, but custom-made timber
pressure bonding
derived by cutting one sheet of plywood. If you use
with A, B, and C.
mass-produced square timber, you are sure to bend
it.
However, it is easier and more efficient to get
curved parts by just cutting the curved stuff from one
sheet of material. The relationship of the milling ma-
chine and plywood is similar to the relationship of
On the other hand, in terms of complex 3D-
scissors and paper. You only have to cut shapes from
milling, you will need a machine such as a 5-axis
the paper every time you need them. In addiction,
milling machine, a robotic arm, or something expen-
because plywood is originally made by laminating
sive. In fact, the using of these machines will most
thin materials, if you need thicker materials, it is possi-
probably increase the cost and time of the process
ble to make your original square timber by cramping
dramatically.
the parts together (Figure 5).

530 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 6 WOODEN FABRIC
Discrete Based on the methodology above, we invented a
Fabrication. 3D new construction method and software, which con-
printed joint parts. verts the user-made surface data to constructible ob-
jects in the real world (Figure 7) (Figure 8). The con-
struction method is based on the module looks like
a snowflake, which is consists of three long members
and three short members (left in Figure 9). The longer
one has male joints, the shorter one has female joints
on its two edge points (right in Figure 9). Each mod-
ule is able to connect to six adjacent modules by in-
terlocking each Thugite.
To use this module-oriented method, in other
By combining these two manufacturing tech- words discretion method, the structure is locally de-
niques, we can decrease the 3D-milling process. Only fined, able to construct locally, and also able to recon-
you have to do is ADD three-dimensional shapes (Fig- figure locally. Thus, if you construct Freeform struc-
ure 6). In addition, if we use the 3D-printed parts as a tures using this system, like Lego bricks, you only
mould, the efficiency of the process should improve have to join each part without needing any complex
even further. blueprints.
Figure 7 Constraints
Wooden Fabric. The beams, which are used in this research, can only
This prototype is bend in one direction. This is an important require-
based on 7 ment of this system. Hence, we have to constrict di-
modules. rections of bending, using congruent isosceles trian-
gle as the smallest unit for dividing the mesh, so that
torsions will not occur on each beam.

Algorithms
First, we make a bounding box from the surface that
user created, deconstruct the box, sort the plane par-
allel to the surface, and define that as the projection
Figure 8
plane. After dividing this projection plane to the ho-
Wooden Fabric.
mogeneous grid, the grid is projected onto the sur-
"Local-
face. Then, vectors from vertices on the projected
reconfigurability"
surface parallel to the normal vector of the projection
plane are generated. (Figure 10).

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 531

 Figure 9
Wooden Fabric.
(left) Each module,
consisting of three
long and short
members, is able to
connect to six
adjacent modules
by interlocking
each Thugite.
(right) 1: Shorter
has female joints 2:
Longer has male
joints on its two
edge points.

Figure 10
Algorithms. (left)
Generated points
are projected to the
imported surfaces.
(right) Surfaces are
generated by
connecting
projected points.

532 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 11 Social implementation through the work-
Fablab Hiroshima- shop
Akitakata. Using this system, we are building the architecture
next August. We are going to insert the Freeform wall
in an old Japanese garage, in order to be used as the
workshop offering digital fabrication (Figure 11). For
this project, we ran the workshop for the users of this
shop. We used the clay as a means of modelling be-
cause most participants do not have enough literacy
to realize what they imagine (Figure 12). However,
using 3D scanners, they can import 3D scanned clay
models to the software and fabricate real Freeform
wooden structures by just pushing the run button of
Figure 12 the machine. Eventually, prototypes, which are pro-
The whole shape of duced by the users through this workshop, are going
the architecture is to integrate and become increasingly sophisticated,
generated by until the final forms are reached. Through this im-
prototyping plementation, we are going to examine the perfor-
through the mance of this system.
workshop with
users. CONCLUSION
In conclusion, the system has acquired a new aspect,
local-reconfigurability (Figure 13). As clothes can be
patched up when they are ripped or spotted, this
character enables the architecture to be updated by
replacing and reconfiguring the module, if an error
occurs in a part of the system. On the other hand, if
you want to build your house, you would get a loan
Moreover, we have been successful in produc-
because it will cost over thirty thousand dollars in
ing a double-layered surface filled with a triangular
Japan. However, using our system, it will able to be-
mesh, made of plywood (left in Figure 9). The skin
gin to build the house depending on your budget be-
of the structure is not only filled with wood, but it
cause the whole structure is defined locally. This is
can also be filled in traditional ways with clay and
to say, the incremental construction is being realized.
earth. Therefore, the system can be used locally in
This will contribute to rebuild areas destroyed by dis-
combination with traditional construction methods.
asters and to develop impoverished areas. Besides,
This will contribute to forming townscapes that re-
addition and subtraction to a building can also come
flect and complement the environment.
easily.

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 533

 Figure 13
We have also
proved that our
system is able to
construct structures
"locally" and
structures is strong
enough.

534 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

REFERENCES
Kanasaki, Kenji and Tanaka, Hiroya 2013 'Traditional
Wood Joint System in Digital Fabrication', Proceed-
ings of the 31st eCAADe Conference – Volume 1, Delft,
The Netherlands, pp. 711-717
Menges, Achim 2011 'Integrative Design Computation:
Integrating material behaviour and robotic manu-
facturing processes in computational design for per-
formative wood constructions', Proceedings of the
31st Annual Conference of the Association for Com-
puter Aided Design in Architecture (ACADIA), Banff (Al-
berta), pp. 72-81
Ohshima, Taisuke, Igarashi, Takeo, Mitani, Jun and
Tanaka, Hiroya 2013 'WoodWeaver: Fabricating
curved objects without moulds or glue', Proceedings
of the 31st eCAADe Conference – Volume 1, Delft, The
Netherlands, pp. 693-702
Sass, Lawrence, Michaud, Dennis and Cardoso, Daniel
2007 'Materializing a Design with Plywood', 25th
eCAADe Conference Proceedings, Frankfurt am Main,
Germany, pp. 629-636

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 535

536 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
 Populating surfaces with holes using particle repulsion
based on scalar fields
Timo Harboe Nielsen1 , Stephen Melville2 , Iain Sproat3
1
Technical University of Denmark 2,3 Ramboll Computational Design
1
timonielsen@me.com 2,3 {stephen.melville|iain.sproat}@ramboll.co.uk

This paper describes the relaxation of charged particles in order to create a

pattern of voids based on a scalar field on any complex polygon mesh. A scalar
field representing stress values or a greyscale image, can be used to create void
patterns of aesthetic or structural character; all with full awareness of the
materiality. Following relaxation, areas with low scalar values consist of large
voids with a small distance between them. Areas of high scalar value consist of
small voids with a greater distance between them. This research has been applied
in the design of a sculpture at Oxford Brookes University, where stress data from
Finite Element Software has been used for the automatic and rational distribution
of holes.

Keywords: Surface perforation, Geometry optimisation, Particle repulsion

INTRODUCTION ners which induce local stress concentrations leading

Topology and geometry optimisation is a topic of to crack propagation (Withey 1997). There have been
great interest for both Architects and Engineers. various previous examples of the use of patterns of
The possibility to quickly create complex shapes in circular voids. One example is Mountain Dwellings
freeform 3D-modelling software has evolved with by BIG Architects, see figure 1.
the need for Architects and Engineers to work to-
Figure 1 gether in close collaboration to enable high aes-
The Mountain by thetic value, structural efficiency and buildability. In
BIG, photographer this matter, topology and geometry have been re-
Carsten Kring searched widely (Chen and Kikuchi 2001; Li et al.
2000; Lewis 2011; Bendsøe 1989).
This paper focuses on the use of circular voids.
The benefits of using circular voids in topology op-
timization are many. For example, with steel, holes
are easier and cheaper to drill than other more com-
plex subtractive shapes, and the production can be
done by automated tools just by providing a cut-
ting or drilling pattern. Circular voids also have the
great structural benefit of being without sharp cor-

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 537

 viding Structural Engineering advice, geometry anal- Figure 2
ysis and Computational Design. The sculpture con- 0-14 tower in Dubai
sists of 15 steel columns with heights ranging from by Reiser +
3m to 6.5m and a single petal at the top of each col- Umemoto, Photo ©
umn. The diameters of the petals range from 2m to Reiser + Umemoto
2.5m, see figure 4. The team wanted to use a perfo-
ration pattern to show a natural and logical illustra-
tion of the nature of the forces in the structure, with-
out compromising either aesthetics or structural ef-
ficiency and to allow light to filter through in a way
reminiscent of a forest canopy. Our work focused
on the distribution of voids in response to a scalar
field representing stress values. Our paper describes
this work and outlines how circular voids can be dis-
tributed on geometrical meshes in regards to scalar
values.
Figure 3
Turbine blade with
perforations for
cooling(Creative
Commons
Attribution-Share
By using a very fine regular grid of points and Alike 3.0 Unported)
assigning a radius to each point the perforation pat-
tern creates an image that can be seen from the dis-
tance. The 0-14 Tower in Dubai by Reiser Ememoto,
Any value can be the key parameter in the dis-
see figure 2, has a double layer facade, where a per-
tribution of the voids and interpolation of values is
forated outer layer with huge voids creates an inter-
also possible, which allow different parameters to in-
esting pattern that both aims to strengthen the ar-
fluence the distribution of voids at the same time
chitectural value while enhancing the environmen-
(with different or equal weight). For the sculpture
tal performance of the building. Perforations are also
at Oxford Brookes University stress values imported
used in other contexts such as in gas turbine blades
from FEA software were used. The stresses were then
for cooling as seen on figure 3. Even though void dis-
converted into scalar values to facilitate the distribu-
tribution is widely used, the materiality is very rarely
tion of voids. Instead of using regular arrays or ran-
taken into account. By distributing voids in regards
dom locations of point, which has been done several
to different factors such as stress values, aesthetics,
times before, we looked into how charged particles
acoustics, transparency, thermal properties or light
could be distributed in regards to any scalar value, on
distribution while taking density and materiality into
any geometrical mesh no matter what complexity it
account, the distribution can work as a key element
might consist of. This way the underlying structure
for both the Engineer and the Architect.
is transformed with an efficient and natural looking
Our research was applied to a sculpture designed
distribution of voids while taking the materiality into
by architecture students and staff at Oxford Brookes
account.
University with Ramboll Computational Design pro-

538 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 4
Oxford Brookes
Rain Pavilion

VOID DISTRIBUTION where F~a is the force vector working on particle a,

It is well recognized that by removing material from pa is the coordinates of particle a, and pb is the co-
areas of low stress, and by adding material to areas ordinates of particle b. For realistic behaviour of the
of high stress, a structure can be highly optimized in particles Newtons Laws of Motion (Young and Freed-
regards to material usage, and structural efficiency man 2012) are used.
(Li et. al. 2000; Yang and Chen 1996). The mini- Newtons Second Law:
mum distance between voids and the size of those
voids has a great influence on the structural strength ~a = m · ~a
F (3)
(Melander and Ståhlberg 1980; Koss and Dubensky
1987; Benson1993). Inappropriate clustering of just where F~a is the force, m is the mass and ~a is the ac-
a few particles, can decrease the strength signifi- celeration
cantly (Ohno and Hutchinson1984). The distribution Newtons Third Law:
of voids must be appropriate to the stress present
in the surrounding material. The correlation is neg- ~a = −F
F ~b (4)
ative between stress value and void size and pos-
itive between between stress value and the mini- where F ~a is the force working on one particle in an
mum distance between voids. Earlier studies have interaction and F ~b is the force working on the other
made use of Coloumb's Law to relax a system of par- particle in an interaction.
ticles and create a regular grid and avoid clustering
of points (Stowell1950; Macdonald 1957). Coulombs
REPULSION OF PARTICLES ON PLANE
Law (Young and Freedman 2012):
In the following the procedure of creating an even
|qa · qb | distribution of voids on a plane is covered. To ensure
F =k· (1) equal distance between different sized voids, parti-
d2
cles have a radius assigned, and the charge works
whereF is the size of the force, k is Coulombs con- from the edge of the void and not the center. If the
stant, qa and qb are the charges of the two interact- two voids intersect a punitive force is applied. This
ing particles and d is the distance between them. As can be described by the following definition derived
the force is working in the line between the two par- from Coloumb's Law (formula 1):
ticles, formula 1 can be translated into a vector by the
following formula ~a = k · |qa · qb | pa − pb
F · (5)
(d − ra − rb )2 d
|qa · qb | pa − pb |q · q | p − p
F~a = k · · (2) ~a = k ·
F
a b
·
a b
(6)
d2 d cf d

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 539

 where formula 5 is for d > ra + rb and formula 6 is
for d ≤ ra + rb . ra is the radius of particle a, rb is
the radius of particle b, cf is a correction factor which
is set close to zero to ensure that the two voids are
pushed apart if intersecting. The value k is set to 1,
as the actual size of the force is not important, but
only the difference between forces working on indi-
vidual particles. The total force working on a particle
can then be described as follows:
∑
N
~t =
F ~i
F (7)
i=1
where F ~t is the total force and N is the number of
particles in the system.
The same unit mass is used for all particles, with-
out taking the size into account. As the repulsion
force is working from the edge of the void, the dis-
tance between different sized voids will be equal if
the masses and charges are equal. This simplifies for-
mula 3 to:
~a = ~a
F (8)
540 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
The system is damped so that the velocity of a
particle is slowly reduced. The velocity of a parti-
cle is described by the following formula (Young and
Freedman 2012)
~vi = (~vi−1 + ~a · t) · kd (9)
where ~vi is the velocity, ~vi−1 is the velocity from pre-
vious iteration, t is the time step and kd is the damp-
ing constant. In the initial step ~vi−1 is assumed to
be zero. An ideal distribution of different sized voids
is shown on figure 5.
In figure 6 a system of particles with equal radii is
shown. The colours reflect the energy of the partic-
ular particle. As the total energy of the system goes
towards zero, the system settles in a regular grid. Fig-
ure 7 shows the same type of system but with ran-
dom sized holes. Both figures only show a part of a
bigger system, which is why the two pictures of each
figure do not necessarily contain the same number
or the same size of particles. It is interesting to see
how clusters of similar sized voids are created. This
is explained by the fact that each void tends to be
surrounded by 6 other voids to create a low energy
hexagonal grid (as seen on figure 6); a pattern that is Figure 5
seen in many places in nature, such as the structure of Ideal distribution of
graphene. It is not possible to create this structure if voids
the size difference between neighboring voids is too
big.
Figure 6
Distribution of
equal sized voids.
Left: Initial position.
Right: Settled
position
Figure 7
Distribution of
random sized voids.
Left: Initial position.
Right: Settled
position
 DISTRIBUTION OF PARTICLES ON MESH
In the following the procedure of distributing voids
in regards to a scalar field is covered. Any value can
be used for the scalar field. We used Von Mises Stress
Values from FEA Software (Sofistik), based on pre-
defined structural loading values. A surface is im-
ported to Sofistik which can then generate a dis-
cretized mesh for finite element analysis. The mesh
and scalar values are then exported to Parametric
Modelling Software (Grasshopper). The colour prop-
erty of mesh vertices in Grasshopper can be used
to store the scalar field. The vertices are coloured
in greyscale values where black (value=0) represents
minimum stress and white (value=1) represents max-
imum stress. The radius and charge of each parti-
cle is changed for every iteration with regards to the
scalar value of their location. The connection be-
tween scalar value and radius and scalar value and
charge is assumed to be linear. The weight of colour
in regards to radius and charge is manually controlled
with two scalar field sensitivity parameters. The cal-
Figure 8
culation of radius and charge can be summarized in
Same total force
the following formulas:
working on
particles on r = rin · ((1 − Sr ) + Sr · (1 − C)) (10)
different faces q = qin · ((1 − Sq ) + Sq · C) (11)
where r is the radius, rin is the initial radius, Sr is
the radius-field-sensitivity parameter, C is the scalar
value at the location, q is the charge, qin is initial
charge and Sq is the charge-field-sensitivity parame-
ter.
A software component has been developed for
use with Grasshopper, to simulate the natural be-
haviour of charged particles in regards to Coulombs
law and Newtons laws of motion. In the following {X}
references to an individual part of the flow chart di-
agram in Figure 9. The mesh {2} is used as input to-
gether with random locations of points {1}. The mesh
is coloured in grayscale values, where 0 represents
minimum scalar value, and 1 represents maximum
scalar value. Every point contains data of what face
it is on, the mesh colour of current location {10} and
its velocity. The radius and charge is calculated based
on the colour of the mesh in the location of the parti-
cle. {11}{12} With the use of Coulombs Law the force
working on the particle is calculated {14}. On com-
plex 3D shapes the vector of the force is not likely to
be in the plane of the face on which the point is lo-
cated. As the particle is constrained to the mesh, any
acceleration added to the particle that is out of the
plane of the face, will contribute to, at the current lo-
cation, non- releaseable energy, which will be added
to the velocity every iteration. If the particle moves
to a face with a different normal vector, some of the
energy that has built up will be released. This makes
it impossible to settle the system. To avoid this, the
force vector is projected onto the face plane {15} to
cull the part of the vector that is not contributing to
releaseable energy. See figure 8 where Ft is the total
force, Fu is the unreleaseable energy and Fr is the
projected vector. For this culling to be appropiate,
only small timesteps should be used.
One of the challenges to be solved is the action
to be taken when a particle reaches the edge of a
face. For every iteration it must be checked if the
particle has reached an edge {19-22}. If the particle
is no longer on the face, it should be moved back to
the edge of the face and either continue movement
on new face or bounce of the boundary maintaining
same velocity, depending on whether the boundary
of the face is also the boundary of the mesh. See fig-
ure 10. {24-31}
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 541
 Figure 9
Flow chart diagram

542 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 10
Boundary actions.
Left: Edge between
neighbour faces .
Right: Face edge
By damping the system {8} the energy of the par-
ticles will slowly reduce. By monitoring the total ve-
locity of the system, the iterations can be stopped
when the total energy of system approaches zero
{34}. An example of a distribution is shown in figure
11.
Figure 11
Distribution of
voids
TEST AND RESULTS
In the following a simple slab is analysed with and
without voids. The slab is 2000 x 5100mm, made of
200mm concrete and constrained against movement
and rotation in both ends. For the perforated version
210 holes ranging from 25 to 75mm in diameter are
created in regards to stress state. Two load combi-
nations are considered. Self-weight and Self-weight
+ 1 kN/m2 variable load. The slab is shown in fig-
ure 12 This distribution led to a material reduction
of 10.9% (from 2,00m3 to 1.78m3 ) and a deflection
reduction of 4.0% (from 0.53mm to 0.51mm for self-
weight + 1kN/m2 ). In the perforated slab the highest
value of the Von Mises Stress is 2.52MPa local to the
edge of a void. In the unperforated slab the value in
the same location is 0.98MPa. This is an increase of
157% (2.57 times the value).
DISCUSSION
The results of this analysis are interesting, yet not
surprising compared to earlier studies on topology
and geometry optimization (Li et al. 2000; Yang and
Chen 1996). By removing material in areas of low
stress the self-weight is reduced and the remaining
material is being used more efficiently. In the test,
stress concentrations at void edges were 2.57 times
the value of the same area without voids, which is a
little lower than expected the 3 times (Stowell 1950).
This can be explained by the fact that the self-weight
and thereby a significant proportion of the load, is
reduced. An important issue is that the calculation
is only done once. As topology is being changed,
so is the stress distribution. As most other studies
show, an iterative process is needed for proper op-
timisation (Bendsøe 1989; Schnack 1974}. The distri-
bution of voids creates an interesting natural pattern
which can be used as a powerful architectural ele-
ment. The script created for the distribution allows
any value, and not only stress values, to be used for
defining the colour of the mesh. If voids are being
used as, for instance, natural light sources, an Archi-
tect might have preferences for void locations that
may differ from the structurally most beneficial loca-
tion. With very few compromises the mix of interests
can be achieved by interpolating two different colour
patterns (with equal or different weight). The process
of distributing voids only requires coordinates of ver-
tices and colour values for every vertex, which makes
further work more doable.
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 543

OXFORD BROOKES RAIN PAVILION
This study was used as a part of the design scheme of
the 'Rain Pavilion' at Oxford Brookes University, which
is to be built in 2014. See Figure 4. The distribu-
tion study was carried out by Ramboll Computational
Design to reduce wind loads, increase structural effi-
ciency and create an interesting pattern that reflects
the nature of the forces in the structure while al-
lowing dappled light through the otherwise opaque
pavilion. The pavilion consists of 15 steel structures
with heights ranging from 3m to 6.5m connected by
a steel frame base. Each structure consists of a ver-
tical mast and a single petal at the top. The diame-
ters of the petals range from 2m to 2.5m. Each petal
has 230 holes with diameters in the range 8mm to
90mm. The largest stem has 200 holes with diame-
ters in the range 6mm to 40mm, and made of 88.9
CHS6.5. The number of holes have been chosen by
testing different numbers and then deciding upon
the option which was considered to be the best com-
bination of high aesthetical value and structural effi-
ciency, using subjective judgement. As the stresses
are higher at the bottom of the stem than at the top,
no voids are present in the bottom. From a height
of 2m small holes start to appear and gradually be-
come larger and closer towards the tip. This reflects
the structural behaviour and creates a visual connec-
tion between the ground and the perforated petals
on top. In this case random numbers were used as
initial radii to create a more vibrant pattern and to
strengthen the architectural concept.
544 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Figure 12
Distribution of
voids on slab
FURTHER WORK
This paper has covered how to make a single distri-
bution of voids, but for better optimisation in regards
to structural efficiency an iterative loop between void
distribution and FEA could be implemented. To allow
an iterative loop, the perforated geometry should au-
tomatically be imported to FEA. At the current stage
the remeshing of the perforated surface is not suffi-
cient for FEA analysis. This restriction must be solved
for an automatic iteration process to be possible. The
initial mesh should be used for every distribution but
the colour can be slightly changed every iteration,
depending on the new stress distribution. The next
step would be to add and remove voids in regards
to the global stress state. Another concern is that
of the stress concentration around edges of holes.
For optimisation the utilisation level or the maximum
deflection should be included in the iteration pro-
cess. If plastification of small areas at void edge is
allowed, the amount of voids could be greatly in-
creased, as stress concentrations are then only ex-
pected to be 1.4 times the stress value if no void
was present (Stowell 1950).The calculations of the
radius/void size could be more accurate with the full
use of the research presented by Stowell (1950) and
the charge calculations could be made more accu-
rate with the implementation of the information pre-
sented by Dubensky and Koss (1987). This implemen-
tation would allow an even more optimised distri-
bution but also require a much finer control of the
individual particles, which would be interesting for
further research. It would also be relevant to carry
out a study regarding an implementation of principal
stress lines. If voids are created so that they are longer
in the direction of the principle stress lines, and nar-
rower in the direction perpendicular to the principle
stress lines, it might eventually result in higher struc-
tural efficiency, and would also be an interesting ar-
chitectural element. A computational optimisation
would be to change formula 7 so a particle is only af-
fected by other particles within a certain distance, as
this would greatly improve the speed with which the
software is running. It would also be instructive to
look at an automatic time step control to ensure that
the system does not become unstable if forces are set
too high, and to automatically find a solution as fast
as possible.
Stowell, EZ 1950 'Stress and strain concentrations at a cir-
cular hole in an infinite plate', Langley advisory com-
mittee for Aeronautics, Technical Note 2073
Winslow, P 2009, Synthesis and Optimisation of Free-
Form Grid Structures, Ph.D. Thesis, University of Cam-
bridge
Withey, PA 1997, 'Fatigue failure of the de Havilland
Comet I', Engineering Failure Analysis, 4, issue 2, pp.
147-154
Yang, RJ and Chen, CJ 1996, 'Stress-based topology opti-
mization', Structural Optimization, 12, pp. 98-105
Young, HD 2012, University Physics with Modern Physics
13 edition, Addison-Wesley

REFERENCES
Bendsøe, MP 1989, 'Optimalshapedesignas a material
distribution problem', Structural Optimization, 1, pp.
193-202
Benson, JB 1993, 'An analysis of void distribution effects
on the dynamic growth and coalescence of voids in
ductile metals', Journal of the Mechanics and Physics
of Solids, 41, Issue 8, pp. 1285-1308
Chen, B and Kikuchi, N 2001, 'Topology optimization
with design-dependent loads', Finite Elements in
Analysis and Design, 37, pp. 57-70
Koss, DA and Dubensky, EM 1987, 'Void/Pore Distribu-
tions and Ductile Fracture', Metallurgical Transac-
tions A, 18A, pp. 1887-1895
Lewis, H 2011, Traversing digital matter states to fill irreg-
ular volumes, Master's Thesis, University of Bath
Li, Q, Steven, GP and Xie, Ym 2000, 'Evolutionary struc-
tural optimization for stress minimization problems
by discrete thickness design', Computers and Struc-
tures, 78, pp. 769-780
Macdonald, WM and Rosenbluth, MN 1957, 'Relaxation
of a system of particles with Coulomb Interactions',
Physical Review, 107 no 2, pp. 350-353
Melander, A and Ståhlberg, U 1980, 'The effect of void
size and distribution on ductile fracture', Interna-
tional Journal of Fracture, 16, pp. 431-440
Ohno, N and Hutchinson, JW 1984, 'Plastic flow localiza-
tion due to non-uniform void distribution', Journal
of the Mechanics and Physics of Solids, 32, issue 1, pp.
63-85
Schnack, E 1974, 'An optimization procedure for stress
concetrations by the finite element method', Inter-
national journal for numerical methods in engineer-
ing, 14, pp. 115-124

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 545

546 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Efficiency in Architectural Geometry Informed by
Materials
Sevil Yazici
Ozyegin University
www.sevilyazici.com
sevil.yazici@sevilyazici.com

Although some studies investigate physics-based dynamic systems to generate

structurally efficient forms by incorporating geometry with performance
requirements, there is a gap in the field questioning on how to link structurally
efficient architectural geometry with mechanical properties of materials. The aim
of this paper is to question the possibility of generating an information loop in
which Young's Modulus, stiffness of the material may both inform the
form-finding process and the structural performance simulation. The proposed
method offers steps including form-finding, series of analyses applied for
architectural geometry and structural performance, as well as optimization.
Based on the simulation results, efficiency values are calculated driven by the use
of different materials. The significance of incorporating material properties in
the early design stage is underlined, by comparing differences, whether the
stiffness of material informs the form-finding process or not.

Keywords: Form-finding, material, architectural geometry, finite element

method, optimization

INTRODUCTION Gaudi's church of Colònia Güell was based on

Using physical models in form-finding has the great-
est potential to understand how systems perform in
general. It was used for many years to create struc-
turally efficient forms. Efficiency is known as a quan-
titatively determined value. Although the term has
various other meanings in different disciplines, it can
be described for structural efficiency as the weight of
the material which can carry a specific load. If a struc-
tural element obtains high efficiency, this means that
the ratio of the strength to its weight is high. Form-
active structures are potentially the most efficient
structures.
a hanging chain model. Gaudi showed in his ex-
periments, that walls or columns are naturally not
vertical. However, they are complex resolutions of
forces expressed such as 3D vectors (Burry, 2006). Al-
though physical models obtain greatest potential to
generate holistic simulation of a problem, there are
some issues related to the accuracy, scaling of mate-
rial and mass. Kilian's digital hanging chain simula-
tion, based on a particle spring system programmed
in Java, aimed to provide a real-time 3D-modeling
environment that allows the design of gravity-based
forms, created via hanging chain principles (Kilian

Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 547

 2006). The digital model also demonstrated how fab-
rication schemas for physical mock-ups of the dig-
itally simulated chain could be linked form-finding
simulation (Kilian, 2004). Later on, B-Processor was
developed that used a particle-spring system, solv-
ing Newton's laws of motion for a number of particles
reacting to forces (Gmelin et. al. 2011).
The most immediate translation of material sim-
ulations as physical models is considered the use of
Finite Element Method (FEM) analysis, in order to
mimic and test the system behaviour. However, FEM
is considered as an inefficient design tool, due to its
demand for material, specified geometry, contextual
definitions and the fact that it obtains a single goal
in computing structural performance of the system
(Ahlquist and Menges, 2011). Design optimization is
used as a decision-making process in building and
construction, in order to solve architectural design
problems through iterative interactions between de-
sign and analysis. In order to create an optimal build-
ing, single optimization criteria may be not sufficient.
Therefore, multi-criteria design and optimization of
buildings is a technical challenge (Aish et. al, 2012).
Some significant work has been undertaken in
the past which explore the relationship between
form, material, performance and optimization pro-
cesses. Evolutionary Structural Optimization works
with the logic of subdividing the surface into discrete
geometries and subtracts pieces based on compu-
tation to achieve optimized solutions by the use of
FEM (Xie et.al, 2005). EifForm gives an opportunity
to explore the relationship between associative ge-
ometry and performance-driven generative design.
The overall structural form is created in response to a
model input by the user and can be altered for vari-
ous design alternatives. The tool can optimize both
the topology and geometry of a structure, such as
material quantity for the given loads, while respect-
ing the constraints (Shea et. al. 2005). Although
these tools are important, there is a gap in exploring
the relationship between material properties, archi-
tectural geometry and structural performance (Yazici,
2013; Yazici and Tanacan, 2012).
548 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
Algorithms are widely used in the rationaliza-
tion process of large-scale architectural free-form
surfaces by generating panels. Nevertheless, it is crit-
ical to incorporate behaviour and production tech-
nology to the panel layout algorithm (Pottman et al.
2008). Although novel methods are currently be-
ing developed for the panelization and rationaliza-
tion processes of forms in any level of complexity
including free-form surfaces, they are mainly based
on operations on global geometry in a later stage of
the design process, which prevents exploring formal
and structural capabilities of material systems. The is-
sue with efficiency needs to be addressed along with
panelization processes.
This paper discusses the question of how to in-
corporate material properties into the form-finding
processes and performance simulation. Young's
Modulus is a material constant, which measures how
much a material stretches when a load is applied. It is
known as the tensile modulus, a measurement of the
stiffness of an elastic material. The aim is to generate
an information loop in which Young's Modulus may
both inform the form-finding process and the struc-
tural performance simulation.
METHODOLOGY
A dynamic model through a hanging chain simula-
tion is generated in parametric design medium. By
creating architectural geometry, specifying material
and imposing loads & boundary conditions, FEM is
operated for the static structural performance simu-
lation for various set-ups, including assigning differ-
ent types of steel to the geometry and adjusting the
arch height of the geometry informed by the stiffness
of the material used in the form-finding process. In-
put and output parameters, including Young's Modu-
lus, thickness of the geometry and equivalent stress,
are specified for the optimization task in the FEM
analysis medium. The method consists of the follow-
ing steps:
 • form-finding
• surface curvature analysis and panelization
• material specification
• loads & boundary conditions
• FEM analysis for the structural performance
• input & output parameters and optimization
Form-finding
A hanging chain simulation is generated with a live
physics engine based on a spring system in the para-
metric design medium. The parameters include stiff-
ness of the spring, damping and plasticity, in order
to reflect material behaviour. The model represents
forms of tension, but can be inverted to create pure
compression shell structures. By reducing or increas-
ing the stiffness of the spring, equilibrium configura-
tion would be changed, along with the arch height of
the geometry. By adjusting this parameter, numer-
ous structural options in equilibrium can be created.
Once the system gets more complex with additional
loads, the model obtains advantages of computing
complex systems in equilibrium.
The proposed geometry is generated via two op-
erations. First, the construction curves of the geom-
etry are generated by the dynamic spring system.
Various curves are created with different stiffness (S)
values of the material. The second operation is to
build structural surfaces to make an approximation
of an efficient structure driven by the construction
Table 1
The relationship
between Young's
Modulus, stiffness
of spring and
height of arch.
curves. A shell structure resulting from five construc-
tion curves with four common vertices is generated
in the 3D CAD modelling environment.
To incorporate different stiffness values for the
materials into the form-finding process results in dif-
ferentiation of the construction curves, more specif-
ically the height of the arches. Stiffness value of
spring (material) is specified in the parametric model
by establishing a proportional relationship with the
Young's Modulus (Table 1). The differences of the
arch height range from 19.77 m to 18.96 m based on
stiffness values range from S=915 to S=1020 by the
use of different types of steel (Figure 1).
Surface curvature analysis and paneliza-
tion
After the generation process, the surface is analysed
with surface curvature tools in 3D CAD modelling
software to map its surface attributes. Gaussian and
mean curvature tools are operated for the geometry.
Through surface curvature analyses, it is possible to
investigate surface attributes and surface faults, if ex-
ist any. This operation plays an important role for as-
sessing the constructability of the surface.
The proposed geometry is considered as a
monocoque (structural skin) made out of panels
which are extruded by 0.3 m. By considering the scale
of the geometry, there will be a necessity to break
down the geometry into smaller components. Once
the surface is defined geometrically, it is subdivided
into smaller components in order to be manufac-
tured in architectural scale. The surface is subdivided
via flat quadrilateral panels, in which the algorithm
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 549
 Figure 1
Different stiffness
values are used
based on Young's
Modulus of
specified materials
in the form-finding
process, resulting in
differences of their
arch height.

determines the panel sizes by evaluating the points Material specification

on U-V curves (divisions in the x and y directions) as-
sociated with the surface topography. By adjusting
these parameters, panel sizes can be altered. The
point numbers of the U-V curves are 20*20 for the
surface (Figure 2).
Mechanical properties of material are critical param-
eters; in order to assess structural performance of ar-
chitectural geometry. The model is investigated fur-
ther through the FEM analysis medium. Different
types of steel, including Steel AISI Grade 18 Ni (200)
Maraging, Steel AISI 1020 and Steel 17-7 PH Stainless
CH900 are used in the simulation.

Figure 2
Surface curvature
analyses and
panelization on the
geometry, in which
quadrilateral panels
are assigned.

550 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Figure 3
A load of -10.000 N
is applied to the
outer surface of the
shell.
Loads & boundary conditions
Loads and boundary conditions need to be incor-
porated to the overall system. Boundary conditions
are imposed where restraints and loads are defined
along with their quantitative values. By considering
that the proposal is a shell structure, the gravity force
(G) influencing the self-weight of the shell is consid-
ered in the simulation. The geometry is restrained on
surfaces, which sit on the ground, in all directions, x-
y-z. A vector force with a magnitude of -10.000 N is
applied to the outer surface of the shell in the direc-
tion of the gravity force (Figure 3).
FEM analysis for the structural performance
By running the FEM analysis, deformations and var-
ious types of stresses on the geometry can be anal-
ysed and evaluated. Total displacements, equivalent
stress and shear stress assessments are undertaken
for this study. Areas shown red in the total displace-
ment diagrams indicate danger areas on the geome-
try which need to be evaluated by adjusting the ge-
ometry, the loads & boundary conditions or the me-
chanical properties of materials, if the values are not
within safety ranges (Figure 4).
Figure 4
FEM analysis
representing total
displacements on
the geometry.
Input & output parameters and optimiza-
tion
Design Explorer is an optimization analysis tool
which is based on a method called Design of Experi-
ments (DOE) in FEM analysis medium. This method
enables to develop an optimized structure by se-
lected input and output parameters. These param-
eters may vary from the mechanical properties of the
materials such as Young's Modulus, Poisson's ratio,
yield strength, ductility coefficient, to the properties
related to the geometry such as thickness, volume,
mass and to the loading & boundary conditions, as
well as stresses and deformations on the geometry.
During the process of assigning the material,
loads & boundary conditions to the simulation, the
input & output parameters are introduced in the FEM.
Six-Sigma optimization from DOE allows setting in-
put parameters as uncertain parameters. Therefore,
the user can investigate how uncertainties in the in-
put variables may affect the results for the output
parameters. Two input parameters are selected as
the Young's Modulus of the material and the thick-
ness of the geometry. The output parameters are de-
fined as maximum equivalent stress (von Mises) and
the total deformation of the geometry. Based on the
optimization outcome, the thickness of the geome-
try is varied from 0.27 to 0.35 m, corresponding to
the maximum stress and deformation. The thickness
of the geometry needs to be greater than 0.3 m, in
order to reduce maximum stress and total deforma-
tions on the geometry when the Young's Modulus is
constant. Altering mechanical properties of the ma-
terial, such as optimizing Young's Modulus means to
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 551
 search the most suitable material based on perfor-
mance requirements. This exposes the limits and ca-
pabilities of materials used in the simulation.
RESULTS
By running structural performance simulation of dif-
ferent set-ups for three steel types including Steel
AISI Grade 18 Ni (200) Maraging, Steel AISI 1020 and
Steel 17-7 PH Stainless CH900, the weight of the ge-
ometry, as well as the von Mises stress values on
the geometry are computed. Therefore, structural
efficiencies that are defined by maximum von Mises
stress value divided by weight of the geometry, are
552 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1
calculated and compared by the use of different ma-
terials (Table. 02).
Efficiency ratio is determined by the efficiency
value 1 (E1), divided by efficiency value 2 (E2). Based
on the simulation results, it is specified that it is crit-
ical to incorporate mechanical properties of the ma-
terial in the form-finding process, in order to increase
efficiency, as in Steel, 17-7 PH Stainless CH900, of
which efficiency is 4.342 times more, if the geometry
is informed by mechanical properties of the materials
in the form-finding process (Table. 03).
Table 2
Structural
performance
simulation results,
material properties
and resulting
heights of arches.
Table 3
Efficiency ratios.
CONCLUSIONS
Although panelization tools aim to create analyt-
ical methods for building complex surfaces, they
usually investigate structurally inefficient forms, like
freeform shells. Efficiency in architectural geometry
needs to be investigated further considering material
properties, architectural geometry and performance
requirements. Although physics-based dynamic sys-
tems aim to generate structurally efficient forms by
incorporating form with performance requirements,
additional studies should need to be undertaken on
questioning of how to link structurally efficient archi-
tectural geometry with mechanical properties of ma-
terials. The aim of this study is to generate an infor-
mation loop in which Young's Modulus, the stiffness
of the material may inform both the form-finding
process and structural performance simulation.
The proposed method offers steps including
form-finding, series of analyses applied for the ar-
chitectural geometry and structural performance, as
well as optimization. Following the generation of a
dynamic system for form-finding through a hanging
chain model by a spring system in parametric design
medium, FEM analysis is operated for the static struc-
tural performance simulation for different set-ups, in-
cluding the usage different types of steel. For the op-
timization process, input and output parameters, in-
cluding Young's Modulus, thickness of the geometry,
equivalent stress and total deformations are speci-
fied in the FEM analysis medium. Based on the op-
timization results, the thickness of the geometry can
be increased or decreased as a basic operation, re-
sponding to the maximum equivalent stress and total
deformation. By optimizing mechanical properties
of the material, such as Young's Modulus, architec-
tural designer may search better materials for geom-
etry to accommodate its performance requirements
or to modify material used by enhancing its proper-
ties in molecular level. Establishing a relationship be-
tween the form-finding and optimization processes
is a challenge to be investigated.
By assessing simulation results, efficiency values
are calculated by the use of different materials. It is
identified that it is critical to incorporate mechanical
properties of materials into the form-finding process,
in order to evaluate efficiency, as well as to generate
awareness related to the relationship between mate-
rial, form, performance and structure.
Buildings are complex systems affected by differ-
ent performance aspects. For the future lines of the
research, the author aims to investigate how archi-
tectural geometry can respond to the material prop-
erties and performance requirements within compre-
hensive boundary conditions, because buildings are
complex systems affected by different performance
aspects. Also, various other input and output param-
eters related to the materials, geometry and fabrica-
tion constraints may be incorporated to the system
at the early stage of the design process.
REFERENCES
Ahlquist, S and Menges, A 2011 'Behavior-based Com-
putational Design Methodologies: Integrative pro-
cesses for force defined material structures', Proceed-
ings of ACADIA 2011, pp. 82-89
Aish, R, Fisher, A, Joyce, S and Marsh, A 2012 'Progress To-
wards Multi-Criteria Design Optimisation Using De-
signscript With Smart Form, Robot Structural Analy-
sis and Ecotect Building Performance Analysis', Pro-
ceedings of ACADIA 2012, pp. 47-56
Burry, M 2006, 'Almost rectangular: Thinking beyond the
Shape, Form and Geometry 1 - Volume 1 - eCAADe 32 | 553
 commonplace', Letter in Architectural Research Quar-
terly, 10, pp. 182-184
Gmelin, S, Agger, K and Henry, M 2011 'Simulation De-
sign Tools: Using Parametric Building Information
Modeling and Physical Simulation for Form Finding
of Double Curved Surfaces', Proceedings of eCAADe,
pp. 215-224
Kilian, A 2004 'Linking Digital Hanging Chain Models to
Fabrication', Proceedings of ACADIA 2004
Kilian, A. 2006, Design exploration through bidirectional
modeling of constraints, Ph.D. Thesis, MIT
Pottman, H, Schiftner, A and Wallner, J 2008, 'Geometry
of Architectural Freeform Structures', Internationale
Mathematische Nachrichten, 209, pp. 15-28
Shea, K, Aish, R and Gourtovaia, M 2005, 'Towards
Integrated Performance-driven Generative Design
Tools', Automation in Construction, 14 (2), pp. 253 -
264
Xie, Y M, Felicetti, P, Tang, J W and Burry, M 2005, 'Form
finding for complex structures using evolutionary
structural optimization method', Design Studies, 26,
pp. 55-72
Yazici, S 2013 'Performance Based Pavilion Design', Pro-
ceedings of eCAADe 2013, Delft University of Technol-
ogy, pp. 127-135
Yazici, S and Tanacan, L 2012 'Material in Performance-
driven ArchitecturalGeometry', Proceeding of Asso-
ciation of Collegiate Schools of Architecture (ACSA)
100th Meeting, Massachusetts Institute of Technol-
ogy , pp. 266-273

554 | eCAADe 32 - Shape, Form and Geometry 1 - Volume 1

Visualisation
Digital Etching
An Alternative Approach to Drawing

Johan Voordouw
Azrieli School of Architecture & Urbanism, Carleton University
johan.voordouw@carleton.ca

This paper explores an analog method of intaglio etching via digital

/computational means. It qualifies an alternative approach to digital
representation through the development of a narrative based architectural
expression. The project, entitled Weerkas, was undertaken as an open call for the
International Architecture Biennale Rotterdam 2014. The drawings used a
combination of 3D computational modelling and AutoCAD drawing to explore
emerging drawing technique derived from historical drawing methods. These
digital drawings were photo-transferred and etched using a polymer intaglio
plate to produce the final set of drawings. While analog / digital connections
have been increasingly explored in architecture, it is rarely used to re-create past
modes of production. While this paper counters a nostalgic approach to
representation, it is anecdotal evidence of a pluralistic method in combining
three-dimensional and two-dimensional drawing technique as an emerging form
of architectural expression.

