Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

Annual Review of Control, Robotics, and

Autonomous Systems
Construction Robotics:
From Automation to
Collaboration
Stefana Parascho
Lab for Creative Computation (CRCL), École Polytechnique Fédérale de Lausanne (EPFL),
Lausanne, Switzerland; email: stefana.parascho@epfl.ch

Annu. Rev. Control Robot. Auton. Syst. 2023. Keywords


6:183–204
construction robotics, architecture, sensing, industrial robots,
First published as a Review in Advance on
November 28, 2022 human–robot collaboration, automation
The Annual Review of Control, Robotics, and Abstract
Autonomous Systems is online at
control.annualreviews.org Over the past decades, robotics has shown great potential to impact the built
https://doi.org/10.1146/annurev-control-080122- environment, from automation to differentiation and efficient construction.
090049 However, construction processes are highly complex and require tackling a
Copyright © 2023 by the author(s). This work is multitude of problems, from safety and robustness to ease of control and
licensed under a Creative Commons Attribution 4.0 interactivity. For this reason, the field of construction robotics is still evolv-
International License, which permits unrestricted
ing, requiring finding solutions for new challenges every day. The present
use, distribution, and reproduction in any medium,
provided the original author and source are credited. review analyzes the role of robotics in construction and architecture over
See credit lines of images or other third-party time and highlights current trends in shifting from pure automation toward
material in this article for license information.
collaborative and adaptive processes that have the potential to fully integrate
robotics into a rigid and challenging industry, such as construction.

183
1. INTRODUCTION
The construction industry is in a moment of crisis. It is one of the largest CO2 producers in the
world and is known to be inefficient and unstructured. Construction labor is usually not valued
highly, and unless workers are explicitly protected, they are often underpaid and work in unsafe
conditions. While there is a clear and urgent need to address these challenges, construction is an
industry that is very slow to change. An example is digitalization and automation: While these
advances have seen rapid development in other industries, such as manufacturing, retail, trans-
portation, and even agriculture, construction is lagging far behind in adopting them (1). Robots
have entered almost all industries in different roles, from automating major tasks, such as pick-
and-place and sorting processes, to engendering new processes through technologies that exceed
our own capacities, such as sensing and artificial intelligence (2).
Whether automation is the answer to the construction industry’s challenges is questionable
and may be too simple of an assessment. The intricacies of such a massive industry that impacts
our planet from the first extraction of a material up until the change of its configuration are too
complex to be solved through automation alone. However, robotics and information technology
have the potential to contribute to solutions to these tremendous challenges. And if we look at the
impact that robotics has had on other industries, it becomes evident that bringing change to the
construction industry will require reflecting on new technologies and considering their potential.
The reasons why the construction industry has not adopted robotics and automation as quickly
as others are multifaceted. The following is an attempt at a nonexhaustive list of possible factors
affecting this shortcoming:

1. Scale: Compared with many other industries, such as automotive and product manufactur-
ing, the construction industry operates at a much larger scale. It thus cannot easily adopt
the machines and processes developed specifically for desktop- or human-scale processes.
While in most industries the robot (or machine) will be larger than the manufactured prod-
uct, construction requires a different perspective to allow it to also produce elements larger
than the production machine.
2. Off-site versus on-site construction: Many construction processes and materials require on-
site work, while most manufacturing industries operate in controlled factory environments.
Although automation can be employed in prefabrication scenarios, this usage covers only
a limited area of the construction industry, missing out on the impact potential of on-site
construction.
3. Predictability: Construction sites are complex, unstructured environments that are prone to
unexpected change and unpredictable situations. This makes the use of autonomous robots
a much greater challenge since the robots need to accurately perceive a largely unknown
environment and be able to react to it. The level of robustness required to achieve this is
far beyond that of a conveyor belt type of process.
4. Human–robot interaction: The complexity of construction processes requires the involve-
ment of human labor. Humans are extraordinarily good at dealing with complex tasks and
making fast evaluations of unpredicted events. As such, they are an inherently necessary part
of a construction site. Employing robots in an unstructured environment around humans
requires a robust and redundant safety system that can react appropriately to unexpected
changes. The challenge of an infallible safety system exceeds by far those of other industries.
5. Lot size and quantity: Each building is unique or is intended to be unique. Of course, stream-
lined prefabricated housing systems exist, but the typical goal of architects is to propose
tailored solutions to each client and site demands. Developing a construction process, im-
plementing and testing it, and adding the necessary layers of safety redundancy and failsafe

184 Parascho
operation means developing a new solution each time. This makes automation inefficient
for individual results in terms of both cost and time.

To sum up, design and construction are best explained as multifaceted ecosystems that are not
comparable to industrial manufacturing processes, which can take advantage of the large number
of identical objects produced, the manageable scale, and the controlled environments.
But if automation turns out to be so difficult to achieve, why has the industry still pushed to
integrate robotics into its processes for the past 40 years? The reasons have changed along with
the discourses and crises that we have been facing. Today, the goal is no longer simply to replace
humans on-site for efficiency, but expands into finding new solutions to existing challenges. Po-
tential advantages or goals of robotics in construction and design are as follows (this is again a
nonexhaustive list, and will hopefully expand as more technology is integrated into construction):
(a) time and cost efficiency; (b) safety for humans (robots can be used for tasks that are dangerous,
tedious, or unhealthy); (c) the capacity to perform tasks that humans cannot do, such as plac-
ing objects precisely, holding heavy objects, and differentiating assembly or material processing;
(d) the capacity to rethink architectural design and its impact (e.g., using robots to save mate-
rial by building complex but efficient shapes); and (e) the potential to leverage the capabilities of
both robots and humans in creative processes in an attempt to broaden the range of solutions that
robotics offer (discussed further in Section 5).
The history of construction robotics is a rather short one, going back only about 40 years.
However, looking at the development of the field over time shows that through close integration
of different disciplines (robotics, engineering, architecture, and computer science), we are pro-
gressing in overcoming the challenges listed above and giving robotics a chance to impact the
construction industry, hopefully for the better.
This review aims at highlighting the way that the focus, intentions, and goals have changed
over time and where construction robotics is situated today. It is thus structured around topics
that evolved naturally, with a clear tendency to move away from simple automation and toward
collaborative and more informed processes.

2. CONSTRUCTION AUTOMATION
Robotics has been introduced into the construction sector mainly with the intention to automate
existing processes, increase efficiency, and decrease costs. As a result, the first three decades of
robotic construction work focused on finding uses for robots in existing construction tasks (3).
However, how designers have taken advantage of the capabilities of robotics beyond au-
tomation should not be underestimated. Even since the beginnings of binary computer systems,
scientists have highlighted the potential of automated machines not only for automating existing
tasks but also for the realization of differentiated structures. Konrad Zuse (4) discussed the au-
tomatic construction of individual objects, particularly in the construction sector, as a main goal
for automation. In architecture, in 1969 Konrad Wachsmann had arguably already conceptual-
ized the first machine for automatic placement, the Location Orientation Manipulator, which was
capable of placing elements in individual locations (although not practically implemented) (5).
The desire to look beyond automating known tasks toward making new designs and applica-
tions possible by making use of robots’ capabilities is thus not a new one—it has accompanied
the development of robotic machines from their very beginning.
Still, the first step in developing a construction robotics process typically implies automating a
known process. From robotic rebar placement on-site (6) to mobile bricklaying robots (7–10), the
history of robotic construction has followed numerous automation processes (Figure 1). Today,

www.annualreviews.org • Construction Robotics 185


Figure 1
Mechanical bricklayer, 1967. Figure reproduced from Reference 8 with permission from British Pathé.

we still see numerous projects focused on automating existing processes in order to make con-
struction processes faster, cheaper, and safer. While fundamental automation processes perform
simple motions (pick and place, milling, bending, etc.), more recent attempts range from these
basic instructions (bricklaying, spatial structures, material placement, and material extrusion) to
highly complex systems requiring sensing and custom control for autonomous functioning.
It should be mentioned here that the term robot is very generic and encompasses a vast palette
of different types of machines. However, this section discusses only the use of industrial robots,
which have become widely popular in construction due to their versatility, availability, and low
cost. The industrial robots used in construction vary in size (from a reach of 0.5 m to more than
4.0 m) and number of axes (from five to nine, if external linear or rotational axes are included), but
they are all anthropomorphic manipulators conceived as a sequence of joints and links that allow
their tool center point to reach multiple positions and orientations in their workspace. Sections 3
and 4 discuss the use of other machines, such as mobile machines and custom robots.
To identify the range of state-of-the art robotics applications in construction automation, it is
important to look at the different types of tasks that can be performed by robots. In additive tasks,
the process is often based on pick-and-place procedures or continuous movements, both of which
are actions that robots are widely used for in other industries. Such processes include bricklaying
(11, 12), assembly of spatial structures (13, 14), robotic welding (15), robotic fiber placement (16),
and 3D printing (17, 18). Subtractive processes are used less widely but are present especially in
timber fabrication, either on their own or combined with additive robotic processes (19). While
industrial robots provide a cheaper alternative to computer numerical control (CNC) milling (es-
pecially five-axis milling) and potentially an increase in the size of the processed component, they
come with limitations in terms of metric precision, applicable force, and dimensional reach. How-
ever, when combined with robotic assembly, the use of the robot overcomes the need to reposition
and reprocess the working piece, which may also decrease tolerances and allow for more design
possibilities (20). Applications of subtractive robotic processes range from cutting and sawing (21–
24), milling (25–27), and drilling (20) to carving (28). Besides additive and subtractive processes,

