Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Cement and Concrete Research 158 (2022) 106772

Contents lists available at ScienceDirect

Cement and Concrete Research


journal homepage: www.elsevier.com/locate/cemconres

Additive Manufacturing using mobile robots: Opportunities and challenges


for building construction
Kathrin Dörfler a, *, Gido Dielemans a, Lukas Lachmayer b, Tobias Recker b, Annika Raatz b,
Dirk Lowke c, Markus Gerke d
a
TT Professorship Digital Fabrication, Technical University of Munich, Germany
b
Institute of Assembly Technology, Leibniz University Hannover, Germany
c
Institute of Building Materials, Concrete Construction and Fire Safety, Technische Universität Braunschweig, Germany
d
Institute of Geodesy and Photogrammetry, Technische Universität Braunschweig, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: In situ Additive Manufacturing (AM) has developed into an active research and industry-transfer area, mainly
Architecture and digital fabrication due to the increasing need for productive and sustainable methods in the concrete construction industry in
Robotic fabrication combination with novel technological enablers. While current systems for in situ AM are often single large-scale
Additive Manufacturing
stationary facilities, Mobile Additive Manufacturing (MAM) systems are an emerging technology that could
Mobile robotics
Computational design
provide scalability for AM processes on construction sites. This scalability is achievable through cooperability of
multiple mobile robots on individual 3D printing tasks, while their mobility and autonomy enable deployment in
both new and existing contexts, and coordination of operations for the direct and indirect interaction with
humans. To ensure applicability and scalability, MAM closely integrates architectural, mechanical, and materials
design, the manufacturing process, sensing, and control. With this paper, we give a comprehensive review of
research trends, open questions and key performance indicators. We support the discussion with potential
architectural application scenarios. Overall, we aim to indicate why addressing MAM is an inherently multi­
disciplinary challenge.

1. Introduction While some developments have been made to improve the efficiency of
these systems, including efficiency gains through the use of multi-nozzle
Digital systems for Additive Manufacturing (AM) in building con­ 3D printing [6], the field could largely benefit from the adoption of
struction have seen rapid developments, with transformations from lab alternative in situ systems accelerating the widespread deployment of
experiments to residential and civil structures [1,2]. Compared to AM technologies in construction today.
traditional manual production processes in the concrete construction In particular, the integration of autonomous mobile robotics in
industry, the integration of in situ AM processes has proven the potential building construction presents new opportunities for in situ AM; Mobile
to extensively increase productivity and resource efficiency [3,4]. Additive Manufacturing (MAM) systems for on-site construction could
Today, in situ AM is characterized by large-scale stationary systems provide an unbounded work area through mobility and thus new
allowing for material deposition processes directly on building sites, the possibilities for architecture in terms of manufacturing building com­
most common of which is the gantry system, which provides high ac­ ponents larger than the static range of the system. They could provide
curacy for end-effector tasks and high payload capacities [5]. However, scalability through interaction (i.e., collaboration, cooperation,
these stationary systems also have certain shortcomings; including, but coordination) within multi-robot systems or with human operators and
not limited to, confined workspaces (i.e., the mechanical devices are site workers present. Furthermore, they could extend the scope of in situ
significantly larger than the structure they build), limited scalability due AM from new construction to constructions in existing contexts for the
to complex parallelization of processes, long material supply distances, purposes of remodeling, repair, and renovation (see Fig. 1).
extensive effort required for the set-up of the system on work sites, and In this paper, we aim to introduce MAM systems for on-site con­
applications are largely constrained to new constructions on vacant lots. struction as both a significant challenge and a huge opportunity. The

* Corresponding author.
E-mail address: doerfler@tum.de (K. Dörfler).

https://doi.org/10.1016/j.cemconres.2022.106772
Received 31 October 2021; Received in revised form 6 February 2022; Accepted 8 March 2022
Available online 10 June 2022
0008-8846/© 2022 Elsevier Ltd. All rights reserved.
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

remainder of this paper is therefore organized as follows. Section 2 in­


troduces the background and motivation for mobile robotics applied to
in situ AM processes. Current research trends, approaches, and re­
quirements in MAM are outlined in Section 3. A description of potential
architectural application scenarios is given in Section 4. Section 5 con­
cludes the paper with an outlook and a motivation for future work in
MAM systems for construction. The scope of this paper is limited to
reviews of methods and approaches for continuous material extrusion
deposition techniques as defined for AM, and largely excludes proced­
ures such as bricklaying, discrete assembly processes, drilling, or
painting.

2. Preliminaries

2.1. Architectural design space of in situ AM processes

To be able to exploit various positive aspects of 3D printing buildings


by layering strands of materials, including but not limited to construc­
tions without formwork, the efficient use of materials, the production of
force-flow and load-optimized components, the integration of different
functions into individual components (e.g. ventilation and insulation)
and the great design freedom for architects and engineers through
seemingly unlimited customization of building structures, a compre­
hensive understanding of the various dependencies is crucial for archi­ Fig. 2. Four aspects that inform the architectural design space of in situ
tectural design. In this paper, the architectural design space is AM processes.
considered to be a result of the interactions between four fundamental
aspects of in situ AM processes, expanding on the aspect classification new in situ AM systems, it is precisely this interaction between the in­
proposed by [7] (1) the structural requirements and morphology of the dividual aspects that leads to a high level of complexity, which this re­
building, (2) its contextual conditions, (3) the material properties and view attempts to take into account; that is, that all aspects are discussed
processing, and (4) the mechanical system, i.e., the fabricating robot in context with each other and that only a joint observation encom­
system(s) (see Fig. 2). While there are many other dependencies char­ passing several aspects and not individual innovations can ultimately
acterizing and defining in situ AM design scopes, the primary ones being lead to meaningful conclusions.
logistics (materials and printer transport), materials (sourcing, access), A decisive factor during the actual manufacturing process with
site conditions, environmental conditions (effect on machines and ma­ concrete is the load-bearing capacity and stability of the structure to be
terials), these are not the primary focus of this paper [8]. printed. In situ material deposition is constrained by specific geometric
As such, the context defines where construction takes place, whether boundary conditions where the final product is a direct consequence of
it is new construction or refurbishment, as well as the accessibility of the the fabrication process, material properties, and contextual parameters,
site. The building structure is the designed and constructed object that both from the green to the cured state of the cementitious material. For
physically exists on the construction site when the AM process is com­ on-site processes, the material is typically deposited in-place with the
plete. Materials include those that will make up the final structure, as orientation as given, ideally without the need for auxiliary structures as
well as ancillary structures and support materials used as part of the support. Hence, there exists a distinct design space, in which potential
building process. The mechanical systems are the robotic devices that failure modes, such as a global instability of the object, a plastic collapse
transport, process and deposit material in specific locations, creating the or a phase change (solid to liquid) and an elastic buckling of the struc­
building structure by depositing material. Each of these aspects — ture, can be avoided [9]. This is particularly true for a rapid construction
context, structure, material and mechanical systems — affects and is process, where stresses in the component increase disproportionately to
affected by the properties of the other. When examining the range of

Fig. 1. Visualization of cooperating mobile robots performing extrusion-based 3D printing for curved wall structures on building sites.