Keywords: Architectural Representation, Etching, Narrative Drawing

PROJECT NARRATIVE throughout the country and the seasons, deviating

The Weerkas project was completed for an open call
to the 6th International Architecture Biennale Rot-
terdam 2014 (IABR 2014). The twelve etchings were
developed over a six months by the author and a
project assistant. The theme of the biennale "Urban
by Nature" was used as the primary narrative for the
project. Each composition is a vignette of a specula-
tive architectural structure, acknowledging a specific
weather condition located in a physical, recogniz-
able place in the Dutch landscape. Furthermore, the
twelve drawing represents one month in one of the
twelve provinces. Therefore, the project is located
from the conventional Dutch focus on the urbanity
of the Randstad (the dense urban corridor from Ams-
terdam to Rotterdam) spreading the project across a
variety of physical conditions.
The narrative of the project is a commentary on
changing weather patterns and its latent affects on
the socio-cultural milieu of the Netherlands. The
project tile - Weerkas - loosely translates to 'Weather-
house'. This is not a house in a domestic sense
(huis), the interiority of which is also a common Dutch
theme, but a translation of a greenhouse (kas), a
place where weather and interior climate is tightly

Visualisation - Volume 1 - eCAADe 32 | 557

 controlled. The underlying narrative noted that while
the Netherlands is overly accomplished at exerting
control and management of its land, water and the
internal climate of their intensive agricultural econ-
omy. Over the course of the coming decades the ex-
ternal weather will be increasingly more variable and
difficult to both predict and control. Therefore,
the small interventions act as a didactic 'follies', nar-
rating the increasing disparity between historic, nor-
mative climate cycles and emerging weather con-
ditions. Given the homogenous geography of the
Dutch landscape, these weather anomalies will have
a dramatic affect on Holland's economic and cultural
future. The expression of each architectural interven-
tion oscillates between structured research and for-
mally expressing these issues through irony, humour
and subversion. Therefore, it is critical to note that
the follies do not attempt to resolve the issues relat-
558 | eCAADe 32 - Visualisation - Volume 1
ing to changing weather patterns. They act as mech-
anisms by which these changes can be recognized -
they are a climatic datum.
The etching technique referenced historic land-
scape etchings produced in Holland's Golden Age of
art in the 17th Century, tying the contemporary nar-
rative of climate change to historic expressions of
the Dutch landscape. The function and didactic va-
Figure 1
lidity of each structure is underscored through tex-
Waterkas,
tual research of academic journal articles in partic-
Rotterdam - One of
ular the archives and data of the KNMI (Koninklijk
the twelve
Nederlands Meteorologisch Instituut - Royal Nether-
completed etchings
lands Meteorological Institute). Three examples from
the set of twelve drawings include changes in water
level either due to increased spring run-off or sum-
mer drought and its affects on shipping in Rotterdam
(Waterkas - Waterhouse) (figure 1), changing precip-
itation patterns in Haarlem (Regenkas - Rain house)
or decreased cloudiness in the province of Friesland
(Wolkenkas - Cloud house). As aforementioned, the
follies do not attempt to solve the problem per se as
these issues are very large, complex and global. The
follies merely express the disparity of historic climate
to current conditions and attempt to establish a rela-
tionship of architecture, landscape and public to ac-
knowledge this incremental, but pervasive, change.
DRAWING DEVELOPMENT
The research sought to explore both the notion
of 'conventional' drawing in AutoCAD and how
one would develop an intaglio etching technique
through digital modes of design.
At the beginning of the project it was impera-
tive to develop a viable mode of drawing in AutoCAD.
Therefore, traditional etchings were researched to
understand general landscape composition, conven-
tional line work, hatching technique and line weight-
ing. A number of techniques were explored includ-
ing Wacom tablets, and various AutoCAD commands.
In the end the author concluded that working us-
ing the mouse and polyline tool offered the highest
speed, accuracy and drawing intention. Ultimately,
this was a personal preference.
 To summarize the general process, the develop-
ment of each drawings occurred over a three-week
period. A number of drawings were developed con-
currently to ensure completion of the project and
similitude within the drawing set. The first phase
of each drawing was the locational and weather re-
search to develop an underlying narrative for the
structure. After a suitable location was found a back-
ground image would be collaged together to ensure
Figure 2
The general
drawing
development of the
Regenkas, Haarlem
a recognizable location for the folly thereby ground-
ing the project in a physical place. This background
was attached in an AutoCAD file and the drawing
process would begin (figure 2). The drawings were
developed using the polyline tool both tracing and
drawing a background to form a cohesive composi-
tion. Due to the scale and complexity of the draw-
ings it was necessary to gauge the density of lines to
ensure one drawings was tonally comparable to its
Visualisation - Volume 1 - eCAADe 32 | 559
 neighbouring drawing. It was critical that each draw-
ing had comparable line density to create a com-
position both within the page, and within the set
of twelve drawings. Once the base layout of the
background was established the three-dimensional
modelling of the folly would commence. The design
of each folly used a number of modelling software.
Projects were developed in Rhino, Rhino Grasshop-
per and 3D Studio Max. After the structure was digi-
tally modelled the file was imported into Rhino in or-
der to use the contour and Make2D function. These
resultant line drawings could then be exported as
a dwg file and integrated into the AutoCAD back-
ground image. After the placement of the 'flattened'
structure within the AutoCAD background the draw-
ing was completed, with a series of iterative prints
made to determine areas that needed additional
work to complete both the drawings and the set. Af-
ter a drawing was finished the dwg file was exported
to Adobe Illustrator to check the line weights. This
step occurred later in the project when the Illustrator
files were sent to the plate maker for printing.
Specific areas of the process that required con-
sideration were the correct alignment of the perspec-
tival angle of the three-dimensional model with the
surrounding background drawing. Alignment of the
perspective angle of the structure to the background
was achieved in one of two ways. One method was to
import the original background image into the Rhino
file and align the angles in the perspective view how-
ever this was cumbersome. While it did achieve satis-
factory results it was not exact. A second method was
to include a small cube in the Make 2D drawing from
which perspective lines could be extracted to deter-
mine the horizon line and this could be aligned to the
horizon of the background image. While this method
was exact in the placement of objects vertically, it did
on occasion shift things horizontally and at times re-
quired adjustment. This was time-consuming given
the nature of the Make-2D function and the complex-
ity of a number of the Rhino models.
Solar consistency was achieved through the
drawing and aligning the folly to existing shadows
560 | eCAADe 32 - Visualisation - Volume 1
within the original background image. After deter-
mining the correct sun position the model was ren-
dered in grey-scale to determine the shaded planes/-
parts of the folly and the shadows cast on the sur-
rounding landscape. This render was exported as a
jpeg and imported into Adobe Photoshop. This file
was broken into a series of layers using the range tool.
This would determine the darkness of the shadow
and therefore the number of hatching layers required
to achieve a comparable tonality. By this method it
was easy to determine if a part of the folly needed
one hatching layer (light) or multiple hatching layers
(dark).
What is relevant is the relative ease with which
three-dimensional modelling was integrated with
two-dimensional drawing. The process of digitally
modelling the architectural follies whilst simultane-
ously drawing the background line drawing is an in-
teresting mode of drawing development. This is not
image making in the sense of rendering, but an un-
derstanding of the dimensional connection between
modelling in perspective view and drawing in a per-
spective view and aligning these two modes in a fixed
drawing. What is important to note however is how
these two modes of working are still very much sep-
arate. The shadows in one mode of working have no
bearing on the other; modelling in Rhino has no im-
plication on drawing in AutoCAD. These conditions
need to be judiciously resolved if not by hand, then
by eye.
PRINTING DEVELOPMENT
Originally it was intended that the author print the
work using conventional plate making techniques. A
series of plates were purchased in cold and hot rolled
steel, aluminium and brass for testing. Additionally a
series of transfer methods were employed to deter-
mine the most efficient and reliable method of trans-
fer. While it was possible to purchase photo-transfer
plates for circuit boards the size of the final drawings
(360 x 510mm) made this method financially unvi-
able, as the plates were too large. Other preliminary
methods included using silkscreen emulsion, which
was partially successful on aluminium but the emul-
sion did not sufficiently adhere to the surface to pro-
duce a reliable transfer. Another method using steel
worked in covering the plate with resist and using the
laser cutter to score the drawing through the resist
and into the steel. The acid was then used to deepen
the cut to improve the line weight when printing. Af-
ter these initial tests it was determined that the com-
plexity of the printing would require professional as-
sistance. The final set of intaglio plates were pro-
duced by Mark Herschede of Haven Press Studio.
In early correspondence Mark suggested a polymer
plate to render the drawings. Polymer plate main-
tained the high level of detail required for the project
with low toxicity. Photo-responsive polymer hardens
when exposed to a specific bandwidth of light. The
plate was exposed to a photo-positive to make the
intaglio matrix. The remaining unexposed, plate was
washed away leaving the recessed intaglio surface.
The plate used in printing the etchings was .73mm
thick. The image was RIP transferred to a 2400 dpi
imagesetter film and exposed to the polymer plate.
The final images were printed on dampened Arches
Cover 250 gsm paper using Portland black ink on
a Takach Etching Press ([E-mail]. Message to: Her-
schede, M. 5th June 2014).
PROJECT SIGNIFICANCE - ALTERNATIVE
DRAWING
The project makes a series of contributions to digi-
tal drawing both in process and the resultant mode
of production. Additionally, the etchings are unique
in connecting emerging drawing technique to past
modes of architectural expression. This historic link
was used to establish a coherent narrative, connect-
ing place and theme to a deeper cultural context.
The drawing process is critical to the project.
Drawing is a speculative act, it contains concurrent
to its execution an act of representation but also, as
Errol Barren notes in his essay 'Drawing in the Digi-
tal Age' research and analysis (Trieb, 2008). The re-
search imbedded into the work is therefore a layer-
ing of different speculations. Firstly, the research re-
garding the project narrative, what each folly is and
how it works. This work is derived from the textual
research from the KNMI research papers. Secondly
is the research of how each structure is formally ar-
ticulated and how it embeds into the landscape, and
lastly and most importantly, the research developed
a larger 'meta' question - how to complete the draw-
ings in the first place. Questions that arose early in
the project process included how does one etch us-
ing Rhino and AutoCAD? Do these drawings look like
historic copies or do they come into their own as
something else? And lastly, what is the implication of
digital etching, is it merely a broadening of an ever-
expanding palette of representational techniques or
does it lead to other computational questions?
The speculative technique of iterative three-
dimensional modelling, drawing in AutoCAD, and
connecting to intaglio etching is novel. This mode of
production required a rethinking of the conventional
approach to both modelling and drafting software.
One anecdote of how one etches using AutoCAD
and Rhino is rethinking how to achieve a suitable
depth in the composition of the drawing through line
weight. In etching it is possible to achieve chang-
ing line thickness by adding pressure when drawing
with the stylus. This is not possible to achieve in con-
ventional AutoCAD drawing. Therefore, it was de-
termined, early in the process, for all lines to have
the same line thickness. Density and depth were
henceforth achieved through the close adjacency
and/or overlapping of line, not from pressure. This
caused issues when using the Make 2-D function in
Rhino to 'hatch' the three-dimensional models. A
sufficient contour distance was determined to de-
velop the necessary darkness, to express the desired
depth. These lines were often times augmented to
achieve the necessary results. An additionally issue
was that the contour function could be visually prob-
lematic when placed in perspective view. Contouring
vertically typically was correct but contouring from
right/left or front/back could have odd perspectival
implications - either lines would double up or disap-
pear making a tartan pattern rather then a consis-
Visualisation - Volume 1 - eCAADe 32 | 561
 tent hatch. Therefore a large number of hatch lines
were amended after contouring was completed to in-
tegrate the image into the background and better ex-
press the perspectival angle in the architectural inter-
vention. Another differentiation between traditional
etching and AutoCAD drawing was the fluidity of the
line. Initial difficulties arose when drawing in splines
in that they were not particularly accurate, difficult
to offset and not as stable when the drawing was
converted to an Adobe Illustrator file. While these
problems could have been mitigated to offer surety
only Polylines were used in the drawings. The re-
sultant process developed a fundamentally different
line type from the hand. In using the polyline tool, a
line perceived as curved from a distance was actually
segmented into a series of straight lines on closer in-
spection. Therefore, these segments scale the draw-
ing, as one inspects the minutiae of the work, a tree
leave as oppose to its trunk, they discover smaller
and smaller segments of straight lines composing the
curves (figure 3). What is extrapolated from these
562 | eCAADe 32 - Visualisation - Volume 1
anecdotal conditions is that they undermine the crit-
icism lodged at the digital as a drawing tool. While
one could continue to discuss the possible discon-
nect of the hand to eye connection, what is critical to
note is that when the intention is to draw, any mode
of production is legitimate, they simply offer different
representational outcomes.
The significance of the narrative aspect of the
project is more difficult and subjective to adjudicate.
It is therefore necessary to underscore a series of ten-
tative assumptions. CJ Lim states that the, "mod-
ern age has been unkind to narrative architecture"
(Lim & Liu, 2011) this is true for contemporary dig-
ital modes of expression as well. In relation to the
use of narrative in contemporary architectural spec-
ulation, these comments are not a critique of the dig-
ital per se but the overarching preference of compu-
tation to fabrication and production; the precision of
making superseding the craft of a 'story'. Addition-
ally it stems that narrative often pulls from history.
Something the current pace of the digital is steadily
Figure 3
A detail of the
polyline work
(Zandkas, Loonse &
Drunense Duinen)
outrunning. In respect to the Weerkas, the drawings
had to express a response to the issues in the land-
scape. Embedding them in a specific time and space.
This is very different from the clarity of the diagram.
Each intervention is a continual layering of simulta-
neous histories from past modes of art, evolving cul-
tural practice, emerging formal expressions and cur-
rent digital practices. This approach towards narra-
tive allows for multiple engagements, the drawings
offer varied topical entry points that connect a series
of related dialogues. These topics include but are not
limited to: weather with cultural practice, historic cli-
mate in relation to current weather, public engage-
ment through architecture, didactic architecture and
a questioning of both urban practice and nature -
particularly in the Netherlands. Narrative drawings
offer a nuanced oscillation between concurrent is-
sues; a drawing set allows issues to ebb and flow, re-
inforcing certain points in one drawing, and others in
adjacent drawings. While this is at once the success
of the project, it is also its main criticism. Drawings
are experientially layered but not always exceedingly
clear. This became apparent in the biennale opening
where the short annotation beside each drawing was
critical for its understanding. However, this should
not negate its validity. Narrative offers an opportu-
nity for digital practice to express the complexity of
its condition; to allow the process percolate through
the drawings. In this sense, all work contains a narra-
tive, however a common critique in relation to digital
work, is that the work rarely reveals its prowess. Fur-
thermore, in moving forward with digital discourse,
narrative is not a total divergence of a growing facet
of current computational thinking. It is merely a re-
definition of what research intends to express. If 're-
sponsiveness' in all its guises is resultantly didactic,
it too will form a narrative. How this narrative con-
nects to cultural context will in part determine how
successful it embeds into the evolving milieu.
CONCLUSION
Drawing remains a critical discourse in architecture.
However, digital projects are typically focused on
three-dimensional modelling in relation to emerging
construction techniques. It is rarely used as a means
to explore new modes of two-dimensional drawing
techniques. This is important because it can clarify
the current misunderstanding between renderings
as image making as oppose to drawing. While both
processes deliver a visual expression, the two modes
are fundamentally different. Additionally, when dig-
ital software is used for two-dimensional drawings it
is focused on technical working drawings or the dia-
gram as oppose to a place of exploration. This pre-
vents the narrative aspects of architecture from en-
tering into the design development process. Work-
ing in the digital offers us the opportunity to ask
meta-questions on all facets of architectural conven-
tions. Using digital programs we are able to ques-
tion the very essence of a line. Tim Ingold (2007) dis-
cusses how lines can be differentiated into multiple
categories, the most relevant to architecture being
defined as either a thread or a trace. Conventionally
architecture has drawn traces on the page, the sur-
face being inscribed by a drawing utensil. Changes
occurring through erasure as oppose to deletion. But
in the digital we work with threads, like a fishing line
cast into a pond. These lines are not necessarily out-
lines; on occasion they are merely guides, giving only
the slight definition to generate form. Therefore, un-
like the trace, which only has a front, the digital line
has a back, top/bottom and left/right. It is not a line
on a surface but a line in space. We can move through
these lines, connect them with surfaces, or extrude
to define form. It is a place of inhabitation that we
can walk through, view and experience. Whether we
pin this down on paper, either through printing or
press, or animate it into continuous movement our
opportunities to explore drawing have exponentially
increased. The digital is not the end of drawing, but
a rewriting of the conventions that have tradition-
ally defined drawing. Therefore, this is an acknowl-
edgement of a plurality of expression, not one of re-
placement but one of in-addition-to. As Mario Carpo
(2014) relates, as we become more attune to a spe-
cific set of digital skills these skills lead to a particular
Visualisation - Volume 1 - eCAADe 32 | 563
 way of digital making. This project expresses an al- riods of record technological development.
ternative approach. The digital remains one of the
great architectural platforms for free expression, as REFERENCES
it becomes normative in architectural pedagogy and Carpo, M 2014, 'Breaking the Curve: Big Data and De-
profession, collectively we should strive for plurality sign', Artforum, February, pp. 169-172
to ensure we build our digital modes of production Grosz, E 2001, Architecture from the Outside: Essays on Vir-
on the widest possible foundation. The plurality is tual and Real Space, MIT Press, Cambridge
not only found in the iterative process of complex Ingold, T 2007, Lines: A Brief History, Routledge, London
Lim, CJ and Liu, E 2011, Short Stories: London in Two-and-
computation but also in the historic modes, which
Half Dimensions, Routledge, London
have since fallen by the wayside. Again, this is not Trieb, M (eds) 2008, Drawing/Thinking: Confronting an
a historicist approach for histories sake, but an ac- Electronic Age, Taylor & Francis, New York
knowledgement that when conducive to the over-
arching framework of the project, historic modes of
representation can engage a deeper cultural context.
Furthermore, what is significant is that it undermines
the conventional critique lodged to drawings pro-
duced 'on the computer'. If we consider Elizabeth
Grosz (2001) position that the computer is yet an-
other virtuality, similar to writing or drawing we note
that as the computer becomes more nuanced we are
able to inscribe greater intention in the process of
the work. In rectifying this position our ability to pro-
duce work becomes even more varied and individu-
alized, countering the concern raised by Carpo's ob-
servation. The ability of software to reconcile varying
dimensions in drawing, oscillating between drawing
and modelling is a benefit that validates its purpose.
What is interesting in the Weerkas project undermin-
ing the conventional use of both drafting and mod-
elling programs is the opportunity to reflect - process
through amendment not autonomous iteration. The
digital offers many opportunities to pursue greater
efficiency. Its improved connection of design to fab-
rication is established practice and pedagogy. How-
ever, the digital also offers a point of reflection, an
exploration of alternative modes of architectural pro-
duction that travel a slower, bumpier road. Digital
modes of production will invariably overshadow and
surpass many analog modes of making, redefining
craft and its procurement, but it should not be nihilis-
tic. As we find new modes to establish our evolving
cultural milieu, periodically checking back might pre-
vent the cultural alienation alluded to in previous pe-

564 | eCAADe 32 - Visualisation - Volume 1

Data Integration In A Visual Mode

Anetta Kepczynska-Walczak
Lodz University of Technology, Poland
anetta@p.lodz.pl

The principal aim of this paper is to discuss data integration issues in the context
of urban scale studies. A special attention is dedicated to built environment,
visual thinking and synthesis of knowledge. The paper is based on literature
studies, professional experience and the outcomes of an experimental students'
project conducted by the author last year.First, the theoretical background and
the current state of research in the area is revealed. Then, the project theme,
goals and organisation are described. So, the main idea of the experiment was to
explore data collection methods at the urban scale without a design goal since the
prime approach was to take into consideration students' perceptions of space and
its multifaceted aspects. Thus, to maintain an open mind about gathering such
information and not to bias participants towards one approach or another were
crucial. Finally, the outcomes of the project are discussed. Students'
presentations showed that they used different approaches not only in terms of
digital tools but also in terms of their understanding of data integration.

Keywords: Built environment, visual thinking, data integration

BACKGROUND ising city changes was described by Kaga and Sug-

Nowadays computer techniques are commonly used
to support design process from early sketching to
building information modelling and management.
What is more, they play a significant role in the phase
of preparing analysis, which precedes planning or
design. A systematic survey and a comprehensive
evaluation of computer software for urban design
process was done by Charlton, Giddings and Horne
(2008). The review concentrated on identifying and
selecting software from six categories: 3D geomet-
ric modelling, pedestrian modelling, environmental
noise mapping, thermal comfort software, wind anal-
ysis software and platforms (VR engines). The inter-
esting work on finding low cost solution for visual-
awara (2008). Another development worth mention-
ing is a GIS tool kit for the socio-economic valua-
tion of urban areas dedicated to designers and deci-
sion makers (Chiaradia, Schwander, Gil and Friedrich,
2008). There is no doubt that the idea of Space Syn-
tax influenced methods of acquiring and perform-
ing specific data crucial to problem solving. Since
it is a support tool for urban analysis, enabling in-
formed decisions, a lot of successful examples of ap-
plications can be found in literature. For example, in
the "Strategic Planning and Design with Space Syn-
tax" (Czerkauer-Yamu and Voigt 2011) the authors
discuss the constraints and opportunities of Space
Syntax and show how it can add value to strategic

Visualisation - Volume 1 - eCAADe 32 | 565

 planning and design for a sustainable and sustaining tion posed in the paper is not how to read the city
built environment. but, moreover, how to present such perceptual expe-
This brief overview only highlights main research rience in a coherent manner. There is no doubt, such
topics in the field of urban studies and analyses in representation would involve not only visual thinking
the context of searching for the best tools for the fu- but also the capability of abstraction.
sion of data. However, it is still difficult to find a com-
prehensive and ready to use methodology for multi- TASK, METHODOLOGY AND OBSERVA-
faceted data integration.
TIONS
Another issue which should be taken into ac-
The task of the project was to search and test differ-
count as a background study is connected with per-
ent tools and methods in order to explore possibili-
ception. The perception of space may be consid-
ties of data representation in a unified view. The sub-
ered through the relationship between a human be-
ject was investigated on the basis of an experimental
ing and a built environment. As Bishop and Lange
project conducted within a Computer Aided Archi-
(2005) stressed, it was estimated that 80 percent of
tectural Design course for undergraduate students
our impression of our surroundings comes from sight
at the Institute of Architecture and Urban Planning
(Bishop and Lange 2005). A further concern is that,
of the Lodz University of Technology. The location
according to Sigmund Freud, a city is a metaphor
chosen was of a municipal character - a part of Naru-
of human psyche. To continue, the perception pro-
towicza Street in Lodz, between Kosciuszki Street and
cess is extremely individual, and actually depends
Dabrowskiego Square. Narutowicza Street is one of
on human psyche. What is more, it is necessary to
the most important streets in Lodz city centre, going
quote here that a city is not only images of the forms
West-East direction. It is perpendicular to the main
given us in the process of visual perception but it
urban axis, which is Piotrkowska Street. Its history,
also means all invisible and nowadays not existing
transformations and not realised projects of its future
forms. As a static picture of a city is not equal to re-
appearance place this street high on the list of urban
ality, the perception of a city needs to consider time
problems in Lodz.
factor (Asanowicz and Asanowicz 2008). According
So, twenty four students formed four teams to
to Ching "since we move in time through a sequence
elaborate on the task. They were expected to work
of spaces we experience a space in relation to where
in a little bit competitive manner between the teams
we've been and where we anticipate going" (Ching
since one of the goals was to result with more than
2007). In other words, we receive lots of raw sen-
one answer to the task. Therefore, each team worked
sory information and our brains use this information
separately to produce a final output. They started
in order to build a model of the world that we may
with site visits in order to be acquainted with all sen-
use to predict and exploit the environment (Asanow-
sorial aspects of the place. Then, more information
icz 2011). The sequence of elements making up the
was gathered by studying maps, plans and docu-
spatial configuration of an urban assumption is a se-
ments concerning the urban area. A variety of col-
quence of spatial pictures which we perceive as we
lected data, in terms of types and also formats, and,
get to know a part of the entire configuration. As
moreover, resulting analyses were to be transformed
Cullen (1961) stated "serial visions are a drama of jux-
in order to achieve an integrated view of data. At
tapositions that come alive when we walk down the
the end, students were asked to prepare presenta-
street and experience the contrasts between, for in-
tions that would visualise the multiple sets of data in
stance, the street and the courtyard, light and dark-
a compact yet comprehensive way. The results of the
ness etc".
task appeared multifaceted and imposed a question:
It is necessary to stress at this point that the ques-
what can be learnt from the project and put forward

566 | eCAADe 32 - Visualisation - Volume 1

 as guidelines or recommendations for future work? Students prepared a diagram simulating the
Completing the task involved exploration of vari- movement from the start point, which was Kosciuszki
ous digital tools and techniques to describe, visualise Street, to the endpoint (Dabrowskiego Square)
and communicate discovered values of an urban tis- recording all information on the way (figure 1). To
sue in a clear and explanatory way. Participants in- achieve that, they supplied the final presentation not
vestigated sensorial aspects, like visual, sound, smell, only with high resolution images of textures - to make
tactile and kinaesthetic. What is innovative and valu- the picture more tangible in a way - but they also
able about the experiment, is the way the task was added a soundtrack to strengthen a holistic impres-
defined: no design, no further purpose to focus on af- sion of the places along the route.
ter having analytical part done. Students were asked
Figure 1
to go and just watch, observe, take some notes, be
A diagram
concentrated on perception of the place. Such ap-
simulating the
proach made the work of analysing different to a typ-
movement with a
ical preparation process preceding planning or de-
record of collected
sign and resulted with very interesting outcomes.
information.
The message behind is that when a designer col-
lects purpose-focused data it is very often affected by
his/her intention and by that it becomes not neutral.
In other words, it is not the raw data any more.
The following paragraphs focus on description of
different approaches of each team.

Team 1.
So, the first team decided to base Narutowicza Street
analysis on the information that they gathered by
means of four senses: vision, hearing, smell and
touch. They agreed that the vision helped identify-
ing the basic 'shape' (morphology) of the urban tis-
sue and enabled them to locate that space in the city.
Students put forward an argument that it supplies us
with the most important information and is crucial in
our functioning. On the other hand, hearing, smell
and touch are giving us the additional information
only that decide if we feel safe and comfortable in the
environment. As to the touch itself, according to the
studies, the group investigated that people perceive Team 2.
textures in the urban space more by their feet than by The second team was asked to make the same
hands, because usually they do not touch each tex- route but, comparing with the first team, in op-
ture, but perceive it by using a sense of vision. Appar- posite direction. So, they started their walk on
ently, pedestrians constantly touch textures by feet Dabrowskiego Square and finished it at the crossing
when they walk. of Zielona and Kosciuszki Street. During the walk
they paid attention to many aspects, namely: his-
tory of the place, textures of surrounding surfaces,

Visualisation - Volume 1 - eCAADe 32 | 567

 smells, noises, safety, revitalised areas, road quality
and general advantages or disadvantages. What is
more, they defined a goal: to focus on the percep-
tion of the overall atmosphere of the place and how
it influenced pedestrians. In order to achieve that
students decided to recall history of the place and
present it in a visual mode. They came up with an
idea to compare photographs of particular buildings
or Narutowicza Street perspectives taken in different
periods (figure 2).
To prepare further analyses they decided to
make the most use of colours. For example,to distin-
guish valuable buildings on a map they marked them
in dark grey. They also gave two other categories,
viz.: buildings of medium value, and buildings not
suitable for surrounding and without any significant
value. To show the road quality they used different
blue tones (figure 3).
568 | eCAADe 32 - Visualisation - Volume 1
In order to highlight negative aesthetic values,
like: blind walls, height differences, excess of adver-
tisements, distorted building line, random and not
composed greenery, poor quality fences, crooked
pavement, rugged parking, space disorder, empty
plots (figure 4) students applied red colour to the pic-
tures taken during the walk, showing by that their at-
titude to the problems very strongly.
Although the team found very interesting as-
pects of the analysed area, and tried different tech-
niques to present them, finally they prepared a linear
Figure 2
presentation with a basic method of showing singu-
Historical and
lar aspects separately. They did not succeed in terms
contemporary
of coming up with a presentation integrating all lay-
views of
ers of collected data in a holistic way.
Philharmonic Hall in
Lodz.
Team 3.
The third team approach to the task was different.
They divided the whole area into six parts: five seg-
ments of Narutowicza Street, separated by nodes
(crossroads), and Dabrowskiego Square - and de-
cided to analyse them separately. What is more, they
divided collected information in two main groups:
objective values and subjective values. The objective
values covered elements of the city structure, such as:
buildings, greenery, people, sounds while the latter
referred to the values recognised and interpreted by
senses. According to students' assumptions it meant
things which influenced their senses and emotions,
which were interpreted individually They assigned
light, atmospheres, colours and smells to this group.
Actually, there are some doubts concerning the
methodology since sounds can be also understood
as subjective values whereas light can be analysed
as the urban structure characteristic. So, some el-
ements do not fit entirely the group they were as-
signed to.This team made the most use of text com-
paring to others and, by making use of graphic pro-
grams they achieved a kind of poetic atmosphere of
the images related to the topics discussed. Unfortu-
nately, what was the most surprising, those images
did not present the analysed place! Although stu-
dents tried to create a matrix of various information,
Figure 3
Road quality
analysis.
the final presentation was composed of single tracks
of observations reflecting six parts of the analysed
route. Since the outcome was not reflecting the true
image of the analysed area it was difficult for the audi-
ence to derive the objective data or conclusions from
it.
Figure 4
Highlighting
negative aesthetic
values.
Team 4.
The forth team was not able to work out one com-
mon solution, and, eventually, they split into three
smaller teams. The first sub-team idea was to analyse
selected issues most important for them and then,
present them in the volumetric form. So, they de-
cided to study the car intensity, pedestrian intensity,
light and safety. Actually, they used 3D modelling
digital tools, basically used for buildings modelling,
to present outcomes (figure 5).
The second sub-team focused on sounds related
to urban zones. With devices available nowadays
it is easy to record sounds, but the problem posed
was of different nature: how to represent them vi-
sually and inform the positive and the negative to
extract conclusions. So, transformations of sounds
into geometry were tested with the application of
a couple of programs to get an animated picture of
the streetscape. As a result, the notes were repre-
sented by circles, the size of which corresponded to
the length of the notes. The notes of each part (a part
understood as a unique track/channel combination)
were connected sequentially by lines (figure 6). And
then, with the picture of the sounds of cars, trams,
people, vehicles on a building site, it was possible to
create a video showing the evaluation of the urban
zone in terms of the quality of life.
The last sub-team decided to focus on the anal-
ysis of accessibility of services in the area in terms
of their variety and in terms of a distance to public
transport nodes. They understood this issue as one of
the most crucial in the context of urban environment
analyses. So, they applied Quantum GIS to create a
database related to a map of the area (figure 7) and,
finally, to be able to perform a number of queries.
What they discovered and appreciated while learn-
ing the digital tool, that the chosen software allowed
not only for effective and reliable visual representa-
tion of the outcomes but appeared very useful for fur-
Visualisation - Volume 1 - eCAADe 32 | 569

ther studies. They noticed, that in a contrast to the
solutions provided by their colleagues, it was a tool
for conducting deep studies and a comprehensive re-
search, due to the capacity for numerous spatial anal-
yses.
570 | eCAADe 32 - Visualisation - Volume 1
Figure 5
A volumetric
visualisation of
collected data.
SUMMARY AND FINAL REMARKS
A fundamental problem posed in this paper is a spa-
tial knowledge acquisition, and furthermore, trans-
forming it into a visual and comprehensive represen-
tation. The focus on the description of urban envi-
ronment perception and understanding made this
Figure 6
project similar to Kevin Lynch and Kazimierz Wejchert
A graphical
approaches (the latter being commonly recognised
interpretation of
by Polish architects as a reference in preliminary ur-
sounds.
ban studies).
First, it is necessary to underline the indepen-
dence of analytical process from any further purpose
or design. The idea was to achieve "clear" results,
liberated from any pragmatic prerequisites and as-
sumptions. What is more, students were allowed to
focus on diverse aspects of urban space according to
their mindfulness and perception abilities. As result,
a creative approach to the subject matter and experi-
menting with diverse representation methods were a
commonplace among project participants. In this re-
spect, it was observed that a variety of computer me-
dia played a prominent role in fulfilling the task. For
Figure 7
A result of a query
in Quantum GIS.
example, there were notable attempts to link visual
representations with sound effects to provide more
comprehensive information. Some students also de-
cided to use video media and animated graphic illus-
trations of the outcomes.
Through all stages of the project there was not a
single leading solution or supportive digital tool cho-
sen. Since the groups worked separately, and on a
competitive basis, it helped to keep separate tracks
of progress and, what is more, to avoid one common
method agreed. Additionally, by applying this ped-
agogical method students were enhanced to be in-
volved in the problem solving process more deeply.
Such freedom of choice of representation means
and computer programs allowed students for cre-
ative exploration, which results were not limited by
the urban-analysis software. What is more, they de-
pended chiefly on participants skills, engagement in
new domains acquisition and personal sensibility. It
is necessary to point out, that results of these ex-
perimentations were not always successful - in some
cases the outcomes represented quite difficult to un-
derstand graphical forms that did not present the
collected data in a clear and legible manner. It also
seems that the greatest difficulty for students was
the meaningful integration of data from different dis-
ciplines and of different character. What is more,
the linear nature of analysed urban space - i.e. a
street - imposed the line graph representations re-
flecting changes over time of particular features, such
as noise level or quality of space. It might be, there-
fore, interesting to repeat this project in relation to a
space of different geometry, and compare the results.
Summing up, the project stimulated students'
awareness of multifaceted quality of urban spaces.
The methodology applied by each group while elab-
orating on the task and final presentations showed
different approaches not only in terms of digital tools
but also in terms of understanding the meaning of
data integration. The author believes that the find-
ings revealed in the paper will contribute to the dis-
cussion on the FUSION - the eCAADe 2014 Confer-
ence theme - and make its goal less elusive.
Visualisation - Volume 1 - eCAADe 32 | 571
 REFERENCES ing "genius loci" of built heritage', in Morello, E and
Asanowicz, A and Asanowicz, K 2008, 'The dynamic per- Piga, BEA (eds) 2013, Envisioning Architecture. De-
ception of digitally created abstract urban space', sign, evaluation, communication, Edizioni Nuova Cul-
Architecture and Modern Information Technolo- tura, Rome, pp. 451-457
gies [online], special EAEA, pp. 1-5. Available at:
http://www.marhi.ru/AMIT/2008/spec08/papers/
Asanowicz/index.php [Accessed 22.04.2014]
Asanowicz, A 2011, 'Digital ‘serial vision’ - new approach
in urban composition teaching', in Zupancic, T, Ju-
vancic, M, Verovsek, S and Jutraz, A (eds) 2011, Re-
specting Fragile Places: Proceedings of the 29th Con-
ference on Education and Research in Computer Aided
Architectural Design in Europe, eCAADe / University
of Ljubljana, Ljubljana, pp. 716-724
Bishop, ID and Lange, E 2005, 'Communication, Percep-
tion and Visualization', in Bishop, ID and Lange, E
(eds) 2005, Visualization in Landscape and Environ-
mental Planning: Technology and Applications, Taylor
& Francis, New York, pp. 3-21
Charlton, JA, Giddings, B and Horne, M 2008 'A sur-
vey of computer software for the urban design pro-
cess', Proceedings of the 9th International Conference
on Design and Decision Support Systems in Architec-
ture and Urban Planning, Eindhoven
Chiaradia, A, Schwander, C, Gil, J, Friedrich, E and Gos-
set, A 2008 'Mapping the intangible value of urban
layout (i-VALUL): Developing a tool kit for the socio-
economic valuation of urban area, for designers and
decision makers', Proceedings of the 9th International
Conference on Design and Decision Support Systems
in Architecture and Urban Planning, Eindhoven
Ching, FDK 2007, Architecture: Form, Space, and Order,
John Wiley & Sons, New Jersey
Cullen, G 1961, The Concise Townscape, Architectural
Press, London
Czerkauer-Yamu, C and Voigt, A 2011, 'Strategic Planning
and Design with Space Syntax', in Zupancic, T, Ju-
vancic, M, Verovsek, S and Jutraz, A (eds) 2011, Re-
specting Fragile Places: Proceedings of the 29th Con-
ference on Education and Research in Computer Aided
Architectural Design in Europe, eCAADe / University
of Ljubljana, Ljubljana, pp. 125-133
Kaga, A and Sugawara, S 2008, 'Research on the Vi-
sualization for Analyzing City Changes', in Muyelle,
M (eds) 2008, Architecture 'in computro' - Integrat-
ing Methods and Techniques: Proceedings of the 26th
Conference on Education and Research in Computer
Aided Architectural Design in Europe, eCAADe / Arte-
sis University College, Antwerp, pp. 939-944
Kepczynska-Walczak, A and Walczak, BM 2013, 'Visualis-

572 | eCAADe 32 - Visualisation - Volume 1

Linkography for evaluating ideas connectivity of
Computer Aided Design-based protocols
Huda Salman
Robert Gordon University, Aberdeen, Scotland
http://www.rgu.ac.uk/dmstaff/salman-huda
h.salman@rgu.ac.uk

This paper presents the analysis of four design protocols, using linkography, to
characterise participants' design interaction and activity while designing with
CAAD programs. The participants were final year students, who have spent at
least four years in a school of architecture and were confident CAAD users. The
conceptual structure and linkability of ideas was assessed using the Linkography
method. The findings demonstrated that, for the same design problem, the
conceptual structure and linkability of ideas can be assessed using the
Linkography method. Furthermore, what characterised the continuity of CAAD
design protocols is the number of ideas in association with design moves along
the timeline of the design process.