186 Parascho
we can also delimit deformative processes, such as wire bending (29, 30), metal sheet bending or
folding (31, 32), and incremental sheet forming (33). These processes are used mostly in academic
research in an attempt to create new and complex shapes.
The ideal of automating existing processes usually comes with a top-down workflow. The de-
sired design of the structure or building is first defined according to the requirements posed by
the specific construction task (site, functional requirements, cost, etc.), after which the specific
robotic instructions are defined (e.g., for bricklaying: pick up a brick from a stack, move it up to
a given height, move it above the placement position, lower it to the final position, and release
it). These instructions are typically predefined for every element to avoid having to calculate the
exact procedures during the construction and keep maximum control of the processes. In a last
step, these instructions are executed by the machine, and the structure is realized.
Top-down processes are used mainly due to their simplicity, the lack of a requirement for
complex sensory procedures such as computer vision, and the minimal expertise required to ex-
ecute them. While this type of procedure can be very robust for certain construction processes,
unexpected problems often arise during fabrication, especially if considering the complexity of
a construction site. Material tolerances can add up, small inaccuracies in the robotic setup can
equally lead to large shifts between the design and the real structure, and changes in the envi-
ronment of the structure might lead to unexpected collisions or inaccuracies. With a completely
predefined process, one does not have the capacity to react to any changes, whether small toler-
ances or large unexpected deviations, and the design and fabrication procedures are 100% bound
to what can be anticipated and safely executed. There are different ways to counteract some of
the issues of a top-down process, from using highly precise materials and site conditions (pre-
fabrication lends itself to that) to using materials that allow for deformation and inaccuracies
(e.g., viscous cement, elastic elements, or glue between bricks) or monitoring the state of the
construction and making adjustments on the spot. The latter option is discussed in Section 4,
which presents ways in which construction processes can be made more reactive and informed.
However, those approaches are no longer considered top-down processes, since with the input of
information, the initially predefined process needs to be altered, by adjusting either the design or
the robotic instructions.
With automation also come questions of labor and human workers. Many automation processes
have the sole purpose of making the construction process faster and cheaper and removing human
labor from the construction site (or the prefabrication environment). An argument in favor of this
type of automation is that it makes construction sites safer, by employing robots for tasks that
are tedious or dangerous for humans, and using them to enable new construction processes that
humans cannot perform due to the required accuracy, payload, or time (34). However, such argu-
ments should not be taken out of context, and they do not constitute any absolute truths. While the
use of robots has enabled new designs and construction methods (e.g., differentiated brickwork,
individual fiber placement, and construction of doubly curved geometries from discrete elements),
this does not release us from asking what effect automating the construction industry has on the
human workforce (35). In addition, pure automation often loses the advantages that humans bring
to a construction process, such as quick reaction to changes by maintaining a clear overview of
the entire construction process and fast adaptability. An argument against human–robot collab-
orations, especially on-site, is the difficulty of ensuring a 100% safe working environment when
working with high-payload machines and dangerous processes. However, as shown in Section 5,
there have been many attempts to bring both humans and robots into construction processes and
take advantage of the complementary skills of each agent to solve the complexity of automated
construction processes.

www.annualreviews.org • Construction Robotics 187


3. OFF-SITE VERSUS ON-SITE
One of the biggest challenges for construction robotics today is moving beyond the controlled
laboratory setup and onto the construction site. Parts of the construction process can be done
in prefabrication processes, such as the fabrication of individual parts of a building, which lend
themselves perfectly to robotics and automation. However, many construction processes cannot
be executed in prefabrication scenarios, due to weight, material, or transport limitations (e.g.,
monolithic concrete constructions, bricklaying, or on-site welding of large components). Arguably
the most challenging part of the construction process for robotics applications is what happens
on-site, whether that is on-site construction or just assembly of prefabricated elements. To see
robotics fully integrated into construction, we need to find solutions for these challenges while
identifying where prefabrication can be used most effectively.

3.1. Off-Site Prefabrication


While early robotics applications in construction focused on automating on-site processes (as
discussed in Section 2), off-site prefabrication is still the predominant modality for employing
robots in construction. The analogy to other manufacturing industries’ setups—such as assembly
lines, for which industrial robots were developed—is evident: Prefabrication can be performed
in a controlled environment of a factory, where every step of the process is precisely planned,
executed, and monitored. Everything from the atmospheric conditions to the number of people
present can be carefully controlled to allow for a safe and efficient process. In addition, a con-
trolled fabrication setup allows for the usage of multiple different machines whose interaction can
be well overseen and planned [e.g., a robot and a stationary table saw (36)]. Examples of robotic
prefabrication range from the processing and assembly of individual timber sheets (37, 38) or
linear (beam or column) elements (13, 20, 39), the prefabrication of modules of buildings (40, 41),
and the assembly of space frame components (42) to the prefabrication of new material systems,
such as textiles or wound fiber elements (43, 44) and bespoke geometry components (45). Of
course, prefabricated modules are limited by logistic constraints (maximum transportable sizes)
and thus can address only a limited part of the construction process. Oftentimes, the assembly is
performed manually in a second step on-site (46, 47).
Another limitation of robotic prefabrication processes is the variation in geometry, material,
or fabrication method that is encountered in a construction project. For this reason, robotics are
employed in specific material industries (e.g., timber manufacturing), especially for projects that
work with similar dimensions of elements and similar processes (48, 49) (Figure 2). In particular,
industrial robots lend themselves to modular construction, in which the same element is produced
multiple times.

3.2. On-Site Fabrication


Prefabrication can address several questions in construction, such as automation of tedious pro-
cesses and local differentiation of geometry. However, it does not impact the entire construction
process, since the assembly needs to be performed on-site (50), and certain material processes
do not lend themselves to prefabrication due to size, weight, or fragility during transport (e.g.,
monocoque concrete structures, brick or stone assemblies, or rammed earth structures). Moreover,
on-site fabrication brings numerous additional challenges that cannot be addressed through com-
mon automation processes, such as preplanned robotic instructions. The sections below classify
ongoing robotics research through these challenges.
3.2.1. Scale and payload of setup. A hard limitation of a robotic construction setup is its di-
mension. A stationary industrial robot can only fabricate something that fits into its reach volume,

188 Parascho
Figure 2
The Sequential Roof, digital assembly of a complex roof structure, Zurich, 2010–2016; copyright Gramazio Kohler Research, ETH
Zurich.

which will typically be at most ∼6 m in diameter. To increase the scale of the possible structures,
the setup can be extended by a gantry system or other extension that increases the reach (51).
A process strongly impacted by the size of the robotic setup is on-site 3D printing. Since this
process typically does not require six-axis freedom of movement, it can be simplified to a gantry
system with three to five axes. However, the gantry must first be set up and exceed the dimen-
sions of the final structure (52–55), which of course does not come without a cost. Other attempts
to increase the reach of the machine include setting up a robot on an extended mobile arm (56)
and utilizing mobile machines (see Section 3.2.2). An interesting approach proposed by Gramazio
Kohler Research involves using a robot to catapult material over a certain distance, thus substan-
tially increasing the dimension of the built structure far beyond the size of the machine (57).
There has been a trend toward a different approach that expands beyond simply increasing the
size of the setup, and instead makes use of existing additional material to address the challenge of
robotic reach. Minibuilders are mobile robots that can climb onto the structure they are about to
build (58); in distributed robotic timber construction, the building material is temporarily used
as a robotic link (59); and in collaborative robotic construction with irregular material, the robots
utilize the construction material (bamboo) to propel themselves to new locations (60). In a similar
approach, Fiberbots climb onto the structure they build in order to grow the final structure
(61). These projects elegantly dissolve the clear separation between robot and environment and
provide a new perspective on how construction sites could contribute to a robotic process rather
than limit it.