2
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

the strength development of the material, or for the printing of canti­ deployable AM systems in the future (see Fig. 5). However, the robotic
levers and horizontally aligned structures, where tensile stresses in the complexity increases significantly with the transition from manual to
material are higher compared to vertically aligned structures. autonomous relocation and generally with increasing mobility and au­
Furthermore, this design space is also a direct consequence of the tonomy of the systems [23]. Due to the lack of a rigid kinematic chain
mechanical deposition system used. For example, the degrees of freedom connecting a robot’s end effector to the ground, the greater its mobility
(DOF) of the system with which the material is deposited determines and flexibility, the more dependent its precise operation is on advanced
whether printing is limited to parallel horizontal planes or if printing sensor and control solutions [24] (see also Section 3.1.1, 3.1.2, and
out-of-plane is possible, which can, amongst other factors, influence the 3.1.4).
degree of cantilever [9,10] or the possibility of force-flow optimized
material placement [11]. AM structures are thus often a result of 2.3. In situ AM within existing contexts
custom-purpose material-robot systems [12], i.e., systems where both
the construction material and mechanical system are designed in coor­ Renovate, retrofit, repair, and maintain — the construction methods
dination with each other. For example, the cylindrical robot reach en­ that are usually associated with “construction in existing contexts” with
velope of the Crane WASP, an in situ AM system with 3 DOF, combined the aim of increasing or maintaining value — are still largely ignored
with a soil-based material, defines the design space of printable geom­ and undervalued in the context of architecture and digital fabrication
etries [13,14] (see Fig. 3). [25]. Long-term care of the built environment is a highly important, but
Another decisive factor for the scope and limitations of possible not as often celebrated aspect of contemporary architecture, where
geometrical shapes is whether the workspace of the mechanical system emphasis is typically put on new constructions. In the field of applica­
is extended by the structure created, as was illustratively shown in the tions within existing contexts, the level of automation is also still far
Minibuilders project [15] in the form of a climbing system (see Fig. 4). below the one for new constructions [26]. While stationary robots are
largely limited to new constructions on vacant lots, mobile robots can
2.2. The relation between scale and mobility in in situ AM processes flexibly access enclosed spaces and could thus be a useful future tool for
renovating, retrofitting, repairing and maintaining our buildings and
The question of scalability in deployable AM on building sites con­ infrastructure. Since these processes are done directly on the existing
cerns both the question of the size of the static workspace (volume), as structure, the deployability and mobility of the systems is a mandatory
well as the possibilities for parallelization (speed), while maintaining a requirement. Hence, such scenarios encourage the development of novel
high degree of flexibility in guiding a nozzle through space building up a platforms, methods, and tools, that are not only context-aware, but can
structure layer by layer [5,16]. Today, especially in the commercial manufacture site and task-specific structures, on-site and in unstruc­
field, this is achieved in a variety of ways, including a) increasing the tured environments.
working area of the 3D printing platform, i.e., using mechanical devices In general, working in existing contexts relies heavily on evaluation
that are significantly larger than the structure they are building [17], b) of the status of the building and additionally requires other processes
manually relocating the 3D printing platform several times, and c) the besides AM, e.g., removing degraded material, cleaning and preparing
parallel use of several 3D printing platforms. surfaces, applying new material and components. To maintain produc­
For example, Cobod [17] and ICON [18] have demonstrated the tivity, the ability of the mobile robots to perform such varying tasks in
construction of whole multi-storey buildings within the workspace of collaboration, as well as the multi-functionality of the systems is of
the 3D-printing systems. US Army Corps of Engineers have developed a crucial importance. Key for robot manipulation in existing contexts for
lightweight deployable 3D-printer system through a cooperative the transformation of existing structures is an extended spatial aware­
research and development agreement with Caterpillar, which can be ness, i.e., geometrically precise and semantically rich internal repre­
assembled and disassembled within 30 min each, can fit in a shipping sentation of the surroundings the robot is operating in. While robot
container, is mounted on retractable wheels, has the ability to be easily mapping is a well-researched problem, existing research is mostly suited
moved on-site without disassembly, and has been used to 3D print for robot navigation (see also Section 3.1.4); the actual manipulation of
separate parts of a building without disassembly [8,19]. APIS Cor [20] the physical space, such as the deposition of material with a high degree
and Cybe [21] have shown projects in which they relocated their of accuracy within found environments, relies on expanding existing
printing arm by a crane to temporary locations in series. Twente AM methods. These methods are subject of extensive further research in this
[22] have developed and demonstrated a deployable 3D-printing plat­ domain, amongst these are object recognition, metrically capturing
form that integrates a robotic arm and a linear axis on a trailer. Crane spatial relationships (e.g., nearness, above, and below), qualitative and
WASP systems [13] are designed such that they can be assembled on site quantitative estimations of the found built structure, as well as the
into a hexagonal configuration to expand the printing area and to print sourcing and reuse of found materials.
with multiple systems in parallel.
Autonomously mobile 3D printing robots could expand manually

Fig. 3. Multiple modular Crane WASP systems can be assembled on site into a hexagonal configuration to expand the printing area and to work together. Their
cylindrical reach envelope in combination with the DOFs and the material in use defines the design space of printable geometries [13,14].

3
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

Fig. 4. The mobile manipulator of the Minibuilders projects is climbing along the structure it is extruding [15].

Fig. 5. A single stationary AM system whose working space must be larger than the structure to be erected (left), compared to several cooperating mobile AM systems
for building construction, whose working space is expandable through their mobility and thus larger than their static workspace (right); thanks to their ability to
cooperate, structures can also be manufactured in parallel.

2.4. Towards human-robot teams: robots as co-builders correlation between autonomy and collaboration: a higher level of au­
tonomy requires higher levels of interaction [34]. Han et al. [35]
At the present, robots in building construction are often programmed therefore imply that a more balanced partnership and collaboration
to work independently of human-specific interactive activities [27]. between humans and robots can only be established with greater au­
Initially however, robotics and automation were invented to assist and tonomy of the machines. The area of applications for highly autonomous
to collaborate with humans in completing tasks, rather than replacing systems could be substantially expanded if they allowed for dedicated
them [28,29]. Dr. Shneiderman, a pioneer in the field of human- human intervention and interaction and if humans were consciously
computer interaction, has argued for decades against system designers kept in the loop [36]. The common understanding could be fundamen­
“who believe they are creating something lifelike and smart” — while tally redefined by considering the role of humans as collaborators and
often, “users feel anxious and unable to control these systems” [30–32]. not just robot operators. Depending on the task, different scenarios with
In industrial manufacturing contexts, robots have already gained both highly autonomous machines would be possible, ranging from highly
cognition and manipulation abilities that allow them to operate in close autonomous (multi-) robotic task completion with minimal human
proximity to humans and even in direct physical interaction, as stated by intervention to highly collaborative human-robot task completion with
Haddadin et al. [28] in their definition of a robot as a co-worker. intuitive task distribution [37].
Human presence on building sites, in relation to robots, happens in New possibilities for digital fabrication could arise by thinking of
various ways; typically, humans on building sites are trained personnel interaction, cooperation, and collaboration, in combination with the
(e.g., operators and experts), but they can also be untrained (e.g., development of highly autonomous machines [38]. As building con­
helpers and spectators), which usually differentiates in either supportive struction is substantially shaped by and dependent on manual skills,
or disruptive interaction. As defined by Schultz [33], humans can regional training conditions, local heritage and craft, autonomous MAM
interact with robots in many ways: as supervisor, operator, teammate, systems should be designed such that they are not only able to coordi­
mechanic/programmer, and bystander. As has been pointed out by Beer nate physical actions amongst themselves, but should further be able to
et al. [34] in the context of human-robot interaction, the main opinion safely share tasks through direct and indirect interaction and collabo­
suggests that higher autonomy leads to less human-robot interaction, ration with humans. As such, they could aid in formulating an alterna­
with research aiming to maximize robotic autonomy with minimal tive vision of a fully automated building design and robotized
human intervention, as seen with many research projects related to construction, proclaiming the idea of a hybrid, mutually augmented
autonomous digital fabrication in architecture. However, there exists a human-robot work team [37].
more interesting perspective suggesting that there is a positive