Keywords: Linkography, Conceptual phase, Design process, Ideas

connectedness, CAAD

BACKGROUND micro level of design activity interaction in the cog-

Design cognition studies claim that Computer Aided
Architectural Design (CAAD) is inappropriate for con-
ceptual design and using it would restrain concep-
tual design. These claims were based on interact-
ing with either sketching only or switching between
sketching and CAAD; these studies employed design
protocols analysis. However, most of these studies
did not examine CAAD actively in the same way as
it recorded sketching.
In comparison to other CAAD oriented studies
within the studio context, one of the emphasised im-
pacts of CAAD on the design process of the macro
level of participants' interaction is efficiency and pro-
ductivity. This suggests that what is actually happen-
ing during a design activity on the macro level of in-
teraction is different from what is experienced on the
nitive studies, which is mainly considered "block of
thought" (Tang 2001) and time consuming (Bilda and
Demirkan 2003).
Design protocol provides a focused investigation
on the design activity (van Someren, Barnard and
Sandberg 1994) that may occur between a student
and the CAAD, therefore, a pragmatic investigation
(Creswell 2009) into the influence of CAAD on design
activity and thinking aspects is still lacking. This study
presents the linkographic analysis of four computer
aided design protocols (for a comprehensive review
on the design protocol study one may refer to the
main research project by Salman, Laing, and Conniff
2014). This study aims to provide a structured anal-
ysis of ideas development through ideas connect-
edness along the design process. The following re-

Visualisation - Volume 1 - eCAADe 32 | 573

 search questions were inquired:
• Does using CAAD appear to affect the design
process productivity and continuity in terms
of ideas association and links?
• Is it a linked process or fragmented, and if so
why?
• What association can the ideas that were ex-
pressed during the protocol have?
• What are the qualitative and quantitative dif-
ferences between the selected protocols?
RESEARCH METHODOLOGY
Design protocol encoding can be approached from a
different analytical perspective to code the concep-
tual dependencies between design ideas. From the
work of Mazijoglou, Scrivener and Clark (1996) on
structuring drawing development, alongside the se-
quential relationship between a given - former draw-
ing and the later drawing, there is a reference to the
development of this drawing and its refinement. At
the same time, the drawing transition network rep-
resents the flow of movements between drawings
during the design sessions (Mazijoglou, Scrivener
and Clark 1996). Again, this flow of conceptual-
perceptual movements could be traced within the
session's linkograph. Linkography is a graphical
structure that retains the conceptual- perceptual re-
lations between "design moves" (Goldschmidt 1995,
1997, 2003) or "design ideas" (van der Lugt 2000;
Bilda 2006; Bilda and Gero 2008). Goldschmidt de-
fines a design "move" as: "an act of reasoning that
presents a coherent proposition pertaining to an en-
tity that is being designed" (Goldschmidt 1994), or "a
step, an act, an operation, which transforms the de-
sign situation relative to the state in which it was prior
to that move" (Goldschmidt 1997). The developmen-
tal aspect of structuring the drawing into a network
is to some extent related to Goel's notion of transfor-
mative conception of refinement and development,
which was utilised recently by Cai, Do and Zimring
574 | eCAADe 32 - Visualisation - Volume 1
(2010) to investigate the impact of precedents on de-
sign and on the micro level of the design process.
ANALYSIS
In the current study, a design move is determined
by any action that has changed the status of the de-
sign into a new status that can be traced externally.
Those actions could be decisions, goals, constraints
that were set and met successfully. The protocol seg-
ments were the main unit for coding design moves.
But the segments included more information that a
design move would entitle. Assigning design moves
was based on Goldschmidt's (1995) definition of a de-
sign move. Design moves were assigned by consid-
ering the visual and verbal context of every segment,
emphasising the context of its being and its potential
connectedness along the design process. Most of the
segments though revealed one to two design moves
at the most, Figure 1. A design move then became
equal to an idea that had been externalised verbally
or graphically during the process.
Figure 1
An example of
coding design
moves and links
(taken from S4
linkograph).
Although Goldschmidt stressed common sense as
the main approach to determining the link between
two or more design moves, she mentions that this, in
practice, means that: "...a link between two moves is
established when the two moves pertain to the same,
or closely related, subject matter(s)...". Other linko-
graphic studies have shown that utilising common
sense seems to be a tentative technique to elicit the
relations of latter/former moves. Other researchers
like van der Lugt (2000) have strengthened Gold-
schmidt's notion by defining indicators for judging
the commonality between design moves. He sug-
gested that the common sense reliability could be
 enhanced by developing indicators for links; in ad-
dition to the similarities in content (Bilda 2006) and
argued that "a link is present, even though the sub-
ject matter between these two ideas is not closely re-
lated". Also by associating links, the potential contri-
bution of every move along the design process (line)
is decided against all the other moves that have been
Figure 2
assigned as design moves.
An example of
However, there are two difficulties in associat-
coding the forelinks
ing the links along the design timeline. The first
and backlins of a
one is how to track the content of each move (each
design move.
segment could have more than one sentence) and
the other difficulty relies in the association between
links that occur further down the line of the design.
To overcome these difficulties Bilda (2006) has sug-
gested a means to aid the analyst's common sense,
by employing a word search to find any commonal-
ity in meaning of such distinct design moves. This in-
cludes more than one run through the questionable
segments and links.
The coded links are categorised into fore-links
and back-links. Each has a different meaning with
regard to the design process. Back-links are links
that are related and considered as a concluding move
"that summarises or evaluates points raised in the ex-
ploration" (Goldschmidt 1995), and Fore-links "record
the path that led to a move's generation, while fore-
links bear evidence to its contribution to the produc-
tion of further moves" (Goldschmidt 1995). The total
number of back-links and fore-links are equal in ev-
ery linkograph but different on the level of a design
move, Figure 2.
Researchers such as Goldschmidt (2003), Bilda
and Gero (2008) and Van der Lugt (2000) devised a
number of measures for evaluating (conceptual) pro-
ductivity and design flow. This study also uses the
same measures of critical moves, link index and clus-
ters. Goldschmidt (2003) defines "critical move" (CM)
"identifies design concepts that are deemed "suc-
cessful" in the sense that the designer values them
enough to devote time trying to develop the con-
cepts or at least to promote them at various points
in the protocol". CM percentage refers to the total
number of CM in a design session in relation to total
number of moves. Link Index (LI) is a measure of how
connected the design ideas are in a design session.
In order to calculate the link index in the overall ses-
sion, the total number of links is divided by the total
number of moves in the design session.
DESCRIPTIVE RESULTS
Coding the design session in terms of design move(s)
and moves association, has revealed the structure of
the conceptual/visual association. Every participant
has revealed a different visual pattern that is related
directly to the design protocol data and the develop-
ment of the student's ideas. As shown in Figures 3, 4,
5, and 6.
Examining the resulting linkographs showed
many differences in the way the student's ideas were
connected during design. These differences are pre-
sented on two levels of analyses: the first level was
carried out using the quantitative method, by which
linkograph is quantified by a number of measures,
that is number of moves, link index (LI), critical moves
(CM), percentages and link clusters. The other level
was the conceptual content of the process from a
methodological point of view, where concept or a
partial idea of a concept is forming along the de-
sign process timeline. With respect to content anal-
ysis, linkograph is seen as a structure of associations
among the design moves, that is coding the relation-
ship between any two dependant design moves.
Visualisation - Volume 1 - eCAADe 32 | 575
 Figure 3
S1 design protocol
Linkograph.

Figure 4
S2 design protocol
Linkograph.

Figure 5
S3 design protocol
Linkograph.

Figure 6
S4 design protocol
Linkograph.

576 | eCAADe 32 - Visualisation - Volume 1

 QUANTITATIVE RESULTS program protocol), while S3 protocol had the small-
The design protocols are described quantitatively est LI value. This means that S2 multi CAAD protocol
through linkographic measures of link indices (aver- is the most productive. This may suggest that switch-
ages), critical moves (CM) and percentages. An av- ing between two programs may be better in terms of
erage of fifty-three moves was decided by the stu- the content and the structure of a CAAD design pro-
dents that created an average of 160 links among cess as he also solved the problem in 2D then 3D.
themselves in an average duration of 59 minutes. As Design reasoning studies (e.g. Goldschmidt and
shown in Table 1, the Link index for S1, S2, S3 and S4 Weil 1998) proved that structure and content corre-
are 3.1, 3.2, 2.8 and 3.1, respectively. late to one another. Accordingly, a productive move
in a Linkograph spreads in two directions, there-
Table 1 fore, identifying the critical moves (CMs) in terms of
Linkographic its conceptual connectivity describes design process
description for four advancement.
students. With respect to critical moves (CM) and percent-
ages, CM is a move which has at least 3 links in one
direction to 8 links. The threshold varies upon the
research purpose and cases (Goldschmidt 2009 per-
sonal communication by email), therefore this was
decided through the percentages of (CM5, CM6, CM7
S2 (multi CAAD programs) had the highest LI and CM8), as shown in Figure 7.
among other protocols. LI value was the same for S1 The percentages of critical moves of five links, (ei-
(single CAAD program protocol) and S4 (single CAAD ther forward or backward CM5), was the highest per-
centage which ranged from 10% to 19%. Likewise,
Figure 7
The critical move by
link level and their
percentages of the
total moves and the
average percentage
for all cases.

Visualisation - Volume 1 - eCAADe 32 | 577

 CM5 percentage was the highest among other CMi
for each student.
LINK DENSITY IN CLUSTERS
To examine the conceptual continuity in terms of the
association of design ideas to reflect on the study
question: What association can the ideas that were
expressed during the protocol have? Cluster density
was computed. A cluster in the design process (linko-
graph) means a successive number of design moves,
more than three (Bilda and Gero 2008) that are linked
and have a serial conceptual dependency. This re-
flects on the "suggestive" design moves of a design
process, which is followed by a series of moves that
explore issue (s) raised by the suggestive move (Gold-
schmidt 1995, 2003; Goldschmidt and Weill 1998;
Dorst and Dijkhuis 1995) or the other related moves
within that cluster (Bilda and Gero 2008). In other
words, this means that the clusters refer to the evolu-
tion of a partial idea towards a tentative design. The
number of design moves that form a cluster, define
a cluster size. For example, if five consecutive links
were in sequence then the cluster size is six, because
six design moves formed five links cluster. These links
spread to relate to other design moves along the pro-
cess in both directions: forward and backwards. To
know what type the cluster is, the links are counted
in both directions to define whether the cluster is a
forward or a backward one. According to Bilda and
Gero (2008) this helps to know the density of a clus-
ter by dividing the total number of links of a cluster
on the size of a cluster.
578 | eCAADe 32 - Visualisation - Volume 1
The results of the total number of forward and back-
ward link clusters, size of clusters, and link density
in the clusters for all participants are shown in Table
2. Six clusters were identified in three of the linko-
graphs of S1, S2 and S3, and seven clusters in S4 linko-
graph. The number of backward clusters in all the
linkographs was higher than the forward clusters.
This suggests that CAAD affected the students
in two ways: (1) made the students revise, evalu-
ate and repeat their ideas more often than gener-
ate or propose, and (2) to an extent made the stu-
dents less inventive in terms of the number of new
ideas, however, the nature of the task and the time
required to finish the experiment should be consid-
ered. These points agree with Goldschmidt's defini-
tion of the backward and the forward critical moves
(1995). Another point to discuss is the observed po-
sitions of the two types of cluster. In each of the
linkographs the first identified cluster was a forward
cluster and the last two or three clusters were back-
ward clusters. This suggests that most of the partic-
ipants from the start of designing started to exter-
nalise their ideas, whether through CAAD or words
to reach a tentative solution to the design problem.
Also, it suggests that the initial ideas are deemed to
be the most successful moves that a designer makes
which enable the designers to further develop their
earlier ideas.
In between the first and the end clusters there
was no common pattern for all participants, but
rather interplay between the two types. Two of
the participants (S2 and S4) showed continuous be-
haviour in constructing forward ideas, the other two,
S1 and S3, inverted their ideas backwards to what Table 2
they proposed in the first cluster to improve and Clusters and
check for two successive clusters ( no 2, 3). Then en- density of links.
gaged in forward thinking (cluster 4), S4 showed a
different pattern of continuous forward thinking and
then spent the later 4 clusters in reverse thinking.
This is shown in Figure 8.
Figure 8
the types of the
clusters in the same
sequence of
occurrence.
Coding the clusters with colours showed the
crossover of clusters in the linkograph, mainly in the
shape of triangles. This suggests that every cluster
spreads in both directions to form a joint smaller
pyramid and as a result, all clusters were intercon-
nected. When the student is satisfied with the result
he mostly inverts his thought backward to relate to
earlier thoughts.
The average size of clusters was higher for the
forward clusters compared to the backward clusters.
It was also higher for all participants in the forward
clusters except S3. It also shows that the fore-link
density is relatively higher than back-link density; as
it was higher for all participants except for S4. When
the fore-link density is higher than the back-link den-
sity it means that the forward clusters initiated more
ideas later on in the process (Bilda and Gero 2008).
High density reflects rich idea development and that
these ideas were potentially successful.
DISCUSSION
The number of coded moves was relatively lower
compared to the initial number of segments. The dif-
ference implied that the segment includes a larger
amount of information than a design move. This is
due to the fact that when a segment is coded under
the design micro strategy, the intention is one, for ex-
ample proposing a solution, or evaluating a solution,
where as this intention could include more than one
idea or more than one design move (act that changes
the design situation). For example, proposing a de-
sign solution could include one or two different ideas
or developed ideas, especially when the segment
was externalised verbally. Normally the student be-
comes more concise when externalising graphically,
thus, the number of segments in the design process-
coding scheme was higher than the number of de-
sign moves in the linkography-coding scheme.
Another observed feature is the qualitative
meaning and extensiveness of the words used in the
verbal subpart of the protocol. On the general level
of design protocol and on the specific level of ideas
connectedness, two aspects were observed. On the
general level of protocols' transcriptions, words such
as 'maybe', 'probably' , 'possibly' and so on were
used which reflect the ill defined nature of the de-
sign situation. On the specific level of ideas connect-
edness, another aspect of design moves' connect-
edness was revealed through tracing the common
words between segments to see whether two distant
moves were connected or not. This process revealed
an interesting character within three of the analysed
protocols (S1, S2 and S4). That is, some of the words
used had become more extensive; the words (which
have conceptual traces) that were verbalised during
designing were changed into more inclusive and spe-
cific meaning along the timeline. This change is due
to the change in the design status and design con-
ceptual maturity, due to the evolving context of the
design process. This was also traceable on the level of
density clusters which appeared to confirm the con-
ceptual progress, but mainly in forward moves and
forward clusters.
CAAD IMPACT ON IDEAS CONNECTIVITY
Furthermore, Linkography analysis suggests that
CAAD affected students' conceptual content in two
ways: (1) participating students revised and repeated
Visualisation - Volume 1 - eCAADe 32 | 579
 their ideas more than proposed new or different
ideas, and (2) for an extent made the students less
inventive in terms of number of options, however,
the nature of the task and the time required to finish
the experiment should be considered. These points
agree with Goldschmidt's definition of the backward
and the forward critical moves (1995), and Bilda and
Gero's (2008) definition of links density.
This suggests that from the start of designing,
most of the participants started to externalise their
ideas, whether through CAAD or words to reach a
tentative solution to the design problem. Also, it sug-
gests that the initial ideas are deemed to be the most
successful, which are often revisited and developed.
This is similar to designing under sketching condi-
tions.
CONCLUSION
This study was an attempt to characterise partici-
pants' design thinking while designing with CAAD
programs. The analysis represents four design pro-
tocols of final year students at a school of architec-
ture. The protocols analysed varied in more than one
aspect. This variation includes: (1) CAAD programs,
(2) the mode of using CAAD programs whether sin-
gle or multiple, (3) protocol segments (total number,
duration and frequencies), and (4) design categories
and total time spent in designing. This gave an up-
dated insight into design thinking (process) while us-
ing CAAD programs in architectural design with fi-
nal year participants. In light of the study results,
the participants demonstrated that, for the same de-
sign problem (brief ), restraining the conceptual de-
sign medium would not necessarily bind the partici-
pants to a certain design strategy or hamper the ar-
rival at a tentative design regardless of its academic
assessment. Instead, the conceptual structure and
linkability of ideas was assessed using the Linkogra-
phy method. Furthermore, what characterised the
continuity of CAAD design protocols is the number
of ideas in association with design moves along the
timeline of the design process. The protocols were
characterised by high linkographic measurements:
580 | eCAADe 32 - Visualisation - Volume 1
Li, CM and Ld. However, the participant who used
more than two software programs showed higher Li
from the other linkographs. This characteristic along-
side other results suggests that using more than one
CAAD program may increase the connectedness of
conceptual ideas along the design time line, and a
higher number of forward clusters. Participants com-
ments were positive about the design situation and
the experiment. However, the ones who used CAAD
early in design were regularly more positive about us-
ing CAAD in concept design. On the basis of these
results, interacting with CAAD visual representations
can be accepted as a conceptual medium and the use
of CAAD alone for externalisation.
REFERENCES
Bilda, Z 2006, The role of mental imagery in conceptual de-
signing., Ph.D. Thesis, Key Centre of Design Comput-
ing and Cognition Faculty of Architecture, University
of Sydney, Australia.
Bilda, Z and Gero, JS 2008, 'Idea development can occur
using imagery only during early conceptual design-
ing', in Gero, JS and Goel, AK (eds) 2008, Design Com-
puting and Cognition, Springer
Cai, H., Do, E and Zimring, C. M. 2010, 'Extended linkogra-
phy and distance graph in design evaluation: an em-
pirical study of the dual effects of inspiration sources
in creative design.', Design Studies, 31, pp. 146-168
Dorst, K and Dijkhuis, J 1995, 'Comparing paradigms for
describing design activity', Design Studies, 18, pp.
441-455.
Goldschmidt, G . 1997, 'Capturing indeterminism: repre-
sentation in the design problem space', Design Stud-
ies, 18, pp. 441-455
Goldschmidt, G 2003 'Expert Knowledge or Creative
Spark? Predicaments in Design Education', Design
Thinking Research Symposium 6
Goldschmidt, G 1995, 'Visual displays for design: im-
agery, analogy and databases of visual images', in
Koutamanis, A, Timmermans, H and Vermeulen, I
(eds) 1995, Visual databases in architecture, Alder-
shot, Avebury., pp. 53-74
Goldschmidt, G 1994, 'Visual Analogy in Design', in
Trappl, R (eds) 1994, Cybernetics and Systems `94,,
World Scientific
Goldschmidt, G and Weil, M 1998, 'Contents and struc-
ture in design reasoning', Design Issues, 14, p.
85–100
Van der Lugt, R 2000, 'Developing a graphic tool for cre-
ative problem solving in design groups', Design Stud-
ies, 21, pp. 505- 522
Maziloglou, M, Scrivener, S and Clark, S 1996, 'Represent-
ing Design Workspace Activity', in Cross, N, Chris-
tians, H and Dorst, K (eds) 1996, Analysing Design Ac-
tivity,, John Wiley and Sons, UK, pp. 389-416
Salman, HS, Laing, R and Conniff, A 2014, 'The impact
of computer aided architectural design programs on
conceptual design in an educational context', Design
Studies, 35(4), pp. 412-439
Van Someren, M W, Barnard, YF and Sandberth, JAC
1994, The think aloud method: a practical guide to
modelling cognitive processes, Academic Press, Lon-
don
Tang, H 2001, Exploring The Roles of Sketches and Knowl-
edge in the Design Process, Ph.D. Thesis, Department
of Architectural and Design Science. Faculty of Ar-
chitecture. The University of Sydney.

Visualisation - Volume 1 - eCAADe 32 | 581

582 | eCAADe 32 - Visualisation - Volume 1
Digital Heritage 1
Interactive Tabletops for Architectural Visualization
Combining Stereoscopy and Touch Interfaces for Cultural Heritage

Bruno Figueiredo1 , Eduardo Castro e Costa2 , Bruno Araújo3 , Fernando Fonseca4 ,

Daniel Mendes5 , Joaquim A Jorge6 , José Pinto Duarte7

1
Escola de Arquitectura, Universidade do Minho
2,6
CIAUD, Faculdade de Arquitetura, Universidade de Lisboa
3,4,5,6
INESC-ID, Instituto Superior Técnico, Universidade de Lisboa
7
http://home.fa.utl.pt/˜jduarte/
1
bfigueiredo@arquitectura.uminho.pt
2,7
{castroecosta|jduarte}@fa.ulisboa.pt
3
brar@vimmi.inesc-id.pt
4,5
{fernando.fonseca|danielmendes}@ist.utl.pt
6
jorgej@tecnico.ulisboa.pt

This paper presents an interactive apparatus to didactically explore Alberti's

treatise on architecture, De re aedificatoria, as generative design systems, namely
shape grammars. This apparatus allows users to interactively explore such
architectonical knowledge in both appealing and informal ways, by enabling
them to visualize and manipulate in real-time different design solutions. The
authors identify the difficulties on encoding the architectural knowledge of a
parametric design model into an interactive apparatus to be used by laypeople.
At last, the authors discuss the results of a survey conducted to users that
interacted with the prototype in order to assess its ability to communicate the
knowledge of an architectural language.

Keywords: Alberti, generative design, multi-modal interfaces, shape grammars,

user experience

INTRODUCTION architecture, making use of a computational frame-

This paper presents an interactive apparatus to di-
dactically explore Alberti's treatise on architecture,
De re aedificatoria (Alberti, 2011), as a generative de-
sign system. This work belongs to a wider project,
Alberti Digital, which aims to determine the influ-
ence of De re aedificatoria on Portuguese Renaissance
work.
We chose to hermeneutically explore the treatise
descriptions on sacred buildings and columnatio sys-
tems by inferring the underlying structure provided
by description grammars and shape grammars (Stiny
and Gips, 1972).

Digital Heritage 1 - Volume 1 - eCAADe 32 | 585

 A synthesis of the project outcomes was shown
as an exhibition, firstly presented in April 2013, at
the Science Museum, University of Coimbra (Figure
1). The exhibition focuses not only on the results
achieved, but also on the tools used, and on their de-
velopment.
It also features a semi-immersive environment
to pedagogically present the knowledge inferred in
defining the shape grammars. This interactive appa-
ratus allows us to explore the architectonical knowl-
edge described on the De re aedificatoria in appeal-
ing and informal ways, by enabling the visitors to vi-
sualize and manipulate in real-time different design
solutions generated interactively.
Simultaneously, the semi-immersive table facili-
tated the comprehension of the combinatory, algo-
rithmic and computational nature of Albertian ar-
chitectural theory, by highlighting explicit relation-
ships between the generative model outcomes, as
well as the morphology and proportional variations
prescribed in the treatise.
METHODOLOGY
The De re aedificatoria descriptions offer enough
knowledge to define an architectural language and
a grammar, expressed in algorithmic terms. Accord-
ingly, the initial stage of our research included four
steps (Figure 2). Since the treatise only contains tex-
tual descriptions, our initial task was to read and
select parts of the text describing the architectonic
components from the columnatio systems and the
586 | eCAADe 32 - Digital Heritage 1 - Volume 1
temples. The second task was to gather and group
the information into parametric schemas that inter-
pret those descriptions. This information can be en-
coded into a grammar that, through the application
of its rules, generates a corpus of solutions that fol-
low the treatise's architectonical rules. Finally, trans-
lating the grammars into a parametric computational
model allows us to visualize the resulting structures.
We developed an interactive setup to explore de-
sign space prerogatives of the treatise in a fashion ac-
cessible to laypeople. We devised a tabletop proto-
type to explore 3D virtual models of existing build-
ings in stereoscopy, as if they were physical models
lying above the tabletop (Figure 5). We also intro-
duced the possibility of customizing an entire model
of a temple according to the rules in our Alberti shape
grammar.
DESIGN DESCRIPTIONS AND DESIGN COM-
PUTATION MODELS: DIDACTIC TOOLS
TO CLASSICAL ARCHITECTURAL KNOWL-
EDGE Figure 1
Establishing design principles can be a complex pro- Alberti Digital
cess, it involves (self ) arguing and defining rules of exhibition, Science
design, using different areas of knowledge (geom- Museum, University
etry, physics, economics and social sciences, etc.). of Coimbra, April
The tradition of classical architecture is an example of 2013 (photo by Do
this feature where standard systems of naming and Mal O Menos)
segmentation contributed to establish a conceptual
framework and its architectural principles. Accord-
ing to Mitchell (1990, p.24), this requires a vocabu-
lary large enough to differentiate parts, symbolic and
functional relationships that enable specifying prop-
erties and interrelations of different parts in the de-
sign. This vocabulary can either be expressed by as-
signing values to variables of a data structure, or by
a set of logical principles and their notations, or even
through a textual description written in natural lan-
guage
Simultaneously, computation can be under-
stood as a series of rational, logic, algorithmic, de-
ductive and inductive processes, whose structure
and systematization explores the possibility of ob-
Figure 2
Diagram
representing the
steps taken to
understand and
decode the treatise

Figure 3
Computation tree
of the temples
shape grammar: to
each step of the
computation, by
the application of a
rule, are visible the
possible
transformations for
the specific left side
shape of the rule.
taining solutions to a given problem. Although GENERATIVE SHAPE GRAMMAR'S IN AC-
(Figueiredo et al,
eminently abstract, these structures allow to relate CORDANCE TO ALBERTI'S THEORY
2013)
knowledge obtained through cognition, simulation Since the treatise has no graphical information, our
and intelligence-based rules, among others. first task was to read it and to systematize knowledge
This research took into consideration the idea that led to the understanding of the column systems
that design descriptions and computation are both and temples architectonical principles. Graphical no-
transversal tools to engage an architectural design tations were drawn from the translation of our inter-
and subsequent language of design. In order to com- pretation of the text. These notations led to the def-
municate this principles to a wider public, most of inition of a set of parameters and conditions which
the times codified in representations that only ar- allowed for the development of a system of schemas,
chitects understand, we have proposed an exhibi- useful to set algorithms that inform both the shape
tion setup that helped the visitor to grasp the ar- and proportions from each part and the relations be-
chitectural principles inscribed in De re aedificatoria. tween the parts of the buildings and their interre-
The layout of the interactive stereoscopic visualiza- lated proportional features.
tion apparatus was structured to explicit the compu- Departing from this knowledge, a set of shape
tational and combinatory relations synthesized in the rules were defined to compose a columnatio systems
shape grammars and their generative outcomes. generative shape grammar and a rectangular tem-
ples generative shape grammar. In both grammars,
by applying their shape rules in a step by step compu-

Digital Heritage 1 - Volume 1 - eCAADe 32 | 587

 tation methodology, are generated design solutions
within the same language. Figure 3 illustrates this
process, a computational sequence, where different
options are shown after the use of each rule of trans-
formation, but only one is subsequently transformed
by the use of the next rule, finally resulting in a design
solution for a rectangular temple.
Based on the developed shape grammars, and
in order to test them, parametric models were de-
veloped in a visual programming interface, imple-
mented into computer programs that automatically
generate digital models in a CAD application, accord-
ing to input parameters (Figure 4). This framework al-
lows for real-time visualization of the results of vary-
ing parameters. These parametric models are highly
detailed and compliant with Alberti's prescriptions
encoded in the shape grammars (Figueiredo et al,
2013), and they served as a direct basis for the devel-
opment of the virtual reality tabletop application.
shown on the table. While a user moves around the
table, the perspective view of the building changes
according to his or her movement. A 3D television set
(D) combined with a 3D active shutter glasses (B) en-
ables stereoscopic visualization, sending 1920x1080
pixel resolution images to each eye 60 times per sec-
ond. This setup allows rendering high definition vir-
tual objects as if they were lying above the table sur-
face. The touch-enabled surface, using a multi-touch
frame (C) capable of detecting up to 10 touches, al-
lows interacting with virtual models through ges-
tures.
Figure 4
Sample of design
solutions generated
by the temples
parametric model.

INTERACTIVE STEREOSCOPIC VISUALIZA-

TION OF ALBERTI ARCHITECTURAL MOD-
ELS: A DIDACTIC TOOL TO COMPREHEND A Figure 5
Tabletop prototype
GENERATIVE DESIGN SYSTEM
hardware setup
The technological advances that have been wit-
nessed in the last few years allowed the development
of new and more interactive applications for all kinds
of scenarios. Multi-touch devices and depth sensors
like Microsoft´s Kinect are clear examples of these ad-
vances, allowing both non-intrusive and inexpensive
user tracking. As for visualization, common 3D dis-
plays allow users to perceive imagery as if it popped
outside the screen. This motivates a fresh look at
tabletop interfaces, towards better support of 3D di-
With this semi-immersive environment, users can
rect manipulations in scenarios until recently con-
explore 3D virtual models of buildings using both
ceivable in the realm of science fiction.
hands. To manipulate the content on the stereo-
Exploiting the aforementioned technological so-
scopic surface the visitor can use several touches to
lutions, we developed a tabletop prototype (Figure
interact with each model at a time. We developed a
5) that consists on a semi-immersive environment
finger-cluster interaction method, which allows users
based on a stereoscopic multi-touch surface com-
to move, rotate and uniformly scale the models. By
bined with a Microsoft Kinect depth camera. This
dragging the fingers anywhere along the surface, the
camera (E) tracks the user's head, enabling a real-time
model will move in the direction of their movement.
personalized 3D perspective view of the contents

588 | eCAADe 32 - Digital Heritage 1 - Volume 1

Figure 6
Tabletop prototype
operating diagram
To rotate the model along the axis perpendicular to
the surface, people need to apply a rotational move-
ment via at least two fingers, but they can use their
entire hand if they want. By changing the relative
position of all fingers a user can uniformly scale the
object. If the distance between the fingers and their
center changes, the model scale will increase or de-
crease accordingly. This technique uses the well-
known algorithm Rotate-N-Translate (Hancock et al.,
2006), available in almost every modern multi-touch
device (Figure 7).
Figure 7
Possible
interactions
through touch:
translation,
rotation, and
scaling
FROM PARAMETRIC MODEL TO THE
TABLETOP INTERFACE
By means of this environment, users can explore 3D
virtual models of Alberti buildings. In addition, we
included a special model, which is a temple that sup-
ports different types of customizations, accordingly
to the shape grammar extracted from the De re aedi-
ficatoria treatise
Similarly to what the generative shape grammars
support, the interactive prototype simulates para-
metric variations on generated solutions by consid-
ering certain proportional, morphological and stylis-
tic attributes in detriment of others, allowing the user
to choose, under certain restrictions, the shape rules
to be applied.
The tabletop prototype, allows people to select
different shape rules using a button-based graphical
user interface (GUI), which controls certain features
of the temple model (Figure 8). Each feature com-
bines specific parameter(s) and condition(s), which
can be changed within a pre-defined set of values,
represented by mutually exclusive toggle buttons.
For each feature, only a button can be activated at a
time, and some buttons might be disabled depend-
ing on selected features. For example, the number
of pairs of lateral chapels that compose the temple
is a feature, whose values can vary between 0, 1, 3
or 5. However, if the cell proportion (another fea-
ture) becomes smaller than 3:1, measuring the rela-
tion between the length and the width, the button
with value five will be disabled (Figure 8, D) since this
proportion does not accommodate five pairs of lat-
eral chapels along the longitudinal inner facades.
The set of customizable attributes, or features
(Table 1), is divided into three modes, related to three
aspects of the temple (Figure 8, A), namely the com-
position of the columnatio system (Figure 8, B), the ar-
rangement of the portico located in the main facade
Digital Heritage 1 - Volume 1 - eCAADe 32 | 589

(Figure 8, C), and the temple itself morphology and
proportions (Figure 8, D).
The main architectural knowledge on temples
that Alberti encoded in his treatise translates to the
three modes of interaction. Although discrete, they
establish interrelations between each other, thus
framing an overall architectural language. Different
combinations from the columnatio system parts will
affect the portico definition as well as the dimen-
sion of the cell, tribune and chapels. Simultaneously,
changes to the cell proportions will constrain the
width, height and depth of the portico
The layout of the interface was designed to pro-
vide people (museum visitors) with a clear under-
standing that, by manipulating one architectural fea-
590 | eCAADe 32 - Digital Heritage 1 - Volume 1
Figure 8
Temple
customization:
menus (Fonseca,
2013)
ture, other characteristics will be affected, as well
the overall expression of the building, thus affording
a better, clearer exploration of the Albertian design
space by laypeople.
This network of proportional and morphological
dependency represents to a certain extent the idea
of concinnitas, the basic principle of the Albertian
thought, described by Rudolf Wittkower (1973, p.33)
as embodying the classical idea of maintaining a uni-
form system of proportion throughout all parts of a
building.
FEEDBACK FROM USERS
The objective of developing such a prototype was
to present an architectural language in a way that
Table 1
Temple
customization:
features
would engage the exhibition visitors, providing an
experience different from the traditional exhibition
media. In order to assess the success of this initia-
tive, a survey was conducted to users that interacted
with the prototype, in which they could express their
opinions about that experience. The survey was con-
ducted to 27 people, with ages over 19. More than
half of the users were between 19 and 24 years old,
and two thirds hold a Bachelor degree or higher.
From such survey, which consisted of closed
Likert-scaled questions, we were able to gather infor-
mation to conclude the following:
• The prototype aroused the immediate curios-
ity of 62 % of visitors, leading them to experi-
ment with it right away.
• 54% of visitors felt that the application in-
terface was easy to use, and only one visitor
could not understand how it worked.
• All visitors agree that the prototype undoubt-
edly enriches the experience of the museum
exhibition, and also agree that this type of ex-
perience can be interesting when applied to
other exhibitions.
• The interactive table provides an engaging
and innovative experience for visitors, thus
enriching the exhibition. The visitors agree
that such an experience is more interesting
than the models and posters also presented
at the exhibition.
• The application enables the visitors to visual-
ize the buildings in 3D through stereoscopy,
and they agree that it enriches the experience.
Only 18 % of visitors did not realize that while
moving around the table the perspective view
of the building changed according to their
movement.
We can say that the interactive system was successful
as a didactic tool, able to convey a specific architec-
tural heritage - architectural knowledge described in
the De re aedificatoria - by allowing users to explore
the system generative outcomes within their vocab-
ulary and syntax content.
CONCLUSIONS
In implementing Alberti's models into the tabletop
prototype, two main difficulties arose. The first dif-
ficulty was technological: the application running on
the tabletop prototype doesn't match the genera-
tive capabilities of the shape grammar formalism, nor
of the corresponding parametric model. The solu-
tion space generated by combining all the parame-
ters in the original models proved too complex to be
implemented into the prototype's application. This
fact forced the selection of a number of parameters
smaller than the ones available in the original shape
grammar and parametric model.
Nevertheless, supposing that all the parameters
could be implemented, the control of such an exten-
sive solution space would require an equally exten-
Digital Heritage 1 - Volume 1 - eCAADe 32 | 591
 sive - and complex - user interface, forcing the use
of even more modes, with more sets of features, and
probably rendering the experience very little engag-
ing to the user.
Therefore, the second difficulty was a cognitive
one. This emphasizes the need to focus research
not only on pushing the technological boundaries
of generative design systems, towards the ability to
model and visualize more accurately and with more
detail, but also on synthesizing the complexity of
the Albertian thought in order to communicate their
essence to a non-specialist audience.
Current research is pursuing a system that can
provide an even deeper engagement of the user. On
one hand, tools are being developed that overcome
the technological constraints that conditioned the in-
terface for the presented prototype. The objective is
to extend the manipulation of geometry beyond the
combination of objects, towards parametric and rule-
based real-time modelling. On the other hand, such
modelling is intended to be as user-friendly as pos-
sible, taking into account both the cognitive and er-
gonomic characteristics of the average user. By com-
bining techniques such as virtual reality, motion cap-
ture and shape grammars, we aim towards a new
generation of interfaces that allow for a smoother in-
teraction between designers, namely architects, and
their creations.
ACKNOWLEDGEMENTS
This work was partially supported by the Por-
tuguese Foundation for Science and Technol-
ogy (FCT): doctoral grants SFRH/BD/69910/2010,
SFRH/BD/88040/2012 and SFRH/BD/91372/2012,
and projects MIVis PTDC/EIA-EIA/104031/2008,
Alberti Digital PTDC/AUR-AQI/108274/2008,
TECTON-3D PTDC/EEI-SII/3154/2012 and Pest
OE/EEI/LA0021/2013.
REFERENCES
Alberti, LB 2011, Da Arte Edificatória, Fundação Calouste
Gulbenkian, Lisboa
592 | eCAADe 32 - Digital Heritage 1 - Volume 1
Araújo, B, Jorge, JA and Duarte, JP 2012 'Combining Vir-
tual Environments and Direct Manipulation for Ar-
chitectural Modeling', Proceedings of eCAADe 2012,
Prague, pp. 419-428
Duarte, JP, Krüger, M, Coutinho, F, Figueiredo, B and
Castro e Costa, E 2013, 'Alberti Digital: Investi-
gando a influência de Alberti na arquitectura por-
tuguesa da contrarreforma', in Brandão, CAL, Caye,
P, Furlan, F and Loureiro, MA (eds) 2013, Na Gênese
das Racionalidades Modernas: Em torno de Leon Bat-
tista Alberti, Editora UFMG, Belo Horizonte, pp. 488-
548
Figueiredo, B, Castro e Costa, E, Pinto Duarte, J and
Krüger, M 2013 'Digital Temples: A Shape Grammar
to Generate Sacred Building according to Alberti’s
Theory', Proceedings of Future Traditions, Rethinking
Traditions and Envisioning the Future in Architecture
Through the Use of Digital Technologies, Porto, pp.
63-70
Figueiredo, B, Duarte, JP and Krüger, M 2013 'Albertian
Grammar Transformations: From the treatise to the
built work in the design of sacred buildings', Proceed-
ings of eCAADe 2013, Delft, pp. 687-696
Fonseca, F 2013, Manipulações no espaço sobre mesas in-
terativas com estereoscopia, Master's Thesis, Instituto
Superior Técnico: Universidade de Lisboa
Hancock, MS, Vernier, FD, Wigdor, D, Carpendale, S and
Shen, C 2006 'Rotation and Translation Mechanisms
for Tabletop Interaction', TABLETOP, Adelaide, pp.
79-88
Mitchell, W 1990, The Logic of Architecture: Design, Com-
putation and Cognition, The MIT Press, Cambridge,
Massachussets
Stiny, G 2011, 'What Rule(s) Should I Use?', Nexus Network
Journal, 13(1), pp. 15-47
Stiny, G and Gips, J 1972 'Shape Grammars and the Gen-
erative Specification of Painting and Sculpture', In-
formation Processing 71: Proceedings of IFIP Congress
71., Amsterdam, pp. 1460-1465
Wittkower, R 1972, Architectural principles in the age of
humanism, Academy, London
A Building Heritage Knowledge Framework using Context
Ontologies
Regenerations of industrial areas - the “Isola del Liri” case study

Silvia Gargaro1 , Elisabetta Leggieri2 , Antonio Fioravanti3

1,2,3
Civil, Building and Environmental Department, Sapienza University of Rome
1,2,3
{silvia.gargaro|elisabetta.leggieri|antonio.fioravanti}@uniroma1.it

The growth of urban areas and the increase of urban migrations caused an
incorporation of industrialized areas in the development of cities. The difficulties
related to city planning in these areas could be coped with more satisfactorily
through the use of Geographic Information Systems (GIS) and Context ontologies
plus reasoning rules to reuse old industrial buildings. The 'Context' entities and
rules are studied using a Context Knowledge Model formalized with Protégé OWL
(Ontology Web Language). The study was carried out on industrial buildings
situated in southern Lazio - Italy - in particular a paper mill network. Applying
GIS instruments to the study of these areas is not enough as designers can only
obtain information, but not knowledge, which depends on the urban, historical,
cultural, economic, environmental and juridical ‘context’. The goal of this
research is to create a combination of these technologies to manage knowledge
for a more aware design to regenerate old industrial areas with new uses.