3.2.2. Mobility and localization. The mobility of the robotic setup is crucial primarily for
transporting the setup on-site and secondarily to allow a robot to access as large an area as possible.
While industrial robots are themselves stationary (unless they are coupled with additional axes
that are tedious to set up on-site), mobile platforms for these robots have been developed (62).
However, these platforms were proposed for precisely leveled and controlled factory sites, not for
rough, quickly changing, uncontrolled construction sites. For this reason, research on relocating
industrial robots before and during the construction process has been going on since the early
2010s.
Possibly the simplest process involves simply installing the robotic setup directly on-site (63).
DimRob was initially developed as a robot that could be moved to a series of fixed locations,
localize itself, and continue fabrication at every movement step (64). Its further development, the

www.annualreviews.org • Construction Robotics 189


In Situ Fabricator, allowed for a more continuous movement through a better tolerance adjust-
ment process (65, 66). In all of these cases, localization was crucial to the successful translation of
the robots from one position to another. Whether this localization is performed manually or by
using computer vision, the goal is typically to identify the exact physical location and orientation
of the machine and the built object with respect to the digital model. While precision is one
of the praised advantages of robots for construction processes, an alternative approach to exact
localization could be the development of soft robotic processes, in which the robot allows for
adjustments in its movement based on the data that it receives locally (67). This local search for
a location could be supported by other types of sensors, as described in Section 4.
Since industrial robots are large, heavy, and generally not designed to relocate themselves,
much recent research has revolved around other types of mobile robots, either preexisting or
custom designed—from drones used to build lightweight structures (68) to smaller, lighter robotic
arms used on mobile movement platforms (69) and large-scale robots developed for landscape
reconfiguration (70) based on the hardware and capabilities of excavators. What is remarkable is
the range in size and hence the variety of construction processes that can be performed by enabling
machines to relocate and occupy multiple positions in space. These processes include assembly
tasks (65), concrete forming (71), and stone stacking and earth displacement (70) (Figure 3).
The alternative approach to turning a generic machine into a mobile one is designing and
constructing custom robots for specific tasks. These robots have the advantage that they can be
streamlined for a given task, thus avoiding the use of unnecessary material or technology, and
can overcome limitations of readily available machines, such as industrial robots (since they can
be smaller, lighter, and more mobile). However, the trade-off is that they will typically not be
easily applicable to different processes and hence need to be cost-efficiently produced or easily
taken apart and reconfigured. Several recent developments have showcased the potential of custom
robots specifically designed to climb up walls, windows, or metal structures (72, 73) or wooden
structures (74).

Figure 3
Construction of a stone wall using an autonomous hydraulic excavator, 2020. The in-progress wall has been overlaid with potential
extensions of the structure generated by a stochastic solver. Figure reproduced from Reference 70 (CC BY 4.0).

190 Parascho
3.2.3. Safety. Construction sites are unstructured environments that are unique and constantly
changing. Imagining a fully automated construction site without any humans involved is not only
unrealistic but also inefficient. Humans are good at reacting to unexpected change and complex
information, while machines need to be programmed specifically to react to any type of exter-
nal information, and thus have more difficulty working with unexpected adjustments. Making use
of human capabilities on-site, for both the construction process itself and the decision-making
process, can increase both efficiency and the capabilities of robotic construction processes. How-
ever, when humans are working alongside machines, safety is an essential concern that has always
limited the application of robots on construction sites (75). The ever-changing conditions and
unexpected adjustments make developing a safety strategy for robots and humans on-site quite
challenging. Of course, there are strategies for human–robot collaboration in general, but they
are not yet robust enough to be applied to an environment as complex as a construction site.
All in all, robotic construction is moving from prefabrication laboratory environments to con-
struction sites. However, addressing all of the complex challenges that such an environment offers
will require improving the way in which robots and humans can interact and developing new con-
trol methods that allow robots to react as spontaneously as humans. Section 5 further discusses
human–robot collaboration scenarios.

4. INCREASING ROBOTS’ CAPABILITIES


Construction automation processes are typically preplanned sets of robotic instructions that are
then executed by robots. However, if the process is unidirectional, it assumes that the desired
and/or simulated action will perfectly match reality. This quickly leads to limitations that result
from material and setup tolerances, changes of the environment, and unpredicted human behavior.
To compensate for the many variables that a construction process brings, some type of check
of the construction process is indispensable. Using sensing capabilities can go beyond a simple
quality check and allow robotic construction processes to be more autonomous, controlled, or
even creative.
One reason for the widespread use of blind industrial robots, besides the simplicity of use, is
the readily available control systems that industrial robot providers offer. These are usually offline
control systems, in which a required movement is defined in simulation software, transferred to
the robot, and executed without feedback. Another barrier in integrating more complex robotic
control systems is the software packages typically employed in the field of construction and design.
Since the fields rely on computer-aided design (CAD) software rather than robotic control and
simulation, the access to the latter is limited to plug-ins and libraries developed by third-party
providers that link robotics software with the CAD software available in construction. Examples
include plug-ins for Rhino and Rhino Grasshopper (76, 77), such as Robots IO (78), HAL (79),
Taco (80), and the open-source libraries compas_fab (81) and compas_rrc (82), which allow for a
simple use of Robot Operating System (ROS)–based algorithms from the comfort of a known 3D
modeling environment. Behavioral robotics is one approach that explicitly does not predefine each
individual movement, but instead defines the behavior of the robot based on input data (83, 84).
Still, even in what might count as simple industrial processes such as pick-and-place procedures,
sensing is being employed more and more to increase precision and speed and decrease the number
of errors. Vision is often used in fast pick-and-place processes to avoid presorting the elements or
to simplify the process of picking up imprecisely placed objects (85). All in all, to tackle the evident
limitations of construction robotics, increasing the information flow between designer/operator
and machine in both directions is absolutely necessary. For this reason, much recent research
in robotic construction has focused on integrating more and more sensing capabilities and

www.annualreviews.org • Construction Robotics 191


developing processes that make use of information exchange rather than improving unidirectional,
top-down automation processes. At the core of ongoing construction robotics research are optical
and visual sensors (such as 2D and 3D cameras), tactile sensors (such as force–torque sensors), and
other feedback approaches.
Optical and visual sensor applications range from simple distance measurements to 3D camera
scans. Point and line distance sensors are often used to identify the exact location of a flat surface,
such as the height of a placed brick (86) or the distance to the ground. That way, material toler-
ances, which are often substantial at a construction scale—a brick’s dimensions can vary by up to
5 mm—can be easily dealt with without needing to increase the precision of the model. Similarly,
this procedure is used to make up for robotic setup tolerances that are a result of either the me-
chanical tolerances of the machine or imprecise measurements of the tool or workspace. Because
the precision at which construction sites operate lies in the centimeter range, the goal cannot be to
develop methods that guarantee submillimeter precision; instead, the construction process needs
to be configured so that it can adjust to those tolerances.
2D and 3D cameras are often used for the localization of a mobile machine or of a built object
(71) and for monitoring the accuracy during a construction process (87). Typically, the impre-
cisions registered by the camera are translated back into the 3D model so that the subsequent
construction steps can be corrected and accurately performed, an approach that also avoids the
uncontrolled accumulation of tolerances; to achieve this, the digital model is adjusted every x num-
ber of steps so that the fabrication process can adapt to those changes. The use of 3D scanning data
also allows the processing of existing objects, such as found or irregular material (88), enabling a
new approach to construction that does not rely on materials being processed into standardized
regular shapes and can instead work with on-site and raw materials (26, 45, 89, 90). Similarly, vi-
sual information allows for adaptive robotic processes in which the exact movements are defined
based only on sensory data of the workpiece (67) or processes that can adapt to changes in the
environment (64, 91). This enables a wider application of robotics that can be adjusted to differ-
ent workpieces and starting scenarios, thus lowering the access to this technology for the typically
individual construction applications. Besides enabling robots to work with irregularly shaped ma-
terials or individual workpieces, access to local 3D data opens up the possibility to directly process
existing structures, for example, for disassembly or reassembly (92) and material reuse (93).
To react to tolerances, whether related to the setup or the materials, we do not necessarily need
a detailed visual capture of the built structure; instead, we can disregard the visual appearance and
use force–torque feedback to assess when the robot has reached the desired pose. For example,
force–torque sensors can be used to place objects robotically without defining the exact height of
the placement (94). Sometimes paired with visual sensors, these force–torque sensors enable a new
approach to precision and tolerances, especially when working with material or structures that do
not provide a digital representation. Instead of requiring a detailed virtual model of the structure
that is being built or modified, the process relies solely on the feedback that the robot receives and
adjusts its movements accordingly (95). While this solution may seem simply necessary given the
lack of precise data, it also opens up the opportunity for more flexible and creative construction
processes that can expand beyond the automated processes that we are used to, as they are not
constrained by an initial predefined plan (96).
Another approach that was enabled by the use of sensors is learning a robotic task based on
generated data and experience. The purpose there is to not have to predefine a rather complex
robotic instruction from scratch but rather to allow the machine to learn the movements and
actions that it needs to perform through reinforcement learning techniques (89, 97–99). While
the learning process is often run in simulation, the sensory feedback is used to transfer the results
of the trained model to the real-life process. Such approaches are one step toward making robotics