4
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

3. Research axes [21], the printstones robotic 3D printing technology [46], or the In situ
Fabricator [47] (see Fig. 8).
3.1. MAM challenges, principles, system requirements, and current With multi-DOF manipulators mounted onto mobile bases, such as
approaches an anthropomorphic robotic manipulator with 6-DOF, extended printing
configurations can be considered that allow for a better transfer of forces
In view of deposition-based MAM on construction sites at full and contact area between layers, one of these techniques is the
architectural scale, we outline and review five primary categories, each tangential continuity method [48]. Manipulating and depositing mate­
with their specific challenges, requirements and principles in Section rial while moving the mobile base (print-while-drive approach) typically
3.1. This categorization lays the foundation for the identification of presupposes the ability of performing a coordinated move between the
performance metrics for research trends and directions in Section 3.2. arm and the base, integrating trajectory planning and whole-body mo­
tion control [49]. Such whole body motion control has for example been
3.1.1. Mobility shown by the YouWasp project [43] and by [50], in which an external
As postulated via the “In situ Fabrication (IF) Challenge” by [16], motion capture system was used to provide mobile base pose estimation
robotic mobility should enable the “arbitrary autonomous mechatronics for closed-loop base control to extrude material in continuous print
tooling capability, anywhere in 3D space, in arbitrary environments, paths exceeding the robot’s reach envelope. Without external motion
without relying on fixed installations”. MAM systems are required to capturing systems, whole-body motion control has so far only been
perform autonomous navigation and continuous material deposition, in shown as point marking on a wall with a mobile construction robot in
which a mobile platform carries an end-effector material deposition [49]. The system performed all state estimations using only on-board
system, in some cases fitted to an extended manipulator. These mobile sensing. However, such a highly autonomous system has not yet been
systems are required to avoid obstacles or other mobile systems, while shown for carrying out AM tasks on a building site.
preventing compromises on the efficiency of the material deposition
process by utilizing effective, minimal movements of the mobile plat­ 3.1.3. Material supply and deposition
form. To ensure the maneuverability of the system inside buildings, the Material supply and deposition options, while posing a challenging
maximum size is limited to, or should be able to be reduced to, stan­ problem in construction overall, have particular implications in the
dardized door openings. context of MAM systems carrying out material extrusion or material
Most existing MAM systems are geared towards ground-based jetting (compare also with the classification framework for defining
mobility to expand the workspace of each system while allowing large Digital Fabrication with Concrete in [51]), and have indirect impacts on
payload capacities, and thus the integration of various material pro­ their mobility metrics.
cessing systems. Current mobile robotic platforms can reach intended The AM process typically consists of the following stages: (a) the
locations for material deposition, either by having the mobile platforms production of the cementitious material by mixing the raw materials -
continuously climbing up the structure they create (i.e., shown in the external or on-board, (b) the delivery of the material to the print head or
case of small-scale mechanisms with severely limited robot reach en­ the raw materials to the on-board mixer — continuously or batchwise,
velopes, [15,39–41]), or by reaching from a nearby location on which (c) the intermediate storage of the mixed or raw materials if required,
the mobile platforms can rest and maneuver during manufacturing (e.g., and (d), the application of individual material strands. For classifying
on floor slabs, [42–44]) (see Fig. 6). the technical design and the mobility of the MAM unit, it is essential
where these individual processes are located. Here, the delivery process
3.1.2. Manipulation always bridges the boundary between the external environment and the
The topic of manipulation is directly linked with the topic of mobile unit. Depending on the location of the mixing, the type of ma­
mobility, in which extended manipulation and positioning compliance terial delivery and any necessary intermediate on-board storage of the
of an AM end-effector requires a mechanical system with one or more materials, a classification can be made between four variants, shown in
DOFs. The smaller the range (static work area) of the robot in relation to Table 1 and Fig. 9.
the work object, the greater the mobility of the robot must generally be. How the material is fed to the printhead or the on-board mixer is an
While robots with a small reach envelope must continually move during essential aspect of MAM systems, as it influences the mobility of the
manufacturing (print-while-driving approach), such as shown in the system and poses greater challenges compared to stationary ones. An
Growing Robot of the Plantoid Project [39] or the FIBERBOTS [40], external material supply will form a tethered connection, where a
robots with an extended reach envelope can also be used in quasi- batchwise material supply combined with on-board storage requires
stationary modes by being stabilized and not in motion during high payload capacities for carrying material containers, each affecting
manufacturing (print-drive-print approach) (see Fig. 7). The latter the mobility of the system in their own way.
approach has, amongst others, been shown in the Digital Construction In external mixed concrete supply systems, i.e., variant 1.1 & 1.2, a
Platform by [45] (see Fig. 7), the CyBe concrete 3D printing platform measured flow of material can be provided by using a digital control

Fig. 6. The Minibuilders robots drive directly on the structure to be extruded (left), as an alternative to robots that typically drive next to the structure, on the same
surface from which the 3D printing is started (right).

5
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

Fig. 7. As the Mediated Matter Group at MIT has shown, the spectrum of mobile Additive Manufacturing can range from robots with a very limited reach envelope
(FIBERBOTS, left) to robots with a wide-range reach envelope (Digital Construction Platform, right). While the robots with a short reach have to move constantly
when depositing material (print-while-drive approach), those with a long reach can be used in quasi-stationary mode (print-drive-print approach) [45].

Fig. 8. Both the mobile construction platform Cybe (top)


[21] and the In situ Fabricator (IF) (bottom) [47] are
comprised of an industrial robot arm with 6-DOF mounted
on hydraulic crawler tracks. While the Cybe platform has
mostly been moved on-site to prefabricate concrete ele­
ments which are later assembled by crane into their final
location, the IF has also fabricated building elements in-
place. The IF has additionally been shown in autono­
mous driving scenarios, while Cybe has been steered via
remote control. Both are used in quasi-stationary manner,
being stabilized with legs during material deposition tasks.

system operating an external pump. Typically, dry material is mixed to system, however, mobility is significantly reduced.
recipe, and the concrete is pumped through a hose to the deposition Another important aspect results from the reactivity of the cement.
device at the end-effector of the mobile system, as for example shown in As soon as the cement binder is in contact with water, hydration begins,
[21,52]. Due to the permanent connection to the material delivery which leads to the setting and subsequent hardening of the concrete.