Keywords: Heritage Building, Design Process, Context, GIS tools, Ontologies

GIS, BIM AND CONTEXT KNOWLEDGE FOR
HERITAGE REUSE DESIGN PROCESSES
The importance of the Context to regenerate disused
industrial buildings, is based on the investigation
about the reasons that have conditioned the land
use, the development, the integration in the context
and changes which influenced the removal.
The most common BIMs did not fill the gaps be-
tween the use of context in architectural design and
the reuse of disused industrial buildings.
The existing GISs represent continuous and dis-
crete territorial data and can be used to represent rel-
evant processes such as pollution and land use activ-
ities and therefore make statistical forecasts of com-
mon interactions possible. Their great potentiality
is not fully used to represent the processes of im-
pact on architectural design. While the interoper-
ability between GIS and BIM software tools has dra-
matically improved in recent years, this process still
remains particularly limited. This shows the limita-
tions of managing the information obtained in order
to have an added value in architectural design con-
text aware.
GIS instruments were used to analyze the sites
and the relations between the factories and urban
system based on different satellite images, past and

Digital Heritage 1 - Volume 1 - eCAADe 32 | 593

 recent cartographies in order to obtain an integrate
overview and comparison.
The sizes of existing files, such as LandXML [1] or
CityGML [2] provide great opportunities for interop-
erability between software tools. LandXML has deep
roots in the development sector of the territory, and
contains classes of valuable items for landscape ar-
chitects. CityGML is much more recent, has been de-
signed to be stored even more classes of items in-
cluding buildings, bridges, tunnels, and mobile site.
While LandXML is writable by many existing software
tools, CityGML is not yet available as the file format
in many existing tools, and this file format is widely
adopted by industry specialists (Nessel, 2013).
The growing number of software tools available,
the growing adoption of digital versus paper-based
exchange of information, the internationalization of
the design process and the growing demand for sus-
tainable information on the whole life cycle of a
building (Gero,1990) requires the use of tools able to
support useful context entities for heritage architec-
tural design.
The study focused on paper mill factories and
showed that GIS technologies are not enough to
have a complete understanding of the existing con-
text for these factories and therefore it can be said
that GIS technologies alone are not enough to sup-
port an informed design for these types of develop-
ments. GIS tools allow to overlap maps, to link tradi-
tional relational databases identify features in the ur-
ban landscape and to attach attribute data in order to
analyze and to inform assessment of urban areas, but
they did not allow to reason on these information;
for this purpose the Context Knowledge Model (Cx-
tKM) (Gargaro and Fioavanti 2013a), developed us-
ing ontologies, was used. The semantic structure of
the model gives opportunities to have a wide range
of analysis on industrial areas.
The core approach of this paper is the relation-
ship between a CxtKM based on ontologies and GIS
tools aimed at the regenerations of the industrial
buildings and areas.
The Context Knowledge Model used for disused
594 | eCAADe 32 - Digital Heritage 1 - Volume 1
industrial sites can be applied to different scenar-
ios. The power of this approach lies in the compu-
tational modeling of the context knowledge for ar-
chitectural design, integrating it with the Semantic
Web technologies that allows interoperability of the
model with different tools, as GIS ones.
HERITAGE BUILDINGS AND THEIR DESIGN
REUSE
The hypotheses concerning possible reuse scenar-
ios for abandoned paper mills, without clashing
with their building's character, led to alternative de-
sign layouts being considered, however it should be
pointed out that the "constraining" aspects of old
structures and industrial spaces needed to be taken
into account, in order to be able to achieve these al-
ternative usages.
Digital models for urban-building design should
be able to represent these different characteristics
and requirements old/new layout, old/new spaces to
compare them, and this frequently is not an easy task.
Nowadays on one hand, traditional researches
highlight links established between industrial build-
ings and their context (urban, industrial, social and
physical); on the other hand, GIS tools are used to
integrate the investigations and information, creat-
ing an information database for the paper mill net-
work (Leggieri, 2013). Notwithstanding research ef-
forts, these two research aspects remain practically
separate. These research efforts were not sufficient
to provide support tools for design knowledge which
is the key point to any modification of existing paper
mills.
Designers need an effective and complete un-
derstanding of the site to support informed design
decisions; nevertheless GIS instruments can only give
site information linked to position; they are not able
to support knowledge that means adding semantics
and reasoning to information. To this end the re-
search defines a model of Context Knowledge spe-
cific for modern industrial buildings.
This approach is based on the recognition that it
is smart and effective to combine CxtKM and GIS to
 regenerate industrial areas. This introduces the no-
Figure 1 tion of a Building Heritage Knowledge Framework -
"Liri" watercourses BHKF - to facilitate public and/or private administra-
and Isola del Liri tion decision-maker (Plume et al., 2011).
and Sora cities The research has defined a model of informa-
tion with the objective to support design process
in a more conscious way taking care of intersection
among individuals - construction - urban context. It
provides a platform for the management of context-
information-knowledge.

PAPER MILLS NETWORK IN CENTER ITALY -

A CASE STUDY
The network of paper mills in Southern Lazio (the
province of Frosinone) can be characterised as major
industrial production site from the begin of the nine-
The unique orographic structure that characterizes
teenth century until the nineteen-fifties. The growth
the dowtown of historical cities is a constant aspect
of a singular industrial infrastructure built along wa-
that characterizes the urban settlements.
tercourses as the hydraulic system was very impor-
Many of these structural features are relevant.
tant for the region's development, involving all the
The cities along the Liri River, which are placed in a
surrounding areas beyond the urban area.
peculiar context, have shaped the environment and
The paper mill districts were stratified over time
natural resources. The features related to the water
so that they characterized the landscape, changing
system influenced the origins of various urban loca-
the original character of the agricultural environment
tions. The origin and evolution of the settlement sys-
and were grafted onto the natural landscape charac-
tem of “Isola del Liri” were related to production activ-
terized by the presence of the close river network.
ities that led to the current appearance of the site and
In these places there are visible traces of the hy-
its special position within the regional settlement.
draulic system derived from the courses of the Liri
The current structure of the city center is built around
and Fibreno rivers, evidenced by the presence of fac-
the small island in the middle of the river. From this
tories, sluices, canals, water intakes, dams, visibly di-
origin the city expanded with its factories adapting
aloguing with nature. This particular intertwining of
them to natural and man-made elements. The pro-
urban and environmental system was the main value
ductive use of the water resources over time has char-
and peculiarity of the context. Nowadays, the urban
acterized the urban landscape marking it with the hy-
area has grown so these paper mills are surrounded
draulic works for the derivation of water for indus-
by city districts. In order to reuse abandoned indus-
trial uses changing and impacting on the environ-
trial areas, it is important to consider their old con-
mental features. The course of the Liri river creates
texts. Therefore the refurbishment design should es-
particular jumps, scenic waterfalls that characterize
sentially be based on the investigation of the reasons
the urban landscape. The use of hydraulic and hy-
behind land use conditioning, development of indus-
dropower production dates back to the oldest paper
trial buildings, their integration into the old context,
mills whose location was influenced by the necces-
changes that have influenced their removal and new
sary presence of the water courses to take advantage
needs that arise.
of the mechanical motion of the river. Paper mills
The morphology of the territory shows some pecu-
require the construction of branch ducts for the use
liarities (Figure1).

Digital Heritage 1 - Volume 1 - eCAADe 32 | 595

 and discharge of waters even when being located on • Populate context knowledge and information
the banks of the river. There are currently ten paper with urban, historical, cultural, economic, en-
mills in the town of "Isola del Liri". Some of them are vironmental and juridical data to raise design-
still working but some are in a neglected state and in ers' awareness;
decay.
• Improve the CxtKM (Gargaro and Fioravanti,
2013b) to formalize historical site evolution,
DESIGN 'INVARIANTS' AND THE REPRE- land and soil use, characteristics and relation-
SENTATION OF CONTEXT KNOWLEDGE ships among them.
The aim of this research is the definition and devel-
opment of tools to help designers better understand BUILDING HERITAGE KNOWLEDGE
design opportunities and automate some design ac-
FRAMEWORK USING CXTKM
tivities (Gero 1990, 2014).
The design could be described as a constrained activ-
The knowledge representation models to define
ity context-dependent.
structured information in specialized domains have
Architectural Design has, among many aims, to
always been present in CAAD tools, but rarely they
integrate designed buildings in their natural and so-
were specifically applied to Context Knowledge for
cial context. This process imposes constraints on the
architectural design (Gero 1990; Gursel et al. 2009;
variables and their values dependent on context. The
Gargaro and Fioravanti 2013a; Gargaro and Fiora-
design activity is influenced by different aspects as
vanti 2013b).
the working environment in which designers work
OIL, DAML and BKM use the concepts introduced
and the 'context' from which the development of
in Resource Description Framework (RDF) extended
the project depends. The designers' perception of
knowledge related to the 'product' with implicit op-
context in architectural design changes in relation to
erators, multiple domains and constraints. The rea-
changing perceptions of the designers. The design
soning to control and infere on context knowledge is
activity can be defined as a decision-making process
organized in a hierarchical set of different OWL. But
in which the exploration of the context and learning
only BKM introduced context entities that are not im-
activities are a goal-oriented work that influences the
plicit, and take account simple properties.
perception of the designers.
The development of the context entities, pro-
The changing environment characteristics lead
posed and implemented a specific model of context
to a reformulation of the building requirements for
knowledge CxtKM (Context Knowledge Model; Gar-
its use. This can be seen also when the evaluation
garo and Fioravanti 2013b) that has been modified
of the comparison between the actual behavior of
taking into account Europeran INSPIRE Directive and
the structure and the expected behavior is insuffi-
Italian GIS Agreement for the GIS databases in order
cient and cannot be satisfied only by manipulating
to allow a better interoperability with standards and
the structure (Gero, 1990). To overcome these diffi-
codes and used for heritage reuse.
culties we propose relational structures to relate con-
The milestones of research are:
text entities to the disused industrial building.
To make this possible each industrial building en-
• Study the peculiarities of modern industrial
tity is related context ones using specific relationship,
buildings;
a Relation Structure - RS- which an inference engine
can use to compute a goal for its regeneration (Car-
• Use GIS technologies to link technology fea- rara and Fioravanti, 2001).
tures of paper mills, hydraulic system, and ur- There are two fundamental classes for the Build-
ban landscape; ing Heritage Information/Knowledge Process: that

596 | eCAADe 32 - Digital Heritage 1 - Volume 1

 of the Building Invariance, which defines the build-
ing spaces that define the industrial building iden-
Figure 2 tity, and that of the context entities which gives the
A Building Heritage contextual peculiarities, defined using European IN-
Knowledge SPIRE Directive and Italian GIS Agreement for the GIS
Framework schema databases in order to allow a better interoperability
of the CxtKM with GIS tools (Fig.2).
For instance, an architect can conceive a building
reuse taking into account his/her own specific Rela-
tion Structure, dependent on the Building Invariants,
that applies its inferential engine rules plus context
rules to have a context-aware decision about indus-
trial building refurbishment. Using ontologies to for-
malize the context knowledge is possible to have a
formal structure of the context entities considered in
a project (i.e. meanings, properties, rules, behavior,
capability, etc.) and formal models (generally mathe-
matical) that allow simulations, verifications and rea-
soning to be performed (Carrara et al., 2009), and fil-
ter mechanism (Fioravanti 2008; [3]) to reduce the
space of possibilities.
The study of the disused building focused on pa-
per mills. The starting considerations were based
on the characteristics and relations between external
and internal spaces to bring out the peculiarities of
these industrial areas.
The starting considerations were based on the
characteristics and relations between external and
internal spaces to bring out the peculiarities of the in-
dustrial areas.
The recognition was made using fillable tables in
which were written the information about each space
of the factory as dimensions, material, preservation
status to obtain data and meanings.
This leads to determine the recurrent values of The "Invariants" of disused industrial areas are repre-
old building layouts, spaces, structures, functional sented by diagrams and tables (Figure 3) in which the
uses, material and their relationships, but also criteria artifact is analyzed in terms of three groups of vari-
to regenerate disused areas and industrial factories in ables: function, properties and capabilities.
view of new needs - in a word the "invariants".

Digital Heritage 1 - Volume 1 - eCAADe 32 | 597


After this phase of exploration and analysis, data and
meanings were imported into GIS tools, giving to
each entity a specific ID and then exported in Con-
text Knowledge Model.
The Context Knowledge Model (CxtK) was imple-
mented following these activities: formulation of the
entities, analysis, synthesis, evaluation, selection and
reformulation of entities hypothesis.
CxtKM was formalized though ontologies devel-
oped in Protégé 4.3 [4]. Then CxtKM is populated us-
ing GIS entities exported in excel and with the same
ID of the context entities of CxtKM (Figure 4). Then
different to-do (and not-to-do) lists and strategies for
regenerating old paper mills were carried out and
compared considering the constraint rules based on
the analysis of typologies and contexts. The context
information are enriched with the context rules and
the building 'Invariance' deduced from the analysis
and they are inferred though the inference engine
JessRules [5], based on First order Logic, that allows
to control all the declared and defined rules in the on-
598 | eCAADe 32 - Digital Heritage 1 - Volume 1
tologies, to maintain the right class hierarchy that is
Figure 3
very important when there are more than one par-
Function, structure
ents, to have design advices and to reduce the possi-
and behavior
ble design choices for a context-aware architectural
'Invariants'
design.
BHKF IMPLEMENTATION FOR HERITAGE
ARCHITECTURES
The CxtKM is built by ontologies (Beetz et al., 2006;
Gursel et al., 2009; Rehman et al., 2008) links GIS tools
through databases in order to support designers and
allow them to model constraints, verify algorithms
and check rules at different levels.
The starting point was the definition of entities
and the creation of rules to link entities, properties,
reports, functions, etc. The semantic structure of en-
tities helps to provide knowledge that can be used to
prune wrong design proposals and discover oppor-
tunities or suggestions for a wide range of analysis
on old industrial areas.
Context was subdivided into sub-entities - Phys-
ical, Cultural, Normative and Economic.The enti-
ties are formalized using Protégé ontology language
(OWL code compliant) and rules have been associ-
ated with each entity to define its behaviours.
The primary source of data was an empirical ex-
ploration of literature review and scientific ones car-
ried out in parallel with the study of data collected
and analysed for the purpose of discovering concepts
and relations which are then organized into a knowl-
edge structure.
A reasoning mechanism was used to reduce the
industrial area regeneration design proposals, prun-
ing inconsistent hypotheses, reducing the number of
possibilities, avoiding the risk of combinatorial explo-
rations (Bickhard an Terween,1995). Hence architects
can concentrate their efforts on attaining goals of
high quality design and innovative building (or part
of it), as they do not have to consider all the possible
strategies but a only reduced number that respects
urban and building Codes and architect's goals and
experiences. As the design process evolves, solutions
and aim are modified and refined until the desired
Figure 4
Conceptual Model
Framework
behaviour or performance is obtained.
The formalization of Context Knowledge Ontol-
ogy is related to environmental, cultural, economic
and juridical entities, specifically aimed at render-
ing explicit the operational meaning through a struc-
tured set of classes / concepts / entities, relationships
and reasoning rules among them.
The prototype of BHKF - Building Heritage
Knowledge Framework - has been defined by the
building functions (F), capabilities (C) and require-
ments (R). The constraints on the functions appear
as required behaviour related to building capabilities.
The capabilities of the building are the unexplored
potentiality of it that can even change the type of
the class. The functions are dependent not only on
the requirements (R) but also on the behaviour of
the building that define the building performances
(P). These functions also need a Context Knowledge
Model that allows to add variables, dependent on
context (Cxt), included the location (L) in which the
building stands.
This prototype could be used to reduce the range
of possibilities for re-use during the design process.
A building design hypothesis can be symboli-
cally represented as a function of:
B = f(F, C, R, Cxt)
Digital Heritage 1 - Volume 1 - eCAADe 32 | 599
 where
C = g(R, P)
Cxt = h(B, L)
The Building function - f (B) - is non-linear, because it
depends on context and on weights applied to build-
ing entities by designers. The weights are assigned
to each entity in relation to the context importance
and to the building requirements. They include, if re-
quired, the approximate evaluation of the expected
costs so that the designers are allowed to choose not
only the most promising or satisfying solutions, but
also the cost-effective ones and the client can be ac-
quainted with that.
PAPER MILLS RULE EXAMPLE:
HYDROPOWER ENERGY PRODUCTION
The case studied for the evaluation of BHKF, as shown
in the previous paragraphs, is located in Central Italy.
The existing industrial sites and buildings gener-
ated by very specific industrial requirements could be
turned into settlements with the help of GIS and con-
text ontologies. The context is characterized by par-
ticular jump shares and waterfalls that characterize
these small towns and influenced by the paper mills
development.
This context can be evaluable to be modelled by
reasoning rules as for instance the high jump and
quantity of water, but this consideration could not
be enough to make the decision to use it in the 're-
generated' building. In this example, the other prob-
lem that needs to be considered is the noise level due
to the production of electricity, as some of the paper
mills are very close to the city centre.
Case Studies
1. Waterfall(n)
2. if jump higher > 3 m
3. and if water quantity >1 mc/h
4. then <hydroelectic energy
,→ production > return true
5. else return false
The described rule was applied to two paper mills,
one located in "Isola del Liri" and the other in "Atina"
600 | eCAADe 32 - Digital Heritage 1 - Volume 1
(a smaller town near Isola del Liri).
The example shows that the paper mill in Atina
does not have the possibility of hydropower energy
generation because a jump of one meter is not eco-
nomically sufficient to produce electricity (figure 5).
However the paper mill in Isola del Liri has the capac-
ity to generate electricity due to fact that the jump is
higher and the quantity is more than one cubic meter
per hour (figure 6).
Applying the model onto other paper mills in
"Isola del Liri" district and in some cases (Figure 7) the
model infers that is not convenient to produce elec-
tricity, even if the jump and the quantity of water is
sufficient, because the paper mill is too much close
the city centre and so the acoustical protections nec-
essary for the surrounding isolation are too much ex-
pensive or there is an incompatibility with the her-
itage landscape.
CONCLUSIONS
The common characteristics and relationships – the
"knowledge invariants" - of modern industrial areas
and buildings are not totally supported by GIS tools
to help architects effectively utilize the knowledge
from these tools. Therefore a Building Heritage
Knowledge Framework - BHKF - has been created to
link a Context Knowledge to GIS.
Figure 5
<hydroelectic energy production > =
,→ FALSE
Figure 6 The development and implementation of the BHKF
has been focused on old industrial area context
knowledge, mainly concerning reasoning rules to
give advice during re-design of these areas. This
model is a promising tool to help designers to man-
age and share heritage context knowledge obtained
from a GIS model and from formalized experience.
A Context Knowledge Model can evaluate a project
a priori so as to evaluate constraints and offer sug-
gestions for alternative design proposals, thereby ex-
plaining motivations or unsatisfied requirements.
In brief, BHKF could:
• Share and spread Context Knowledge to find
better design solutions to reuse industrial
buildings;
<hydroelectic energy production > =
,→ TRUE • Inter-operate various common tools to refur-
bish industrial areas;
• Facilitate dialogue between designers and
public administrations;
Figure 7
Inferred Rule • Help public administration to verify design
Example proposals with reuse or refurbishment Codes.

ACKNOWLEDGEMENTS
The research was partially funded by MIUR (the Italian
Ministry of Education University and Research), 2012
Research Project: "Built Heritage Information Mod-
elling/Management - BHIMM".

REFERENCES
Beetz, J, Leeuwen, JP and Van de Vries, B 2006 'Towards
a Topological Reasoning Service for IFC-Based Build-
ing -information Models in a Semantic Web Context',
Proceedins of joint International Conference on Com-
<hydroelectic energy production > = puting and Decision Making in Civil Building Engineer-
,→ TRUE ing, Montreal , pp. 3426-3435
... Bickhard, MH and Terween, L 1995, Foundational Issues
27. If <Industial Building Area > < 10 in Artificial Intelligence and Cognitive Science, Elsevier
,→ mq from <Downtown > Science Publishers B.V., Amsterdam, NL
28. then <hydroelectic energy Carrara, G and Fioravanti, A 2001 'A Theoretical Model
,→ production > of Shared Distributed Knowledge Bases for Collabo-
29. return FALSE rative Architectural Design', Strategic Knowledge and
Concept Formalization III, Heron Island, Sidney, pp.
129-143

Digital Heritage 1 - Volume 1 - eCAADe 32 | 601

 Carrara, G, Fioravanti, A, Loffreda, G and Trento, A
2009 'An Ontology-based Knowledge Representa-
tion Model for Cross-Disciplinary Building Design: A
General Template', Proceedings of eCAADe 2009, Is-
tanbul (Turkey) , pp. pp. 367-374
Carrara, G, Loffreda, G and Fioravanti, A 2013 'A Proac-
tive Platform for Knowledge Management in Cross-
Disciplinary Building Design', eChallenges e-2013
Conference Proceedings, IIMC International Informa-
tion Management Corporation Ltd 2013, Dublin, pp.
1-10
Fioravanti, A 2008 'An e-Learning Environment to En-
hance Quality in Collaborative Design. How to Build
Intelligent Assistants and ‘Filters’ Between Them', Ar-
chitecture and Modern Information Technology, 4(5),
online: see [3].
Fioravanti, A, Loffreda, G and Trento, A 2011 'Computing
Ontologies to Support AEC Collaborative Design',
Proceedings of eCAADe 2011, Ljubljana, pp. pp.177-
186
Gargaro, S and Fioravanti, A 2013a 'Traditions based on
context. How context ontologies can help archaeo-
logical sites', Future Traditions: eCAADe International
Workshop Proceedings 2013, Porto, pp. 105-114
Gargaro, S and Fioravanti, A 2013b 'A Context-
Knowledge Model for Architectural Design', Pro-
ceedings of eCAADe 2013, Delft, pp. 81-90
Gero, J 1990, 'Design Prototypes: A Knowledge Repre-
sentation Schema for Design', AI Magazine, 11(4), pp.
26-36
Gruber, TR 1993 'Toward Principles for the Design of On-
tologies Used for Knowledge Sharing', Proceedings of
the International Workshop on Formal Ontology 1993,
Padova
Gursel, I, Sariyildiz, S, Stouffs, R and Akin, O 2009 'Contex-
tual Ontology Support as External Knowledge Rep-
resentation for Building Information Modeling', Pro-
ceedings of CAAD Futures, pp. 487-500
Kelly, N and Gero, JS 2014 'Interpretation in design: mod-
elling how the situation changes during design ac-
tivity', Research in Engineering Design, 25(2), pp. 109-
124
Leggieri, E 2013 Analisi e prospettive per i complessi indus-
triali dismessi. Il caso delle ex Cartiere di Isola del Liri
in provincia di Frosinone (Analysis and perspectives for
disused industrial buildings. The case study of the pa-
per mills at “Isola del Liri” – Frosinone- Italy), Ph.D. The-
sis, Sapienza University of Rome
Nessel, A 2013 'The Place for Information Models in Land-
scape Architecture, or a Place for Landscape Archi-
tects in Information Models', Proceedings of Digital
602 | eCAADe 32 - Digital Heritage 1 - Volume 1
Landscape Architecture 2013, Berlin, pp. 65-72
Fridman Noy, N and Hafner, CD 197, 'The State of the Art
in Ontology Design A Survey and Comparative Re-
view', AI Magazine, 18 (3), pp. 53-74
Peter, A, Newton, B and Wills, C 2009 'Visualising Archi-
tecture: A field study in rural England', Proceedings
of 11th Bogus Conference, London, pp. 12-14
Plume, J and Mitchell, J. 2011 'An Urban Information
Framework to Support Planning, Decision-Making
& Urban Design', Proceedings of CAAD Futures 2011 :
Designing Together, pp. 653-666
Rehman, F and Yan, XT 2008, 'A Case Study to Support
Conceptual Design Decision Making Using Context
Knowledge', in Yan, X.T., Jiang, C. and Eynard, B. (eds)
2008, Advanced Design Manufacture to Gain a Com-
petitive Edge, Springer, pp. 13-22
Russel, S and Norving, P 2010, Artificial Intelligence: A
Modern Approach 3rd Edition, Pearson Education Inc.
Stevens, R, Horrocks, I, Goble, C and Bechhofer, S 2001
'Building a Reason-able Bioinformatics Ontology Us-
ing OIL', Proceedings of the IJCAI-2001 Workshop on
Ontologies and Information Sharing, Seattle, USA, pp.
81--90
Wurzer, G 2009 'Systems – Constraining Functions
Through Processes (and Vice Versa)', Proceedings of
the 27th eCAADe Conference, Computation: the new
realm of architectural design, Istanbul, pp. 659-664
[1] http://www.landxml.org/
[2] http://www.citygml.org/
[3] http://www.marhi.ru/AMIT/2008/4kvart08/
Fioravanti/Fioravanti_paper_AMIT_5.pdf
[4] http://protege.stanford.edu/
products.php#desktop-protege - v4.3.
[5] www.jessrules.com/jess/index.shtml - v7
Gamification of Shape Grammars
Collaborative and Participatory Mass-Housing Design for Kashgar Old
Town

Serdar Aydin1 , Tian Tian Lo2 , Marc Aurel Schnabel3

1,2,3
The Chinese University of Hong Kong
1,3
{serdar|marcaurel}@cuhk.edu.hk 2 skyduo@gmail.com

This paper describes the framework of an ongoing research, titled

'quasiGRAMMARS', seeking a participatory mass-housing approach. In the
context of the city of Kashgar, China, where the convergence of
Islamic-Chinese-Turkic cultures has been shaped within a unique style since the
10th century, mass-housing becomes a 3D puzzle that requires each piece to be
placed with full of care, motivation, participation, analysis, strategy, art and
finally design. Gamification is about designing collaboration and participation
for mass-housing, whereas shape grammars are meant for analysis and design.
This game finally turns into a strategic game to be scrutinised further in relation
to game theory that is mathematically concerned with the economics too.
However, the present study aims at proving a participatory design strategy that
incentivises valuable action through gamification techniques. Focusing on its
specific design development, it reveals some of these techniques to gamify
mass-housing for Kashgar in eight steps. While unveiling gamification term for
use in architecture domain, the paper discusses the limitations and future
directions of the research.

Keywords: Shape grammars, gamification, mass-housing, participatory

decision-making, Kashgar

1. INTRODUCTION consequences of lacking sufficient effort to take the

By the time, you have been reading this sentence,
nearly 2 million messages was created and shared
via WhatsApp (pennystocks.la). Of the available data
in the world, more than 90% has been accumu-
lated in the last two years, with the social media use
being the most prolific data source (Hudson-Smith
2014, 42). And these figures were unimaginable only
three years ago. Despite its fascinating speed, the
straightforwardness of social media devices as an op-
portunity to make use of these data lead to informa-
tion overload. Simply designed, a gamified platform
can be a key to open up willingness for massive con-
tribution. Gamification is a key to activate dynamics
for a meaningful purpose and, in particular, with fun.
quasiGRAMMARS is a research project looking for pos-
sibilities of such a scheme in mass-housing, and, at a

Digital Heritage 1 - Volume 1 - eCAADe 32 | 603

 greater extent, in architecture (Lo et al. 2014).
This paper's definition, literally, is that gamifi-
cation is to provoke and make use of higher-level
intelligence via a collective manner that draws on
big data available online through sharing. Through
gamification, potentials of social network and fun-
environments are utilised by designing collective in-
telligence for mass-housing instead of computation-
ally expensive solutions. In other words, gamification
plays the designed role of an ombudsman that is ap-
pointed to bring varying demands over a table based
on voluntary and intrinsic participation.
More specifically, the underlying purpose of the
present work is to introduce some of gamification el-
ements in the context of the development of Kash-
gar's old-city. Starting with a brief introduction of
Kashgar and its value in this research, the paper dis-
cusses the stages through which gamification be-
comes a diagram to map an open-collaborative sys-
tem for mass-housing in Kashgar. Finally it presents a
preliminary work which was conducted in an elective
course with 15 students grouped into two.
preserved example of a traditional Islamic city to be
found anywhere in Central Asia (Michell et al, 2008,
79)." And it is ironically the most-endangered one at
the moment. Large-scale urban projects being im-
plemented with the incorporation of local and cen-
tral governments have already torn down nearly the
two-thirds of the old-city. The ongoing project de-
scribed as a 'precautionary' measure against earth-
quakes by the authorities is worth $4.39 billion, re-
quiring 50,000 households, in other words more than
200,000 people, to be resettled (www.rfa.org).
Figure 1
The satellite map of
Kashgar old-town
which undertakes a
major urban
renovation project
requiring nearly
200,000 people to
be replaced
(Google Earth).

2. KASHGAR: JEWEL IN CENTRAL ASIA

Kashgar, the westernmost city in China, is located
where many different cultures and religions have jux-
taposed thanks to its geographical position. The an-
cient city is described as "the heart of one of the
most lovely and bountiful oases in all Central Asia
The process of the resettlement is oriented without
(Starr 2013, 307)." Kashgar politically belongs to the
an inclusive decision-making process. Families are
Xinjiang-Uygur Autonomous Region whose capital,
compelled to move in identical units 'prefabricated'
Urumqi, is the world's most landlocked city (Dong
for efficiency and affordability. Since they are not in-
and Zhang 2011). This locational information con-
volved in decision-making, the projects cast doubts
cisely explains the environmental conditions that
though, allegedly, legalised by the Chinese govern-
have formed the regional architecture.
ment on the basis of earthquake safety. And therein
Situated in the Tarim Basin in the periphery of
lies political and ethnic tensions between the local
Taklimakan Desert, Kashgar's enduring examples of
Uyghur people and the Han Chinese who dominates
Uygur, a Turkic people of Central Asia, architecture
the Chinese state that wants to have more author-
are threatened with oblivion and unbridled devel-
ity over its gateway to Central Asia. Unlike politi-
opment which is invariably everywhere in China as
cal/ideological problems, Kashgar's richness lies un-
a consequence of neo-liberal policies (Michell et al.
derneath its distinct Uygur identity which is a mul-
2008, 147; Florenzano et al. 2010; Aydin 2012).
tifaceted composition of Islamic, Central Asian, Mid-
The historical urban fabric in Kashgar is "the best-
dle Eastern, Chinese, Mongol, Tibetan, Persian and