192 Parascho
control more accessible and less limited to specialists and experienced users, which is necessary in
order to open up their use in an industry with no robotics expertise.
Besides employing sensors on the robot itself, other approaches have also been implemented
to retrieve information about a robotic process. Optical tracking is often used when working with
flying machines (68) but is also useful for industrial robotic setups with low precision (100). While
this technique simplifies the localization process, it requires a fixed setup and controlled condi-
tions (no obstruction, regular light conditions, etc.). As such, its application and calibration on-site
is not easy to implement, and it has therefore been used primarily in laboratory setups. Still,
the process allows for the tracking of the exact movement of the robot or even of a human in-
side the robotic workspace, and it opens up the way for human–robot interaction processes (101,
102) (see Section 5).
To sum up, sensors have become indispensable for robotic processes, regardless of whether the
goal is higher fidelity for automation processes or more potential for interaction and adaptability.
While many sensor approaches have been implemented in laboratory environments, the challenge
remains to bring them on-site in a robust and reliable way and enable more robotic autonomy for
construction. At the same time, their use and application rely strongly on the control setup and
interface and its accessibility to nonexpert users, particularly when moving toward human–robot
collaborative processes. This topic is discussed further in the next section.

5. MULTIAGENT COLLABORATION
The sections above have discussed the variety of construction robotics processes, from predefined
fabrication processes to sensor use and mobile machines. Another factor that has strongly im-
pacted the variety of applications is the number of agents involved in the construction process.
The use of the term agents is deliberate, so as to include not only multirobotic processes but also
processes that involve humans and machines. The collaboration of multiple agents not only speeds
up construction but also enables more complex processes with multiple tasks to be performed si-
multaneously. The increase in possibilities is thus not linear with the number of agents, unlocking
entirely new possibilities where agents can support each other and complement their tasks and
skill sets.

5.1. Multirobotic Fabrication


The use of multiple machines working side by side can be first attributed to the goal of increas-
ing the reach or the speed of production. Machines working in parallel by executing the same
task will multiply the production speed and dimensions of a built structure, as seen in other
industries, such as the automotive one. In construction, for example, the Brick Labyrinth used
multiple robots to place bricks simultaneously (103). However, allowing the robots to work to-
gether, either in the same workspace or in separate workspaces, enables new processes that are not
possible using a single machine. Two robots can, for example, be employed to cooperatively work
on the same structure or element while not sharing the workspace, thus avoiding collision or path-
planning issues. Some processes utilize one robot as a leading agent and have a second one mirror
its movement, for example, to cooperatively wind fiber elements (43, 104). Others make use of
the individual movement ranges of the robots to differentiate the geometric outcome, for exam-
ple, for wire-cutting processes (105, 106) or sheet metal bending (31). In all of these examples,
the use of multiple robots allows for a larger geometric solution space and more control in the
geometries of the fabricated elements.
A closer cooperation can be achieved by allowing the robots to work in the same workspace ei-
ther simultaneously or sequentially. The robots can support each other by taking on different tasks

www.annualreviews.org • Construction Robotics 193


Figure 4
LightVault, CREATE Laboratory, Princeton University, 2020. The glass vault was assembled cooperatively by two industrial robots.

at different moments of the fabrication, such as placing an element and acting as a support when
not used for placement (107). The result is a process that allows for more geometric differenti-
ation and precision without the need of additional support structures (14). Of course, such tasks
come with a larger number of constraints, as the robots can easily limit each other’s movement
space, requiring the use of robotic sequence- and path-planning algorithms (108, 109).
Given the advantages of multirobotic assembly processes, these processes have been suc-
cessfully applied to different materials and processes, such as brick construction (Figure 4),
timberwork, and concrete processing (40, 92, 110). It is important to note that while these pro-
cesses utilize multiple machines, they are rarely aimed at full automation of the construction
processes and employ robots only for tasks that are inherently easier for robots to perform than
humans, such as precise spatial placement and supporting of structures.
Another approach to multirobotics foregrounds the use of multiple small machines, often mo-
bile, instead of large static industrial robots to achieve the construction of structures larger than
themselves. To this end, machines can pass on material from one to another to increase the size
of the built structure (111) and climb onto the structures they built to continue building (112).
The challenges here remain mostly working with the limited payloads of small robots and in-
creasing the sizes of the potential results through smart interactions between the machine and the
environment or construction material. Section 3 discussed similar processes.
A natural progression is from using similar machines cooperatively to using heterogeneous
teams of robots and making use of the different capabilities of different machines. Applications
range from the separation of parallelized tasks [e.g., using smaller robots to prepare stacks of
bricks and larger robots to individually place them (103)] to using mobile machines together with
stationary ones to increase the built structure’s volume [e.g., using flying machines to pass on
material from one large-scale robot to another (44)].
Employing multiple robots also raises questions of control. While the robots are often con-
trolled centrally, with movements or behaviors defined before their interactions with each other,
other control strategies such as distributed control (113) and stigmergic processes (114) are
promising for construction tasks. These strategies allow a robotic system composed of multiple
machines to make decisions autonomously, based on the information they share or information

194 Parascho
from the environment, which makes the entire process more adaptable and robust to changes and
unexpected results (115, 116).

5.2. Human–Robot Collaboration


As discussed in Section 2, the role of humans in construction processes cannot be disregarded.
While full automation might be possible, although extremely challenging, trying to exclude
humans entirely from construction processes would come with a great loss of craft and decision-
making capabilities. Instead, recent research has focused on strengthening the connection between
humans and robots, whether this includes designers, operators, or construction workers. Lacking
humans’ decision-making and adaptation capacities, machines need to be programmed to pro-
cess a given dataset, making it highly challenging to prepare them for the complex interactions in
construction processes. However, robots are clearly an advantage for specific tasks, such as pre-
cisely moving along a trajectory with a controlled speed, placing objects in exact 3D positions
and orientations, and holding objects without moving for defined lengths of time. Collaborative
human–robot processes range from having humans work in the same area as a robot—each per-
forming a specific task—to having humans and robots manipulate elements together and allowing
humans to intervene during construction. Most robotic construction processes typically involve
a human in some way (if only to press an emergency stop button), even if this is not explicitly
described as human–robot collaboration. Robots are often used for placement, while humans per-
form connections between elements (14, 63, 117). In concrete manufacturing, robots are used
for 3D printing or spraying a structure, while humans might apply a finish to the surface or place
reinforcements—although given the right setup, these processes can also be automated (118, 119).
However, we almost always find tasks that are more easily performed by humans, and enabling that
interaction often simplifies and speeds up the process.
Targeted research efforts have aimed to improve the information flow between design, hu-
man, and machine, particularly using augmented or virtual reality (AR/VR) technology. AR/VR is
utilized to visualize the desired target geometry and perform quality checks on tolerances and dif-
ferences between the planned model and the construction outcome (87, 120, 121), but it can also
be used to inform the designer/operator of changes occurring on-site and thus enable an initial
model to be adapted where needed. For processes where humans and robots work next to each
other, AR/VR technology can help visualize the robotic movement and make the workspace safer
(122).
A different method of direct interaction between robots and humans can be achieved through
motion-tracking systems, whether through highly precise systems that use multiple cameras and
scanners or through more accessible but less precise commercial systems. Motion tracking can
be used to enable the operator to interact with the machine based on gestures or simply based
on their movement (102), thus enabling a much more intuitive approach to robot control. So
far, these approaches have been used mostly for art installations and performances showcasing
the technology (Figure 5). A next step would be to employ them for robotic applications that
require high precision. Motion control was also used for interactive systems of custom robots
that go beyond employing the robot as a fabrication agent, toward full-scale robotic structures
that become architecture themselves (123).
Involving humans and robots concurrently in the design decision-making may be the closest
type of collaboration possible in a construction setting. As an emerging approach to both design
and construction, research is focusing on how designers can work directly with the machine in a
dialogue (the robot performs a task, to which the human can react), such that every action influ-
ences the other agent’s next move (96). Based on game theory (124) and incremental design (125),

www.annualreviews.org • Construction Robotics 195


Figure 5
Quipt human–robot collaboration through gesture control, 2017. Project and photo by Madeline Gannon.

this approach allows humans and robots to maximize their creative freedom through interaction
with the constraints set by the other agent. While human–robot collaboration may look very dif-
ferent with respect to task distribution and safety setups, the current trend is to bring humans and
robots closer together to maximize the use of complementary skills to avoid deadlocks and hard
constraints on-site.