6
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

Table 1 precise deposition of the concrete, which necessitates the control of


Material mixing and delivery classification in MAM. several parameters. These parameters include: the deposition speed,
Production Material feed nozzle size and shape, and the possibility to precisely stop the flow of
material to allow for non-continuous material deposition to enable
1.1 Mixing Continuous delivery of mixed material from external to
(external) on-board transitions between statically fabricated elements and open print-path
1.2 Mixing Batchwise delivery of mixed material to on-board storage configurations.
(external)
2.1 Mixing (on- Continuous delivery of raw material from external to on- 3.1.4. Localization, positioning and context-awareness
board) board
2.2 Mixing (on- Batchwise delivery of raw material to on-board storage
Precise material deposition is critical to the success of AM processes
board) [61], where marginal deviations in position or orientation will decrease
the stability of the printing process due to one-sided loads, and could
lead to printing failure [42]. Mobile robots especially suffer from posi­
This means that the rheological properties of the concrete change irre­ tional inaccuracies [62] and are restricted in terms of localization ca­
versibly over time, thus the processing of the material becomes a time- pabilities, as a result of the provided sensor systems and the localization
critical process. The concrete should be sufficiently deformable during method. However, given the variety of algorithms [63], specific re­
transport as well as at the print head to ensure pumpability and strictions can be derived for particular construction-site developments.
discharge from the nozzle. In this respect, short delivery distances, e.g. By their very nature, construction sites are constantly and rapidly
through intermediate storing of the ready-mixed concrete on the mobile changing environments. With direct and indirect human intervention,
unit (variant 1.2), or even on-board mixing would be advantageous varying lighting conditions, and harsh environmental conditions,
(variant 2.2). The storage container or the mixer can be arranged on the localization methods developed are required to be robust and precise for
mobile unit itself or on a cooperating mobile unit. On-board interme­ the application of AM on construction sites [64]. The localization
diate storing of the material has been proven in small scale experiments method applied to ground-based mobile systems needs to take into ac­
using FDM or clay extrusion mobile printers [42,43,53] as well as for count the probable minor ground surface quality of an on-site situation,
cement-based materials, e.g. Baubot X1 beta [46,54]. Current research compared to a lab or level hall floor [65]. This requires full dynamic 6-
for stationary concrete printing systems also shows the possibility of DoF localization, which is not commonly available in commercial sys­
mixing of the concrete immediately in front of the nozzle to further tems. Construction components are subject to tolerances, although these
reduce the delivery distance of the fresh concrete to the print nozzle occur in all manufacturing tasks, traditional construction allows values
[55–57], with the aim of reducing unintentional variability in the ma­ in the range of centimeters, at least one order of magnitude above the
terial as well as integration of material gradation. recommendations for ordinary metal processing [66,67]. In AM, these
After the concrete has been deposited, it should sufficiently transi­ tolerances have to be reduced for the localization task to ensure suc­
tion into a setting or hardening state, i.e., macroscopically develop a cessful printing. This has the particular consequence that the localiza­
green strength, in order to be able to successively support the load of the tion must take place in relation to the reference or model coordinate
subsequent layers and ensure a rapid printing progress. In order to system, though requires adaptivity to the actual, current state of reali­
control the contradicting requirements, i.e., good workability during zation of the object.
delivery and depositing and sufficient structural build-up after depos­ In general, localization algorithms can be classified into two cate­
iting, the material development of fresh concrete can be accelerated. gories: relative and absolute [68]. Relative localization, also called dead
This can be done, for example, by adding an accelerator or a further reckoning or inertial localization, is solely based on the temporal inte­
binder through an inline mixing process directly at the nozzle. Besides gration of sensor information [63,69,70]. While relative localization
first applications of inline mixing on mobile printers [46], there are algorithms, e.g., by wheel odometry, IMU, or visual odometry, enable
numerous developments in progress in the field of 3D concrete printing autonomous operation and provide suitable performance for mobile 3D
that transfer well to MAM [58–60]. printing on a small scale [42], the compounding measurement errors
The main requirement of the material deposition system is the and noise will lead to an ever-increasing deviation. Limitations of in­
ertial localization are: only the position change relative to the referenced
starting point is measured, whose position must also be referenced [71],
and velocity integration of this form of inertial localization assumes a
level floor, resulting in an additional error for uneven floors on con­
struction sites [72]. Wheel odometry is particularly affected by this
noise.
This leads to the necessity of “enabling a mobile robot to determine
its position and orientation… independent of assumptions about previ­
ous movements” [73]. In this regard, absolute localization is currently
refined to improve the tool center point (TCP) positioning [70]. With an
absolute approach, using a global reference frame, the manufacturing of
a structure according to a planned CAD model is more straightforward
compared to the relative positioning approach. If parts are constructed
in isolated locations or collaboratively with multiple robots, absolute
positioning seems indispensable. External reference devices such as total
stations, laser reflector combinations, or optical markers perform
excellent concerning absolute localization and provide positional in­
formation within an global reference coordinate system (see Fig. 10), as
evidenced by recent mobile 3D printing operations [43,50,61,74].
However, since they require a constant line of sight, the robot workspace
is constricted, and the usability under previously described construction
site influences is limited [16]. Instead, environment-based localization
by on-board sensing and referencing to the local surroundings provides
Fig. 9. Illustration of material mixing and delivery in MAM. higher flexibility.

7
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

Fig. 10. Mobile positioning for 3D printing through external vision-based localization using an externally mounted camera tracking ArUco markers on the base (left),
and using a camera on board of the base tracking Aruco markers on the floor (right). Here, the marker tracking provides sufficient accuracy for demonstrating both
parallel printing and printing-while-moving approach.

For localization within static construction environments and when various levels of scene representation, also referred to as spatial AI [84],
using a relative positioning approach, robot sensor point clouds are visual real-time SLAM has gradually improved in recent years, starting
iteratively fitted to pre-defined maps, e.g. by Iterative Closest Point ICP from sparse features, evolving to dense maps, and recently allowing for
[44]. Due to the variable environment of construction sites however, increasingly semantic labels [85]. However, due to the high accuracy
pre-defined maps will be rapidly out-dated. More robust and dynamic requirements and continuity of deposition-based AM and the necessity
algorithms can be implemented by combing a pre-scanned map with to be able to deal with part tolerances, the accuracy of visual SLAM
odometry data, resulting in the adaptive Monte Carlo Localization algorithms must still be heavily improved. As such, prior knowledge in
(AMCL) [75]. Lastly, Simultaneous Localization and Mapping (SLAM) is the form of CAD or BIM data, which is typically available for digital
most beneficial for fully autonomous operation as it continuously up­ building sites, can be utilized to improve the mapping and localization
dates the map of the environment [76–78]. Here, it is necessary to [86]. Moreover, the combination of referencing the robot in a global
consider that widespread 2D SLAM algorithms based on laser range map with referencing the robot relative to the manufacturing part, as for
measurements do not generally predict moving obstacles or uneven example demonstrated by the In situ Fabricator for the assembly of
surfaces due to 2D approaches [78]. Therefore, SLAM algorithms with discrete blocks into a tower [86] (see Fig. 11), provides also a promising
the generation of 3D maps are required [79,80]. Handling dense 3D solution to be applied for precise and continuous printing operations.
point clouds requires time-consuming computational effort, which is In sum, to attain an accurate enough localization system for addi­
why a separation between map construction and real-time localization is tively manufactured structures and to also allow for cooperative oper­
proposed [81]. ation between multiple units, mobile systems require a two-level system.
Out of the available methods, specifically visual SLAM has proven The first level of the system is the absolute localization of the mobile
successful localization at construction sites [82]. When an autonomous platform within the context of the construction environment, finding the
system must operate for an extended period of time, carry out a wide relation between the platform, work site, collaborators, and other ob­
variety of tasks (not all of which are necessarily known at design time), jects. The second level, which is used in combination with the first,
and communicate with other entities, including humans, its real-time should provide the estimation of the end-effector location relative to the
visual SLAM system should build a human-understandable, general, work piece, with which the required accuracy of continuous layered
and persistent scene representation, which is close to metric 3D geom­ material deposition should be attained (see Fig. 12).
etry, at least locally. Real-time visual SLAM, as used for localization and
context-awareness, is capable of dealing with changes in the environ­ 3.1.5. Control, coordination and task allocation
ment, human interaction, or robot cooperation [83]. Hence, for enabling Given robust localization, positioning, and context-awareness, the

Fig. 11. The In situ Fabricator (left) performing autonomous end effector positioning by reference to the workpiece via 3D laser scans (right) for the assembly of a
vertically stacked tower.

8
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

Fig. 12. Two-level localization system as proposed to attain the required accuracy for mobile 3D printing tasks.

Fig. 13. Snapshots of a simulated animation of two AMBOT robots cooperatively 3D printing an airplane model using an FDM process. A chunk-based slicer was used
to discretize an object into parts, amongst other taking into account the ratio of the nozzle height and width, which can then be printed in parallel by the two
participating units.