604 | eCAADe 32 - Digital Heritage 1 - Volume 1

even Indian influences. Therefore, mass-housing in
Kashgar gains a methodological peculiarity akin to
an open-ended 3D puzzle game demanding mani-
fold means for an inclusive optimised solution.
In addition, the vernacular organisation of archi-
tectural and urban space in Kashgar is worth men-
tioning. The houses in Kashgar's old town invariably
have a courtyard regardless of the householder's so-
cial or financial statue. Accessed via narrow lanes, the
courtyard typology, by which vertical and horizon-
tal allocation of different functions are composed,
displays variation based on affordability. As a reli-
gious condition, which is prevalent in other Islamic
historical cities, the balance between privacy and sis-
ter/brotherhood defines visual and spatial perme-
ability in the old town via courtyards and lanes. As a
consequence the old city is a product of interwoven
arrangements, whereas strong social relationship is
fundamental for its agglomeration that grows from
the mosque as a de facto standard.
Apart from the data gathering and design as-
pects of this research, the third focus is Kashgar's
digital heritage by making use of gamification tech-
niques. Considering that large-scale projects under
construction, it is a bit late to save Kashgar's archi-
tecture in real terms. By digitally recording and dis-
playing its history through different media, quasi-
GRAMMARS is projected to bring out creative presen-
tation and exhibition means via gamification. But this
aspect belongs to the proceeding stages of this re-
search.
3. GAMIFICATION AS A MEANS FOR USE
OF SHAPE GRAMMARS
3.1. What gamification is not
The term of gamification might sound novel however
its use dates back to the beginning of the last century.
In 1912 Cracker Jack, a snack brand, started to give
toys as a gift in their boxes. Although gamification
in this research is meant to be 'collectively produc-
tive', with this instance its focus might seem on the
marketing domain that is to encourage for more 'con-
sumption'. However, by and large, gamification is de-
scribed as the use of game design elements in non-
game contexts (Deterding et al. 2011). Depending
on familiarity, there are varying common assump-
tions towards gamification. The following paragraph
informs us on what gamification is and is not.
First of all, gamification is not turning everything
into a game. The purpose of gamification is not to
pull us out of reality but rather finding what is not
boring in an activity that usually requires collabora-
tion and engagement. This is neither to say that they
are 'serious games' which are used as training and
learning environments such as in military and edu-
cation. The focus of simulations in serious games is
on testing the abilities of learners and on improv-
ing their skill sets in a virtual environment similar to
real conditions. On the other hand, game theory is
to mathematically analyse decision-making 'strate-
gies' or individual 'choices' (Kelly 2003, 1), whereas
gamification may be helpful to improve collaboration
for 'a choice' and encourage involvement in 'a strat-
egy' (Kapp 2012, 9). Also, points, badges and leader-
boards (PBLs) are irrepressibly penetrating into every
aspects of our daily lives in tandem with the grow-
ing use of social media. PBLs are one of the most
common game elements however they are not suf-
ficient with regards to what games and game design
can provoke. We can gamify a situation by thinking
like a game designer, which is different than being
a game designer. With gamification, we break down
games into elements that are explained below.
3.2. Game elements: How?
The largest LAN (Local Area Network) party
with around 11,000 participants was recorded
at DreamHack Winter 2007 in Sweden
(www.guinnessworldrecords.com). What is intrigu-
ing about it is that the focus is on "everything you can
do with computers" combining fun activities with
learning and sharing, such as gaming, communica-
tion, programming, designing, music composing,
etc. (www.dreamhack.se). Admittedly, games play
a major role in societies being shaped by the 21st
century culture of gadgets and devices. Use of on-
Digital Heritage 1 - Volume 1 - eCAADe 32 | 605
 line games is constantly in increase as a business and
marketing strategy to motivate people in engage-
ment and sharing (Zichermann and Linder 2010,
47). Huizinga's description names the boundaries
of engagement in play and play environments as the
Magic Circle in which once you enter, "it is sacrosanct
for the time being", i.e. the game rules matter most
not the real world (Huizinga 1955, 77).
In order to invite the player into the Magic Cir-
cle, game elements should be designed properly to
prompt engagement together with aesthetics that
contributes to the whole experience which we look
for. Game elements are classified differently but we
use here the one (Werbach and Hunter 2012) which
separates them as dynamics, mechanics and compo-
nents. It does not necessarily require using all el-
ements but integrating ones that will be most effi-
cient in motivating player engagement. Mapped as
a pyramid, dynamics are placed on top while me-
chanics and components step behind them respec-
tively. A list of dynamics may include constraints,
emotions, narrative, progression and relationships
(Werbach and Hunter 2012). This paper further looks
into dynamics as a map to use in explaining a concep-
tual framework for gamifying an open-collaborative
design systems based with shape grammars. Further
studies will look into mechanics and components un-
der which the number of elements is higher.
3.2.1. Game elements for mass-housing in Kash-
gar. As a designed constraint, shape grammars are
the rules of the design game. Participants are given
priorities once they make a decision which conform
predefined criteria relying on Kashgar's grammar.
The game is open to changes, i.e. new shape gram-
mar rules can be added to generate different results.
However, the more resemblance to Kashgar's char-
acter, the more priority is given to get involved in
new decisions. This is also constrained by a voting
system guided by environmental evaluations such as
daylight and wind. At the end of the game, those
who lead the leaderboard pay less since their collab-
oration points are high. Though risky, creative ideas
that lead to innovative design alternatives can also be
606 | eCAADe 32 - Digital Heritage 1 - Volume 1
gained through design rules not alike the initial.
Notice that the concept of this proto-game in-
cludes at least three game elements which are exter-
nal rewarding, feedback and cooperation. Besides,
we could add a time constraint, levels and story-
telling which would be related to intrinsic rewards
that has authentic meaning behind. Intrinsic mo-
tivators have a more significant place in gamifica-
tion, e.g. saving the character of a historical city,
participating its development and contributing to its
digital heritage. Game elements, such as badges
and points, do not inherently become external or
internal motivators. Instead the context makes the
distinction, e.g. feelings gained from mastering a
challenge that causes increase in self-confidentiality.
Gamification explains intrinsic reward structures by
means of self-determination theory (Deci and Ryan
2000). Autonomy, competence and relatedness are
the elements of self-determination theory. Auton-
omy is corresponds to the feeling of doing something
meaningful, while competence means mastery from
achieving a seemingly difficult task. Relatedness is
anchored to connection and interaction within so-
cial circles (Werbach and Hunter 2012). Through so-
cial connection, designing for such a collective intelli-
gence has the same goal as many other complex sys-
tems such as ant colonies which nestle individual ac-
tions of many agents in order to harness the best or
most relevant outcome (Porter 2008, 128).
Gamification in the Kashgar project works for
data collection via social media on which there is a
lot of Kashgar photos and individual stories waiting
to be identified. They contribute to knowledge in
analysing Kashgar and its recent past to understand
the pattern of urban life that is often both personal
and impersonal. Therefore, it is such a gamified plat-
form that uses the potential of game elements to col-
lect these data in order to expose hidden values and
meanings into decision-makers' attention, i.e. the cit-
izens, designers, planners and developers.
3.3. Gamifying Shape Grammars
Based on Kashgar's case, this research seeks a nar-
rative articulation of mass-housing design with en-
hanced participation. Inhabitants' involvement is
complemented with alternatives in the design stage.
To produce these alternatives, this research uses
shape grammars which are invented by Stiny and
Gips (1972). For many, shape grammars obstruct cre-
ativity because of its rule-based approach. However,
it is widely acknowledged that shape grammars are
exceptionally potent when the problem needs peda-
gogical, historical, synthetic and/or analytical design
solutions. Given that mass-housing projects in his-
torical towns require retrospective and tactful solu-
tions, analysing Kashgar Old Town with shape gram-
mars is particularly important to attain a contextual
end-product with which the local can identify.
Recent works (Beirao and Duarte 2005; Benros et
al. 2007; Aydin and Schnabel 2013) examine shape
grammars to achieve a flexible urban design tool fo-
cusing on both descriptive and generative sides by
taking advantage of their parametric essence. The
research of City Induction (Duarte et al. 2012), and
the City Maker thesis (Beirao 2012) provide rigorous
insight on the problem of city planning, the lack of
flexible design tools and integration of shape gram-
mars in urban planning.
Being developed as an urban modelling tool at
ETH Zurich, CityEngine is used for the research (Ha-
latsch et al. 2008; Parish and Muller 2012) on quick
visualisation of urban geometries derived from built-
in and customisable shape grammar rule sets.
Marrakesh (Duarte et al. 2007), Porto (Coimbra
and Romao 2013) and Maputo (Barros et al. 2013) are
amongst the historical cities that are analysed and/or
modelled via shape grammars. These works heav-
ily rely on site analysis to understand their unique
architecture style to propose further urban develop-
ment, whereas Pompeii work (Haegler et al 2009) is
a procedural modelling of an ancient city that is vi-
sualised as a means of digital heritage. To highlight
the Marrakesh work (Duarte et al. 2007), its resultant
shape grammars are evaluated in response to energy
efficiency based on evolutionary optimisation tech-
niques such as genetic algorithms (Caldas and Santos
2012).
One common aspect of these works is that they
point out the strength of shape grammars in being
generative and descriptive, and more importantly
that their shape grammar use is in parallel with proce-
dural information and evaluation. This research takes
shape grammars for granted because they are so
practical in solving the complexity of mass-housing
projects.
The creativity diagram correlates creativity
and designing to playing through copying that is
summed up by two activities; recursion and embed-
ding (Ham 2013). The diagram's focus is on learning
for beginners who are, without exception, asked to
be creative as soon as enrolled for architecture or de-
sign education. This is not always easy for a designer
who is constantly pushed for time to finish a project
on time. However our focus is on a novel mass-
housing design workflow which necessarily needs
playing designed with gamification techniques.
Two earlier scientific projects use gamification
techniques to predict possible protein combinations
(fold.it) and DNA sequences (phylo.cs.mcgill.ca). En-
abling for a meaningful contribution, Foldit targets
to identify new proteins that could help prevent and
treat important diseases by crowdsourcing the re-
search through a puzzle game. Aiming at discover-
ing DNA sequences, Phylo is a research in molecu-
lar biology to identify new genes. In this puzzle, the
players are asked to make multiple sequence align-
ment so that the research team could decipher the
data driven from a heuristic algorithm to be used for
the research in genetic disorders like cancer. By mak-
ing use of collective intelligence, both research ben-
efit from game design since it is computationally ex-
pensive to achieve an optimal solution. In spite of
a cumbersome process, the two studies integrate a
novel approach that is based on the potentials of
data collection and management via willingness and
fun activities. In other words, it is literally a win-win
game connecting instantly available data with scien-
Digital Heritage 1 - Volume 1 - eCAADe 32 | 607
 tific studies.
Similarly, the potentials of shape grammars are
to be tested with this research gamified pattern
which describes the connection between the partici-
pant and the rules and tools. In a sense, this pattern is
the gameplay of quasiGRAMMARS, to be designed in
a narrative way (Aydin and Schnabel 2014). Exploring
a new way of story-telling for mass-housing projects,
it demonstrates the flexibility of shape grammars as
well as their readability by lay-persons. The levels of
'gamification of shape grammars' are as following:
1. Context: A design without context loses sig-
nificance. As a design research, this research
starts with the gathering of contextual infor-
mation. The designer does not have to be
on site to collect information but a gamified
social web-platform enables everyone who is
concerned with the social aspect of the design
to contribute to constant data collection. The
context of the Kashgar study is its old-town
which is a very popular destination for the ad-
venturous. Massive information is to be easily
identified through the contribution of these
people who are concerned with saving Kash-
gar as well as their still-fresh stories.
2. Site Analysis: This research is about mass-
housing design proposed to be as transpar-
ent as possible. Apart from narrative and pho-
tographic dataset collected through its online
platform, the research requires analysing the
site more precisely. The site is already under
demolition, whereas the fifteen per cent of
the old-town is planned to remain as an open-
air architectural exhibition. Site analysis is to
aggregate on the content collected at the pre-
vious step so that a feedback cycle can be pa-
rameterised inside the loop.
3. Shape Grammars: Having collected various
data sets, shape grammar rule sets are to be
defined in this stage. The end-target is to
achieve three types of shape grammars; an-
ticipated, possible and unanticipated (Knight
608 | eCAADe 32 - Digital Heritage 1 - Volume 1
and Stiny 2001). The first is meant to be a
result that is reached with pre-defined rules
while the possible shape grammars refer to
as the use of additional rule sets. And the
last one, the unanticipated shape grammars,
is to search new designs that do not necessar-
ily rely on the site analysis, thus rule sets.
4. Initial Design: In order to initialise the game,
a seed is required. Initial design is provided
by the designer who is in charge of the shape
grammar analysis.
5. The Game: Engagement, motivation,
decision-making, negotiation and anything
that may be possible in participative mass-
housing design take place in this stage. Gam-
ification elements are exploited at upmost
but efficiently. The actual framework of this
section is the focus of the next stage in this
research.
6. Optimisation: It is a genetic algorithm that
maps the optimisation of decided plans with
a pareto algorithm. In order to test the poten-
tials of collective intelligence, this stage can
be further developed in a way optimisation
can be searched by contribution via an exten-
sion of the game.
7. Last Decisions: At this stage the game pri-
vatises itself only for the dual discussions be-
tween its designers and future-users.
8. Evaluation: Further evaluation is meant to list
the performance of the overall product(s).
4. PARTICIPATORY MASS-HOUSING
Mass-housing projects are carried out for the soci-
ety native to complex relationships from economical
level to aesthetical. Yet, the AEC industry is globally
in such a top-down manner that the occupants of the
society are mostly limited to participate only in the
marketing stage. They can only choose from what
is predefined and select the one that is hardly suit-
able indeed. Instead of what a family may need, it
is at the moment the other way around where fami-
lies have to adapt to the units usually far away from
dealing with family-specific demands. The ideology
of 'having a house' has changed from planning and
designing the house to fit the individual families to
choosing the 'container' units like a product in a mall
that the family could best adapt to. By adopting such
an open-source system and open-collaborative de-
sign strategy, this research examines the need to de-
velop a platform for a relatively bottom-up design ap-
proach that allows the participation of its inhabitants
giving most of the control back to the people.
Open Building is an approach for building de-
sign that was promoted by John Habraken (1961).
Habraken proposed two main domains of actions -
the action of the community and that of the inhab-
itants. Without the individual inhabitant, the result
is usually uniform and brutal, which we can see in
most mass-housing projects nowadays. On the other
hand, the community is necessary however, without
a design control mechanism, resultant spontaneous
decisions may be chaotic. Use of shape grammars of-
fers a control mechanism by means of which partic-
ipants contribute without knowing how it is run but
with a feeling of high-autonomy. Therefore gamifica-
tion breeds fun, engagement and motivation insofar
as to design a platform that increases intelligence by
using the power of collective working, like a win-win
game.
The Ökohaus (Eco-home) is a project conducted
by Frei Otto and Herman Kendell in 1988 for
the Internationale Bau Ausstellung (IBA) exhibition
(www.laciudadviva.org). It is a collective housing
which exercises user participation and open design.
Frei Otto sees this as an opportunity to consider new
ways of living in high-density urban context. Occu-
pants are selected only if they are willing to spend the
time and effort to participate in the design process. In
exchange, the cost of the unit are much lower than an
average house in the city.
Next21 (Kim et al. 1993), is an experimental
housing project that consist of 18 individual housing
units. For this project, the focus is more on the build-
ing system itself instead of the collaboration process
as compared to the previous example. Specific de-
sign strategies are generated from the framework of
two principal concepts, the system building and the
two-stage building. In 1996, they did an experimen-
tal remodeling of one unit with the participation of its
residents and it was a great success (Sasakura 2005).
This provides precedence for the possibility of a par-
ticipation/collaboration design method and also flex-
ible building system for collective housing.
From the two examples, the participatory pro-
cess is made possible not only by the designer but
also the habitants themselves. The architect pre-
pared the framework or infrastructure for the partic-
ipation and provide incentives to attract habitants
to contribute willingly. In the case of Okohaus, Frei
Otto only gave two simple rules, the design had to in-
corporate greenery and that every space must have
enough sunlight. Next21, on the other hand, had unit
elements such as facades and partitions set for the
habitants to mix and match. The gamification is ac-
tually happening in a disguised form which can be
improved to enhance the engagement.
5. PRELIMINARY WORK: GAMIFICATION
WITH QUASI-GRAMMARS
A preliminary work as a part of an optional course
with fifteen MArch students was conducted to see
different attitudes in such a "game" without the con-
straints from the "Gamification of Shape Grammars."
A design environment was set up, examining the re-
sults that granted significant findings for the frame-
work of this research (Table 1).
The objective of this studio was to observe the
struggle between flexibility and control, the conflict
of the top-down versus bottom-up approach. To limit
the variables, this studio addressed only the collabo-
rative design amongst different designers and users
rather than the involvement of the users and the
stakeholders. The aim was to investigate the possible
problems faced when a group of collaborators come
together to design a part of the building.
The purpose was to create an environment that
Digital Heritage 1 - Volume 1 - eCAADe 32 | 609

each individual designer would imagine herself or
himself living in. Under the condition of a fixed plan
layout by the main architect, the designers were each
given a number of unit cubes to fill up the plan (Fig-
ure 2). Each cube did unnecessarily function as a spe-
cific unit, thus being a public space, a void or a green
plot. The designers then worked together to gener-
ate a circulation such that each space would be ac-
cessible. Together with the cubes this rule defined
the quasi-grammars in this work. The main architect
would then collate the data for the designers to move
to the next stage of planning. As they went into de-
tails, they would realize some problems and would
need to shift their cubes around, which would then
require further discussions. After a few rounds of
discussion, the designers had discussions about the
building outlook, i.e. façade. The main architect col-
lated the plans and models to check if there are prob-
lems with the overall model.
The 'grammar' of the building in this case is set by
the main architect (the cubes) while the façade is set
by the designers themselves. However, every façade
design has to follow a main theme set by the main
architects which is 'verticality'. The outcome of the fi-
610 | eCAADe 32 - Digital Heritage 1 - Volume 1
Table 1
Constraints given to
participants.
nal product is a collage of unique façades yet they are
following the required rule.
The final designs were evaluated with respect to
the quasi-grammars. As the design process was very
open ended, the designers could change the con-
straints as long as the whole group agreed and the
main architect's confirmation was given.
By comparing the process and outcomes with
the quasi-grammars, we realise that instead of giving
specific parameters, it will be more efficient for the
main architect to just provide general guidelines and
limitations. However, this would also need the main
architect to keep close inspection of every designer
so that they do not divert too far away from the main
objectives.
Figure 2
Building broken
down into units.
Figure 3
Final product.
6. CONCLUSION
Kashgar's vernacular architecture, its cultural con-
text, local demands, development, political sensi-
tivity, harsh environmental factors and so on, they
all comprise a complex intricacy to design a large-
scale mass-housing. 'Gamification of Shape Gram-
mars' takes the advantage of each of these attractors
within a gamified environment where a participatory
and collaborative design strategy is sought so that
everybody wins. Gamification is nothing more than
a means compared to the design aspect of this re-
search. Efficiently designed, gamification techniques
will be subtly integrated into the loop in which stan-
dard and non-standard information types will be pa-
rameterised with each other. Though predicted, limi-
tations of gamification such as 'pointsification' will be
explained in detail in future works. Briefly, it is to say
that the design of gamification elements play a ma-
jor role in creating an engaging experience which is
the aim of this project. Another of most significant
targets in quasiGRAMMARS is to contribute to digital
heritage studies in the context of Kashgar. Through
gamification techniques, novel presentation meth-
ods will be produced. In this paper, the framework
of the quasiGRAMMARS project is presented. A pre-
liminary study is conducted to draw a draft for the
next stage in Kashgar mass-housing research. Pro-
ceedingly, an online platform for data gathering will
be initiated and field-trips will be carried out as a part
of site analysis in Kashgar.
REFERENCES
Aydin, S 2012, A parametric quasi-grammar for generat-
ing a flexible urban design tool: Shape grammars and
genetic algorithms, Master's Thesis, University of Liv-
erpool
Aydin, S and Schnabel, MA 2013 'Coding Shape Gram-
mars: Hints for generating a parametric design tool
for large-scale urban renewal projects', Proceedings
of the 47th International Architectural Science Associ-
ation Conference (formerly ANZAScA), Hong Kong, p.
177 – 186
Aydin, S and Schnabel, MA 2014 'A Survey on the Visual
Communication Skills of BIM Tools', Proceedings of
CAADRIA 2014, Kyoto
Barros, P, Beirao, JN and Duarte, JP 2013 'The Lan-
guage of Mozambican Slums', Proceedings of the 31st
eCAADe Conference, Delft, pp. 715-724
Beirao, JN 2012, City Maker. Designing Grammars for Ur-
ban Design, Ph.D. Thesis, TU Delft
Beirao, JN and Duarte, JP 2005 'Urban Grammars: To-
wards Flexible Urban Design', Proceedings of the 23rd
Conference on Education in Computer Aided Architec-
tural Design in Europe, eCAADe, Lisbon
Benros, D, Duarte, JP and Branco, F 2007 'A System for
Providing Customized Housing', Proceedings of the
12th International Conference on Computer Aided Ar-
chitectural Design Futures, Sydney
Caldas, LG and Santos, L 2012 'Generation of Energy-
Efficient Patio Houses with GENE_ARCH: Combining
an Evolutionary Generative Design System with a
Shape Grammar', Proceedings of the 30th Conference
on Education in Computer Aided Ar-chitectural Design
in Europe, eCAADe, Prague
Coimbra, E and Romão, L 2013 'The Rehabilitation De-
sign Process of the Bourgeois House of Oporto:
Shape Grammar Simplification', Proceedings of the
31st eCAADe Conference, Delft, pp. 677-685
Deterding, S, Dixon, D, Khaled, R and Nacke, L 2011 'From
Game Design Elements to Gamefulness: Defining
“Gamification”', Proceedings of the 15th International
Academic MindTrek Conference, (MindTrek '11), Tam-
pere, pp. 9-15
Dong, W and Zhang, X 2011, 'Urumqi', Cities, 28, pp. 115-
125
Duarte, JP, Beirao, JN, Montenegro, N and Gil, J 2012,
'City Induction: A Model for Formulating, Generat-
ing, and Evaluating Urban Designs', in Arisona, SM,
Aschwanden, G, Halatsch, J and Wonka, P (eds) 2012,
Digital Heritage 1 - Volume 1 - eCAADe 32 | 611
 Digital Urban Modeling and Simulation, Springer,
Berlin Heidelberg, pp. 73-98
Duarte, JP, Rocha, JM and Soarez, GD 2007, 'Unveiling
the structure of the Marrakech Medina: A shape
grammar and an interpreter for generating urban
form', Artificial Intelligence for Engineering Design,
Analysis and Manufacturing, 21, p. 317–349
Fabien, EH, Janssen, P and Lo, TT 2013 'Group Forming:
Negotiating Design Via Web-Based Interaction and
Collaboration', Proceedings of CAADRIA 2013, Singa-
pore
Florenzano, M, Courel, MF and De Domenico, F 2010,
'Digitally Conserving an Endangered Built Heritage
in Kashgar, an Oasis City of the Taklimakan', in
Schneier-Madanes, G and Courel, MF (eds) 2010, Wa-
ter and Sustainability in Arid Regions, Springer, pp.
165-179
Haegler, S, Muller, P and van Gool, L 2009, 'Procedural
Modeling for Digital Cultural Heritage', Journal on
Image and Video Processing, 2009, pp. 1-11
Halatsch, J, Kunze, A and Schmitt, G 2008, 'Using Shape
Grammars for Master Planning', in Gero, JS and
Goel, AK (eds) 2008, Design Computing and Cogni-
tion, Springer, p. 655–673
Ham, D 2013, 'Restructuring Beginning Design Curricu-
lums with Visual Calculation', in Botía, JA and Char-
itos, D (eds) 2013, Workshop Proceedings of the 9th
International Conference on Intelligent Environments,
IOS Press BV
Hudson-Smith, A 2014, 'Tracking, Tagging and Scanning
the City', Architectural Design, 84, pp. 40-47
Huizinga, J 1955, Homo Ludens: A Study of the Play-
element in Culture, Routledge & Kegan Paul
Kapp, KM 2012, The Gamification of Learning and
Instruction—Game-Based Methods and Strategies for
Training and Education, John Wiley & Sons, San Fran-
cisco
Kelly, A 2003, Decision Making Using Game Theory :
An Introduction for Managers, Cambridge University
Press, Cambridge
Kim, J, Brouwer, R and Kearney, K 1993, NEXT 21: A
Prototype Multi-Family Housing Complex, Ann Arbor,
Michigan: University of Michigan, College of Archi-
tecture and Urban Planning
Knight, T and Stiny, G 2001, 'Classical and non-classical
computation', arq: architectural research quarterly,
5(4), pp. 355-372
Lo, TT, Aydin, S and Schnabel, MA 2014 'Collabora-
tive Design with Quasi-Grammars', Proceedings of
CAADRIA 2014, Kyoto, p. 941–942
612 | eCAADe 32 - Digital Heritage 1 - Volume 1
Lo, TT and Schnabel, MA 2013 'Definition of Smart Para-
metric Model for Collaborative Design of Mass Hous-
ing', Proceedings of the 47th International Conference
of the Architectural Science Association (ANZAScA)
Michell, G, Vicziany, M, Yeshu, T and Gollings, J 2008,
Kashgar: Oasis City on China's Old Silk Road, Frances
Lincoln, London
Narahara, T and Terzidis, K 2006 'Multiple-constraint Ge-
netic Algorithm in Housing Design', Proceedings of
the 25th Conference of the Association for Computer-
Aided Design in Architecture World
Parish, Y and Muller, P 2012 'Procedural modelling of
cities', Proceedings of SIGGRAPH '01: The 28th An-
nual Conference on Computer Graphics and Interac-
tive Techniques, p. 301–308
Ryan, RM and Deci, EL 2000, 'Self-determination theory
and the facilitation of intrinsic motivation, social de-
velopment, and well-being', American Psychologist,
55(1), pp. 68-78
Sasakura, H 2005 'Variable Infill System Rearrangement
Experiment for Residence 405 at Osaka Gas Experi-
mental Housing', Proceedings of the 2005 World Sus-
tainable Building Conference
Starr, SF 2013, Lost Enlightenment: Central Asia's Golden
Age from the Arab Conquest to Tamerlane, Princeton
University Press
Stiny, G and Gips, J 1972, 'Shape Grammars and the Gen-
erative Specification of Painting and Sculpture', In-
formation Processing, 71, p. 1460–1465
Werbach, K and Hunter, D 2012, For the win: How game
thinking can revolutionize your business, Wharton
Digital Press, Philadelphia
Zichermann, G and Linder, J 2013, The Gamification Rev-
olution: How Leaders Leverage Game Mechanics to
Crush The Competition, McGraw-Hill
[1] http://pennystocks.la/internet-in-real-time/
[2] http://www.rfa.org/english/multimedia/
kashgardemolition-03232010120925.html
[3] http://www.guinnessworldrecords.com/
world-records/3000/largest-lan-party
[4] http://www.dreamhack.se/DHS11/event-information/
what-is-dreamhack/
[5] http://fold.it/portal/
[6] http://phylo.cs.mcgill.ca/
[7] http://www.laciudadviva.org/blogs/?p=14164
B(H)IM - Built Heritage Information Modelling
Extending BIM approach to historical and archaeological heritage
representation

Davide Simeone1 , Stefano Cursi2 , Ilaria Toldo3 , Gianfranco Carrara4

1,2,3,4
Sapienza University of Rome
1,2,3,4
http://www3.dicea.uniroma1.it
1,2,4
{davide.simeone|stefano.cursi|gianfranco.carrara}@uniroma1.it
3
ilaria.toldo@gmail.com

This paper describes the context and the proposal for the extension of Building
Information Modelling to built heritage in order to enhance information
management during the investigation and restoration activities. The core of the
presented model is the integration of a BIM-based modelling environment and a
knowledge base developed by means of ontologies, in order to represent all the
semantics needed for a comprehensive representation of the historical artefact.To
test its features, the model has been applied to the real archaeological
investigation process of the Castor and Pollux temple at Cori, Italy.

Keywords: BIM, Built Heritage, Ontology-based systems, Knowledge

management, Archaeological investigation

INTRODUCTION: INFORMATION MANAGE- blurred-outlined knowledge domain and any lack of

MENT AND BUILT HERITAGE INVESTIGA-
TION
In Built Heritage investigation, the accuracy and the
completeness of the artefact knowledge representa-
tion is a key factor that deeply influences the follow-
ing activities of investigation, intervention and main-
tenance: in fact, any decision made by the different
actors involved is based on "what is known" of the
object, and any lack of knowledge or inconsistence
can lead to errors and even irreparable damages.
Differently from the construction industry,
where the building and all its elements are very
clearly modelled in all their components and fea-
tures, in this field the heritage artefact is usually a
documentation is considered to be equivalent to the
destruction of discovered remains. In addition, un-
derstanding an heritage artefact is more than under-
standing just its history, physical configuration and
condition; It involves a large amount of knowledge
about a high number of social, political, economic,
and cultural issues relating to the external environ-
ment (Letellier et al, 2007).
Every heritage object can also be seen as a
unique historical archive, a precious source of pri-
mary information from which any research, investi-
gation or conservation activity will retrieve new data
(Bedford, 2009). This information is collected, docu-
mented and made virtually available by the different
professionals. Nevertheless, this knowledge is usu-

Digital Heritage 1 - Volume 1 - eCAADe 32 | 613

 ally inaccessible since it is spread among different
disciplines and restricted to several not-integrated
"knowledge islands"; useful data about the artefact
and its context - potentially able to support the pro-
cess of interpretation and intervention- are buried in
tons of different data repositories. Even when part
of this knowledge is introduced in the investigation
and intervention phases, this is not effectively rep-
resented in the Built Heritage Model and, therefore,
is not fully accessible or usable by the specialists in-
volved in the different process phases. As a result,
useful data about the artefact, its context and the
possible methodologies of interpretation and inter-
vention are buried in tons of documents, databases
or data repositories.
Even the introduction of digital technologies to
the built heritage field has left this problem unsolved:
while several efforts have been made to develop
virtual reality techniques oriented to building origi-
nal appearance simulation or knowledge acquisition
technologies, very little research has been focused
on heritage knowledge modelling and representa-
tion (Kalay et al., 2007). Some approaches (i.e. the
Virtual Reality GIS -VRGIS) have started to integrate
Virtual Reality models with GIS systems (Dore et al.,
2012) but, at present, they are only partially able to in-
clude not-geometrical data in the artefact represen-
tation.
To solve these shortcomings, the research pre-
sented in this paper aims at extending the Build-
ing Information Modelling approach to built her-
itage field, providing a modelling environment able
to effectively support knowledge representation and
management for historical/archaeological artefacts.
In fact, while information about historical/ar-
chaeological artefacts is currently disseminated in
different data repositories with a relevant issue in
terms of partiality and data duplication, BIM can con-
centrate and organize such data in a single knowl-
edge base that acts as an electronic reference dossier
of the object during the entire process of investi-
gation and conservation. Since a large amount of
knowledge related to the object is independent from
614 | eCAADe 32 - Digital Heritage 1 - Volume 1
the artefact but still necessary to be introduced in
its representation (i.e. historical context, social in-
formation, environmental resources, other heritage
artefacts information, etc.), we have decided to inte-
grate Building Information Modelling with a knowl-
edge management system based on ontologies. In
this way, it is possible to enclose several crucial in-
formation that would be difficult to represent in the
component-based BIM approach, providing a more
comprehensive and accessible model of the artefact.
In order to test the real effectiveness of the pro-
posed modelling approach, we have chosen to apply
the BIM+ontologies system - that we defined as Built
Heritage Information Modelling- to the process of in-
vestigation and documentation of the roman temple
of Castor and Pollux in Cori, Italy (Figure 1, 2).
STATE OF THE ART
Over the past decade, a wide number of researches
concerned with the digital acquisition of exist-
ing buildings and the construction of 3D models
integrated with different heritage documentation
databases. Most recently, attention has been moved
towards the development of information manage-
ment instruments considered suitable to the analysis
and maintenance of historical artefacts.
This interest can be due to the recognition of
Building Information Modeling as a technology that
can improve the representation of the relationships
between tangible and intangible heritage knowl-
edge.
Despite its widespread adoption for the design
and management of new buildings, few projects
have looked at Building Information Modeling as a
technology that can improve the representation of
the relationships between tangible and intangible
heritage knowledge.
In fact, while the purpose of a 3D model is limited
to a geometrical visualization, a BIM representation
contains all the data associated to the elements that
make up the artefact, with the result of an instrument
which can allow different kinds of reasoning.
Figure 1
Case study:
Corinthian Temple
of Castor and
Pollux, 1st century
B.C., Cori, Italy.
Remains of the
original structure:
two Corinthian
columns, 10 meters
high, decorated by
20 grooves which
support an
architrave (left);
central cella and
side wings (right).

Figure 2
Heterogeneity of
knowledge related
to the data
acquisition,
modelling and
representation
process: metric
survey (left); Luigi
Rossini, sketching
of the temple of
Castor and Pollux’s
ruins, 19th century
(right).

Digital Heritage 1 - Volume 1 - eCAADe 32 | 615

 In spite of the apparent disinterest, BIM in her-
itage conservation field is not without past instances.
Recent researches have focused on modelling his-
torical buildings in order to create components li-
braries from the laser scanning to the BIM environ-
ment (Arayici, 2008). Penttila et al. (2007) provided
a case study that evaluates utilization of BIM for the
retrofit of historical buildings by means of an "inven-
tory model" as database about both past and present
of an existing building condition. Nevertheless, these
experiments have been focusing more on database
creation and extension rather than considering BIM
as a documentation hub to support and enhance
specialists' activities and collaboration. Some au-
thors (Attar et al., 2010; Fai et al., 2011) explored the
potential of BIM in lifecycle management and sim-
ulation such as energy saving and fire evacuation.
Besides explicit semantic description, BIM also facili-
tates the theoretical and historical study via historical
documents enrichment (Pauwels et al., 2008).
In built heritage field, the process of cataloging
is often done with the creation of large databases
mainly characterized by a vast heterogeneity related
to the typology of media and formats used, to their
level of accessibility, and to the related logical mod-
els of representation (Volk et al., 2014).
This clashes with the limited amount and the low
level of semantics that can be attributed to the ob-
jects by means of current BIM platforms: the mere
description of their features through the use of a list
of properties generates information which is often re-
dundant or overly simplified in many separate slots,
highlighting the inability of such data models to rep-
resent logical associations among entities with other
sections of data related to different disciplines. This
aspect clashes with the requirement that, as stated
before, the information related to the cultural her-
itage have, by nature, a strong need for correlation
in order to fully express their true worth.
For this reason, integration of BIM with informa-
tion management techniques like semantic reason-
ing is essential to enhance representation of archae-
ological sites knowledge, requiring a modelling en-
616 | eCAADe 32 - Digital Heritage 1 - Volume 1
vironment in which data and entities are linked in
a relational system, revealing some information that
would otherwise remain "hidden".
Such considerations lead to a possible solution in
a modelling approach that meets the different needs
listed above and plays the role of representing the se-
mantics of objects. At present, ontology-based sys-
tems allow the representation of entities not only
through the description of their properties but also
by formalizing the relations that exist between them
(Di Mascio et al., 2013; Carrara et al., 2009; Beetz,
2005).
On these bases, we propose a BIM model inte-
grated with a knowledge base developed by means
of ontologies, in order to enclose a sufficiently accu-
rate and computable formalization of the knowledge
related to the artefact and its components, and to
expand current BIM abilities of inference and rules-
based reasoning on this knowledge structure.
BIM AND ONTOLOGIES FOR A COMPRE-
HENSIVE REPRESENTATION AND DOCU-
MENTATION OF THE BUILT HERITAGE
A correct representation of an historical artefact re-
quires a large amount of information that are pro-
gressively created during the different phases of the
investigation process and stored in dedicated data
repositories. The investigation operations are usually
not synchronized in time and often several years are
needed to reach a sufficient comprehension and doc-
umentation of the artefact. For this reason, the infor-
mation management platform and protocols play a
key role in the whole investigation process, allowing
the different operators to provide, access, edit, man-
age, use and update such information in accordance
to their role, task and expertise.
In AEC (Architecture Engineering and Construc-
tion) field, BIM has taken on this task, providing a
modelling environment where geometrical informa-
tion and some not-tangible data are structured in a
coherent database. At the same time, its component-
based perspective still leaves out of the model a large
amount of knowledge, not directly relatable to arte-
Figure 3
The
conceptualization
of Built Heritage
Information Model:
the BIM
environment and
the knowledge
base are integrated
to represent the
data collected
during the
investigation
(SKB-Survey
Knowledge Base)
phase and derived
from the actors
interpretation (ISB - fact's components but still needed to fully represent
Interpretation the object. For this reason, the BIM environment has
Knowledge Base). been integrated with an ad hoc knowledge base de-
veloped by means of ontologies. This database, inte-
grate with an inference engine, allows to include and
make computable all the knowledge needed in the
investigation process and provided by the different
actors (Figure 3).
In the investigation process supported by the
Built heritage Information Model, the first logical step
is the discretization and identification of the differ-
ent components of the artefact to be progressively
enriched with the data collected and provided by
the operators. This phase can be mainly supported
by the BIM environment, creating classes and in-
stances of the different components, making explicit
their properties and creating hierarchical relation-
ships among them (part-of, is-a).
Differently from the AEC industry where each el-
ement is defined and modeled in all its parts and
properties, in built heritage often several compo-
nents are not clearly identifiable (especially at the be-
ginning of the investigation process). To overcome
this problem, some classes have been developed in
order to represent "unknown objects" in terms of
their geometric and material features, waiting for
a later interpretation that should associate them to
the actual classes. In this way, each actor can see
which entities are still unidentified and, on the bases
of their knowledge and experience, provide sugges-
tions for their interpretation gradually improving the
completeness of the model of the artefact.
During the site survey phase, a wide amount of
data is associated to each of these entities in terms
of properties, values and relationships. This knowl-
edge (defined as Survey Knowledge Base - SKB) rep-
resents all the information that can be directly gained
through the different investigation activities depend-
ing on the different disciplines considered (for in-
stance dimensions, morphology, topology, materi-
als information, temperature, deterioration rate, car-
bon dating, etc.). When an element is associated to a
class, it inherits all the properties that it is necessary
to provide in order to reach an adequate level of rep-
resentation, helping the operators in understanding
which information they have to collect from the site
and which ones are missing.
In addition, Building Information Modelling pro-
vides a single database where the actors can add and
edit data collected along the entire investigation pro-
cess. As a result, the information modelled is always
up-to-date and the model itself can be considered as
a track record of the results of investigation and inter-
pretation activities.
Digital Heritage 1 - Volume 1 - eCAADe 32 | 617
 In the presented case study, the Dioscuri's tem-
ple and its elements have been modelled in the Au-
todesk Revit environment; geometrical data were
mainly derived through a survey campaign carried
out in a traditional way (hand measuring and pho-
togrammetry) but supported by an information man-
ager (Figure 4). This shows how the Built Heritage In-
formation Model is useful to support also traditional
investigation processes, without forcing changes in
specialists' work methods but only integrating in the
platform the data resulting from their operations.
The following survey activities have focused on
other physical aspects of the artefact such as mate-
rials diagnostics and environmental data collection
in order to provide information for the later restor-
ing interventions. All this data were modelled in the
BIM environment quite easily, developing classes for
each element type (columns, walls, stairs, etc.) and
adding as properties and rules as needed. Since this
was a first case study, the entire elements inventory
was created from the beginning but when a standard
template has been created, the same inventory can
be reused in other survey campaigns, deeply reduc-
618 | eCAADe 32 - Digital Heritage 1 - Volume 1
ing the related amount of work.
After modelling the current state of artefact, a
historical investigation has been carried out in order
to gain useful data from different archives, modelling
different configurations of the area in accordance to
the different historical periods.
In fact, built heritage usually bears with its his-
tory a lot of morphology mutations, use changes
and even destructions or relocation of its parts. For
instance, in this case study some columns where
moved from their original position during time and
reused in the medieval age as part of new walls. This
aspect increases the complexity of the artefact and of
its representation, since its different states need to be
represented and visualized. For this purpose, a spe-
cific set of properties has been implemented in order
to represent if the element is in place or not, the age,
the period of realization, and the different phases of
the heritage object life it was in.
Although sometimes the data can be collected
on site with a high degree of accuracy, the Survey
Knowledge Base is not sufficient to provide a com-
prehensive representation of a built heritage object.
Figure 4
The Dioscuri’s
temple modelled in
the BIM
environment of
Autodesk Revit. For
each component
identified on site, a
set of properties
and values has
been implemented
in order to enrich
the artefact
representation with
not-geometrical
data (i.e. material,
historical period,
degradation level,
etc.).
 For this reason, in fact, it is crucial to include, in the
model, immaterial information that can document
others of its aspects such as its evolution during time,
its historical, social and technological context, its in-
tended use, materials caves and sources, etc.
This knowledge base is derived by other knowl-
edge domains and, when introduced and made ac-
cessible in the proposed model, it can be used to sup-
port the following actions of interpretation, investi-
gation and intervention. For instance, information
about the historical period and the pits available at
that time can help in understanding materials typol-
ogy and construction techniques, while social con-
text data can provide useful information for identify-
ing possible uses of spaces. In order to include exter-
nal knowledge in the artefact model, we have cho-
sen to integrate the BIM representation environment
with a knowledge management system based on on-
tologies.
An ontology, as defined in the ICT field, is
an explicit specification of concepts that includes
within the same descriptive system the concepts of a
knowledge domain and the relations between them
(Gruber, 1993); and only recently those tools have
been proposed as a method to formalize immaterial
knowledge about cultural heritage (Pauwels, 2008).
In the proposed model, ontologies allow struc-
turing and managing concepts related to the differ-
ent domains that are needed to fully comprehend
the historical artefact, and any kind of data and doc-
umentation that can be useful for its interpretation
such as external links, textual documents, images,
modeled objects, bibliographic references, etc. (Fig-
ure 5).

Figure 5
A building
component
modeled by means
of the
ontology-editor
“protégé”. The
entity
‘Fusto’(Part_Of
_Column) has been
formalized with a
set of properties
related to the
historical
documentation of
the Dioscuri’s
temple and
provided by the
different specialists
during the
investigation
process.