6. CHALLENGES AND OPPORTUNITIES


This review has discussed the applications of robotics in architecture and construction as a pro-
gression from automation processes to multiagent interactive ones that come with the promise
of seeing these technologies implemented on-site and beyond. Analyzing the progression of the
most representative terms in research papers over the past 22 years shows a clear trend toward
collaborative processes and human–robot collaborations (Figure 6), which implies that pure au-
tomation is not the sole purpose of the research field. However, there is still a long way to go
to overcome the challenges that robotic construction tasks bring, and many open questions re-
main. The aims of future construction robotics research will likely be to achieve more generalist
processes that can be easily implemented and applied to different tasks, to enable fast adoption
of new robotic technologies in construction, and to make it easier to deploy and test those new
technologies. This is not a challenge lacking solutions in the field of robotics, where many detailed
approaches have already been developed. The true challenge is rather bridging the gap between
different disciplines, enabling better communication among fields, and providing means to easily
translate robotics technology for new applications such as construction.
To achieve these goals, future research will likely need to tackle several different areas:

 Accessibility: Can robotic technology become easier to implement, test, control, and in-
teract with? Are there methods for intuitive control of machines that allow designers and
construction workers to explore the range of possibilities that robotics offers without re-
quiring them to become experts in the field? This does not mean that the construction field
should be completely unskilled with regards to robotics; rather, the question is whether we

196 Parascho
Cobot
1,900 Collaborative robot
1,700 Robot artificial
intelligence
1,500 Human–robot
Number of papers

interaction
1,300 Creative robot
1,100 Robot
Automation
900

700

500

300

100
2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021
Year
Figure 6
Analysis of the most representative terms used in construction robotics research, showing the number of papers published each year on
ScienceDirect that included each term.

can provide designers, engineers, and construction workers with an environment and inter-
face that makes it easy to learn a new technology and experiment with it such that they can
more quickly understand its limits and possibilities and thus its range of applications.
 Versatility: Can we develop processes that are easily adaptable to new materials, environ-
ments, or working conditions? Doing so would drastically reduce the development time and
effort required to set up a single process, since that process would rely on existing adapt-
able ones; however, given the vast range of parameters in a construction task, accomplishing
this goal is still extremely challenging. Making processes more versatile means simplifying
the implementation workflow, increasing the feedback input from the environment, and
developing more generic end effectors and robotic setups.
 Upscaling: The limited scale of robotic setups compared with the scale of construction re-
mains a large challenge to overcome. There is no consensus on whether the research should
focus on developing larger and larger machines or small machines that can extend their own
reach. This is natural, of course, given the vast variety of processes and requirements, such
as payload, precision, and speed.
 Creativity: Allowing for more creativity within robotic processes is a key factor in enabling
the design community to make use of these tools and fully engage with them already in
the design conceptualization phase. Lowering the barrier of interaction with machines and
improving the sensing and communication between humans and machines can lead to new
co-design approaches in which humans and robots support each other at every decision step.

We may find ourselves at one of the most critical points in the field’s history, when it be-
comes clear whether robotics technology can truly impact the design and construction industry
or will remain a niche research area for robot-enthusiastic academics, architects, and engineers.
The key question is whether we will be able to move beyond the most straightforward advantage
that robotics offer, of precisely constructing new and complex geometries, toward overcoming
current challenges of sustainability and resource scarcity. Numerous questions remain. For exam-
ple, can robots be used for applications other than new construction of buildings, and would this
bring true contributions to the question of resource use? Can we make robotics more accessible

www.annualreviews.org • Construction Robotics 197


economically, in terms of energy use, and in terms of the expertise required to increase its availabil-
ity and impact? And can we find ways to combine this technology with the current architectural
and design discourse, so that solutions go beyond solving a single engineering problem and instead
engage with the social, economic, and physical challenges of the field?

DISCLOSURE STATEMENT
The author is not aware of any affiliations, memberships, funding, or financial holdings that might
be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS
I would like to thank Maxence Grangeot for his support in editing this article.

LITERATURE CITED
1. Manyika J, Ramaswamy S, Khanna S, Sarrazin H, Pinkus G, et al. 2015. Digital America: a tale of the haves
and have-mores. Rep., McKinsey Glob. Inst., Washington, DC
2. Int. Fed. Robot. 2021. World robotics 2021: industrial robots. Rep., Int. Fed. Robot., Frankfurt, Ger.
3. Bock T, Linner T. 2016. Site Automation: Automated/Robotic On-Site Factories. New York: Cambridge
Univ. Press
4. Zuse K. 1993. The Computer—My Life. Berlin: Springer
5. Bock T, Lauer WV. 2010. Location Orientation Manipulator by Konrad Wachsmann, John Bollinger
and Xavier Mendoza. In Proceedings of the 27th International Symposium on Automation and Robotics in
Construction, ed. T Brno, pp. 704–12. London: Int. Assoc. Autom. Robot. Constr.
6. Bock T, Linner T. 2016. Reinforcement production and positioning robots. In Construction Robots:
Elementary Technologies and Single-Task Construction Robots, pp. 39–46. New York: Cambridge Univ. Press
7. Andres J, Bock T, Gebhart F, Steck W. 1994. First results of the development of the masonry robot
system ROCCO: a fault tolerant assembly tool. In Automation and Robotics in Construction XI: Proceedings
of the Eleventh International Symposium on Automation and Robotics in Construction, ed. DA Chamberlain,
pp. 87–93. Amsterdam: Elsevier
8. Br. Pathé. 2014. Mechanical bricklayer (1967). YouTube. https://www.youtube.com/watch?v=
4MWald1Goqk
9. James HH. 1967. Brick-laying machine. US Patent 3,659,392
10. Pritschow G, Dalacker M, Kurz J, Gaenssle M. 1996. Technological aspects in the development of a
mobile bricklaying robot. Autom. Constr. 5:3–13
11. Bonwetsch T, Kobel D, Gramazio F, Kohler M. 2006. The Informed Wall: applying additive digital
fabrication techniques on architecture. In Synthetic Landscapes: Proceedings of the 25th Annual Conference
of the Association for Computer Aided Design in Architecture, ed. GA Luhan, P Anzalone, M Cabrinha,
C Clarke, pp. 489–95. Dover, DE: Assoc. Comput. Aided Des. Archit.
12. Bonwetsch T, Bearth & Deplazes, Gramazio & Kohler. 2008. Gantenbein vineyard façade, fläsch. In
SIGGRAPH ’08: ACM SIGGRAPH 2008 Art Gallery, p. 52. New York: ACM
13. Apolinarska AA. 2018. Complex timber structures from simple elements. PhD Thesis, ETH Zürich, Zürich,
Switz.
14. Parascho S, Gandia A, Mirjan A, Gramazio F, Kohler M. 2017. Cooperative fabrication of spatial
metal structures. In Fabricate 2017: Rethinking Design and Construction, ed. A Menges, B Sheil, R Glynn,
M Skavara, pp. 24–29. London: UCL Press
15. Ariza I, Mirjan A, Gandia A, Casas G, Cros S, et al. 2018. In place detailing: combining 3D printing
and robotic assembly. In Recalibration: On Imprecision and Infidelity: Proceedings of the 38th Annual Confer-
ence of the Association for Computer Aided Design in Architecture, ed. P Anzalone, M Del Signore, AJ Wit,
pp. 312–21. Dover, DE: Assoc. Comput. Aided Des. Archit.
16. Prado M. 2020. Skeletal composites: robotic fabrication processes for lightweight multi-nodal structural
components. Constr. Robot. 4:217–26