Fig. 14. The YouWasp system has demonstrated mobile multi-robot material deposition, both in simulation by investigating and evaluating various task decom­
position, sequencing and allocation approaches, and in an experimental test scenario, depositing ceramic material in an tracked arena.

use of multi-robot systems offers increased efficiency through produc­ additional schedulers must generate a printing scheme and assign sub-
tion time reduction and collaboration [87]. When considering multi- tasks to individual robots [53]. The resulting seams, however, will
agent-printing operations on large-scale components, additional colli­ have a direct effect on the structural stability. Furthermore, structural
sion avoidance and task allocation are required at an early stage of considerations must be taken into account when printing with time-
planning. Dividing components into substructures without workspace dependent materials such as concrete. Thus, a flexible control
overlap is the standard way to make use of multi printer systems [88–90] approach involving robot interaction or collaboration is required [91].
(see Figs. 13 and 14). Centralized control methods are generally based on a single
While slicing and path planning are based on existing algorithms, controller and require a holistic a priori planning [92]. These have

9
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

proven easy to implement due to entirely preplanned processes, but are is traversing over the area of the printed structure, it is also limited to the
just recommended for few robots with a small number of tasks and printed structure with marginal expansion of the work area possible
minor complexity [12,93]. In contrast, decentralized architectures through overhang. The area of movement for a ground-based system can
provide individual controllers for each unit, which enables individual be seen as its opposite, where movement is only possible where no
task assignment and independent robot behavior [94]. A major draw­ printed structure has been constructed, resulting in a much larger area of
back of fully decentralized systems is the absence of direct communi­ mobility, though constrained by the limitations of the static workspace,
cation between the units, so all issues have to be solved implicitly by the most prevalent in the vertical axis.
individual robot programs [95]. The context in which the system is able to operate is of high rele­
However, the complexity of MAM in construction demands global vance with the transfer of mobile systems to construction sites, where a
process awareness by each robot. Especially during in situ applications, direct relation can be made to the localization of the system in un­
capabilities of both approaches are needed for compensation of material structured environments. The methods of localization can be described
application errors, collision avoidance related to printed components by as a single localization procedure using environmental mapping, a
other robots, and environmental disturbances. Therefore, hybrid control localization system which consists of multiple devices providing
architectures based on a virtual master, reasonably a virtual robot [96], different aspects of sensory perception, or localization with a method of
a component or a construction site model, and networking robot con­ refinement using a leveled system for global and then local position
trollers are recommended for further development. estimation. Autonomy metrics can be described through the type and
As proposed in [97,98], a CAD model can perform as such a virtual amount of involvement of human operators and collaborators and the
master. By decomposing the described building component into com­ ability to adapt to changes in the environment, a modified ranking
parable subtasks with similar processes, finite discretization of the specified for construction was suggested by [101].
manufacturing is achieved. This is especially necessary when printing
with time-dependent materials such as concrete and stabilizing rebar 4. Architectural application scenarios
integration or supporting robots [99]. The task assignment itself can be
based on a variety of algorithms, with the simplest being random-choice Through use of multiple MAM systems, structures of unbounded size
[100]. However, since not all robots have to be able to perform the same can be made using parallelization of tasks. Through the use of a high
tasks, restrictions are necessary according to the capabilities of the in­ DOF system, a high freedom in deposition is made available through
dividual unit. In sum, hybrid control architectures with adaptive task alternative nozzle orientations and transformations, which leads to
assignment do not only enable scalability and flexibility through the increased freedom of form and structural advantages and a variety of
opportunity of integrating additional functions, but also ensure the application scenarios in building construction. In particular, the inter­
redistribution of tasks in the event of robot failure. action and combined effort of several robots opens up new possibilities
compared to stationary and single-nozzle systems. Therefore, we present
a pair of newly developed collaborative mobile robots aiming to bring
3.2. Summary and performance metric for a classification AM to the construction site (see Fig. 15) [103]. They will act as research
platforms to study the proposed architectural application scenarios as
A classification of several existing mobile system AM processes both follows below.
from academia and industry is made comparing their material, process,
contextual and mechanical aspects (see Table 2).
While the process is characterized by the material used, the method 4.1. Cooperative MAM for new constructions
of deposition can vary, and the review shows that multiple instances of
extrusion can provide differing results based on mechanical components The manufacturing of vertical typologies in buildings, in particular of
such as the nozzle, delivery length and pressure, actuated extruders, and walls and columns, has a limited complexity in terms of their in-place
layer compression. printability, both for stationary or mobile AM systems. However, the
The mobility of the system can be described by the surface traversed, integration of static and building physics functions still makes high
combined with the definition of the work area. Where a climbing system demands to production. The integration of functions can be achieved

Table 2
Comparison of existing mobile systems for AM processes.
Name Developer Material and process Mobility Site Autonomy Localization

Growing robot [39] CMBR, IIT Polymer FDM Uniaxial climbing extended through In Conditional Relative
AM situ automation
Fiberbots [40] MIT Mediated Matter Fiber depositions Uniaxial climbing extended through In Conditional Relative: IMU
AM situ automation
Minibuilders [15] IAAC Concrete Extrusion Climbing extended through AM In Partial automation Absolute: external
situ guiding lines
Koala 3D [41] Univ. of Chile Polymer FDM Climbing while AM Off- Partial automation –
site
AMbots [42] AM3 Lab Polymer FDM Mobile platform with vertical axis Off- Conditional Relative: optical sensors
site automation
AGV with industrial LS2N Polyurethane Automated Ground Vehicle with In Conditional Absolute: external laser
robot [61] 6DOF arm situ automation reflectors
YouWasp [43] UCL Clay Extrusion Mobile platform with 6DOF arm Off- Conditional Absolute: external
site automation optical sensors
Baubot [102] Printstones Concrete Extrusion Mobile platform with 6DOF arm In Conditional –
situ automation
Robot Printers [50,52] Singapore Centre for 3D Concrete Extrusion Mobile platform with 6DOF arm Off- Partial automation Absolute: external
Printing site optical sensors
Cybe [21] Concrete Concrete Extrusion Mobile platform with 6DOF arm In Partial automation –
situ
DCP [45] MIT Mediated Matter Polyurethane foam Mobile platform with 6DOF arm In Conditional Absolute: external laser
spraying mounted on boom-arm situ automation sensor

10
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

rectangular door and window openings, or the integration of other


standardized components.