Digital Heritage 1 - Volume 1 - eCAADe 32 | 619

 In particular, we chose to rely on an ontology-
based model oriented to the representation of enti-
ties (concrete or abstract; part of the artefact or ex-
ternal) in terms of three knowledge components de-
fined as meanings, properties and rules. This tem-
plate has been previously developed and tested by
the research group to provide a BIM-based mod-
elling environment - defined Building Knowledge
Modeling-, in which actors can share not only data
but also all related concepts and all knowledge at-
tached to any entity involved in new buildings design
process (Carrara et al., 2009).
The first component - the meaning- provides a
domain-dependent description of the entity; prop-
erties, instead, represent all the descriptive aspects
related to the concerned element (e.g.: geometrical,
physical, historical, technological features); at last,
Rules represent the connections among entities, by
expressing relationships or reasoning links.
MPR template allows a dynamic and recursive
approach to the formalization of knowledge being
as effective in the representation of simple entities
as in those of greater complexity, focusing the de-
scription of concepts on the analysis of their struc-
ture and effective representation and, therefore, im-
proving cooperation among the actors of the investi-
gation process. This structure allows not only to effi-
ciently define and manage the knowledge belonging
to the built heritage modeled entities but also to per-
form activities of reasoning about the concepts rep-
resented.
In the ontology-based system, an inference en-
gine checks the formalized entities and the relation-
ships among them, assuring the consistency and
the coherence of the information represented in the
model. In addition, some data are automatically in-
ferred by executing operations and algorithms on the
introduced properties values, providing a complete
knowledge base to support actors' decisions and op-
erations. This is very useful in archaeological heritage
investigation, since many historic layers usually over-
lap, many elements are re-used with different func-
tions over time and several interpretation inconsis-
620 | eCAADe 32 - Digital Heritage 1 - Volume 1
tences can emerge.
As well known in information management prac-
tice, the availability of large amount of information is
not sufficient to assure collaboration among different
specialists: integrating all this semantics in a single
model could even be counter-productive, since the
needed information can be hidden in this complex
system of entities and data. For this reason, a seman-
tic filter has been implemented in the modeling envi-
ronment, letting each actor access, edit and visualize
only the information that he needs in accordance to
his field of expertise and its tasks in the heritage pro-
cess.
In the proposed case study the knowledge base
has been created by means of the ontology editor
Protégé [1], representing each entity through the
Meaning-Properties-Rules structure. In this way, all
the information related to the historical documen-
tation of the Dioscuri's temple has been embedded
in the model of the artefact. Paintings and pictures
from the first excavation phase (1940-50), paintings
and drawings representing the heritage site in differ-
ent times (as the one shown in fig.2), ancient texts
about the temple and the city of Cori were modelled
and made accessible in the knowledge base, provid-
ing a comprehensive representation of the artefact.
In order to connect the BIM environment with
the knowledge base, particular attention has to be
given in creating a correspondence between the on-
tologies structure and the entities network in the BIM
environment. For this reason, specific relationships
(i.e. Is-Part-of for representing encapsulation of enti-
ties, or Is-A to formalize features inheritance between
father-son classes) were implemented in the knowl-
edge base.
In our experiment, the bridge between the BIM
and the ontology-based databases has been imple-
mented by means of the Revit DBLink plugin.
Differently from previously knowledge manage-
ment models, the integration of the ontology-based
model with the Building Information Modeling en-
vironment allows to have at the same time a geo-
metrical representation of the artefact and its com-
ponents, and all the semantics (material and immate-
rial information) provided in both the survey and in-
terpretation phases. This database comprehensively
represents "what is known" of the object and "what is
needed to know" for further proceed with the process
of investigation, conservation and communication.
CONCLUSIONS
In this paper, we presented a knowledge modelling
approach aimed at extending the Building Infor-
mation Modelling approach to represent built her-
itage artefacts in order to support the information
management during investigation and restoration
phases. Its structure, developed by integrating a
BIM environment with an ontology-based system, al-
lows accurate representation and effective integra-
tion of all the knowledge related to the artefact, ei-
ther gained from the archaeological survey or de-
rived from other knowledge areas. Similarly to its im-
pact on the AEC sector, Building Information Mod-
elling introduction into the Built Heritage field sup-
ports the management of the information collected,
modelled, used and provided by the different actors
involved in the investigation/conservation process.
In fact the Built Heritage Information Model improves
the availability and accessibility of all the knowledge
related to a historical/archaeological artefact, mak-
ing easier to interpret its nature, monitor its changes
and document each investigation and intervention
activity. As a result, more aware decisions can be
made by relying on the knowledge formalized in the
proposed model in order to effectively identify emer-
gency situations, schedule intervention activities and
plan routine management and maintenance.
REFERENCES
Arayici, Y 2008, 'Towards building information modelling
for existing structures', Structural Survey, 26(3), pp.
210-222
Attar, R, Prahbu, V, Glueck, M and Khan, A 2010 '210 King
Street: A Dataset for Integrated Performance Assess-
ment', SimAUD 2010 Conference Proceedings, Sympo-
sium on Simulation for Architecture and Urban Design,
Orlando, Florida, pp. 27-30
Bedford, J, Blake, B, Bryan, P, Andrews, D, Barber, D and
Mills, J 2009, Metric Survey Specifications for Cultural
Heritage, English Heritage, London
Beetz, J, Van Leeuwen, JP and de Vries, B 2005 'An ontol-
ogy web language notation of the industry founda-
tion classes', Proceedings of the 22nd CIB W78, Con-
ference on Information Technology in Construction,
Dresden, Germany, pp. 193-198
Carrara, G, Fioravanti, A, Loffreda, G and Trento, A
2009 'An Ontology-Based Knowledge Representa-
tion Model for Cross-disciplinary Building Design -
A General Template', Computation: the new Realm
of Architectural Design, Proceedings of eCAADe Con-
ference, 27th eCAADe Conference Proceedings, Instan-
bul, Turkey, pp. 367-373
Dore, C and Murphy, M 2012 'Integration of Historic
Building Information Modeling and 3D GIS for
Recording and Managing Cultural Heritage Sites',
18th International Conference on Virtual Systems and
Multimedia: “Virtual Systems in the Information Soci-
ety, Milan, Italy, pp. 369-376
Fai, S, Graham, K, Duckworth, T, Wood, N and Attar, R
2011 'Building information modeling and heritage
documentation', Conference Proceedings of XXIII CIPA
International Symposium, Prague, Czech Republic, p.
8
Gruber, T 1993, 'A translation approach to portable on-
tology specifications', Knowledge Acquisition, 5(2),
pp. 199-220
Kalay, YE, Kvan, T and Affleck, J 2007, New Heritage – New
Media and Cultural Heritage, Taylor & Francis Group,
London & New York
Letellier, R, Schmid, W, LeBlanc, F, Eppich, R and Chabbi,
A 2007, Documentation, and Information Manage-
ment for the conservation of Heritage Places. Guiding
Principles, The Getty Conservation Institute, Los An-
geles, US
Di Mascio, D, Pauwels, P and De Meyer, R 2013 'Improv-
ing the knowledge and management of the histori-
cal built environment with bim and ontologies: the
case study of the book tower', Proceedings of the
13th International Conference on Construction Appli-
cations of Virtual Reality, London, UK, pp. 427-436
Pauwels, P, Verstaeten, R, De Meyer, R and Van Camp-
enhout, J 2008 'Architectural Information Modelling
for Virtual Heritage Application', Digital Heritage –
Proceedings of the 14th International Conference on
Virtual Systems and Multimedia, Lymassol, Cyprus,
pp. 18-23
Penttilä, H, Rajala, M and Freese, S 2007 'Building Infor-
mation Modelling of Modern Historic Building', Pre-
Digital Heritage 1 - Volume 1 - eCAADe 32 | 621
 dicting the Future - 25th eCAADe Conference Proceed-
ings, Delft, Netherlands, pp. 607-613
R, Volk, Stengel, J and Schultmann, F 2014, 'Building In-
formation Modeling (BIM) for existing buildings —
Literature review and future needs', Automation in
Construction, 38, pp. 109-127
[1] http://protege.stanford.edu

622 | eCAADe 32 - Digital Heritage 1 - Volume 1

Digital fabrication as a tool for investigating traditional
Chinese architecture
A case study of the dou gong

Di Li1 , Michael Knight2 , Andre Brown3

1,2,3
University of Liverpool, UK
1,2,3
{di.li2|mknight|andygpb}@liverpool.ac.uk

This paper presents part of a broader research project in which a set of

complementary computational techniques have been applied to investigate and
interpret traditional Chinese architecture. The techniques involve digital
modelling, algorithmic representation and digital fabrication. The ancient rule
books that describe the traditional Chinese design styles and construction
technologies are the basis for the parametric rules and algorithms and the
application to the modelling and fabrication process. The focus of this paper is
the study of a typical Chinese joint structure, the dou gong. The precedent case
study and the components of the dou gong are reviewed and analysed. Then the
digital representation and fabrication processes that have been employed are
demonstrated. Practical problems were found in this process and modifications
were consequently made. At the end of the paper, we discuss the achievements
and benefits that arise from this investigation, and potential applications in
practice. Conclusions from work presented in this paper and for the broader
research are drawn respectively. Significant prospective work is suggested.

Keywords: dou gong, Ying Zao Fa Shi, digital fabrication, traditional Chinese
architecture

INTRODUCTION With the rapid development of digital tech-

China is one of the Four Great Ancient Civilizations
which has a span of over 5000 years and its traditional
Chinese architecture plays a significant role in the
world's architectural heritage. Unfortunately, almost
all the buildings constructed before the Tang Dynasty
(618-907) have disappeared. However there are
some buildings from the period from the Song Dy-
nasty (960-1279) to the Ching Dynasty (1616-1912)
remaining, but in differing states of repair.
niques in architecture over the last 25 years, it is possi-
ble to re-discover the value of the lost or damaged ar-
chitecture by transforming two dimensional physical
information recorded in the historical literature into
three dimensional digital representations (Webb and
Brown, 2011). Using parametric and shape grammar
techniques (Li, 2001; Wu, 2005) it has been shown
possible to describe the relationships in a digital de-
sign environment that describe the traditional de-

Digital Heritage 1 - Volume 1 - eCAADe 32 | 623

 signs in an algorithmic way (Li et al. 2013). Conse-
quently, further analysis can be made together with
a better level of understanding and interpretation.
In turn, this can significantly aid understanding, in
restoring and recreating examples of traditional Chi-
nese architecture, or in designing contemporary ex-
amples that respect traditional geometries and val-
ues.
This paper is part of a broader research investiga-
tion into the digital interpretation and recreation of
traditional Chinese architecture. We do so by turn-
ing ancient rule book information into algorithms
and applying these in a contemporary parametric
modelling environment. This paper concentrates on
the understanding, representation and digital fabri-
cation of a typical Chinese joint structure, the dou
gong. This is a particularly complex aspect in the de-
sign of traditional Chinese architecture, as there are
many elements to the joint that have a sophisticated
relationship with each other, as shown in Figure 1
(Liang, 1985), and there are many variations accord-
ing to the scale, location, function, decoration and
624 | eCAADe 32 - Digital Heritage 1 - Volume 1
historical period. The joint plays a multi-functional
role that covers structure and construction require-
ments as well as being an essential component of
the aesthetic qualities of the whole construction. As
Liang (1985) points out, "the dou gong plays the lead-
ing role, a role so important that no study of Chi-
nese architecture is feasible without a thorough un-
derstanding of this element, the governing feature of
the Chinese 'order'." Figure 1
dou gong (Liang,
1985)
CASE STUDY OVERVIEW AND PRECEDENT
There are the only two surviving ancient rule-books:
Ying Zao Fa Shi (Building Standards) from Song Dy-
nasty (960-1279) and Ching Dynasty: Kung-ch'eng
tso-fa tse-le (Structural Regulations) from Ching Dy-
nasty (1616-1912), which are referred to as the "two
text books of Chinese ancient architecture" by Liang
(1985). According to Li (2001), Ying Zao Fa Shi is rule-
based and indicates parametric characteristics. The
joint, called the dou gong, is described in one of the
carpentry volumes in Ying Zao Fa Shi (Li et al. 2013),
which is the bracket set on top of the columns and
under the beams and the roof.
To understand the dou gong parametrically, two
approaches can be taken. First, the dou gong is
treated as a whole rather than focusing on each in-
dividual element and the modelling process concen-
trates on building up the whole structure and set-
ting up the position of all the elements. Then de-
tailed elements are dealt with later. This is the top-
down method. The second involves the construction
of each element first and then connecting them to-
gether. This is the bottom-up method and the con-
centration of this paper, which will be discussed in
detail in the following sections.
Following the first approach, a graphical rep-
resentation is made as shown in Figure 2 using
Grashopper. Since the main structure of dou gong is
doubly symmetric, in order to construct the model ef-
ficiently, one fourth of a complete dou gong is built
up. By mirroring twice, a complete dou gong could
be carried out. This is a conceptual model which
is based on a commonly used concept of the tree
 branches. Several layers of the tree branches grow rameters. This enables the possibility of the achieve-
and expand outwards, which generates the extended ment of the aesthetic function of the joint, that is, to
tree canopy. In this process, the iterative layers and treat the joint as a pure classic decoration. And in-
extension of each layer are the key factors for sim- deed, since the Ming Dynasty, the structural function
ulating the bracket set. To achieve this, first of all, has become less important due to the development
a series of points are set up along the Z axis. This of other constructional technologies while the aes-
provides the initial partition of the layers, which are thetic value became more emphasized.
the basis for gong (the long arms). The number of On the other hand, several problems were ob-
the layers is controlled by an input slider which can served. The accurate shape of the joint is not repre-
be changed based on the need of layers. The Graph sented well in the experiment; as well as its volume,
Mapper is an effective method in the controlling. an extra set of logic relationship which defines the
Then the parabola function is selected to generate all precise shape and volume of the joint is needed. This
the points locating the tree branches (representation is necessary to guarantee the achievement of other
of gong). The strongpoint of parabola function is to important functions of the joint (which will be intro-
be able to easily define all the layers together. But duced in the following sections). In addition, the ac-
from the later process, it is found that this might lead curacy of each element of the joint is also omitted.
to the loss of control of the branches shape within Apart from these matters, there is a key component
each layer. To improve this shortage, the Gaussian which acts as a lever in the function, omitted from
function is introduced to replace the parabola func- this model. The above problems lead to the adop-
tion in this process. As shown in Figure 3, the shape tion of the second approach described in the follow-
of the branches is closer to the gong described in the ing sections.
rule book. Following this, a rectangular grid of points
Figure 2 is set up within each layer. The area of layers reduced
The top-down gradually from the top to the bottom alone the Z axis.
approach of dou Here Path Mapper is used to arrange the location of
gong points within each layer. Then, building up all the
branches by lofting the curves based on the grid of
points. In this process, it is noticed that although the
position of the gong (tree branches) is defined, the
position of the dou (the small rectangular blocks) is
omitted without a clear definition and consideration.
But actually from Figure 2 (top view), it is shown that
Figure 3 the small square of grid within each layer could be
The top-down used as the position of dou.
approach of dou From this experiment, we note three points of
gong using the merit. First, the layers and extension within each
Gaussian function layer are achieved. This is a basis to recreate differ-
ent scales of the joint without the need for another
set of logic definitions. In addition, this attempt can
fulfil some of the main functions stated in the rule
book, such as the potential to enlarge the roof area
both outwards and inwards. Thirdly, the configura-
tion can be adjusted by controlling all the input pa-

Digital Heritage 1 - Volume 1 - eCAADe 32 | 625

 DOU GONG (THE JOINT)
In order to investigate the detailed elements of the
dou gong, an understanding of its development and
structure is helpful. From the first appearance of dou
gong in period of the Warring States (approx. 475BC-
221BC), it developed and changed up to the time
when Ying Zao Fa Shi was written. This guide outlined
a relatively systematic description and established a
standard. Later in the Ming Dynasty (1368-1644) and
Ching Dynasty, it was further developed into a more
complex artefact. Apart from the format, the name of
the dou gong components also changed during the
development. Since the two rule books are the only
available ancient references, the name labelling sys-
tem currently used follows the earlier rule book, Ying
Zao Fa Shi. In this way the names used in this paper
will also be consistent with the broader research work
of the authors.
To gain an overall impression of the joint, Liang's
research work (2001) is significant. According to
Liang (2001): The base of the joint is a large square
block (dou) on the top of the column. There are set
into that block crossed arms (gong) spreading in four
directions. These in turn bear smaller blocks (dou)
that carry still longer arms spreading in the four direc-
tions to support upper members in balance. The jut-
ting arms (hua gong) rise in tiers (jumps) and extend
outward in steps from the large-block fulcrum to sup-
port the weight of the overhanging eaves. This exter-
nal pressure is countered by internal downthrusts at
the other ends of the bracket arms. Intersecting the
hua gong in the bracket set are transverse gong that
parallel the wall plane. Long cantilever arms called
ang descend from the inner superstructure, balance
on the fulcrum, and extend through the bracket sets
to support the outermost purlins. This outer bur-
den is countered by the downthrust of upper inte-
rior purlins or beams on the "tails" of the ang. The
extruding "beaks" of the ang easily identify them in
the bracket sets.
There are five main functions in terms of the as-
pects of structural, mechanical and aesthetic func-
tions. Firstly, it is located between the column and
626 | eCAADe 32 - Digital Heritage 1 - Volume 1
the roof as shown in Figure 1, transferring the roof
load evenly through the column to the foundation.
Second and most distinctive one is to provide an
earthquake-resistant function. Since all the horizon-
tal and vertical components insert into each other
following a prescribed order and rule without any
mechanical fixings or glue, its elastic performance
can absorb both horizontal and vertical earthquake
waves. Previous records show that many traditional
buildings with dou gong such as the Forbidden City
perform very well during earthquake disasters (Liang,
1985). Third and most practical the dou gong can sup-
port the outwards enlarged roof area to better pro-
tect the walls and windows from heavy rains but at
the same time prevent the blockage of the daylight-
ing from the pendulous roof edge. It is especially im-
portant because in ancient periods traditional win-
dows are made of paper and the main structure of
the buildings are made of wood, both require wa-
ter proof mechanism. An example discovered by
Liang (1985) is Fuoguang Temple major hall (built
at 857), the extent of eaves project about four me-
ters out from their supporting columns. "Their im-
portance in sheltering the wooden structure from
weather damage for over 1100 years is, they throw
away from the building the rainwater that courses
down the tile troughs of the concave roof" (Liang,
1985).Fourthly,the dou gong also extents inwards to
shorten the span of the beam and fang so the stresses
on the beam and fang are reduced. Apart from
the structural and mechanical functions,the dou gong
also has aesthetic functions, which are not dispens-
able although it is normally hidden under a large roof.
Factors such as the size and complexity, as well as the
style and colour of the dou gong indicates the level
of importance of the building itself, and that further
implies the social status and richness of the building
owner. For example, the royal style dou gong was
strictly forbidden in civic buildings in ancient dynas-
ties.
Depending on the location within the building,
there are mainly three types of dou gong as shown in
Figure 4. Zhuan Jiao Pu Zuo (dou gong at the corner,
Figure 4 connect with column, support the corner beam and
Three main types of roof corner, and this is the major structure compo-
dou gong (Liang, nent), Bu Jian Pu Zuo (dou gong not at the corner, not
2001) connect with column, this is the assistant supportive
component) and Zhu Tou Pu Zuo (dou gong not at the
corner, connect with column, this is the main sup-
port of the eaves load). The measurement of the size
of dou gong is Pu Zuo and normally there are four to
seven Pu Zuo dou gong. Within each dou gong, fifteen
key elements can be summarized in three categories:
dou, gong and ang. Dou is the blocks under and to
give support to gong, or on top of and at the two ends
of gong in order to provide vertically extension to an
upper layer of the gong. Gong is the long bracket arm
supported by dou and to extend horizontally the area
of the whole structure. Ang performs as a balance
lever inside the joint. Different types and Pu Zuo of
Figure 5 dou gong has different prescribed order to insert se-
Components of the rials of dou and gong and other minor elements to-
dou gong (Edited gether. Details of fifteen key elements and the over-
by the authors all assembly orders are illustrated in Figure 5. The ex-
based on Liang, ample used here is a five Pu Zuo dou gong in Song
2001) Dynasty based on Ying Zao Fa Shi, since five Pu Zuo
dou gong has the suitable complexity which clearly
shows all the major components, but does not have
too many repetitions of each component. Therefore
it is used as a typical case study example for the re-
search at the starting point. Once it is clearly under-
stood, other types and sizes of the joint could be sim-
ilarly treated.
Figure 6
The component DIGITAL REPRESENTATION AND FABRICA-
parts TION
Based on the description of the dou gong elements
above and detailed rules, a set of fifteen digital mod-
els have been generated as shown in Figure 6. At this
point, since the digital model is parametric, the real
units are omitted. Instead, all the measurements are
made based on the proportions, or in other words,
the relationships between each segment. With the
proportions, eight grades of the buildings have dif-
ferent units and absolute values to use in construc-
tion period (Li et al. 2013).

Digital Heritage 1 - Volume 1 - eCAADe 32 | 627

 Due to the complexity of the dou gong, the main is-
sue identified during the modelling process is how
to guarantee the accuracy of all the components, in-
cluding the connections between different surfaces,
the precise curvature of the shapes, the correct di-
mensions of the components which may partly omit-
ted in the rule book. To solve these issues, several
experiments were undertaken. To merge the sur-
face connections together and prepare for the later
printing process, inspections were made both auto-
matically and manually. The gaps on the outer sur-
faces were normally easy to find while the overlap-
ping of surfaces inside a volume was difficult to de-
tect. If this occurred, the target component must be
"opened" and problem areas must be fixed manu-
ally. But in general, since the software used (i.e. Rhi-
no/Grasshopper) is parametrically based, the man-
ual fix work was only done when necessary, in a rate
of about less than ten problem areas out of thou-
sands of surfaces. Therefore, the overall process is still
much advanced than the traditional way of check-
ing the carpentry work surface by surface solely by
the carpenter. Regarding the precise curvature of the
shapes, there are mainly two different types. The cur-
vature of the dou at the lower part of the block is
a real curve where the curvature is based on other
dimensions of the block. But the "curvature" of the
gong at the both ends is made using several straight
lines rather than a real curve. Apart from these, some
omitted details or unclear descriptions also occurred
in the rule book which brings difficulties in the mod-
elling process. An essential way of solving this is
to confirm the information of all the related nearby
components and then derive the absent information
then get a reasonable assumption. After this, put
the assumption component back to the whole model
and check whether it is consistent with the nearby
components or not. Several components such as the
hollow parts of ang, the slope of li shua tou and wai
shua tou, the middle supportive part of hua tou zi
were carried out by this method.
From the above evaluation, a revised version of
the models was produced. Conventional fabrication
628 | eCAADe 32 - Digital Heritage 1 - Volume 1
of the joint would be in wood. In digital fabrica-
tion, many options were open to us such as pow-
der, wax, plastic, metal and, less common, paper.
In this research, a relatively recent 3D digital fabri-
cation technique using built up sheets of paper (3D
Printer Matrix 300+) was selected for the following
reasons. Paper, compared with plastic, is a sustain-
able material which does not pollute the environ-
ment and when laminated together in multiple lay-
ers, it becomes very strong and rigid compared to
powder and wax. At the same time, compared with
metal, the micro space between sheets can be com-
pressed when needed, which maintains the elastic
character of the traditional wood material and also
the function of the joint. And this feature gives ad-
vantage when assembling the components: even
without pre-defining the tolerance, different compo-
nents can insert into each other tightly but without
damaging the contact surfaces. This advantage is
also discussed later in the tolerance discussion. For
comparison, a powder model version was used as
a comparison but showed that it is less appropriate
than the one made using paper because the pow-
der components were relatively weak and fragile and
therefore not suitable for assembly and re-assembly.
In order to make full use of the 3D printer and
produce the component parts effectively, all the 43
pieces of the fifteen elements (due to the repetition
of some elements) digital models are carefully orga-
nized according to the geometrical shapes so that
they can be composed within one stack of paper
which is the raw material that the MC or 3D printer
uses. Consequently the joint could be completed in
one printing, as shown in Figure 7.
PRACTICAL PROBLEMS, MODIFICATIONS
AND ASSEMBLY
During the process of extracting the components
from the paper block and peeling off the spare ma-
terial, practical problems arose with the jiao hu dou,
qi xin dou (two ears), qi xin dou (four ears) and san dou
constructions. In the paper stack, in order to save
space to a maximum level, these small scale dou are
 arranged in a position such that they partially inter- DISCUSSION AND CONCLUSION.
lock. However in practice it proved very difficult to After assembly, because the construction is ex-
extract each individual component without damag- tremely closely fitting, an analysis of the degree of er-
ing the faces of an adjacent component. ror in the fabrication process was made by accurately
measuring the dimensions in-plane and across-plane
Figure 7
using a digital caliper, to ensure that it was within an
The components
acceptable level. The tolerance is shown in the Table
are composed in
1. The majority of the tolerances measured proved to
the digital model to
be good and the average tolerance was 0.94%.
allow fabrication
Compared with conventional handmade tech-
from a single block
niques, obvious advantages were demonstrated.
When producing the joint by digital fabrication, the
modelling process is efficient such as the compo-
nents can be reused in other types of the joints if
they share some parts the same, without doing the
"analysis, assumption, checking" process again. And
when the overall scale needs to be changed, the dig-
Figure 8 ital models can easily achieve this. What is more im-
Modifications of the portant, industrialization is the trend to gradually re-
models place handmade method to increase the productiv-
ity, standardize the process and optimize the out-
come products. And indeed, today there are very few
knowledgeable and experienced carpenters who can
be found in this area which has led to a gradually dis-
appearance of this kind of traditional construction.
Thus our technique using digital representation and
fabrication could be used as an alternative to main-
tain the traditional style and skills for the future gen-
eration and research purposes.
Modifications were made to these five elements in
fabrication process. They are arranged as a single
Figure 9
layer without overlap. With this arrangement, the
The digitally
small delicate parts could be extracted without dam-
fabricated paper
age, as shown in Figure 8. This is also the quickest
components
way to reproduce these pieces considering the pa-
per 3D printer makes the fabrication layer by layer
(sheet by sheet). And indeed, this single layer mod-
els could also be properly arranged into the previous
composed models of one paper stack and be printed
once. After this, all the elements are assembled to
complete the dou gong. Figure 9 shows all the phys-
ical models of components and Figure 10 shows the
assembly process of the physical models.

Digital Heritage 1 - Volume 1 - eCAADe 32 | 629

 education, although almost no institutions have such Figure 10
module outside China, and many institutions inside Assembly of the
China do not produce effective teaching materials fabricated
on this. And according to author's experience, many components into a
young students have difficulty to understand such complete joint
a complicated structure. So a relatively cheap and
easily made physical model could help them greatly.
In this process, everything is not very difficult or
abstract, meanwhile safe enough for those without
any carpentry skills, but the accuracy level is high
enough. So the students can thus gain a direct un-
derstanding of such a complicated structure.
By completing the digital fabrication process,
several difficult and unclear aspects of the dou gong
have been better understood. The important out-
comes of the digital representation and digital fab-
rication techniques can be summarized as:
1. The order of assembly is better understood.
2. Clarity was brought to the understanding of
role and modus operandi of some elements
which were not fully described in the rule
books.
3. The vital importance of the small supportive
Benefits from this research are twofold: the findings
parts such as ears of the dou in the whole
could be used in future research and practise, also,
structure were better appreciated.
it could be instructive in teaching and education. To
understand traditional Chinese architecture is impos- 4. In using a paper 3D printer, differences were
sible if the dou gong is not closely studied. During examined in the modelling tolerances com-
the process of digital representation, fabrication and pared to using other forms of 3D modelling
assembly, a detailed 3D joint has been derived from materials. Also evident from this result was
the 2D pictures and text descriptions. This is a great that, paper, as a modelling material is very
help to understand how it is built up from its compo- appropriate, and for this purpose better than
nent parts. other materials used in digital fabrication and
This is the basis of future work on how to gen- may give a better option in terms of being a
erate new designs that respect the old traditions. sustainable material choice.
And it was only through the physical model assem-
bly process, that the importance of the traditional 5. What is more significant is that the 3D model
sun mao (small connection parts on wood) could be allows development of the other aspect of
found, which cannot be detected in any virtual en- our study, that of a parametric interpreta-
vironments. Apart from these points, study and un- tion of dou gong by clearly understanding the
derstanding of traditional Chinese architecture and inter-relationships of the individual elements.
Chinese architecture history should be a core in the Therefore it is an important foundation for the
further development of our research.

630 | eCAADe 32 - Digital Heritage 1 - Volume 1

Table 1
Tolerances;
established by
comparing the
actual size versus
the digital
dimension

Digital Heritage 1 - Volume 1 - eCAADe 32 | 631

 6. This is an alternative way to appreciate and
maintain knowledge and understanding of
traditional value and styles.
7. The findings of this research could also be use-
ful in architectural education and education
more generally.
By approaching the design and construction of the
dou gong in this bottom-up approach, problems
raised up in the top-down approach were also suc-
cessfully resolved. This was an important step in
the overall research of digital interpretation and
recreation of traditional Chinese architecture. Addi-
tionally, the assembly of a physical joint from dig-
itally fabricated components allowed a greater un-
derstanding which complemented the purely digi-
tal modelling that had been undertaken in the early
phases of the research. Overall the value of the dou
gong has been better understood through different
tools that contemporary digital technology can offer.
Consequently, the understanding of the dou gong is
the basis for our next step in the research of inter-
pretation of the unique roof, such as why the extent
of eave projection could achieve up to 4 meters and
why the raising degree of the roof permits light to
penetrate to the interior of the building despite the
wide overhang (Liang, 1985). In this way, the tradi-
tional Chinese architecture will be better understood
and the traditional value will be well respected in
contemporary time. This is a firm foundation for in-
spiration of the new Chinese style designs.
FUTURE WORK
Related future works could include:
1. Comparing the wood and paper densities and
other inner characterises to further investi-
gate the possibility of using paper material in-
stead of wood in practise.
2. If the above point could be achieved, the de-
velopment of a recycled paper building in a
traditional Chinese style could be a direction
of new design.
632 | eCAADe 32 - Digital Heritage 1 - Volume 1
3. Because paper expands with increased hu-
midity, water proof mechanism in the joint
is a follow-up research area, and similarly for
fire resistance, but environmentally friendly
chemicals are available already to do this.
4. The mechanical performance of the paper
joint should also be further studied.
What is more important, with the latest development
of four dimensional printing technology, the joint
could be a prospective research target. In addition
to the tradition 3D concept, a fourth dimension of
time is added into the printing technology which al-
lows the outcome product has a flexibility to change
its shape (as a respond) under certain condition or
environment, in order to fulfil some new functions
that fit the new environment. The dou gong is a po-
tential research target because under the earthquake
condition, the vibrational waves will be the "chang-
ing condition or environment", while the flexibility
of self-adjustment of the shape and inter relation-
ships between all the components could be studied
as a respond, in order to continuously fulfil the anti-
earthquake requirement, which is the most unique
and significant function of dou gong.
REFERENCES
Li, A 2001, A shape grammar for teaching the architectural
style of the Ying Zao Fa Shi, MIT, Cambridge, Mass.
Li, D, Knight, M and Brown, A 2013 'A Parametric Recre-
ation of Traditional Chinese Architecture-- A case
study on the floor plan', Proceedings of the 31st
eCAADe Conference, Delft University of Technology,
Delft, The Netherlands, pp. 697-703
Liang, S 1985, A pictorial history of Chinese architecture:
a study of the development of its structural system and
the evolution of its types, MIT Press, Cambridge, Mass.
Liang, S 2001, The Collected Edition of Liang Sicheng,
China Architectural Industry Press, Beijing
Webb, N and Brown, A 2011 'Digital Forensics as a tool
for augmenting historical architectural ayalysis', Pro-
ceedings of CAADRIA 2011, Hong Kong, pp. 505-515
Wu, Q 2005, 'Bracket teaching program: A shape gram-
mar interpreter', Automation in Construction, 14, pp.
716-723
Health and Safety Design by means of a Systemic
Approach
Linking Construction Entities and Activities for Hazard Prevention

Armando Trento1 , Antonio Fioravanti2 , Francesco Rossini3

1,2,3
Sapienza, University of Rome
1
www.armandotrento.it
2
www.dau.uniroma1.it/sito_consiglio_di_area/docenti/Fioravanti.htm
1,2
{armando.trento|antonio.fioravanti}@uniroma1.it
3
arch.francesco.rossini@gmail.com

Education and Research in Computer Aided Architectural Design in Europe faces

many urgent tasks. Among the Architecture, Engineering and Construction (AEC)
international scientific societies, only few researches systematically investigate
on how to integrate the design solutions with Health and Safety (HS) planning
measures, enhancing a collaborative "fusion" of all involved actors in Design and
Construction decision making. Process automation cannot be enhanced until
design/management tools, such as Building Information Models, can rely only on
entities formalised "per se", geometrical items fulfilled by isolated-object specific
information. To face complex problems, BIM models should be able to implement
and manipulate multiple sets of entities, qualified by clearly established
relationships, belonging to organically structured and oriented (sub-) systems.
This paper reports on an early stage research project, focused on the
identification of operative rules for Health and Safety design. Implementation on
the unique case study of Palazzo della Civiltà Italiana functional refurbishment
faces two main objectives: one, more pragmatic, is concerned with boosting
workers education about non-standard operative tasks, by means of accurate
ad-hoc construction narrative visualisation; another one, more challenging and
theoretically complex, consists in modelling "judgment-based" rules, aimed at
supporting automated reasoning in Safety Design.

Keywords: Construction Hazard Prevention, Project construction management

and visualization, Health and Safety Management, Risk Modelling, Knowledge
Representation