198 Parascho
17. Hack N, Lauer WV, Gramazio F, Kohler M. 2014. Mesh mould: differentiation for enhanced per-
formance. In CAADRIA 2014: Rethinking Comprehensive Design: Speculative Counterculture, ed. N Gu,
S Watanabe, H Erhan, MH Haeusler, W Huang, R Sosa, pp. 139–48. Hong Kong: Assoc. Comput.-Aided
Archit. Des. Res. Asia
18. Anton A, Jipa M-A, Reiter L, Dillenburger B. 2021. Fast complexity: additive manufacturing for be-
spoke concrete slabs. In Second RILEM International Conference on Concrete and Digital Fabrication: Digital
Concrete 2020, ed. FP Bos, SS Lucas, RJM Wolfs, TAM Salet, pp. 1067–77. Cham, Switz.: Springer
19. Krieg OD, Dierichs K, Reichert S, Schwinn T, Menges A. 2011. Performative architectural morphology:
robotically manufactured biomimetic finger-joined plate structures. In eCAADe 2011: Respecting Fragile
Places, pp. 573–80. Ljubljana, Slov.: eCAADe
20. Thoma A, Jenny D, Helmreich M, Gandia A, Gramazio F, Kohler M. 2020. Cooperative robotic fabri-
cation of timber dowel assemblies. In Research Culture in Architecture: Cross-Disciplinary Collaboration, ed
C Leopold, C Robeller, U Weber, pp. 77–88. Basel, Switz.: Birkäuser
21. Pawlofsky T, Weissmahr T. 2014. 7xstool. kkaarrlls. http://www.kkaarrlls.com/index.php?feature=
editions,7Xstool
22. Self M, Vercruysse E. 2017. Infinite variations, radical strategies. In Fabricate 2017: Rethinking Design and
Construction, ed. A Menges, B Sheil, R Glynn, M Skavara, pp. 30–35. London: UCL Press
23. Pawlofsky T. 2014. Star puzzle cut by robot. Craftwise. http://www.craftwise.ch/star-puzzle-cut-by-
robot
24. Xydis A. 2018. RobArch 2018 workshop. Achilleas Xydis. https://www.achilleasxydis.com/RobArch-
2018-workshop
25. Fleischmann M, Menges A. 2012. ICD/ITKE Research Pavilion: a case study of multi-disciplinary col-
laborative computational design. In Computational Design Modelling: Proceedings of the Design Modeling
Symposium, Berlin 2011, ed. C Gengnagel, A Kilian, N Palz, F Scheurer, pp. 239–48. Berlin: Springer
26. Vestartas P, Rezaei Rad A, Weinand Y. 2021. Robotically-fabricated nexorades from whole timber. In
Proceedings of the International fib Symposium on Conceptual Design of Structures, ed. C Fivet, P D’Acunto,
M Fernández Ruiz, PO Ohlbrock, pp. 539–46. Lausanne, Switz.: fib
27. Bucklin O, Menges A, Amtsberg F, Drexler H, Rohr A, Krieg OD. 2022. Mono-material wood wall: novel
building envelope using subtractive manufacturing of timber profiles to improve thermal performance
and airtightness of solid wood construction. Energy Build. 254:111597
28. Brugnaro G, Hanna S. 2019. Adaptive robotic carving: training methods for the integration of material
performances in timber manufacturing. In Robotic Fabrication in Architecture, Art and Design 2018, ed.
J Willmann, P Block, M Hutter, K Byrne, T Schork, pp. 336–48. Cham, Switz.: Springer
29. Hack N, Lauer WV, Gramazio F, Kohler M. 2015. Mesh mould: robotically fabricated metal meshes
as concrete formwork and reinforcement. In FERRO-11: Proceedings of the 11th International Symposium
on Ferrocement and 3rd ICTRC International Conference on Textile Reinforced Concrete, ed. W Brameshuber,
pp. 347–59. Bagneux, Fr.: RILEM
30. Smigielska M. 2018. Application of machine learning within the integrative design and fabrication of
robotic rod bending processes. In Humanizing Digital Reality: Design Modelling Symposium Paris 2017,
ed. K De Rycke, C Gengnagel, O Baverel, J Burry, C Mueller, et al., pp. 523–36. Singapore: Springer
31. Saunders A, Epps G. 2016. Robotic lattice smock: a method for transposing pliable textile smocking tech-
niques through robotic curved folding and bending of sheet metal. In Robotic Fabrication in Architecture,
Art and Design 2016, ed. D Reinhardt, R Saunders, J Burry, pp. 78–91. Cham, Switz.: Springer
32. Moussavi SM, Svatoš-Ražnjević H, Körner A, Tahouni Y, Menges A, Knippers J. 2022. Design based on
availability: generative design and robotic fabrication workflow for non-standardized sheet metal with
variable properties. Int. J. Space Struct. 37:119–34
33. Kalo A, Newsum MJ. 2014. An investigation of robotic incremental sheet metal forming as a method
for prototyping parametric architectural skins. In Robotic Fabrication in Architecture, Art and Design 2014,
ed. W McGee, M Ponce de Leon, pp. 33–49. Cham, Switz.: Springer
34. Hu R, Iturralde K, Linner T, Zhao C, Pan W, et al. 2020. A simple framework for the cost–benefit
analysis of single-task construction robots based on a case study of a cable-driven facade installation
robot. Buildings 11:8

www.annualreviews.org • Construction Robotics 199


35. García de Soto B, Agustí-Juan I, Hunhevicz J, Joss S, Graser K, et al. 2018. Productivity of digital
fabrication in construction: cost and time analysis of a robotically built wall. Autom. Constr. 92:297–311
36. Eversmann P, Gramazio F, Kohler M. 2017. Robotic prefabrication of timber structures: towards
automated large-scale spatial assembly. Constr. Robot. 1:49–60
37. Rogeau N, Latteur P, Weinand Y. 2021. An integrated design tool for timber plate structures to generate
joints geometry, fabrication toolpath, and robot trajectories. Autom. Constr. 130:103875
38. Robeller C, Weinand Y, Helm V, Thoma A, Gramazio F, Kohler M. 2017. Robotic integral attach-
ment. In Fabricate 2017: Rethinking Design and Construction, ed. A Menges, B Sheil, R Glynn, M Skavara,
pp. 92–97. London: UCL Press
39. Leung PYV, Apolinarska AA, Tanadini D, Gramazio F, Kohler M. 2021. Automatic assembly of jointed
timber structure using distributed robotic clamps. In CAADRIA 2021: Projections, ed. A Globa, J van
Ameijde, A Fingrut, N Kim, TT Sky, pp. 583–92. Hong Kong: Assoc. Comput.-Aided Archit. Des. Res.
Asia
40. Thoma A, Adel A, Helmreich M, Wehrle T, Gramazio F, Kohler M. 2019. Robotic fabrication of bespoke
timber frame modules. In Robotic Fabrication in Architecture, Art and Design 2018, ed. J Willmann, P Block,
M Hutter, K Byrne, T Schork, pp. 447–58. Cham, Switz.: Springer
41. ABB Robot. 2021. Transforming construction with Autovol. YouTube. https://www.youtube.com/
watch?v=9rkLPEWwFv4
42. Søndergaard A, Amir O, Eversmann P, Piskorec L, Stan F, et al. 2016. Topology optimization and robotic
fabrication of advanced timber space-frame structures. In Robotic Fabrication in Architecture, Art and Design
2016, ed. D Reinhardt, R Saunders, J Burry, pp. 190–203. Cham, Switz.: Springer
43. Doerstelmann M, Knippers J, Menges A, Parascho S, Prado M, Schwinn T. 2015. ICD/ITKE Research
Pavilion 2013–14: modular coreless filament winding based on beetle elytra. Archit. Design 85:54–59
44. Duque Estrada R, Kannenberg F, Wagner HJ, Yablonina M, Menges A. 2020. Spatial winding:
cooperative heterogeneous multi-robot system for fibrous structures. Constr. Robot. 4:205–15
45. Devadass PD. 2016. Robotic fabrication of non-standard material. In Posthuman Frontiers: Data, Design-
ers, and Cognitive Machines: Proceedings of the 36th Annual Conference of the Association for Computer Aided
Design in Architecture, ed. K Velikov, S Ahlquist, M del Campo, G Thün, pp. 206–13. Dover, DE: Assoc.
Comput. Aided Des. Archit.
46. Apolinarska AA, Bärtschi R, Furrer R, Gramazio F, Kohler M. 2016. Mastering the “sequential roof”
computational methods for integrating design, structural analysis, and robotic fabrication. In Advances
in Architectural Geometry 2016, pp. 240–58. Zürich: vdf Hochschulverlag AG an der ETH Zürich
47. Wagner HJ, Alvarez M, Groenewolt A, Menges A. 2020. Towards digital automation flexibility in large-
scale timber construction: integrative robotic prefabrication and co-design of the BUGA Wood Pavilion.
Constr. Robot. 4:187–204
48. ERNE. 2022. Effizienz am Bau dank Digitaltechnologie. ERNE. https://www.erne.net/de/
leistungen/technologien/digitale-fertigung
49. Augustynowicz E, Smigielska M, Nikles D, Wehrle T, Wagner H, Michel R. 2021. Collaborative design
of prefabricated façade systems. CoDeFa. https://codefacades.ch
50. Linner T. 2013. Automated and robotic construction: integrated automated construction sites. PhD Thesis, Tech.
Univ. München, Munich, Ger.
51. ETH Zürich. 2014. RFL. ETH Zürich. https://ita.arch.ethz.ch/archteclab/rfl.html
52. PERI. 2021. 3D construction printing. PERI. https://www.peri.com/en/business-segments/3d-
construction-printing.html
53. ICON Technol. 2020. Meet ICON’s next generation Vulcan construction system. ICON Technology.
https://www.iconbuild.com/vulcan
54. Apis Cor. 2022. Home page. Apis Cor. https://apis-cor.com
55. COBOD. 2022. The BOD2. COBOD. https://cobod.com/products/bod2
56. Keating SJ, Leland JC, Cai L, Oxman N. 2017. Toward site-specific and self-sufficient robotic fabrication
on architectural scales. Sci. Robot. 2:eaam8986
57. Dörfler K, Ernst S, Piškorec L, Willmann J, Helm V, et al. 2014. Remote material deposition. In What’s
the Matter? Materiality and Materialism at the Age of Computation, ed. M Voyatzaki, pp. 361–77. N.p.: Eur.
Netw. Heads Sch. Archit.