4.2. Cooperative MAM in existing contexts

MAM systems are of particular importance for building in existing


contexts, reaching places that gantry robots cannot. As existing com­
ponents need to be selectively removed, treated, exchanged or
augmented from a structural, aesthetic or functional point of view, the
use of mobile robots could enable a combination of subtractive and
additive processes, which could be carried out through cooperative skills
of multiple actors. Attaining or adapting digital models through reality
capturing by the mobile systems is thus of high importance for site-
specific work, where the replacement of components is required to fit
seamlessly with the existing construction. An example of this is reno­
vation or retrofitting of a space where interior walls are removed, and a
new layout is designed and manufactured in place, taking into account
inaccuracies in the building, e.g. inclination and skew. While replacing
Fig. 15. A pair of collaborative mobile robot research platforms for in situ AM, existing components are often required in maintaining the building’s
consisting of a wheeled mobile base, an integrated linear axis, and a collabo­ aesthetics and function, integrating additional functions such as heating,
rative 6-axis robot arm, here one of the systems equipped with a clay 3D ventilation, electronics or passive insulation could be beneficial for
printing system.
component enhancement. With MAM systems, these functions could be
integrated by extending the structure through addition of a secondary
though geometrical design, e.g., the integration of voids or gradation in layer, manufactured on-site, against the existent structure.
the density of the print paths, or the integration of additional mechan­
ical systems through the placing or co-extrusion of secondary materials 5. Conclusion
for reinforcement. This approach can be extended by materials design,
which allows for further grading of the component by using a selection This paper has examined mobile autonomous systems for AM in
of mixtures, in the case of a composite material, which can be applied in research and industry, which encompass a wide range of solutions but
specific locations within the building component optimized for the share the same goal of increasing efficiencies in the architecture, engi­
function required. Examples for this are the use of a high strength ma­ neering, and construction (AEC) sector. This efficiency manifests as a
terial used in parts where the building component is highly stressed, or a reduction of time required to fabricating structures, a decrease in ma­
lightweight, insulating material used as part of a façade to allow for a terial used, and an increase in accuracy through improvement building
varying thermal mass and insulation value. Cooperating mobile systems resolution as a result of decreasing margins of tolerance, without the
could extend the existing possibilities already provided by stationary worry of fitment.
AM systems, e.g., by distributing tasks between different units, and thus A large challenge for MAM systems remains the material delivery,
enabling selective grading through delivering multiple tailor-made that is, with increasing scale, more material is required, thus either
material mixtures in building components. limiting the mobility through a tethered connection or requiring mobile
While we see multiple instances of vertical components being pro­ systems to resupply onboard material containers.
duced with in situ AM techniques, the inclusion of horizontal compo­ While mobility and material deposition are required, the critical
nents is still mostly omitted. With vertical components, the structure component in MAM systems is the context-awareness, allowing fabri­
that is printed is supported and stabilized by the previous layers, as the cation of building components in the correct location and with the
structure is not defined by the area it occupies but the height it attains. required relation between them. The absolute localization within the
Conversely, in horizontal components, the ideal is to cover large areas environment using global landmarks through marker or marker-less
while the height is aimed to be minimized to reduce unneccesary weight. tracking, combined with a relative localization to the workpiece, has the
The aim of spanning large areas requires the filament to be deposited potential to allow for accurate continuation and collaboration on
eccentricly, creating overhangs and thus the introduction of concen­ building components with one or multiple mobile systems.
trated stresses and possible structural instability. Horizontal typologies, With the introduction of multiple MAM systems, building tasks can
such as floors and ceilings, are thus often prefabricated for installation be distributed amongst actors, often through discretization of building
on-site. Alternatively, tensile stresses could be avoided for in situ AM, for components. A deep understanding of movement of the mobile systems
which the geometrical design could be optimized to ensure a is required, while a flexible control layer should be considered that can
compression-dominant structure with the introduction of curvature. redistribute tasks based on unforeseen circumstances. In extension to
Here, cooperative abilities of MAM could be leveraged, e.g., by intro­ large scale stationary systems, the increased contextual awareness
ducing temporary support by one robot stabilizing a structure and combined with mobility provides new opportunities for use of the
enabling the material deposition by another. Other options could technology in existing contexts. The autonomy of the MAM system could
include the material deposition of a secondary material to create also enable collaboration with humans on site, which would allow for
removable support structures where necessary. These support structures more freedom in the design of AM buildings. Highly complex and
would allow for an extended form-freedom by removing some con­ cognitively challenging tasks involving skills and craft could be carried
straints on the process, including structural or process-related out by humans, which could ideally be complemented by robots per­
limitations. forming tasks that require, for example, high geometric precision, high
Cooperation with humans on-site could additionally include a com­ force-adaptivity, or require a high payload capacity.
bination of processes, where the robotic system manufactures a building
component through additive means, and the human places an element as CRediT authorship contribution statement
part of a prescribed or interactive process, whereafter the MAM system
perceives the object placed and adapts its process to include it. A prac­ Kathrin Dörfler: Conceptualization, Writing - original draft, Writing -
tical example of this process is the inclusion of lintels to create regular review and editing.