Digital Heritage 1 - Volume 1 - eCAADe 32 | 633

 HEALTH AND SAFETY IN AEC
Education and research in computer aided architec-
tural design in Europe faces many urgent tasks. To
correctly decline the most critical aspects of this sec-
tor, it must be observed that, compared to other in-
dustries, AEC is one of the most hazardous (Carter
and Smith, 2006; Wang et al., 2006; Camino et al.,
2008).
The issue of safety in workplaces is particularly
felt especially in the AEC sector, which still appears
to be among the industries the one with the highest
number of injuries, with a high fatal accident rate (Eu-
rostat, 2012).
Despite rigorous efforts to improve the construc-
tions working environment in the European Union,
the fatal accident rate is approximately 13 workers
per 100,000 as against 5 per 100,000 for the all sec-
tors average (Aulin and Capone, 2010).
Italy has the highest rate of workplace accidents
and deaths than any other E.U. country. According
to Giaccone (2009), the reported work-related acci-
dents for the manufacturing and construction sector
for in the year 2008 is 367,132 cases. Giaccone (2009)
claimed that these records do not include work-
related accidents among undeclared workers, who
account for 13.4% of the total labour force, which
would contribute to a much higher figure.
Italy, in the overall context of the European
Countries, shows one of the highest frequency of
accidents, denouncing a certain backwardness on
the supervision procedures and the diffusion of pre-
vention. In Italy, according to Ciribini (2010) the
poor performances and the lack of managerial skills
affecting the construction sites are due to an ab-
sence of the co-operation amongst the various ac-
tors (Client Organizations, Main Contractors, Sub-
Contractors, and Suppliers). Moreover, the current
economic situation is worsening the trend and com-
panies are forced to work with low profit margins, try-
ing to save on all aspects of the construction, neglect-
ing the safety.
Designers, architects, engineers and contractors
have an influence on the health and safety of building
634 | eCAADe 32 - Digital Heritage 1 - Volume 1
site employees (Gambatese and Hinze, 1999; Behm,
2005; Frijters and Swusste, 2008; Gambatese et al.
2008; Toole and Gambatese, 2008).
From the designers' point of view, there is a dif-
ferent attitude depending on the extent of work to
be done: it's well known that for small / medium con-
tracts, safety design is seen as a bureaucratic exercise
or, at worst, a matter to be entrusted to experience
and habits of business and / or individual workers on
the building site.
Conversely, with regard to the works of higher
complexity and costs, the need to integrate the de-
sign solutions with safety planning measures is re-
alised in few cases. Only sporadically, advanced ICT
automated techniques are employed to support the
heterogeneous competences and duty required by
the Health and Safety Coordinator (HSC) professional
profile.
Since the adoption of European Union Frame-
work Directive (89/391/EEC) and the Construction
Site Directive (CSD) (92/57/EEC) - in Italy transposed
in D.Lgs 81/08 - European building designers are
legally required to consider health and safety in their
designs.
However, previous studies have shown that de-
signers in general - not just in the construction in-
dustry - fall short of satisfying this obligation (Behm,
2005; Fadier and De la Garza, 2006; Frijters and
Swuste, 2008); mainly because the most of contrac-
tors often neglect the implementation of their health
and safety plans (Wang et al., 2006; Saurin et al.,
2008).
In addition, the importance of safety economic
evaluation is often underestimated: international sci-
entific literature demonstrates that frequently hap-
pens that HSC, because of a contextual construction
planning uncertainty, ends up with identifying gen-
eral solutions to fit loosely, without considering any
alternative, at comparable cost, even if producing
safety improvement measures.
Ultimately, the combination of the inefficiencies
between designers and executors leads to an erro-
neous Safety Evaluation and Estimation, which re-
sults in an inefficient use of resources.
PROJECT SAFETY DESIGN AND PLANNING
Safety standards control and management is an ex-
traordinarily resources-consuming task, as it is re-
quired to monitor a dynamic 'status quo' of great
complexity, which depends on the association and
fusion of different professional domains, operating
within the context of building sites characterised by
heterogeneous risks.
The lack of interest from the companies involved
in participating in early stages, before the definition
of solutions to apply, lead them to serially reuse de-
vices and working methods, not considering that
each site is a unique prototype of himself.
Analysis and causation of accidents and histori-
cal data provide valuable but general information for
safety planning. Commonly, this information is used
to descript textually a possible phenomenon, sup-
ported by very accurate statistic charts on probability
and risk about the event occurring in a certain oper-
ational site area.
Text, charts and evocative pictures are not suffi-
cient to predict when and where accidents occur on
unique construction projects. This represents a slow-
down of the editing and updating process, a sharp
boundary from the point of view of quantity / quality
of information provided.
This has led to the advent of information tech-
nology enabled approaches for construction safety
using virtual designs and simulations of construction
operations.
The kind of tool generally used to support these
comprehensive competencies includes the class of
process management systems and the class of build-
ings information modelling.
Among the AEC international scientific society,
only few researches systematically investigate on
how to make a step forward in digital tools to sup-
port HSC, enhancing a collaborative "fusion" of all in-
volved actors in Design for Safety.
Research efforts in recent years have focused dif-
ferent approaches for improving site safety with dig-
ital models. Khoshnava et al. (2009) stated the main
goals as follows:
• to plan and model proactively the sequences
of tasks together with the needed safety ar-
rangements and utilities;
• to ensure that all constructions can be built
without any safety threat and necessary joints
exists for fixing the safety utilities;
• to document planned safety solutions in de-
tail and self-explaining way, and, this informa-
tion to be conveyed throughout all key play-
ers until the last worker in chain.
This paper's work takes off from observing how diffi-
cult is trying to design construction safety by means
of systems that have been developed for supporting,
alternatively, on one side the architectural building
design or, on the other side, the construction process
management.
INFORMATION MODELLING AND THE IS-
SUE OF SAFETY IN COMMON PRACTICE
In most advanced economies across the globe, over
the last several years, designers started to use Build-
ing Information Models, sharing databases in order
to be used in synchronous between companies, de-
signers and inspectors.
BIM aims to an efficient exchange of quality in-
formation allowing operations systematization, by
simultaneously updating objects that describe the
project.
BIM-based software packages have well-
established positions, and are used by AEC-FM pro-
fessionals. Such tools also form a natural starting
point for BIM-based site layout and safety planning.
The most important BIM software features 3D/4D
safety planning, management and communication
include: 3D modelling and viewing capabilities, 4D
tools and features, tools for analysing risks or safety
of the designs and plans, and data exchange capabil-
ities.
Digital Heritage 1 - Volume 1 - eCAADe 32 | 635
 Commercial IFC-compliant BIM models allow
software interoperability, phase-dependent project
analysis, clash detection among components. Re-
cently, in addition to the above capabilities, new plat-
forms have been introduced to platforms to define
and manage 4D models. These new environments
are focused on assisting Construction Managers in
describing the sequence of operations to be per-
formed, by linking a narrative timeline to building en-
tities.
In today's AEC practice for safety planning and
management, what can be observed in terms of most
common application methods is parallel use of var-
ious - non compliant and non interoperable - soft-
ware.
For instance, site layout drawings or specific il-
lustration about tasks to be performed, are two-
dimensional representations, only implicitly linked -
just for analogy - to the description of risks, hazards
and safety prescriptions of what is expected to hap-
pen in reality.
Moreover, with regards to descriptive docu-
ments, the use of spreadsheets or pre-designed
forms for facilitating data entry procedures, are not
adequately supplied by, and linked to, a detailed li-
brary and structured databases.
SAFETY REQUIREMENTS AND BIM: STATE
OF THE ART
In order to inform different domains actors, elimi-
nate hazard and reduce risks, nowadays, several com-
panies and organisations focused on building infor-
mation models are researching on the implemen-
tation of and the connection to tools oriented to
worker's safety training and education, design for
safety, safety planning (job hazard analysis and pre-
task planning), accident investigation, and facility
and maintenance phase safety.
Nowadays, in order to inform the different do-
mains’ actors, eliminate the hazards and reduce risks,
several companies and organisations have focused
on BIM. These companies are investigating how to
implement these tools and utilize them for their
636 | eCAADe 32 - Digital Heritage 1 - Volume 1
workers' safety training and education, design for
safety, safety planning (job hazard analysis and pre-
task planning), accident investigation and facility and
maintenance phase safety.
Zhang et al. (2013) and Qi et al. (2011) devised a
design for safety tool, making design for safety sug-
gestions available to designers and constructors by
formalizing collected design-for safety suggestions
and checking the building model.
Bansal (2011) uses GIS based navigable 3D ani-
mation in safety planning for predicting places and
activities which have higher potential for accidents;
he links the information between the CPM schedule
and safety recommendation database.
The VTT Technical Research Centre of Finland
(2010) developed a manual procedure of using BIM
technology for safety planning, management, and
communications. As part of the 4D-construction
safety planning, VTT visualized BIM-based 4D safety
railings for fall/edge protection in Tekla Structures.
In 2010 the Georgia Institute of Technology intro-
duced the first preliminary results of an automated
safety rule checker for BIM.
The literature shows that (building) information
modelling has enabled virtual safety controls to be
used to identify safety hazards.
However, most of the existing efforts in safety
planning either largely rely on human input or of-
fer knowledge-based/semi-automated implementa-
tion.
Further automation of the tool and better visual-
ization are new options to be explored.
The U.K. Government has commissioned a work-
ing group to integrate regulations into BIM in order to
simplify compliance [1]. BIM4Regs aims to incorpo-
rate Building Regulations, planning and health and
safety requirements into BIM models and has been
commissioned by the BIM Task Group.
It is being executed by a cross section of govern-
ment and industry organisations including the De-
partment for Communities and Local Government,
the Health and Safety Executive, the planning portal,
BRE, software vendors and industry representatives.
 Regulatory information will be embedded within
BIM software allowing architects and engineers to
see if designs comply.
It is important to underline that many rules in-
volve geometrical domain of the representation: for
instance, regulations based on minimum distance
measurements could be easily incorporated into BIM.
Although P. Caplehorn, chair of the BIM4Regs
working group, believes that some 30% of regula-
tions could be easily incorporated into BIM software,
the majority of general design requirements fall into
the "judgment-based" category.
It will be more difficult to incorporate the
"Judgmentjudgement-based" responses to into reg-
ulations, since these needs to meet the safety targets.
In order to have a high quality site production
plan, all the required arrangements for a site layout
model needs to be carefully studied.
Exploring how hazards, risk and solutions might
be built into BIM, activating automated reasoning, in
a way that designers, constructors and users find use-
ful, is an urgent task to face for international research.
SAFETY REQUIREMENTS AND BIM: POTEN-
TIALS AND LIMITS
Since safety rules, guidelines, and best practices al-
ready exist, they can be used in conjunction with ex-
isting three-dimensional design and schedule infor-
mation to formulate an automated safety rule check-
ing system. Safety conditions appear, then are re-
solved within the construction process, as a project
progresses.
Compared to existing BIM application such as
clash detection and BIM-based quantity take-off, a
basic requirement for a rule-based checking sys-
tem is that each building object carries information:
for example, object name, type, attributes, relation-
ships and metadata including object identification
(ID) number, date, and author creating model ele-
ments. It is important to point out that the schedule
data needs to be linked to the building object data
since the assigned protective system needs to be up-
dated accordingly.
There are rule-based platforms available that can
apply rules to IFC building model data. Especially
the Solibri Model Checker (Solibri Inc.) is one of the
software packages that can be used for combining,
viewing and examining the content of various BIM-
models. This software provides special tools for rule-
based automated checking, analysis as well as quan-
tity and other data take-offs. Users can also edit and
create new rule sets.
However, effective automation cannot be en-
hanced until design/management tool can rely only
on entities formalised and structured "per se", fulfilled
by isolated information.
BIM models should be able to implement and
manipulate multiple sets of entities, qualified by
established relationships, belonging to organically
oriented (sub-) systems. More sophisticated infor-
mation modelling structures are needed in order
to allow querying and computing Design/Construc-
tion/Safety entities at a higher levels of abstractions.
Even if the 3D/4D BIM approach is certainly more
advanced and valuable than methods and tools tra-
ditionally used by HSC, it shows some critical issues:
• classes and detail accuracy of existing para-
metric entities is oriented mainly to Building
design, less to Construction design and not
to Safety design. Construction and Safety are
still on the fringes of the world of BIM.
• Software manage information not enough
structured to model assemblies relations
among entities (e.g. hierarchical or topolog-
ical) or - very critical theme in AEC practice -
between building entities (e.g. components,
spaces, equipment, machines, etc.) and ac-
tors.
• Consequently, 4D BIMs support a low-
effective level of rules modelling, with limited
possibilities of rule based automatic checking,
and still far from "judgment-based" reasoning.
It is believed that the area of “real time project
information on safety risks and therefore measures
Digital Heritage 1 - Volume 1 - eCAADe 32 | 637
 to be taken by the workers” is still a very urgent and
open research problem.
HOW TO ENHANCE REPRESENTATION OF
SAFETY ENTITIES AND RULES
Authors developed a compact Knowledge Represen-
tation Structure, which is able to represent all the
concepts involved during the design process, thus
improving mutual understanding among all the ac-
tors involved and supporting (cross)-domain reason-
ing: this knowledge structure has been implemented
as a Meanings-Properties-Rules ontology template,
related to each represented entity.
Depending on the main target of the specific de-
sign framework, the universe of the entities included
in any kind of design knowledge would have several
sets. Referring to the building design these can in-
clude spaces, components and furniture. Referring
to construction design and these can include phases,
activities, actors, operative spaces, building compo-
nents, formworks, structures, technical installations,
machines, context, and so on. Referring to safety
design these can include the site equipment, safety
railings, guardrails, falling prevention elements, pre-
scription and alert signals, etc.
All concepts considered to be relevant in a de-
sign framework are linkable to well defined homoge-
neous sets of entities by means of suitable sets of re-
lationships.
Any designed object is, or can be, made of enti-
ties at different levels of complexity: differently from
BIM entities, and IFC classes, the proposed knowl-
edge model allows the formalisation of the relation-
ships that combine entities at a lower level of com-
plexity in more complex one called assembly, in a re-
cursive way.
Any assembly-entity will have a Made-Of rela-
tion with its components or, using a formal imple-
mentation language, a Has link; in addition, as pre-
viously mentioned, it is characterized by a set of its
own meanings properties and rules not characteriz-
ing its component entities.
By being recursive, even "complex" entities may
638 | eCAADe 32 - Digital Heritage 1 - Volume 1
be constituent parts of other entities having a higher
layer of complexity, that therefore are assemblies of
assemblies.
By going deeper into the analysis of design and,
consequently, in the later stages of it, there is an in-
creasing need to further specify the entities involved
and to increase the array of requirements and perfor-
mance specification necessary to represent each en-
tity.
This kind of increasing specification approach
establishes a relationship of father/son kind among
the general and specific entities considered, imple-
mentable by means of the formal link: Is-A.
For implementing this theoretical model, we
are using the ontology technologies in order to
model the design entities, physical or abstract, and
their space-time relationships structured by means of
meanings, properties (defining their state) and rules
(relations, reasoning rules, consistency, best prac-
tices).
Analysis, checking, evaluation and control of
concepts associated to specific entities is performed
by means of inferential engines, with deductive 'If-
Then' type procedures.
A system of engines -matching rules among the
ontologies - will work on a deductive layer over-
lapped at the actual BIM level, allowing the design-
ers to use in a coherent manner different levels of ab-
straction, or to exploit a conceptual interoperability.
APPLICATION IN A NON-STANDARD CASE
OF STUDY: PALAZZO DELLA CIVILTÀ ITAL-
IANA
Given the state of the art and the potential not yet
fully expressed by 4D BIM systems in the field of
Health and Safety management, this paper reports
about an on-going research that is oriented to for-
malise Safety Design knowledge based on field ex-
perience.
This research is made possible by a particularly
positive convergence of two events. In one side,
the authors have been contractually designated by
Fendi Srl, an international private company known
 in the fashion sector, for leading Health and Safety
Coordination in the challenging functional restora-
tion works of a relevant historical building known as
"Palazzo della Civiltà Italiana" (Figure 1), also known
as "Squared Colosseum". The company intends to
move there its offices headquarter. On the other side,
Sapienza, University of Rome assigned to the authors
a research fellowship grant in order to investigate on
HS support systems.
Figure 1
Palazzo della Civiltà
Italiana, 2014
Figure 2
Palazzo della Civiltà
Italiana, 1940
This public property is an iconic building of particu-
lar historical interest and protected by the Minister of
Cultural Heritage rules. In 1935 a design competition
was established in order to have this building for the
1942 World Exhibition. The Palazzo was designed by
the architects G. Guerrini, E. B. La Padula and M. Ro-
mano and constructed between 1938 and 1943 (Fig-
ure 2). It was inaugurated in November 1940 as the
centrepiece of the Esposizione Universale di Roma
neighbourhood (EU42) and continues to be its most
iconic building. The structure is also considered one
of the most representative examples, in the modern
history of architecture, of rational synthesis of classi-
cal linguistic elements.
The Squared Colosseum consists of a box that
stands on a wide base connected to the street level
by two opposing staircases. The scale is imposing:
the base covers an area of 8,400 square meters, and
the building has volume 205,000 cubic meters with a
height 68 meters (50 meters from the base; each level
is 8 meters high).
The four faces of the prismatic solid, covered with
slabs of travertine stone, are characterized by the
strong rhythm of the arches in sequence (nine re-
peated openings for six floors), strong architectural
motif that makes it instantly recognizable as one of
the symbolic monuments of Rome.
Main objective of this sophisticated intervention
is to ensure, in respect of the triad Time - Cost - Perfor-
mance, a product of high architectural quality, per-
formed in compliance with the history of the build-
ing, planning the work phases in such a way to make
them closer to serial standards of production.
In this context, complicated by the need to pro-
tect the finish and the existing structures, special-
ist workers are going to assemble the technical el-
ements for the construction of mezzanines, 4 meter
high, and interior sub-structures built with the tech-
nology of steel and glass in order to frame the enor-
mous open space in more functional human scale
ones.
It seemed immediately necessary to arrive at a
real fusion between the design solution for architec-
tural model of the project, the protection of the listed
building, the implementation of safety measures, al-
lowing a continuous exchange / upgrade informa-
tion between architectural design, project and con-
struction site safety.
Digital Heritage 1 - Volume 1 - eCAADe 32 | 639
 METHODOLOGY IMPLEMENTATION PATH
The overall research aim is to develop a method for
enhancing the quality of information modelled/man-
aged and defining an implementation path toward
a desirable dynamic and holistic knowledge-based
support system.
Acting in the context of this unique case of study,
by applying a reverse engineering approach, it is be-
ing possible to benefit from this on-going significant
field experience for pushing the limits and potentials
of exploring the existing systems.
This paper reports on an early research work ori-
ented to define and implement a systematic method-
ology for helping in HSC activities, namely to calcu-
late the safety-related performance of a project, pro-
viding a consistent basis for comparisons between
different safety design solutions.
The main objectives can be subdivided in two
levels:
• The first is targeted at educating workers
about non-standard operative tasks;
• another one, more theoretically complex,
oriented to model "judgment-based" rules,
aimed at supporting automated reasoning in
Safety Costs' Evaluation and Estimation.
On one side, a 4D BIM-based site plan is being mod-
elled and used to produce illustrative representations
of the site organisation and of the safety arrange-
ments.
The 3D views together with 4D narrative, can be
used for orientation of site workers, task guidance
and instructions, for informing about risks.
On the other side, to manage safety solutions,
both predicting during the building design phase
and evaluating - in real time - during the construction
phase, we are formalising few representative rules to
be checked by the system and some related opera-
tive actions to be suggested consequently.
In order to have automatic judgement, as a start-
ing task a clear definition of rules to be checked is
needed. Specifically, in the safety design phase we
640 | eCAADe 32 - Digital Heritage 1 - Volume 1
are trying to address the prevention of hazard as a
case study.
According to the Italian D.Lgs. n.81/2008, inher-
iting EU Directive (CSD) (92/57/EEC), safety risk as-
sessments are function of physical entities on one
side and, on the other side, they can be linked to
more abstract coordination process activities.
Primarily, we analysed how to represent a safety
condition in terms of objects, attributes and relations
needed. Afterwards we formally defined a routine to
run the risk evaluation, intended as complex rules to
be applied to the objects, attributes and relations in-
volved.
The formal representation of Safety categories
and rules to be applied have been implemented by
using a compact Knowledge Representation Struc-
ture and managed by an ontology based technolog-
ical pipeline developed by the authors.
In order to assist human decision makers in
safety planning and scheduling activities, the out-
come of the hazard prediction assessment is an au-
tomatically generated alert which would inform the
necessary parties.
Future work will be oriented to implement the re-
porting delivery and the safety solution suggestion,
for instance by recommending correct geometry, lo-
cation, materials and time of installation of the pro-
tective equipment that must be installed to prevent
a hazard.
FORMALISING DIFFERENT LEVEL OF
SAFETY DESIGN KNOWLEDGE
In order to avoid a typical limitation in the evalua-
tion of health and safety risks, authors implement
by means of the previously presented knowledge
modelling approach, some classes of a more general
Health and Safety design ontology.
"Judgement based" evaluation can be per-
formed by using quantitative indicators formalised
by Gangolells et al. (2010), which are based on data
available in the project documents/models. Indica-
tors measure physical property values of construc-
tion/safety related entities, not available in current
Figure 3 BIM/IFC models.
4D BIM logistics The implementation steps are namely:
model
1. Represent Safety Risk related to Construction
Activity (e.g. expressed in OWL language by
means of ontology editors, e.g. Protégé)
2. Represent an extended library within BIM of
construction/safety related entities (by means
of API), with special emphasis of Risk Indicator
physical properties
3. Link Construction Activity (time/space instan-
tiated) with actual BIM/IFC safety related en-
tity (e.g. in Autodesk Navisworks).
After the previous steps are successfully imple-
mented, the system will allow safety designer and HS
decision-makers, by supporting them automatically
calculating Risk Indicator of the instantiated project
Construction Activities.
The proposed ontology will allow the classifica-
tion of all the terms (aspects, impacts, risks, and pro-
cedures) related to the Health and Safety "judgement
based" evaluation as well as the relationships that ex-
ist among sets of design objects.
Then again, each class will be enriched with dif-
ferent properties which will be used by the decision-
making tool to identify the main significant Health
and Safety conditions in each design/construction
process, and moreover, to evaluate their impact in a
specific construction project in order to provide pro-
cedures.
DISCUSSIONS AND FUTURE WORK
Safety planning can be a part of 4D production plan-
ning. This can create a safety planning practice that is
undertaken earlier than traditionally in construction
projects, and furthermore it can capture a more de-
tailed planning level (Figure 3).
A BIM-based site plan can be used to produce illus-
trative representations of the site and safety arrange-
ments, and the views can be used for orientation of
site workers, task guidance and instructions, for in-
forming about risks.
Safety control and evaluation is based on both
proactive and reactive performance indicators rely-
ing on percentage of safe work packages and actual
accident data.
The proposed methodology and implementa-
tion path is oriented to support designers and
construction planners in visualizing and evaluating
Safety Risks, related to construction activities by
means of Risk indicators.
Vice versa, safety designers need to spend more
resources in order to utilize the construction work
breakdown structure and therefore enhancing the
geometric model with construction/safety related
entities by including more details which will be re-
sulting an ever-changing work site schedule.
In summary, the research expected general re-
sults are a step forward towards:
• Enhancing safety related detail of the BIM
which is typically modelled by the project de-
signers. Specifically: accuracy of construc-
tion site entities and building construction
sub/phases linked to Safety work breakdown
structure.
• A reusable safety knowledge model, introduc-
Digital Heritage 1 - Volume 1 - eCAADe 32 | 641
 ing a system of hierarchical relationships in
time and space for representing parametrical
Safety conditions.
• Establish relationships between activities and
tasks, hierarchizing relationships between op-
erators and devices and / or safety equipment.
• Predict site hazard and define corrective mea-
sures by means of a stored library populated
by alternative resolutions. Support HS design
by proposing realistic solutions to resolve the
identified issues.
REFERENCES
Aulin, R and Capone, P 2010 'The Role of Health and
Safety Coordinator in Sweden and Italy Construc-
tion Industry', 18th CIB World Building Congress 2010,
W099 - Safety and Health in Construction, The Lowry,
Salford Quays, UK, pp. 93-106
Bansal, VK 2011, 'Application of geographic informa-
tion systems in construction safety planning', Inter-
national Journal of Project Management, 29(1), pp.
66-77
Carrara, G, Fioravanti, A, Loffreda, G and Trento, A
2009 'An Ontology-based Knowledge Representa-
tion Model for Cross Disciplinary Building Design. A
general Template', Computation: the new Realm of
Architectural Design, eCAADe 27, Istanbul, pp. 367-
373
Fioravanti, A, Loffreda, G and Trento, A 2011, 'An innova-
tive comprehensive knowledge model of Architec-
tural Design Process', International Journal of Design
Sciences &Technology, 18(1), pp. 1-18
Fuertes, A, Casals, M, Gangolells, M, Focada, N and Roca,
X 2008, 'An ontology for Environmental and Health
and Safety risk', Frontiers in Artificial Intelligence and
Application, 174, pp. 100-108
Gambatese, J, Behm, M and Hinze, J 2005, 'Viability of
Designing for Construction Worker Safety', Journal of
Construction Engineering and Management, 131(9),
pp. 1029-1036
Gangolells, M, Casals, M, Forcada, N, Roca, X, Fuertes, A,
Macarulla, M and Vilella, Q 2010 'Idetifying Poten-
tial Health and Safety Risks at the Pre-Construction
Stage', Proceedings of the 18th CIB World Building
Congress W099, Salford, UK, pp. 59-73
Jørgensen, K, Sander, D and Staghøj, A 2010 'Integration
of Safety in the Building Delivery System', Proceed-
642 | eCAADe 32 - Digital Heritage 1 - Volume 1
ings of the 18th CIB World Building Congress W099,
Salford, UK, pp. 46-58
Lew, JJ and Lentz, TJ 2010 'Designing for Safety - Appli-
cations for the Construction Industry', Proceedings of
the 18th CIB World Building Congress W099, Salford,
UK, pp. 37-45
Qi, J, Issa, RRA, Hinze, J and Olbina, S 2011 'Integration
of safety in design through the use of building infor-
mation modeling', Proceedings of the 2011 ASCE inter-
national Workshop on Computing in Civil Engineering,
Miami, Florida, USA, pp. 698-705
Sulankivi, K, Kähkönen, K, Mäkelä, T and Kiviniemi, M
2010, 'no title given', in Barret, P, Amaratunga, D,
Haigh, R, Keraminiyage, K and Pathirage, C (eds)
2010, 4D-BIM for Construction Safety Planning, CIB,
Manchester, UK
Trento, A. and Jeong, Y.W. 2008, 'Interleaving Semantics
for Multi-Disciplinary Collaborative Design in AEC',
in Cunningham, P. and Cunningham, M. (eds) 2008,
Collaboration and the Knowledge Economy: Issues,
Applications, Case Studies, IOS Press, Amsterdam, ,
pp. 649-656
Zhang, S, Lee, J, Venugopal, M, Teizer, J and Eastman,
CM 2011 'Integrating BIM and safety: an automated
rule-based checking system for safety planning and
simulation', Proceedings of CIB W099, Washington,
DC, USA
Zhang, S, Teizer, J, Lee, J, Eastman, CM and Venugopal,
M 2013, 'Building information Modeling (BIM) and
Safety: Automatic Safety Checking of Construction
Models and Schedules', Automation in construction,
29, pp. 183-195
[1] http://www.vtt.fi/sites/
bimsafety/?lang=en [last access 2014-06-15]
[2] www.bimtaskgroup.org [last access 2014-06-03]
Digitally-Assisted Stone Carving on Canada's Parliament
Hill
James Hayes1 , Stephen Fai2 , Phil White3
1,2
Carleton Immersive Media Studio 3 Public Works and Government
Services Canada
1,2
cims.carleton.ca
1,2
{jhayes|sfai}@cims.carleton.ca 3 phil.white@tpsgc-pwgsc.gc.ca

In this paper, we discuss the results of a collaboration between the Carleton

Immersive Media Studio (CIMS), the Dominion Sculptor of Canada, and the
Heritage Conservation Directorate (HCD) of Public Works and Government
Services Canada (PWGSC), on the development of a digitally-assisted stone
carving process. The collaboration couples the distinguished skill of the
Dominion Sculptor with digital acquisition and digital fabrication technologies in
the reconstruction of a stone relief sculpture on the façade of the East Block
building of the Canadian Parliament in Ottawa, Canada. A variety of digital
technologies were used including, hand-held laser scanning, digital
photogrammetry, 3d-printing, CNC milling, and robotic stone milling, in initial
research for the fabrication of maquettes and the collaboration with the
Dominion Sculptor.

Keywords: Digital Heritage, Digital Fabrication, Masonry Conservation

While there is much interest in the use of digital
fabrication technologies in the design and construc-
tion of new buildings, there is little research being un-
dertaken for the application of these same technolo-
gies in the conservation and rehabilitation of her-
itage buildings. At the same time, the digital datasets
generated by laser scanners and photogrammetry,
which can be utilized by digital fabrication equip-
ment, have become integral to heritage documenta-
tion (Gilboa et al. 2013; Guarnieri et al 2010; Fatuzzo
et al. 2011; Staley et al. 2009). Our research at
the Carleton Immersive Media Studio (CIMS) devel-
ops new workflows and methodologies for computer
assisted acquisition and fabrication in the rehabilita-
tion of significant heritage assets. Ultimately, it is in-
tended that these new methods of working will not
only contribute to heritage conservation, but more
generally undermine the hegemony of standardiza-
tion that currently dominates the architecture, engi-
neering and construction (AEC) industry. To that end,
we are developing protocols for meaningful collabo-
ration between architects and the craftspeople and
fabricators that actually construct the elements of a
building.
In this paper, we discuss the results of a col-
laboration between CIMS, the Dominion Sculptor
of Canada, and the Heritage Conservation Direc-
torate (HCD) of Public Works and Government Ser-
vices Canada (PWGSC), on the development of a
digitally-assisted stone carving process. This collab-

Digital Heritage 1 - Volume 1 - eCAADe 32 | 643

 oration couples the distinguished skill of the Domin-
ion Sculptor with digital acquisition and digital fabri-
cation technologies in the reconstruction of a deteri-
orated stone relief sculpture on the façade of the East
Block of the Parliament Hill National Historic Site in
Ottawa, Canada. The Dominion Sculptor is responsi-
ble for creating significant architectural stone, wood,
and bronze sculptures on Parliament Hill.
CONTEXT
Canada's Parliament buildings were conceived in
1859 after Ottawa was chosen as the capital of the
then Province of Canada. The three gothic revival
buildings sit atop a rugged escarpment along the Ot-
tawa River in the centre of the city. The Centre Block,
which contains the House of Commons and the Sen-
ate, is flanked on either side by the West Block and
East Block - each containing offices for senators and
members of parliament. Over the next decade, a
long-term work plan will see all three buildings reha-
bilitated.
The West Block was vacated in 2011, and in-
terior finishes were removed to facilitate asbestos
abatement and masonry consolidation. The HCD
has been tasked with documenting the exposed ma-
sonry structure of the West Block during the reha-
bilitation using laser-scanning, photogrammetry and
record photos. CIMS has been working with HCD to
explore novel applications for the extensive point-
cloud datasets being generated during the docu-
mentation process. One of these applications - the
focus of this paper - explores the use of digital fabri-
cation for the rehabilitation of traditionally cut sculp-
tural elements.
TRADITIONAL WORKFLOW
There are a number of possible workflows for replac-
ing a sculptural stone element on a heritage build-
ing. These are determined by the specific location,
condition and context of the sculptural element and
the building more generally - as well as the heritage
conservation standards of that specific jurisdiction.
While some of these workflows include removal of
644 | eCAADe 32 - Digital Heritage 1 - Volume 1
the sculptural element for the replacement or repair
process -in the case of this project - the existing sculp-
ture will remain in the East Block façade until the re-
placement has been carved. Further, the new sculp-
ture is to be as exact a copy of the original as is rea-
sonably possible. The traditional workflow described
below, therefore follows these parameters.
The traditional process of re-carving a damaged
or weathered stone sculpture begins with creating a
copy of the existing stone sculpture. A mould is cre-
ated of the sculpture that is to be replaced by apply-
ing a layer of silicone rubber to the stone, followed
by a layer of plaster. The silicone rubber captures the
fine details of the stone surface and the plaster pro-
vides rigidity for the mould. The two layers of the
mould are removed from the stone, and filled with
plaster to create the copy of the sculpture. The dam-
aged or eroded areas are rebuilt on the plaster copy
using modelling clay. The sculptor's work in clay may
be based on existing similar sculptures, historic pho-
tographs of the sculpture prior to the damage, or in
the absence of any definitive evidence, the repairs are
guided by the sculptor's artistic judgement. When
the sculptor's work is completed, the moulding pro-
cess is repeated and a second plaster cast is made
to create a homogenous model that incorporates the
sculptor's work.
To transfer the geometry of the plaster model to
stone, the sculptor uses a device called a pointing
machine (Figure 1). Formalised in the 18th century,
a pointing machine uses a large needle at the end
of an adjustable armature to locate a single point in
space on the surface of the plaster copy. The point-
ing machine is transferred to the uncut block of stone
and the sculptor chisels away the stone until the nee-
dle reaches the same point, or more typically a few
millimeters proud of the point. The sculptor roughs-
out the stone by repeating this process of transfer-
ring a single point from the plaster copy to the stone
over and over. The pointing machine is set aside after
enough points have been transferred and the sculp-
tor uses a variety of tools to add the final details to
the stone sculpture.
Figure 1
Pointing machine
on a block of stone
in the Dominion
Sculptor’s
workshop.
Another common process in the rehabilitation
of sculptural stone elements is to repair the ele-
ment rather than replace the entire sculpture (Gim-
mer 1984). This is typically performed where a piece
of the sculptural element has broken off and is miss-
ing. A model of the replacement piece is sculpted us-
ing modelling clay and fit into place on the sculptural
element. A stone copy of the clay model is carved and
affixed to the sculptural element using non-corrosive
pins and epoxy.
DIGITALLY-ASSISTED WORKFLOW
Through the Carleton University School of Architec-
ture, two main digital fabrication technologies are
available to CIMS: a Dimension SST 1200es 3D printer
and a large AXYZ three-axis CNC router. The 3D
printer uses fused deposition modeling technology
to build up thin layers of ABS plastic to create physical
objects. In opposition to the 3D printer which builds
up material, the CNC router removes material with a
rotary cutting tool to shape an object from a block
of material. The two different technologies have spe-
cific advantages and limitations. The CNC router can
cut large objects up to 2440mm x 1220mm, but is lim-
ited to three axes of movement and therefore is not
capable of creating "under cuts". The 3D printer has
essentially no limits to the complexity of geometry
that it can produce, but it is limited to objects smaller
than 254mm x 254mm x 305mm. Relatively small ob-
jects of complex geometry are therefore more suited
to the 3D printer. As a quick proof of concept to
demonstrate to the various parties involved in the
rehabilitation process what digital fabrication is, a
gargoyle from the courtyard of the West Block was
digitized using photogrammetry. The gargoyle was
scheduled for repair because the wings were miss-
ing and required replacement. A maquette of the ex-
isting conditions was fabricated at 1:10 scale using
a 3D printer. The ABS plastic gargoyle was well re-
ceived, but it was learned that the masons working
on the West Block work with 1:1 maquettes of dam-
aged sculptural elements.
The next experiment was to fabricate, at a scale
of 1:1, two sculptural elements - both in states
of disrepair - with the intention that the masons
could make use of these digitally fabricated maque-
ttes. The first was a 1200mm wide sandstone tym-
panum that had eroded significantly since its cre-
ation in the 1860s and required a complete replace-
ment. The tympanum was digitized using Photo-
scan, photogrammetric software and due to its sub-
stantial size and regular geometry was milled us-
ing the AXYZ CNC router from medium-density fibre
board. One side of the tympanum maquette was fab-
ricated to show the current weathered condition and
the other side was speculatively reconstructed using
Rhinoceros modeling software to show what it might
have looked like when originally constructed (Fig-
ure 2). The second sculptural element was a pilaster
capital from the courtyard of the West Block, with
a broken volute that required only repairing of the
missing portion of the volute. The capital was digi-
tized using a Creaform VIUscan handled laser scan-
ner, and due to its freeform geometry and smaller
scale was fabricated using the 3D Printer. The size
Digital Heritage 1 - Volume 1 - eCAADe 32 | 645
 Figure 2
The digital model of
the tympanum in
Rhinoceros (left)
and the CNC-milled
medium-density
fibre board
maquette (right).

of the build area of the printer meant the capital
was divided and printed in four pieces (Figure 3).
While these modelling technologies and workflow
were discussed with project managers and contrac-
tors working on the West Block rehabilitation project,
the masons, while impressed with the technological
feat of digitizing and fabricating the tympanum and
capital, argued that their traditional methodologies
were better suited for the task at hand.
OWL AND THISTLE RELIEF SCULPTURE
The West Block project is a wholesale rehabilitation
with significant construction infrastructure in place
including extensive scaffolding. Removing stones
from the façade and even dismantling portions of the
building is part of the rehabilitation process. The East
Block on the other hand, is operating under normal
day-to-day conditions without any construction in-
frastructure in place. As part of the ongoing preser-
vation work undertaken by PWGSC, a single sand-
stone relief sculpture was identified on the façade of
the East Block to be replaced due to its deteriorating
condition. The sculpture features an owl among this-
tles carved into a 1725mm x 1120mm piece of Berea
sandstone (Figure 4). It is found above the entrance
to the courtyard on the 1910 wing. CIMS was ap-

Figure 3
The capital in situ
on the West Block
(left) and the 3D
printed maquette
of the capital (right).

646 | eCAADe 32 - Digital Heritage 1 - Volume 1

Figure 4
The owl and thistle
relief sculpture in its
current weathered
state.
proached to collaborate with the Dominion Sculptor
and HCD on the task of creating a replacement carv-
ing, building on the initial investigations of the gar-
goyle, tympanum and capital, with the aim of extend-
ing the workflow beyond the fabrication of maque-
ttes to include digital fabrication of the stone itself.
The owl and thistle sculpture is in poor condi-
tion, and actively deteriorating. The vertical face-
bedding of the stone, combined with decades of ero-
sion and freeze-thaw cycles have resulted in much
of the carved detail in the stone de-laminating and
spalling off. This is most evident on the face of
the owl, and the lower right corner of the sculpture,
which has completely disappeared.
Building on the workflow developed for the West
Block projects, the digitally-assisted stone carving
process developed for the East Block begins with cre-
ating a copy of the extant stone carving. Rather than
using a plaster casting process to create a copy, how-
ever, the relief sculpture was digitized using pho-
togrammetry. A series of photographs were taken
with a D-SLR from a man-lift elevated to access the
sculpture on the second story of the East Block. Us-
ing PhotoScan photogrammetric software, a poly-
gon mesh model of the relief sculpture was gener-
ated. The mesh was optimized in Geomagic Studio
in preparation for digital fabrication. From that dig-
ital model, a copy of the sculpture was milled from
high-density polyurethane foam using the AXYZ 3-
axis CNC router at the Carleton University School of
Architecture. The foam maquette, taking the place
of the plaster copy, served as the basis for the Do-
minion Sculptor to rebuild the damaged areas with
modelling clay (Figure 5). With the modelling clay-
overlay applied, the maquette was digitized again
using a Creaform VIUscan close-range laser scanner.
Digital Heritage 1 - Volume 1 - eCAADe 32 | 647
 Figure 5
The foam maquette
of the relief
sculpture with
modelling clay
applied.

Figure 6
The digital mesh
model inclusive of
the sculptor’s work
and digitally
constructed
moulding.

648 | eCAADe 32 - Digital Heritage 1 - Volume 1

 Knowing in advance that the clay work would be digi-
tized the sculptor repaired two narrow sections of the
surrounding moulding at the top and bottom of the
moulding, rather than the entire moulding. Profiles
of the narrow sections were extracted from the scan
data and were used to digitally reconstruct the entire
moulding.
From this second digital model (Figure 6), which
incorporates the clay work of the sculptor and the
digitally reconstructed moulding, the relief sculpture
is being milled from sandstone using a 6-axis indus-
trial robot. The robot will mill the sandstone to within
1.5mm of the surface of the digital model (Figure
7). This provides material for the Dominion Sculp-
tor to remove using traditional stone carving tools
in the process of adding details and textures to the
entire surface of the relief sculpture. To ensure the
end result of the robotic milling has the appropriate
level of detail - enough for the sculptor to work with-
out a pointing machine, but not unnecessarily ex-
cessive and time consuming - a series of test pieces
were milled from foam to determine suitable milling
parameters. After the sculptor's hand finishing is
complete, the relief sculpture will be installed in the
façade of the East Block building on Parliament Hill.
Figure 7
Robotic milling of
the sandstone in
progress.
RESULTS
The integration of digital technologies into the work-
flow brings many benefits in the context of the East
Block relief sculpture. One of the benefits of the
process is in creating the initial copy of the existing
stone sculpture. The traditional mould-making pro-
cess is difficult to execute in-situ, on a vertical sur-
face and requires the erection of scaffolding for sev-
eral days. The digitization process requires only a
man-lift for a few hours to take photographs or use
a handheld laser-scanner. Additionally there is no
physical contact with the stone if using photogram-
metry and minor contact of small adhesive targets
if using the Creaform VIUscan. These non-contact
technologies eliminate residues left on the surface of
the stone from the silicone rubber moulding mate-
rial. For a sculptural element that is too brittle and
may be damaged during the moulding process, digi-
tal acquisition technologies provide an alternative.
The work of the Dominion Sculptor was reduced
in the reparation of the surrounding moulding, as de-
scribed previously. The digital construction of the
moulding also allowed for a quick revision to the pro-
file of the bottom section of the moulding after con-
cerns about water drainage were raised. The mould-
ing was revised digitally to ensure water sheds off
the bottom of the moulding instead of collecting and
contributing to the deterioration of the stone.
Most importantly, the digital technologies ex-
tend the capabilities of the Dominion Sculptor with-
out supplanting them. All of the processes in the
workflow that require a high level of skill or hand-
craft remain in the domain of the sculptor. The clay
sculpting work and the appearance of the missing or
damaged elements such as the feathers and the face
of the owl, both require the same level of skill and
knowledge of stone carving required whether digi-
tal technologies are used or the traditional workflow
is followed. The interpretation of the original tool
marks and the final finish applied to the stone us-
ing traditional stone carving tools are not replaced by
the robot. The rebuilding work in clay by the sculp-
tor provides the geometry of the sculpture for the
Digital Heritage 1 - Volume 1 - eCAADe 32 | 649
 robotic milling process, but the texture and pattern-
ing of the clay is not intended to be the finish of the
stone. The final tool marks can only be practically
added by hand.
DIGITAL ANALOGIES
The production of stone sculpture has a long history
of sculptor's employing others, whether apprentices
within their own studio or artisans outside their stu-
dio, to perform a variety of tasks including plaster
casting and stone carving (Putnam 1939; Rich 1956).
A typical workflow for a sculptor might be to create a
clay maquette of a sculpture, have a stone carver per-
form the roughing out process with a pointing ma-
chine, and then the sculptor finish the stone carving
process themselves. The Dominion Sculptor works
essentially on his own, without the benefit of appren-
tices within his studio. Without apprentices to per-
form the roughing out process, the robotic milling
process is seen as a digital surrogate. Similarly, laser
scanning can be seen as a digital surrogate for the
use of the pointing machine. A laser scanner is essen-
tially a digital pointing machine measuring the posi-
tion of millions of discrete points on the surface of the
stone or maquette, whereas the pointing machine re-
quires manual input to measure the position of a sin-
gle point.
Another process that is analogous to a traditional
stone carving technique is the digital construction of
the moulding surrounding the owl and thistles. Tra-
ditionally, the replacement of a stone moulding be-
gins with the creation of a profile traced from the ex-
isting stone moulding. The profile is used to make
a template which guides the stone carving process
of the new moulding. In the case of the digitally-
modelled moulding, the first step was to extract a
profile from the scan data from the two narrow sec-
tions of the moulding repaired by the Dominion
Sculptor. The extracted profile was then used to gen-
erate the new moulding via the "sweep" command in
Rhinoceros.
650 | eCAADe 32 - Digital Heritage 1 - Volume 1
CONCLUSION
The work on the East Block relief sculpture used a
hybrid digital/analogue workflow that takes advan-
tage of the benefits of digital technologies, without
inhibiting the skill of the Dominion Sculptor. The pro-
cess of applying the hand finish has not begun, but
the workflow has thus far proven to be a viable one.
Additionally, the workflow need not be carried out in
its' entirety to be of benefit. Based on the available
technologies and the level of skill and knowledge of
sculptors and/or stone carvers available for a given
project more emphasis may be placed on the digi-
tal techniques or on the traditional ones. For exam-
ple, based on available craft-labour, creating a copy
of the existing stone using the traditional silicone-
mould and plaster casting process might be more
feasible than the digital process, but at the same time
there are no stone carvers available for the roughing
out process. In this case, a plaster copy is created tra-
ditionally, repaired using clay by the sculptor, but the
stone is milled robotically and hand finished by the
sculptor. Alternatively there may be a scenario that
the stone is extremely brittle and cannot be phys-
ically contacted, but robotic milling of the replace-
ment stone is not feasible. The workflow can be mod-
ified so that the existing stone sculpture is digitised
using photogrammetry, and a maquette is created
using 3D printing or CNC routing, but the complete
stone carving process is done traditionally by hand.
The workflow developed for the East Block
project is intended for the replacement of sculptural
stone elements in their entirety. Further research is
planned in the area of sculptural stone repairs, where
portions of the sculpture are repaired by attaching
newly carved pieces of stone to the existing sculp-
ture using metal rods and epoxy. This was the case of
the gargoyle and capital described previously. Other
areas of further research include exploring the limits
of digital reconstruction prior to fabrication similar to
the example of the tympanum, and the surrounding
moulding on the East Block project.
REFERENCES
Fatuzzo, G, Mussumeci, G, Oliveri, SM and Sequenzia, G
2011, 'The “Guerriero di Castiglione”: reconstructing
missing elements with integrated non-destructive
3D modelling techniques', Journal of Archaeological
Science, 38, pp. 3533-3540
Gilboa, A, Tal, A and Shimshonic, I 2013, 'Computer-
based, automatic recording and illustration of com-
plex archaeological artifacts', Journal of Archaeolog-
ical Science, 40, p. 1329–1339
Grimmer, AE 1984, A Glossary of Historic Masonry Deterio-
ration Problems and Preservation Treatments, Depart-
ment of the Interior National Park Service Preserva-
tion Assistance Division, Washington
Guarnieri, A, Pirotti, F and Vettore, A 2010, 'Cultural Her-
itage Interactive 3D Models on the Web: An ap-
proach Using Open Source and Free Software', Jour-
nal of Cultural Heritage, 11, p. 350–353
Miller, A 1948, Stone and Marble Carving, Alec Tiranti Ltd.,
London
Putnam, B 1939, The Sculptor's Way: A Guide to Modelling
and Sculpture, Farrar & Rinehart Inc., New York
Rich, JC 1956, The Materials and Methods of Sculpture, Ox-
ford University Press, New York
Staley, RD, McGrath Jr, HT, Overholt, DC and Whitmore,
TR 2009, 'Making the Project Real: Changes in De-
livery of Concept to Construction', Association for
Preservation Technology International, 40, pp. 51-58

Digital Heritage 1 - Volume 1 - eCAADe 32 | 651

652 | eCAADe 32 - Digital Heritage 1 - Volume 1
Investigating a narrative architecture
Mackintosh's Glasgow School of Art

Danilo Di Mascio1 , Tom Maver2

1
Department of Architecture, G. d'Annunzio University, Chieti-Pescara, Italy 2 Mackintosh
School of Architecture, The Glasgow School of Art, Glasgow, Scotland
1
http://www.danarchitect.com 2 http://www.gsa.ac.uk/
1
ddimascio@danarchitect.com 2 T.Maver@gsa.ac.uk

In this paper a theoretical approach and a methodology to investigate and

document a narrative architecture will be presented. In the architecture field the
word narrative is often used in reference to ways of representing or telling the
story of a project. Within the context of this research with the expression
"narrative architecture" the authors mean an architecture that, like a book, tells a
story through its material and immaterial characteristics. In order to analyze the
selected characteristics of a narrative architecture, a possible approach is
represented by a digital 3D reconstruction and a critical analysis of the digital
model produced. The digital reconstruction process and the theoretical
background have been applied to a case study, a masterpiece in the history of
architecture: Mackintosh's Glasgow School of Art. Several graphic works have
been created to communicate the main characteristics of this narrative
architecture. The research project attempts to explore ways to study and explain
existing buildings from new and innovative points of view, but at the same time it
can trigger important reflections in architectural design and education.