200 Parascho
58. Jokic S, Novikov P, Maggs S, Sadan D, Jin S, Nan C. 2022. Minibuilders. Institute for Advanced Architecture
of Catalonia. http://robots.iaac.net
59. Leder S, Oguz OS, Kim HG, Hartmann VN, Toussaint M, et al. 2020. Co-design in architecture: a modular
material-robot kinematic construction system. Paper presented at the IEEE International Conference on
Intelligent Robots and Systems, virtual, Oct. 25, 2020–Jan. 24, 2021
60. Kalousdian NK, Łochnicki G, Hartmann VN, Leder S, Oguz OS, et al. 2022. Learning robotic ma-
nipulation of natural materials with variable properties for construction tasks. IEEE Robot. Autom. Lett.
7:5749–56
61. Kayser M, Cai L, Falcone S, Bader C, Inglessis N, et al. 2018. Fiberbots: an autonomous swarm-based
robotic system for digital fabrication of fiber-based composites. Constr. Robot. 2:67–79
62. KUKA. 2022. KUKA Mobile Platform 1500. KUKA. https://www.kuka.com/en-de/products/
mobility/mobile-platforms/kmp-1500
63. Han IX, Bruun EPG, Marsh S, Tavano M, Adriaenssens S, Parascho S. 2020. From concept to con-
struction: a transferable design and robotic fabrication method for a building-scale vault. In Distributed
Proximities: Proceedings of the 40th Annual Conference of the Association for Computer Aided Design in Archi-
tecture, ed. B Slocum, V Ago, S Doyle, A Marcu, M Yablonina, M del Campo, pp. 614–23. Dover, DE:
Assoc. Comput. Aided Des. Archit.
64. Gramazio F, Kohler M, Helm V, Ercan S. 2012. Mobile robotic fabrication on construction sites:
DimRob. In 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 4335–41.
Piscataway, NJ: IEEE
65. Giftthaler M, Sandy T, Dörfler K, Brooks I, Buckingham M, et al. 2017. Mobile robotic fabrication at
1:1 scale: the In situ Fabricator. Constr. Robot. 1:3–14
66. Dörfler K. 2018. Strategies for robotic in situ fabrication. PhD Thesis, ETH Zürich, Zürich, Switz.
67. Lundeen KM, Kamat VR, Menassa CC, McGee W. 2019. Autonomous motion planning and task
execution in geometrically adaptive robotized construction work. Autom. Constr. 100:24–45
68. Mirjan A, Augugliaro F, D’Andrea R, Gramazio F, Kohler M. 2016. Building a bridge with flying robots.
In Robotic Fabrication in Architecture, Art and Design 2016, ed. D Reinhardt, R Saunders, J Burry, pp. 34–47.
Cham, Switz.: Springer
69. Dielemans G, Dörfler K. 2021. Mobile additive manufacturing: a robotic system for cooperative on-site con-
struction. Paper presented at the International Conference on Intelligent Robots and Systems, Prague,
Czech Repub., Sept. 27–Oct. 1
70. Johns RL, Wermelinger M, Mascaro R, Jud D, Gramazio F, et al. 2020. Autonomous dry stone: on-site
planning and assembly of stone walls with a robotic excavator. Constr. Robot. 4:127–40
71. Lussi M, Sandy T, Dorfler K, Hack N, Gramazio F, et al. 2018. Accurate and adaptive in situ fabrication
of an undulated wall using an on-board visual sensing system. In 2018 IEEE International Conference on
Robotics and Automation, pp. 3532–39. Piscataway, NJ: IEEE
72. Yablonina M, Menges A. 2019. Towards the development of fabrication machine species for filament
materials. In Robotic Fabrication in Architecture, Art and Design 2018, ed. J Willmann, P Block, M Hutter,
K Byrne, T Schork, pp. 152–66. Cham, Switz.: Springer
73. Yablonina M. 2021. Small robots and big projects. In Realignments: Toward Critical Computation: Proceed-
ings of 41st Conference of the Association for Computer Aided Design in Architecture, ed. K Dörfler, S Parascho,
J Scott, B Bogosian, B Farahi, et al., pp. 464–71. Dover, DE: Assoc. Comput. Aided Des. Archit.
74. Wagner HJ, Garufi D, Schwinn T, Wood D, Menges A. 2021. Three-dimensional fibre placement in
wood for connections and reinforcements in timber structures. In Proceedings of the IASS Annual Sym-
posium 2020/21 and the 7th International Conference on Spatial Structures. Madrid: Int. Assoc. Shell Spat.
Struct.
75. Bock T, Linner T. 2016. Site Automation: Automated/Robotic On-Site Factories. New York: Cambridge
Univ. Press
76. Robert McNeel Assoc. 2022. Rhinoceros. Robert McNeel and Associates. https://www.rhino3d.com
77. Rutten D. Davidson S. 2022. Grasshopper: algorithmic modeling for Rhino. Grasshopper. https://
www.grasshopper3d.com
78. visose. 2022. Robots. GitHub. https://github.com/visose/Robots

www.annualreviews.org • Construction Robotics 201


79. Schwartz T. 2017. HAL | robot programming & control. Food4Rhino. https://www.food4rhino.com/
en/app/hal-robot-programming-control
80. Sheng Y-T, Wang S-Y, Frank F. 2016. Taco ABB. Food4Rhino. https://www.food4rhino.com/en/
app/taco-abb
81. Rust R, Casas G, Parascho S, Jenny D, Dörfler K, et al. 2022. compas_fab. GitHub. https://github.
com/compas-dev/compas_fab
82. Fleischmann P, Casas G. 2022. compas_rrc_start. GitHub. https://github.com/compas-rrc/compas_
rrc_start
83. Vasey L, Prado M, Koslowski V. 2015. Behavioral design and adaptive robotic fabrication of a fiber com-
posite compression shell with pneumatic formwork. In Computational Ecologies: Design in the Anthropocene:
Proceedings of the 35th Annual Conference of the Association for Computer Aided Design in Architecture, ed.
L Combs, C Perry, pp. 297–310. Dover, DE: Assoc. Comput. Aided Des. Archit.
84. Brugnaro G, Baharlou E, Vasey L, Menges A. 2016. Robotic softness: an adaptive robotic fabrication
process for woven structures. In Posthuman Frontiers: Data, Designers, and Cognitive Machines: Proceedings
of the 36th Annual Conference of the Association for Computer Aided Design in Architecture, ed. K Velikov,
S Ahlquist, M del Campo, G Thün, pp. 154–63. Dover, DE: Assoc. Comput. Aided Des. Archit.
85. ABB. 2022. Vision systems. ABB. https://new.abb.com/products/robotics/application-equipment-
and-accessories/vision-systems
86. Dörfler K, Sandy T, Giftthaler M, Gramazio F, Kohler M, Buchli J. 2016. Mobile robotic brickwork:
automation of a discrete robotic fabrication process using an autonomous mobile robot. In Robotic Fab-
rication in Architecture, Art and Design 2016, ed. D Reinhardt, R Saunders, J Burry, pp. 204–17. Cham,
Switz.: Springer
87. Mitterberger D, Dörfler K, Sandy T, Salveridou F, Hutter M, et al. 2020. Augmented bricklaying. Constr.
Robot. 4:151–61
88. Wermelinger M, Johns R, Gramazio F, Kohler M, Hutter M. 2021. Grasping and object reorientation
for autonomous construction of stone structures. IEEE Robot. Autom. Lett. 6:5105–12
89. Wu K, Kilian A. 2019. Designing natural wood log structures with stochastic assembly and deep learning.
In Robotic Fabrication in Architecture, Art and Design 2018, ed. J Willmann, P Block, M Hutter, K Byrne,
T Schork, pp. 16–30. Cham, Switz.: Springer
90. Larsson M, Yoshida H, Igarashi T, Spencer S. 2019. Human-in-the-loop fabrication of 3D surfaces with
natural tree branches. In SCF ’19: Proceedings of the ACM Symposium on Computational Fabrication, ed. SN
Spencer, pap. 1. New York: ACM
91. Feng C, Xiao Y, Willette A, McGee W, Kamat VR. 2015. Vision guided autonomous robotic assembly
and as-built scanning on unstructured construction sites. Autom. Constr. 59:128–38
92. Bruun E, Adriaenssens S, Parascho S. 2022. Structural rigidity theory applied to the scaffold-free
(dis)assembly of space frames using cooperative robotics. Autom. Constr. 141:104405
93. Lukka TJ, Tossavainen T, Kujala J, Raiko T. 2014. ZenRobotics recycler – robotic sorting using ma-
chine learning. In Sensor Based Sorting 2014, ed. T Pretz, U Waschki, pp. 1–8. Clausthal-Zellerfeld, Ger.:
GDMB
94. Blomdell A, Dressler I, Nilsson K, Robertsson A. 2010. Flexible application development and high-
performance motion control based on external sensing and reconfiguration of ABB industrial robot
controllers. In 2010 IEEE International Conference on Robotics and Automation, pp. 62–66. Piscataway, NJ:
IEEE
95. Wolf M, Kaiser B, Hügle S, Verl A, Middendorf P. 2022. Data model for adaptive robotic construction
in architecture. Procedia CIRP 107:1035–40
96. Han IX, Parascho S. 2022. Improv-structure: exploring improvisation in collective human-robot construction.
Paper presented at the International Society for Intelligent Construction Conference, Guimarães, Port.,
Sept. 6–9
97. Apolinarska AA, Pacher M, Li H, Cote N, Pastrana R, et al. 2021. Robotic assembly of timber joints
using reinforcement learning. Autom. Constr. 125:103569
98. Belousov B, Wibranek B, Schneider J, Schneider T, Chalvatzaki G, et al. 2022. Robotic architectural
assembly with tactile skills: simulation and optimization. Autom. Constr. 133:104006