11
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

Gido Dielemans: Writing - original draft, Writing - review and [19] E.L. Kreiger, M.A. Kreiger, M.P. Case, Development of the construction processes
for reinforced additively constructed concrete, Addit. Manuf. 28 (Aug. 2019)
editing.
39–49, https://doi.org/10.1016/J.ADDMA.2019.02.015.
Lukas Lachmayer: Writing - original draft, Writing - review and [20] Apis Cor Construction, 2021. http://apis-cor.com/en/. (Accessed 29 November
editing. 2017).
Tobias Recker: Writing - original draft, Writing - review and editing. [21] “CyBe Construction.” n.d. https://www.cybe.eu/cyberc3dp/ (accessed Nov. 29,
2017).
Annika Raatz: Writing - review and editing. [22] “Project gallery : Twente Additive Manufacturing.” n.d. https://www.twente-am.
Dirk Lowke: Writing - original draft, Writing - review and editing. com/projects/ (accessed Feb. 05, 2022).
Markus Gerke: Writing - review and editing. [23] R. Duballet, O. Baverel, J. Dirrenberger, Classification of building systems for
concrete 3D printing, Autom. Constr. 83 (2017) 247–258, https://doi.org/
10.1016/j.autcon.2017.08.018.
[24] K. Dörfler, et al., Architectural-scale mobile robotic fabrication beyond structured
Declaration of competing interest factory conditions: case study mesh mould wall of the DFAB HOUSE, Constr.
Robot. 2 (2019), https://doi.org/10.1007/s41693-019-00020-w.
[25] S. Çetin, C.De Wolf, N. Bocken, Circular digital built environment: An emerging
The authors certify that they have no affiliations with or involvement framework, Sustainability 13 (11) (2021), https://doi.org/10.3390/su13116348.
in any organization or entity with any financial interest (such as hono­ [26] Y. Huang, L. Alkhayat, C.De Wolf, C. Mueller, Algorithmic Circular Design With
raria; educational grants; participation in speakers’ bureaus; member­ Reused Structural Elements: Method and Tool, 2021, https://doi.org/10.3929/
ethz-a-010025751.
ship, employment, consultancies, stock ownership, or other equity [27] T. Salmi, J.M. Ahola, T. Heikkilä, P. Kilpeläinen, T. Malm, in: Human-Robot
interest; and expert testimony or patent-licensing arrangements), or Collaboration and Sensor-Based Robots in Industrial Applications and
non-financial interest (such as personal or professional relationships, Construction, 2018, pp. 25–52, https://doi.org/10.1007/978-3-319-70866-9_2.
[28] S. Haddadin, M. Suppa, S. Fuchs, T. Bodenmüller, A. Albu-Schäffer, G. Hirzinger,
affiliations, knowledge or beliefs) in the subject matter or materials in: Towards the Robotic Co-worker, 2011, pp. 261–282, https://doi.org/
discussed in this manuscript. 10.1007/978-3-642-19457-3_16.
[29] L. Wang, S. Liu, H. Liu, X.V. Wang, Overview of Human-robot Collaboration in
Manufacturing May, Springer International Publishing, 2020.
Acknowledgment [30] “A Case for Cooperation Between Machines and Humans - The New York Times.”
n.d. https://www.nytimes.com/2020/05/21/technology/ben-shneiderman-
This research was supported by the Deutsche For­ automation-humans.html (accessed Oct. 31, 2021).
[31] B. Shneiderman, Human-centered artificial intelligence: reliable, safe &
schungsgemeinschaft (DFG, German Research Foundation) – Pro­ trustworthy 36 (6) (Apr. 2020) 495–504, https://doi.org/10.1080/
jektnummer 414265976 – TRR 277 Additive Manufacturing in 10447318.2020.1741118.
Construction. [32] P. Maes, B. Shneiderman, J. Miller, Intelligent Software Agents vs. User-
Controlled Direct Manipulation: A Debate, 1997.
[33] M.A. Goodrich, A.C. Schultz, Human-robot interaction: a survey, Found. Trends
References Human-Computer Interact. 1 (3) (2007) 203–275, https://doi.org/10.1561/
1100000005.
[34] J.M. Beer, A.D. Fisk, W.A. Rogers, Toward a framework for levels of robot
[1] F. Bos, R. Wolfs, Z. Ahmed, T. Salet, Additive manufacturing of concrete in
autonomy in human-robot interaction, J. Human-Robot Interact. 3 (2) (2014) 74,
construction: potentials and challenges of 3D concrete printing, Virtual Phys.
https://doi.org/10.5898/jhri.3.2.beer.
Prototyp. 11 (3) (2016) 209–225, https://doi.org/10.1080/
[35] I.X. Han, F. Meggers, S. Parascho, Bridging the collectives: a review of collective
17452759.2016.1209867.
human–robot construction, Int. J. Archit. Comput. 19 (4) (2021) 512–531,
[2] A. Paolini, S. Kollmannsberger, E. Rank, Additive manufacturing in construction:
https://doi.org/10.1177/14780771211025153.
a review on processes, applications, and digital planning methods, Addit. Manuf.
[36] M. Johnson, J.M. Bradshaw, P.J. Feltovich, Tomorrow’s human-machine design
30 (2019), 100894, https://doi.org/10.1016/j.addma.2019.100894.
tools: from levels of automation to interdependencies, J. Cogn. Eng. Decis. Mak.
[3] T. Wangler, et al., Digital concrete: opportunities and challenges, RILEM Tech.
12 (1) (2018) 77–82, https://doi.org/10.1177/1555343417736462.
Lett. 1 (Oct. 2016) 67, https://doi.org/10.21809/rilemtechlett.2016.16.
[37] L. Atanasova, D. Mitterberger, T. Sandy, F. Gramazio, M. Kohler, K. Dörfler, in:
[4] A.U. Rehman, J. Caron, Cement and Concrete Research The Realities of
Prototype as Artefact - Design Tool for Open-ended Collaborative Assembly
Additively Manufactured Concrete Structures in Practice, 2022.
Processes, ACADIA, 2020, pp. 1–11.
[5] B. Khoshnevis, Automated construction by contour crafting – related robotics and
[38] N. Mavridis, T. Bourlai, D. Ognibene, The human-robot cloud: situated collective
information technologies, Autom. Constr 13 (1) (2004) 5–19, https://doi.org/
intelligence on demand, in: Proc. - 2012 IEEE Int. Conf. Cyber Technol. Autom.
10.1016/j.autcon.2003.08.012 [Online]. Available.
Control. Intell. Syst. CYBER 2012 May, 2012, pp. 360–365, https://doi.org/
[6] “Autodesk Project Escher Finds First Commercial Release in Cronus |
10.1109/CYBER.2012.6392580.
Engineering.com.” n.d.
[39] A. Sadeghi, A. Mondini, B. Mazzolai, Toward self-growing soft robots inspired by
[7] L. Cai, On-Site Autonomous Fabrication at Architectural Scales, Massachusetts
plant roots and based on additive manufacturing technologies, Soft Robot. 4 (3)
Institute of Technology, 2018. https://dspace.mit.edu/handle/1721.1/119078.
(2017) 211–223, https://doi.org/10.1089/soro.2016.0080.
[8] E. Kreiger, B. Diggs-McGee, T. Wood, B. MacAllister, M. Kreiger, Field
[40] M. Kayser, et al., FIBERBOTS: an autonomous swarm-based robotic system for
Considerations for Deploying Additive Construction, vol. 28, Springer
digital fabrication of fiber-based composites, no. 1, in: Constr. Robot. 2018 21 2,
International Publishing, 2020.
2018, pp. 67–79, https://doi.org/10.1007/S41693-018-0013-Y. Dec.
[9] P. Carneau, R. Mesnil, N. Roussel, O. Baverel, Additive manufacturing of
[41] M. Vélez, E. Toala, J.C. Zagal, Koala 3D: A continuous climbing 3D printer, in:
cantilever - from masonry to concrete 3D printing, Autom. Constr. 116 (2020),
Robot. Comput. Integr. Manuf. 64, 2020, https://doi.org/10.1016/j.
https://doi.org/10.1016/j.autcon.2020.103184.
rcim.2020.101950. January.
[10] H. K., N.H.Hendrik Lindemann [Online]. Available, in: “Gradual Transition
[42] L.G. Marques, R.A. Williams, W. Zho, A mobile 3D printer for cooperative 3D
Shotcrete 3D Printing,” Poster August, 2018, p. 1, https://www.researchgate.net/
printing 2. Mobile printer design, in: Solid Free. Fabr. Symp., 2017,
profile/Norman_Hack/publication/326735829_Gradual_Transition_Shotcrete_3
pp. 1645–1660.
D_Printing/links/5b6189970f7e9bc79a731ea6/Gradual-Transition-Shotcrete-3
[43] J. Sustarevas, K.X. Benjamin Tan, D. Gerber, R. Stuart-Smith, V.M. Pawar,
D-Printing.pdf.
YouWasps: towards autonomous multi-robot Mobile deposition for construction,
[11] H. Kloft, M. Empelmann, N. Hack, E. Herrmann, D. Lowke, Reinforcement
IEEE Int. Conf. Intell. Robot. Syst. (2019) 2320–2327, https://doi.org/10.1109/
strategies for 3D-concrete-printing, Civ. Eng. Des. 2 (4) (2020) 131–139, https://
IROS40897.2019.8967766.
doi.org/10.1002/cend.202000022.
[44] K. Dörfler, T. Sandy, M. Giftthaler, F. Gramazio, M. Kohler, J. Buchli, Mobile
[12] B. Jenett, A. Abdel-Rahman, K. Cheung, N. Gershenfeld, Material–Robot system
robotic brickwork: automation of a discrete robotic fabrication process using an
for assembly of discrete cellular structures, IEEE Robot. Autom. Lett. 4 (4) (2019)
autonomous mobile robot, in: Robotic Fabrication in Architecture, Art and Design
4019–4026, https://doi.org/10.1109/lra.2019.2930486.
2016, 2016, pp. 204–2017, https://doi.org/10.1007/978-3-319-04663-1.
[13] M. Lombarda I. C. Wasp A. S. Project M. C. Architects I. Calvinon.d.“TECLA –
[45] S.J. Keating, J.C. Leland, L. Cai, N. Oxman, Toward site-specific and self-sufficient
Technology and Clay The First Eco-sustainable House 3D Printed From Raw Earth
robotic fabrication on architectural scales, Sci. Robot. 2 (5) (2017) 15, https://
is Now a Reality.”.
doi.org/10.1126/scirobotics.aam8986.
[14] “ 3D Printer House | Crane WASP - 3D Printers | WASP.”n.d. https://
[46] “Home | Printstones.”n.d. https://www.printstones.eu/ (accessed Sep. 30, 2021).
www.3dwasp.com/en/3d-printer-house-crane-wasp/ (accessed Sep. 26, 2021).
[47] A. Thoma, A. Mirjan, V. Helm, F. Gramazio, M. Kohler, Rock print: an
[15] “Minibuilders.” n.d. http://robots.iaac.net/ (accessed Aug. 02, 2017).
architectural installation of granular matter, Act. Matter (2019), https://doi.org/
[16] J. Buchli, et al., Digital in situ fabrication - challenges and opportunities for
10.7551/mitpress/11236.003.0045.
robotic in situ fabrication in architecture, construction, and beyond, Cem. Concr.
[48] C. Gosselin, R. Duballet, P. Roux, N. Gaudillière, J. Dirrenberger, P. Morel, Large-
Res. 112 (2018) 66–75, https://doi.org/10.1016/j.cemconres.2018.05.013.
scale 3D printing of ultra-high performance concrete – a new processing route for
[17] COBOD, 2021 [Online]. Available, https://cobod.com/. (Accessed 9 January
architects and builders, Mater. Des. 100 (Jun. 2016) 102–109, https://doi.org/
2021).
10.1016/J.MATDES.2016.03.097.
[18] “Home | ICON.” n.d. https://www.iconbuild.com/ (accessed Feb. 05, 2022).