Keywords: Digital reconstruction, narrative architecture, representation and

visualization, critical analysis, 3D modeling

MOTIVATION ergy savings) at the expense of other considerations,

In any good work of architecture (including vernac-
ular architecture, hence not only works designed by
professional architects) the aesthetic, constructive,
functional and cultural aspects should always be bal-
anced. Superficial implementation of the theories
elaborated during the modern movement by many
architects after the Second World War and the recent
focus on only a few sustainability issues (such as en-
have caused a gradual aesthetic and cultural impov-
erishment of many new buildings. A building can be
functionally efficient but at the same time uninhab-
itable because it may create discomfort to its inhabi-
tants. The aesthetic-cultural aspects of a building are
certainly more difficult to evaluate compared to com-
putable parameters such as energy consumption be-
cause they require a holistic approach and do not

Digital Heritage 1 - Volume 1 - eCAADe 32 | 653

 present well defined boundaries, therefore a unique
interpretation and solution does not exist. Following
Bernard Tschumi, we believe that even if architecture
has to do with the building activity, it can't be re-
duced to it; it is instead a series of concepts added
to the building (De Michelis 2003). The best archi-
tecture designed and built throughout the course of
history on any continent, in addition to responding
to specific functional requirements, bears aesthetic-
cultural qualities that communicate meanings and
ideas that make the architectural experience mean-
ingful, exciting and pleasant. A possible way to com-
prehend these qualities, in order to reproduce them
in new buildings, is through an in-depth study of
architectural precedents. This research path should
avoid the pitfalls of approaches mainly focused on
architectural styles and shapes. In this paper a differ-
ent way of reading buildings is proposed, focusing on
their narrative qualities.
NARRATIVE ARCHITECTURE
The word "narrative" is usually used within fields such
as literature and cinema and it has been diffused
in the architecture field since the mid-1980s (Coates
2011). Recent research and publications have ex-
plored its connections with architecture (Psarra 2009;
Coates 2011), with slightly different interpretations.
In the architecture field the word "narrative" is of-
ten used in reference to ways of representing or dis-
cussing a project. Within the context of this re-
search project, with the expression "narrative archi-
tecture" the authors mean an architecture that, like a
book, tells a story through its material and immate-
rial characteristics such as the morphology, the con-
structive choices, the materials, the arrangement of
spaces, the lighting system, the colours, the decora-
tions and so on. The story can be intentionally told by
a piece of architecture, namely the architect through
the architecture gives hints of his/her intention for
the story. All the elements chosen by the architect
should be functional and coherent to the story that
he/she wants to tell. However, the visitor can inter-
pret the building, and hence the story, with a cer-
654 | eCAADe 32 - Digital Heritage 1 - Volume 1
tain degree of freedom, because the building can
also tell something that wasn't originally planned by
the architect and the same elements can be inter-
preted in slightly different ways. Moreover, over time,
a building may be subjected to interventions that can
slightly change the original design of the architect, al-
terations that can also influence the narrative of the
piece of architecture. Many similarities exist between
a book and a building, in effect both have an author
and a visitor/reader. Furthermore, an architecture,
like a tale, is also realized to provoke reactions and
emotions.
OBJECTIVES OF THE RESEARCH
The main objectives of the research are the follow-
ings:
• to define a theoretical framework, methods
and tools to investigate a narrative architec-
ture;
• to test the theoretical approach and the
methodology previously elaborated on a case
study and document its narrative characteris-
tics.
The investigation, documentation and communica-
tion of narrative architecture is interesting for several
reasons:
• to improve and deepen our understanding
of aesthetic, constructive, functional and cul-
tural values of architecture;
• to represent, communicate and disseminate
the aforementioned values and increase pop-
ular awareness of this approach to architec-
ture. The understanding and appreciation of
these aspects can help to make people aware
of the importance of safeguarding architec-
ture for present and future generations.
Furthermore, this research and its outcomes will
be useful in education and practice, representing a
methodology to study different kinds of architecture
and to design new buildings. To reach the intended
objectives, digital technologies will be used because,
as demonstrated in other research by the authors,
they allow us to reveal aspects of the buildings im-
possible to perceive from traditional drawings and
models alone.
OUTLINING A THEORETICAL FRAMEWORK
TO STUDY A NARRATIVE ARCHITECTURE
A building is a complex system, a fusion of material
and immaterial characteristics. Much like a book, a
piece of architecture can be read and interpreted in
many ways but, considering its complexity, it is nec-
essary to use a holistic approach, a fusion of theo-
ries, methods and tools to properly analyze and doc-
ument it. Hence, in this research paper the word
"fusion" is interpreted in two different, but interre-
lated, ways. It should be noted that the aim is to
read mainly those aspects that generally can be per-
ceived through direct experience of a building. How-
ever, it is clear that the final result of a work of ar-
chitecture is also influenced by many other aspects
linked to society, politics, economics, religion, cul-
tural movements and so on. Taking into account our
objectives, which material and immaterial character-
istics of an artifact should be analyzed? Consider-
ing the definition of narrative architecture that we
have previously stated, in which way and through
which elements can an architect tell a story and com-
municate specific meanings? According to what has
been stated and written by various scholars, includ-
ing Bruno Zevi (1957), space has been considered as
the main characteristic of architecture. It is possible
to experience a building's space by moving through
the various spatial units that constitute it. This se-
quence of spaces can be explored using various tra-
jectories. Everything looks very much like the con-
cept of narrative path. In fact, a narration is charac-
terized by the exposition of an event and its tempo-
ral sequence that can be chronological or not. In our
case the event is represented by the piece of archi-
tecture (in all its aspects) while its temporal sequence
is constituted by the spatial arrangement of the var-
ious rooms. The spaces are discovered and experi-
enced in sequence, then the architecture is revealed
like the plot of a story and conveys emotions while it
discloses itself. But unlike a written story, a narrative
architecture usually has an open structure that does
not require a specific sequence to be followed in or-
der to read it. The same concept of narrative archi-
tecture suggests an operational methodology to an-
alyze it.The first analyses should affect all the aspects
strictly linked to space and movement, namely:
• Spaces: dimension and shape of the various
spatial units, this characteristic directly affects
the shape of the envelope and of its inner par-
titions (for example, walls);
• Connectivity: connections between the vari-
ous spatial units;
• Functional arrangement: organization and
placement of the various spaces;
• Movements: dynamic of movement that uses
horizontal and vertical paths to connect the
various spaces.
To these characteristics linked to the spaces, other el-
ements that help to highlight, strengthen or enrich
the story that the designer wants to tell and suggest
are added. All these new aspects can be observed
and/or experienced through direct knowledge of the
piece of architecture. The characteristics considered
most relevant in this case are the following:
• Constructive system: the constructive tech-
nology used (the assembling system and so
on);
• Materials: the materials used, their position
and their surface treatment;
• Light: the natural and artificial lighting
sources, shadows;
• Decorations: the various decorative elements
used;
Digital Heritage 1 - Volume 1 - eCAADe 32 | 655
 • Colours: the kind of colours used and where
they are employed.
The selected characteristics represent the main aes-
thetic, constructive, functional and cultural aspects
that contribute to telling the building's story. The
importance during an architectural experience of
some of the previously mentioned characteristics (to-
gether with solids and cavities, scale and propor-
tion, rhythm, and other architectural aspects) have
been described in Rasmussen (1962). To study a
narrative architecture hence it is necessary to also
study its structure, the elements that constitute it, as
both aspects (narrative and structure) are strictly con-
nected to each other. Generally the constructive so-
lutions should be used to translate an idea in a spa-
tial conception. Taking into consideration the quan-
tity and the variety of the required analysis to study
a narrative architecture, which approach should be
adopted?
DIGITAL RECONSTRUCTION
In order to analyse the selected characteristics of a
narrative architecture, a solution is represented by
a digital 3D reconstruction and a critical analysis of
the digital model produced. The validity of this ap-
proach and methodology for comprehending vari-
ous aspects of buildings has already been proven by
one of the authors in different case studies (Di Mas-
cio 2013).The objective of the digital reconstruction
is the realization of a 3D model, that offers the pos-
sibility of analyzing a high number of aspects of the
piece of architecture. But the digital reconstruction is
a process that has autonomous value, independent
from further analysis. As Frischer (2008, available at
[1]) states:<<Models not only illustrate what we knew
when we started creating them, they also have the
potential of revealing new knowledge that was al-
ways lurking below the surface of the facts but which,
to emerge and be grasped, needed to be visualized in
3D.>> Hence, the same reconstruction process con-
tributes to deepening and broadening knowledge of
the artifact. This is an interactive process, because,
after the analysis and interpretation of the available
656 | eCAADe 32 - Digital Heritage 1 - Volume 1
documentation related to an artifact and during the
digital reconstruction phase, the scholar can verify
the correctness of his interpretations: hence he re-
ceives feedback from the model and reacts accord-
ingly. During the realization of a digital 3D model it
is essential to select a particular level of abstraction
useful to achieve the selected objectives. A synthetic
model allows us to exclude unnecessary information
in order to focus only on the narrative characteristics.
The digital reconstruction process and the theoreti-
cal background previously described have been ap-
plied to a case study: Mackintosh's Glasgow School
of Art.
THE CASE STUDY: THE GLASGOW SCHOOL
OF ART
General description
The art school is located in Glasgow, Scotland, and it
was designed in two different phases. Even if a visitor
is unaware that the building was designed and built
in two different periods, the reading/observation of
the exterior can trigger his/her interest and curios-
ity, as the building tells something about this aspect:
each of the four façades are completely different from
one another. For the design of many architectural el-
ements (windows, doors, etc.), Mackintosh took in-
spiration from Scottish vernacular houses and cas-
tles, but he also got design ideas from other sources
such as Japanese architecture. The building repre-
sents both a school of art and the Glasgow School of
Art, namely a specific building, located in a particular
place in a certain city, in a particular culture at a given
period of time (MacMillan 1988). The building is con-
siderable in size and it is composed of five main floors
(from the sub-basement up to the second floor), plus
mezzanines. From the outside it appears as a myste-
rious fortress and inside it presents a magical atmo-
sphere, suited to a place where art is produced. The
variety of constructive solutions, spaces, decorations
and types of light, make the GSA a building rich in
moments of experience (Figure 1) and worth an in-
depth analysis with new methods and tools. Even
if the overall atmosphere and characteristics of the
Figure 1
Pictures of the
Glasgow School of
Art (from left to
right): the West
Facade, the
Museum and the
East Facade (Source:
personal archive of
the author).
building remain close to what they were when it was
built, there are a few differences that influence the ex-
perience of the art school, for instance non-original
doors inserted at the beginning and at the end of the
corridors (Buchanan 2012) that interrupt spatial and
visual continuity.
In this paper, the first step of the research project
will be presented, namely the digital reconstruction
(still ongoing) of the building's spaces, paths and
technical elements. Unfortunately, because of an un-
lucky event (the explosion of a projector in the base-
ment of the building) that caused a ferocious fire on
the 23rd of May 2014, the west wing of the School
has been heavily damaged. During the fire, particu-
larly important rooms were destroyed, such as the fa-
mous library (an art-nouveau jewel), the lecture the-
atre and Studio 51. Therefore work on this case study
will be even more valuable. The authors had the op-
portunity to directly experience the whole building
throughout many inspections during one of the co-
authors' period as Visiting Researcher at the Mackin-
tosh School of Architecture (MSA) in 2013, looking at
key areas and details, including many that were lost
in the fire, in a critical way and documenting them
with sketches and pictures.
Methodology
The methodological process can be synthesized in
the following main phases:
• Collection of documentation;
• Analysis of the collected documentation;
• Digital reconstructions (interpretation, mod-
eling and organization of the elements);
• Representation and organization of the infor-
mation.
Collection and analysis of documents
One of the main issues that arose at the beginning
of the research was in regards to which version of the
building should be modelled: as designed, or as built,
and at what date. Taking into account the main ob-
jective of the research (that also requires a physical
experience to understand and document the percep-
tual characteristics), we decided to model the GSA as
close as possible to its current state. The absence of
either CAD drawings or a complete set of traditional
drawings reflecting the current status of the building,
led the authors to use and match various sets of draw-
ings collected from different sources, namely: plans,
elevations and sections of the design proposal, dat-
ing back to 1910; a new set of 2D drawings made for
Digital Heritage 1 - Volume 1 - eCAADe 32 | 657
 a publication in 1993 (Macaulay 1993), three longi-
tudinal sections (hand drawn) created for a disserta-
tion at the Mackintosh School of Architecture, and a
physical 3D model currently located in the GSA shop
and described in Robertson (1995). This documenta-
tion has been complemented by continuous on-site
inspections. To organize and manage the collected
documents a metafile has been created.
Digital reconstruction of the Glasgow
School of Art
The collected and scanned drawings were imported
into AutoCAD to redraw them and create a vector
basis for the digital 3D reconstruction. In the 2D
CAD environment, after placing the drawings in cor-
respondence between them, thanks to the support
of vertical and horizontal axes, it was realized that
many discrepancies were presented. Taking into con-
sideration these issues and the different selected ele-
ments that compose a narrative architecture, instead
of drawing the architectural/technical elements, the
authors first chose to draw the volumes of all the
spaces. This choice was made in order to obtain a first
general idea of the morphology of the whole build-
ing, and an overview of the volumes of the spatial
units and their relationships and hierarchy. This pro-
cess was accomplished creating a sort of virtual pa-
per model of the building, putting together a set of
2D planes, each one with a cut off scanned draw-
658 | eCAADe 32 - Digital Heritage 1 - Volume 1
ing (plans, elevations and sections), arranged in the
proper position. The completed paper model was
used as a basis to create the volumetric model (Fig-
ure 2). During the next step, different colours were
assigned to each individual spatial unit to highlight
the different functions of the spaces (Figure 3).
The modeling process then focused on the indi-
vidual spatial units and precisely on those elements
that due to their shapes, materials, and constructive
charateristics can catch visitors' attention and hence
attract their interests during the path. These ele-
ments can be defined as 'narrative episodes'. The
CAD drawings produced and prepared in AutoCAD,
specifically the first floor plan and one section, were
imported into Rhino and used as a reference dur-
ing the modeling phase. A 2D-3D drafting software
package like Rhino was used instead of a 3D modeler
like 3DSMax, because more precision was needed as
well as better control of the geometric shapes dur-
ing the redrawing process and because during the
digital reconstruction phase, the 2D drawings in the
same working environment are being continually up-
dated. During the digital reconstruction process, be-
cause of the complexity of the building, it has been
essential to undertake many inspections of techni-
cal elements and diverse unclear details and to docu-
ment them with pictures, sketches and metrical sur-
veys. Hence the authors decided to concentrate on
Figure 2
(left) the digital
model realized with
the scanned
drawings; (right)
the volumetric
model under
construction.
Figure 3
(left) the volumetric
model with the
colours that
highlight the
different functions
of each single
spatial unit; (right) a
section of the
volumetric model
that shows the
different shapes of
some spatial units.
the core space of the GSA building, namely the Mu-
seum located on the first floor. The Museum in ad-
dition to being a connection node is a spatial and
narrative episode. The visitor goes into the entrance
hall that is a narrow space, a bit oppressive with its
vaults, and relatively dark, but immediately his/her
attention is caught by the light that floods down the
central stairwell. This light is an invitation to come up
and discover a huge space that the visitor could never
imagine before. The main element that characterizes
this huge exhibition space is its ceiling, and in partic-
ular its roof timber trusses. This distinctive element,
like all the other roof trusses presented in the other
places in the school, present a specific design solu-
tion; Mackintosh, in fact, avoided standard engineer-
ing iron components. The precision tools in Rhino
were very useful to model the different timber com-
ponents, 37 in total, that constitute each roof truss.
The elegant shape of the roof trusses does not repre-
sent a simple reply to structural needs but it evident
that they also have an aesthetic quality expressed
through their details (hearts cut into them), shape,
materials and constructive technique. With the lat-
ter aspect we specify how the single pieces are con-
nected together, because this aspect also affects the
aesthetic quality of the technical element and hence
of the space and of the narrative experience. These
details are like the sentences in a novel, the quality
of writing affects the quality of the narration. The
same approach has been used to model and inves-
tigate the paths, the other spaces and their charac-
teristics.
Representation of the information
A piece of architecture cannot be described through
a single method and a single tool. Every represen-
tational method fills the gaps presented in the oth-
ers. The 3D models have also been created taking
into consideration which graphic works have to be
prepared. All the spaces have been classified in a
scheme with information about their dimensions and
shapes. Through this representation it is made ev-
ident that the story told by the shape of the sec-
tions of the spaces is much more varied and excit-
ing compared to the one told by the shape of the
plans, mainly rectangles. Many diagrams and draw-
ings were prepared to represent and communicate
paths and movements inside the building. The main
paths have been represented with both 2D plan view
and 3D rendered views. For technological details
such as the roof trusses, different analytical draw-
ings have been prepared, including axonometric ex-
ploded views, to appreciate the formal and construc-
tive characteristics inspired by the Japanese joints,
and a rapid prototyping model has also been man-
ufactured directly from the Rhino file (Figure 4). With
all the graphic works produced, a visual narrative will
be prepared.
Digital Heritage 1 - Volume 1 - eCAADe 32 | 659

DISCUSSION OF THE RESULTS
The concept of narrative architecture, together with
the methodological framework that has been elab-
orated to analyse it, has proven to be effective to
comprehend in a proper way the main qualities of
the Glasgow School of Art designed by C.R. Mack-
intosh. The direct experience of a piece of archi-
tecture is an invaluable moment because it allows
one to comprehend the building through the senses,
hence to enhance the understanding of many impor-
tant aspects before and during the process of anal-
ysis. The digital reconstruction process of the GSA
(still ongoing) and the digital 3D models produced
thus far allowed for an in-depth study of some of the
main aspects that characterize a narrative architec-
ture, namely: spaces, functional arrangement, con-
660 | eCAADe 32 - Digital Heritage 1 - Volume 1
Figure 4
(upper left) 3D
digital model of the
whole roof
structure of the
museum; (upper
right) rendered
view of the roof
truss; (lower left)
rapid prototyping
model of the roof
truss; (lower right)
exploded view of
the roof truss.
nectivity, movements and constructive characteris-
tics of a few important technical elements. The 3D
model gives a comprehensive overview of the inte-
rior and the exterior of the whole building, which is
impossible to have with direct experience alone; in
fact through a 3D model, it is possible to have a visual
overview of the entire piece of architecture. In this
way, the model is confirmed as a precious interpre-
tative tool and not a mere representation. The volu-
metric model that shows all of the spatial units, gives
an immediate view of the shape, dimensions, posi-
tion and function of each individual space and proves
its usefulness in understanding the relationship be-
tween the various spaces located on different floors.
During the digital reconstruction process it was pos-
sible to hypothesize and test constructive/technical
solutions adopted in certain points, defined narra-
tive episodes (for instance the roof timber trusses and
the main window in the director's room), which could
be difficult to analyze and document in other ways
because of their difficult geometry and shape. Dur-
ing the digital reconstruction work, major technical
and management issues have also been overcome,
thanks to the high number of 3D elements in the
model. The 3D models, together with the other an-
alytical drawings, document and communicate new
and in-depth information on the Art School that de-
spite the vast number of publications on Mackin-
tosh and his masterpiece, have never been analysed
or shown before, therefore it is original content. It
is impossible to make a single physical model that
presents at the same time a high number of interior
and exterior details. Furthermore, a physical model is
difficult to modify and update and it limits the num-
ber of analyses that can be undertaken, except for
the ones planned at the beginning of the work. In
contrast, a digital 3D model does not present any
limit, it is open to any new modification and update
and it can be enriched with different kinds of infor-
mation. The digital model representation can also
serve as the base to produce other digital models and
analytical drawings; in this way there will always be
a metrical correspondence between all the graphic
works, hence a higher precision and clarity in the doc-
umentation and communication of the different as-
pects.A 2D - 3D drafting software program like Rhino
turned out to be a proper solution in this kind of dig-
ital reconstruction. This software presents a rich set
of tools that are very useful for managing and cre-
ating every kind of shape; for this reason Rhino is
very frequently used in fields such as industrial de-
sign and marine design. Various aided tools (such as
a wide number and type of snaps), absent and not ef-
fective inside pure 3D modeling software packages,
have been very useful to overcome the discrepancies
between the different drawings and the geometric
complexity of many technical elements in the Mack-
intosh building. Furthermore, Rhino allows the cre-
ation of rapid prototyping models directly from the
digital 3D model with the aim of exploring and com-
municating the building characteristics for a possi-
ble future exhibition. The set of models and graphic
works (both those already produced and those yet to
be produced) will be used to create a visual narrative
to tell the narrative architectural story of this build-
ing. In past centuries, the model has always been
the most understandable representation of a build-
ing for people outside the field of architecture, for
this reason it also has its own inherent value as au-
tonomous work. It is planned to use the same model
within a game engine to investigate and document
the experience of the space. As already stated and
investigated by one of the authors, videogames with
a first person viewpoint allow for movement inside a
space and observation in real time, providing an ex-
perience and a freedom of movement much closer to
reality (Di Mascio 2010). This aspect will be further
developed in future papers.
CONCLUSIONS AND FUTURE DEVELOP-
MENTS
Within this research paper, theories, methods and
tools to analyse narrative architectures have been
investigated, elaborated and proposed. The whole
methodological process requires objective and sub-
jective interpretations. With this paper the authors
aim to highlight the importance of understanding
the narrative aspects of architecture, including per-
ceptual aspects in addition to material ones, in or-
der to improve the quality of the architectural expe-
rience and hence of our daily life. The cultural as-
pects will be explored in greater depth in future pa-
pers. The core idea is the re-reading of the architec-
ture in our historical and contemporary built environ-
ment using a new angle, not limited to the functional
aspects. The creation of a 3D digital model opens
new opportunities in the analysis and communica-
tion of architecture. The digital reconstructions make
evident the main elements of the narrative architec-
ture namely: paths, spaces, the hierachical relation-
ship between the various spatial units, the aesthetic-
constructive qualities of some elements and so on.
Digital Heritage 1 - Volume 1 - eCAADe 32 | 661
 Traditional representational methods and tools alone
are insufficient for analysing and communicating the
characteristics of a narrative architecture appropri-
ately, whereas when starting from the space, new
tools made available by digital technologies allow
various paths of analysis. As the case study, Mackin-
tosh's famous Glasgow School of Art, demonstrates
so well, masterpieces pertaining to the history of ar-
chitecture are like the classics of literature: they al-
ways have something new to say (Calvino 2013). In
fact during a re-reading of a masterpiece it's always
possible to discover something meaningful. The re-
search project attempts to explore ways to study and
explain existing buildings from new and innovative
points of view, but at the same time it can trigger
important reflections in architectural design and ed-
ucation. The same methodology will be applied to
two other buildings of the GSA campus, the Bourdon
building (current home of the Mackintosh School of
Architecture) and the new Reid Building designed
by Steven Holl, officially inaugurated on the 9th of
April 2014. A comparison between these three build-
ings is also part of the research schedule. During the
next phases of the research endeavour a further in-
depth analysis and development of the theoretical,
methodological and experimental approaches will
be considered. The theoretical and methodological
approach that has been developed in this research is
suitable to analyze the narrative of other architectural
works around the world. The same approach can be
also used to analyse damaged, unbuilt or lost build-
ings.
ACKNOWLEDGEMENTS
D. Di Mascio would like to thank his co-author Prof.
Tom Maver and Prof. Chris Platt (Head of the Mack-
intosh School of Architecture) for the opportunity
to undertake a research period at the Mackintosh
School of Architecture, Glasgow School of Art. The
authors also wish to thank all the following people,
from various departments and academic institutions,
for making so much useful information on the Mack-
intosh building available: Mark Baynes, Jared Ben-
662 | eCAADe 32 - Digital Heritage 1 - Volume 1
jamin, David Buri, Paul Chapman, Paul Clarke, Stu-
art Jeffrey, Robert Proctor, Pamela Robertson, Johnny
Rodger, and Peter Trowles.
REFERENCES
Buchanan, W (eds) 2012, Mackintosh's Masterwork. The
Glasgow School of Art, Bloomsbury Publishing, Lon-
don, United Kingdom
Calvino, I 2013, Why Read the Classics, Penguin, London,
United Kingdom
Coates, N 2011, Narrative Architecture, John Wiley & Sons
Ltd, Chichester, United Kingdom [Kindle version] Re-
trieved from Amazon.it
Frischer, B 2008, 'The Rome Reborn Project. How Tech-
nology is helping us to study history', OpEd, Novem-
ber 10, 2008. University of Virginia, PDF document,
pp. 1-5
Macaulay, J 1993, Glasgow School of Art. Charles Rennie
Mackintosh, Phaidon Press Limited, London, United
Kingdom
MacMillan, A 1988, 'Architetto e Designer', in Laganà,
G (eds) 1988, Charles Rennie Mackintosh 1868-1928,
Electa, Milano, Italia, pp. 9-14
Di Mascio, D 2010 'Learning from Videogames Level De-
sign: An Educational Point of View for Architec-
ture', Proceedings of the 10th International Confer-
ence on Construction Applications of Virtual Reality
(CONVR2010), Sendai, Japan, 4 -5 November 2010,
pp. 199-208
Di Mascio, D, Pauwels, P and De Meyer, R 2013 'Im-
proving the knowledge and management of the his-
torical built environment with BIM and ontologies.
The case study of the book tower', in CONVR2013
(eProceedings of the 13th International Conference on
Construction Applications of Virtual Reality), London,
United Kingdom, 30 - 31 October 2013, pp. 427-436
De Michelis, D 2003, 'Intertestualità. Intervista con
Marco De Michelis', in Damiani, G (eds) 2003, Bernard
Tschumi, Rizzoli, Gruppo Skira, Milano, Italy, pp. 19-
24
Psarra, S 2009, Architecture and Narrative. The formation
of space and cultural meaning., Routledge, Abingdon
(Oxon), United Kingdon
Rasmussen, SE 1962, Experiencing Architecture, MIT Press,
Cambridge (Massachusetts), United States
Robertson, P 1995, 'Anatomy of the Glasgow School of
Art', MAC Journal, Mackintosh School of Architecture,
2, pp. 59-63
Zevi, B 1957, Architecture as space: how to look at archi-
tecture, Horizon Press , New York, United States
[1] http://www.romereborn.virginia.edu/rome_re-
born_2_documents/papers/Frischer_OpEd_final2.pdf

Digital Heritage 1 - Volume 1 - eCAADe 32 | 663

664 | eCAADe 32 - Digital Heritage 1 - Volume 1
Index of Authors
A
Abdel salam, Hassan 2-89 C
Abdelmohsen, Sherif 2-473 Çağdaş, Gülen 1-201, 1-227
Achten, Henri 1-181, 2-23, 2-237, Carey, Brad 2-395
2-411, 2-621 Carrara, Gianfranco 1-613
Agkathidis, Asterios 2-71 Carreiro, Miguel 2-271
Agnew, Brian 1-119 Carvalho, Nelson 2-271
Akiyoshi, Koki 1-527 Castro e Costa, Eduardo 1-585
Aksoy, Yazgı 1-227 Çekmiş, Aslı 1-277
Al Faleh, Omar 2-577 Celani, Gabriela 1-33, 2-315
Alborghetti, Paolo 2-97 Charlton, James 1-109
Alwan, Zaid 2-437, 2-515 Chatzikonstantinou, Ioannis 1-287, 2-39
Anagiotou, Despo 1-153 Christensen, Jesper Thøger 1-239
Anderson, Ian 1-313 Cifcioglu, Ozer 2-217
Androutsopoulou, Eirini 1-191 Clayton, Mark 2-385, 2-445
Araújo, Bruno 1-585 Cokcan, Baris 2-131
Asut, Serdar 1-401 Colakoglu, Birgul 1-305
Avis, Charles 1-173 Coutinho, Filipe 1-515
Aydin, Serdar 1-603 Coyne, Richard 2-185
Azevedo, Sérgio 2-271 Cruz, Joana 2-271
Cursi, Stefano 1-613
B
Baerlecken, Daniel 2-107 D
Baldassari, Matteo 2-587 D'Alpuim, Jorge 2-271
Barekati, Ehsan 2-385 Dalton, Ruth 1-213
Beetz, Jakob 2-375 Danker, Fabian 2-525
Beirão, José Nuno 1-33 Davidová, Marie 2-139, 2-153
Bergin, Michael 2-455 de Castro Fernandes Abreu
Bielik, Martin 2-303 ,→ Gonçalves, Afonso Maria 2-327
Blümel, Ina 2-557 de Vries, Bauke 1-455
Booth, Geoffrey 2-445 Decker, Martina 2-179
Brath, Mads 3-365 Di Mascio, Danilo 1-653
Braumann, Johannes 2-131 Dieckmann, Andreas 2-417
Brell-Cokcan, Sigrid 2-131 Dinçer, Ahmet Emre 1-201
Brown, Andre 1-623 Ding, Yan 1-137
Brune, Markus 1-109 Diniz, Nancy 1-63
Burry, Jane 1-137, 1-165, 1-267 Donath, Dirk 2-303
Burry, Mark 1-137, 1-165 Dörstelmann, Moritz 1-423, 2-61
Buš, Peter 1-181 Doulgerakis, Adam 1-53

Index of Authors - Volume 1 - eCAADe 32 | 665

 D'Uva, Domenico 2-81 Hanna, Sean 2-539
Hansen, Ellen Kathrine 2-613
E Hayes, James 1-643
Economou, Athanassios 2-357 Heller, Alfred 1-49
El Ahmar, Salma 2-595 Hess, Trevor 1-71
ElBanhawy, Eiman Yehia 1-213 Holden Deleuran, Anders 2-29
Eloy, Sara 2-271 Holgate, Peter 2-437
Erbas, Irem 2-39 Hollberg, Alexander 2-303
Erdine, Elif 1-145, 1-373 Hubers, Hans 2-39
Erhan, Halil I. 1-247
Erioli, Alessio 2-97 I
Evers, Henrik Leander 2-29 Ibrahim, Magdy 2-427
Ibrahim, Mohamed 2-337
F
Fai, Stephen 1-643 J
Fereos, Pavlos 1-433 Janssen, Patrick 1-79, 2-547
Figueiredo, Bruno 1-585 Jensen, Mads Brath 1-297
Fink, Dietrich 1-43 Jensen, Rasmus Lund 2-117
Fioravanti, Antonio 1-593, 1-633, 2-595 Johansson, Mikael 2-261
Fjeld Edvardsen, Dag 2-375 Jones, Oliver 2-525
Florián, Miloš 2-23 Jones, Paul 2-437
Foged, Isak Worre 1-297, 2-117, 2-365 Jorge, Joaquim 1-585
Fok, Wendy W 1-391 Juvancic, Matevz 1-99
Fonseca, Fernando 1-585
K
G Kachri, Georgia (Zeta) 2-539
Gargaro, Silvia 1-593 Kallegias, Alexandros 1-373
Genganagel, Christoph 2-29 Kalogianni, Dorothea 2-185
Georgiou, Michail 2-169 Kepczynska-Walczak, Anetta 1-565
Georgiou, Odysseas 2-169 Kerkidou, Maria 1-53
Gerber, David 1-463 Kim, Jong Bum 2-445
Gledson, Barry 2-515 Kjems, Erik 2-495
Gokmen, Sabri 2-107 Knight, Michael 1-623
Grasl, Thomas 1-11, 2-357 Kocaturk, Tuba 2-71
Greenberg, Evan 1-145 König, Reinhard 2-567
Grigoriadis, Kostas 2-123 Kontovourkis, Odysseas 1-411, 1-153
Gronier, Guillaume 2-227 Kreutzberg, Anette 2-253
Gsandtner, Martin 2-139 Krijnen, Thomas 2-375
Guerriero, Annie 2-227 Kruger, Mario 1-515
Kulcke, Matthias 2-245
H Kuma, Taichi 2-61

666 | eCAADe 32 - Index of Authors - Volume 1

Kurilla, Lukáš 1-181, 2-23 Negendahl, Kristoffer 2-49
Kwiecinski, Krystian 2-207 Nielsen, Timo Harboe 1-537
Kwok, Theresa 2-169 Norell, Daniel 2-145
Kyjanek, Ondrej 1-423 Nováková, Kateřina 2-237

L O
Lancia, Ruggero 1-313 Ochmann, Sebastian 2-557
Larsen, Andreas Lund 2-117 Oliveira, Sancho 2-605
Laskari, Anna 2-283 Osório, Filipa 2-605
Latifi, Mehrnoush 1-267
Leggieri, Elisabetta 1-593 P
Leitão, António 1-257 Paio, Alexandra 2-405, 2-605
Li, Di 1-623 Pak, Burak 1-89, 1-365
Liang, Hai-Ning 1-63 Pantazis, Evangelos 1-463
Lo, Tian Tian 1-603 Paredes, Miguel 1-333
Lorenz, Wolfgang E. 1-495 Pasold, Anke 2-365
Love, Peter 2-395 Patlakas, Panagiotis 2-505
Pechlivanidou-Liakata,
M ,→Anastasia 1-53
Marques, Sibila 2-271 Pedro, Tiago 2-271
Martens, Bob 1-11 Pedro Sousa, José 1-475
Martins, Pedro Filipe 1-475 Pereira, Vasco Alexandre 2-405
Matejovska, Dana 2-411 Perkov, Thomas 2-49
Matthews, Jane 2-395 Petzold, Frank 1-43
Maver, Tom 1-653 Pinto Duarte, José 1-515, 1-585
Melville, Stephen 1-537 Prado, Marshall 1-423, 2-61
Mendes, Daniel 1-585 Proença, Pedro 2-271
Mendes, Letícia 1-33 Proença, Sara 1-257
Meneghel, Rafael 2-315 Prohasky, Daniel 1-137, 1-165
Menegotto, José Luis 2-195
Menges, Achim 1-423, 2-61 Q
Menter, Adam 2-455 Qian, Zi 1-119
Morton, David 1-343 Quinn, Gregoory 2-29
Mostafavi, Sina 1-445
Moural, Ana 2-271 R
Moya, Rafael 1-137, 1-165 Rahmani Asl, Mohammad 2-455
Mühlhaus, Michael 1-43 Ramsgard Thomsen, Mette 1-381
Mullins, Michael Finbarr 2-613 Remy, Nicolas 1-487
Myrup, Morten 2-375 Rodhe, Einar 2-145
Rossini, Francesco 1-633
N Roupé, Mattias 2-261

Index of Authors - Volume 1 - eCAADe 32 | 667

 Russell, Peter 2-417 Tong, Hakan 1-201
Rutherford, Peter 1-323 Trento, Armando 1-633
Triemer, Michel 2-303
S Tryfonos, George 1-411
Sagias, Alexandros 1-53 Tschetwertak, Julia 2-303
Salehi, Mahsa 1-267 Tsiliakos, Marios 1-433
Sales Dias, Miguel 2-271 Turrin, Michela 2-39
Salet, Theo 1-455
Salman, Huda 1-573 V
Sanchez, Rodolfo F. 1-247 van der Zee, Aant 1-455
Sarhan, Ahmed 1-323 Varoudis, Tasos 2-293, 2-505
Sariyildiz, Sevil 2-217 Ventura, Davide 2-587
Sauda, Eric 1-71 Verbeke, Johan 1-89
Scherer, Raimar J. 2-485 Viklund Tallgren, Mikael 2-261
Schmitt, Gerhard 2-567 Vilar, Elisângela 2-271
Schnabel, Marc Aurel 1-603 Vinsova, Ivana 2-411
Schneider, Sven 2-303 Vock, Richard 2-557
Schubert, Gerhard 1-43 Voordouw, Johan 1-557
Schwinn, Tobias 1-423 Vrouwe, Ivo 1-365
Sedrez, Maycon 2-315 Vyzoviti, Sophia 1-487
Seifert, Nils 1-43
Šichman, Martin 2-139 W
Simeone, Davide 1-613 Wang, Tsung-Hsien 2-465
Sing, Chun-pong 2-395 Watkins, Simon 1-137, 1-165
Słyk, Jan 2-207 Wessel, Raoul 2-557
Soni, Shubham 1-129 White, Phil 1-643
Sproat, Iain 1-537 Wojcik, Marcin 2-159
Standfest, Matthias 1-505, 2-567 Wurzer, Gabriel 1-11
Stapleton, Kieran 2-515
Stasiuk, David 1-381 Y
Stouffs, Rudi 1-79, 2-347 Yan, Wei 2-455
Strumillo, Jan 2-159 Yazici, Sevil 1-547
Yunis, Leyla 1-423
T
Tamke, Martin 2-29, 2-375, 2-557 Z
Tanaka, Hiroya 1-527 Zarrinmehr, Saied 2-445
Tanti, Matthew 1-445 Zarzycki, Andrzej 1-357, 2-179
Tapias, Estefania 1-129 Zhou, Jingyang 2-395
Tauscher, Helga 2-485 Zubairu, Tijjani 2-217
Thompson, Emine Mine 1-213, 1-119 Zupancic, Tadeja 1-99
Toldo, Ilaria 1-613

668 | eCAADe 32 - Index of Authors - Volume 1