202 Parascho
99. Wibranek B, Liu Y, Funk N, Belousov B, Peters J, Tessmann O. 2021. Reinforcement learning for
sequential assembly of SL-blocks: self-interlocking combinatorial design based on machine learning.
In eCAADe 2021: Towards a New, Configurable Architecture, ed. V Stojaković, B Tepavčević, pp. 27–36.
Ljubljana, Slov.: eCAADe
100. Stadelmann L, Sandy T, Thoma A, Buchli J. 2019. End-effector pose correction for versatile large-scale
multi-robotic systems. IEEE Robot. Autom. Lett. 4:546–53
101. Gannon M. 2017. Human-centered interfaces for autonomous fabrication machines. PhD Thesis, Carnegie
Mellon Univ., Pittsburgh, PA
102. Clark L. 2016. The robot whisperer who tames giant industrial machine ‘monsters’ to do her bidding.
Wired, May 11. https://www.wired.co.uk/article/madeline-gannon-robotics
103. Piškorec L, Jenny D, Parascho S, Mayer H, Gramazio F, Kohler M. 2019. The brick labyrinth. In Robotic
Fabrication in Architecture, Art and Design 2018, ed. J Willmann, P Block, M Hutter, K Byrne, T Schork,
pp. 150–64. Cham, Switz.: Springer
104. Parascho S, Knippers J, Dörstelmann M, Prado M, Menges A. 2015. Modular fibrous morphologies:
computational design, simulation and fabrication of differentiated fibre composite building components.
In Advances in Architectural Geometry 2014, ed. P Block, J Knippers, NJ Mitra, W Wang, pp. 29–45. Cham,
Switz.: Springer
105. Rust R, Gramazio F, Kohler M. 2016. Force adaptive hot-wire cutting: integrated design, simulation,
and fabrication of double-curved surface geometries. In Advances in Architectural Geometry 2016, ed.
S Adriaenssens, F Gramazio, M Kohler, A Menges, M Pauly, pp. 288–305. Zürich: vdf Hochschulverlag
AG an der ETH Zürich
106. Søndergaard A, Feringa J, Nørbjerg T, Steenstrup K, Brander D, et al. 2016. Robotic hot-blade cut-
ting. In Robotic Fabrication in Architecture, Art and Design 2016, ed. D Reinhardt, R Saunders, J Burry,
pp. 150–64. Cham, Switz.: Springer
107. Parascho S. 2019. Cooperative robotic assembly: computational design and robotic fabrication of spatial metal
structures. PhD Thesis, ETH Zürich, Zürich, Switz.
108. Gandia A, Parascho S, Rust R, Casas G, Gramazio F, Kohler M. 2019. Towards automatic path planning
for robotically assembled spatial structures. In Robotic Fabrication in Architecture, Art and Design 2018, ed.
J Willmann, P Block, M Hutter, K Byrne, T Schork, pp. 59–73. Cham, Switz.: Springer
109. Huang Y, Garrett CR, Ting I, Parascho S, Mueller CT. 2021. Robotic additive construction of bar
structures: unified sequence and motion planning. Constr. Robot. 5:115–30
110. Parascho S, Han IX, Walker S, Beghini A, Bruun EPG, Adriaenssens S. 2020. Robotic vault: a cooperative
robotic assembly method for brick vault construction. Constr. Robot. 4:117–27
111. Yablonina M, Ringley B, Brugnaro G, Menges A. 2021. Soft Office: a human–robot collaborative system
for adaptive spatial configuration. Constr. Robot. 5:23–33
112. Napp N, Nagpal R. 2014. Distributed amorphous ramp construction in unstructured environments. In
Distributed Autonomous Robotic Systems: The 11th International Symposium, ed. MA Hsieh, G Chirikjian,
pp. 105–19. Berlin: Springer
113. D’Andrea R, Dullerud GE. 2003. Distributed control design for spatially interconnected systems. IEEE
Trans. Autom. Control. 48:1478–95
114. Rubenstein M, Cornejo A, Nagpal R. 2014. Programmable self-assembly in a thousand-robot swarm.
Science 345:795–99
115. Petersen KH, Napp N, Stuart-Smith R, Rus D, Kovac M. 2019. A review of collective robotic
construction. Sci. Robot. 4:eaau8479
116. Han IX, Meggers F, Parascho S. 2021. Bridging the collectives: a review of collective human–robot
construction. Int. J. Archit. Comput. 19:512–31
117. Bärtschi R, Knauss M, Bonwetsch T, Gramazio F, Kohler M. 2010. Wiggled brick bond. In Advances in
Architectural Geometry 2010, ed. C Ceccato, L Hesselgren, M Pauly, H Pottmann, J Wallner, pp. 137–47.
Vienna: Springer
118. Ercan JS, Lloret-Fritschi E, Gramazio F, Kohler M. 2020. Crafting plaster through continuous mobile
robotic fabrication on-site. Constr. Robot. 4:261–71

www.annualreviews.org • Construction Robotics 203


119. Placzek G, Brohmann L, Mawas K, Schwerdtner P, Hack N, et al. 2021. A lean-based production ap-
proach for shotcrete 3D printed concrete components. In Proceedings of the 38th International Symposium
on Automation and Robotics in Construction, pp. 811–18. London: Int. Assoc. Autom. Robot. Constr.
120. Fologram. 2022. Home page. Fologram. https://fologram.com
121. Allner L, Kaltenbrunner C, Kröhnert D, Reinsberg P. 2021. Augmented reality. In Conceptual Joining:
Wood Structures from Detail to Utopia: Holzstrukturen IM Experiment, ed. L Allner, C Kaltenbrunner,
D Kröhnert, P Reinsberg, pp. 172–75. Basel: Birkhäuser. Ed Angew.
122. Johns RL, Anderson J, Kilian A, Bieg K, Briscoe D, Odom C. 2019. Robo-Stim: modes of human robot
collaboration for design exploration. In Impact: Design With All Senses: Proceedings of the Design Modelling
Symposium, Berlin 2019, ed. C Gengnagel, O Baverel, J Burry, MR Thomsen, S Weinzieri, pp. 671–84.
Cham, Switz.: Springer
123. Kilian A. 2018. The flexing room architectural robot. An actuated active-bending robotic structure using
human feedback. In Recalibration: On Imprecision and Infidelity: Proceedings of the 38th Annual Conference
of the Association for Computer Aided Design in Architecture, ed. P Anzalone, M Del Signore, AJ Witt,
pp. 232–41. Dover, DE: Assoc. Comput. Aided Des. Archit.
124. Myerson RB. 1997. Game Theory: Analysis of Conflict. Cambridge, MA: Harvard Univ. Press
125. Wohlin C. 1995. Improving through an incremental approach. In Proceedings of the 2nd European
Industrial Symposium on Cleanroom Software Engineering. Lund, Swed.: Q-Labs

204 Parascho

You might also like