12
K. Dörfler et al. Cement and Concrete Research 158 (2022) 106772

[49] A. Gawel, et al., A fully-integrated sensing and control system for high-accuracy LNCS, 2018, pp. 254–269, https://doi.org/10.1007/978-3-319-91635-4_14.
mobile robotic building construction, IEEE Int. Conf. Intell. Robot. Syst. (2019) January.
2300–2307, https://doi.org/10.1109/IROS40897.2019.8967733. [77] C. Feng, Y. Xiao, A. Willette, W. McGee, V.R. Kamat, Vision guided autonomous
[50] M.E. Tiryaki, X. Zhang, Q.-C. Pham, Printing-while-moving: a new paradigm for robotic assembly and as-built scanning on unstructured construction sites, Autom.
large-scale robotic 3D Printing [Online]. Available, 2018, http://arxiv.org/abs/1 Constr. 59 (2015) 128–138, https://doi.org/10.1016/j.autcon.2015.06.002.
809.07940. [78] J. Leonard, H. Durrant-Whyte, Simultaneous Map Building and Localization for
[51] R.A. Buswell, et al., A process classification framework for defining and an Autonomous Mobile Robot, 1991, https://doi.org/10.1109/
describing digital fabrication with concrete, Cem. Concr. Res. 134 (March) IROS.1991.174711.
(2020), https://doi.org/10.1016/j.cemconres.2020.106068. [79] S. Leutenegger, S. Lynen, M. Bosse, R. Siegwart, P. Furgale, Keyframe-based
[52] X. Zhang, et al., Large-scale 3D printing by a team of mobile robots, Autom. visual–inertial odometry using nonlinear optimization, Int. J. Robot. Res. 34 (3)
Constr. 95 (August) (2018) 98–106, https://doi.org/10.1016/j. (2015) 314–334.
autcon.2018.08.004. [80] C. Houseago, M. Bloesch, S. Leutenegger, KO-fusion: dense visual SLAM with
[53] L. Poudel, et al., Architecting the cooperative 3d printing system, Proc. ASME tightly-coupled kinematic and odometric tracking, in: International Conference
Des. Eng. Tech. Conf. 9 (Nov. 2020), https://doi.org/10.1115/DETC2020-22711. on Robotics and Automation (ICRA) 2019, 2019, pp. 4054–4060.
[54] “Printstones Mobile Concrete 3D Printing - YouTube.” n.d.https://www.youtube. [81] A. Nüchter, K. Lingemann, J. Hertzberg, H. Surmann, 6D SLAM-3D Mapping
com/watch?v=CZZVvtAjjmM (accessed Oct. 28, 2021). Outdoor Environments, 2007, https://doi.org/10.1002/rob.20209.
[55] TRR 277 AMC A03. https://www.tu-braunschweig.de/trr277/projects/amc-struc [82] R. Bao, R. Komatsu, R. Miyagusuku, M. Chino, A. Yamashita, H. Asama, Cost-
ture/a-projects/a03. (Accessed 28 October 2021). effective and robust visual based localization with consumer-level cameras at
[56] Y. Tan, M. Dahlenburg, S. Kessler, J. Fottner, Virtual Prototyping mit DEM zur construction sites, in: 2019 IEEE 8th Glob. Conf. Consum. Electron. GCCE 2019
Entwicklung eines Near-Nozzle-Mixing Verfahrens für den additiven 3D Oct, 2019, pp. 983–985, https://doi.org/10.1109/GCCE46687.2019.9015417.
Betondruck für den Roboter Einsatz, in: 25. Fachtagung Schüttgutfördertechnik [83] C. Ntanos, C. Botsikas, G. Rovis, P. Kakavas, D. Askounis, A context awareness
2021, OVGU, Magdeburg, 2021. framework for cross-platform distributed applications, J. Syst. Softw. 88 (1)
[57] N. Zhang, M. Xia, J. Sanjayan, Short-duration near-nozzle mixing for 3D concrete (2014) 138–146, https://doi.org/10.1016/j.jss.2013.10.018.
printing, Cem. Concr. Res. 151 (Jan. 2022), 106616, https://doi.org/10.1016/J. [84] A.J. Davison, FutureMapping: The Computational Structure of Spatial AI Systems,
CEMCONRES.2021.106616. arXiv, 2018.
[58] L. Reiter, T. Wangler, A. Anton, R.J. Flatt, Setting on demand for digital [85] J. McCormac, A. Handa, A. Davison, S. Leutenegger, SemanticFusion: dense 3D
concrete–Principles, measurements, chemistry, validation, Cem. Concr. Res. 132 semantic mapping with convolutional neural networks, Proc. - IEEE Int. Conf.
(2020), 106047. Robot. Autom. (2017) 4628–4635, https://doi.org/10.1109/
[59] S. Muthukrishnan, S. Ramakrishnan, J. Sanjayan, Technologies for improving ICRA.2017.7989538.
buildability in 3D concrete printing, Cem. Concr. Compos. 122 (Sep. 2021), [86] T. Sandy, M. Giftthaler, K. Dörfler, M. Kohler, J. Buchli, Autonomous
104144, https://doi.org/10.1016/J.CEMCONCOMP.2021.104144. repositioning and localization of an in situ fabricator, in: Proceedings - IEEE
[60] Y. Tao, A.V. Rahul, K. Lesage, Y. Yuan, K. Van Tittelboom, G. De Schutter, International Conference on Robotics and Automation (ICRA 2016), 2016,
Stiffening control of cement-based materials using accelerators in inline mixing pp. 2852–2858, https://doi.org/10.1109/ICRA.2016.7487449.
processes: possibilities and challenges, Cem. Concr. Compos. 119 (May 2021), [87] Z. Feng, G. Hu, Y. Sun, J. Soon, An overview of collaborative robotic
103972, https://doi.org/10.1016/J.CEMCONCOMP.2021.103972. manipulation in multi-robot systems, Annu. Rev. Control. 49 (Jan. 2020)
[61] K. Subrin, T. Bressac, S. Garnier, A. Ambiehl, E. Paquet, B. Furet, Improvement of 113–127, https://doi.org/10.1016/J.ARCONTROL.2020.02.002.
the mobile robot location dedicated for habitable house construction by 3D [88] J. McPherson, W. Zhou, A chunk-based slicer for cooperative 3D printing, Rapid
printing 51 (11) (2018) 716–721, https://doi.org/10.1016/j.ifacol.2018.08.403. Prototyp. J. 24 (9) (2018) 1436–1446, https://doi.org/10.1108/RPJ-07-2017-
[62] H. Blum, J. Stiefel, C. Cadena, R. Siegwart, A. Gawel, Precise Robot Localization 0150.
in Architectural 3D Plans [Online]. Available, 2020, http://arxiv.org/abs/2 [89] H. Shen, L. Pan, J. Qian, Research on large-scale additive manufacturing based on
006.05137. multi-robot collaboration technology, Addit. Manuf. 30 (Dec. 2019), 100906,
[63] S.G. Tzafestas, Mobile robot control and navigation: a global overview, J. Intell. https://doi.org/10.1016/J.ADDMA.2019.100906.
Robot. Syst. 91 (1) (2018) 35–58, https://doi.org/10.1007/s10846-018-0805-9. [90] S. Elagandula, L. Poudel, Z. Sha, W. Zhou, Multi-robot path planning for
[64] K. Asadi, et al., An integrated UGV-UAV system for construction site data cooperative 3D printing, in: ASME 2020 15th Int. Manuf. Sci. Eng. Conf. MSEC
collection, Autom. Constr. 112 (Apr. 2020), 103068, https://doi.org/10.1016/J. 2020 1, Jan. 2021, https://doi.org/10.1115/MSEC2020-8390.
AUTCON.2019.103068. [91] H. Bui, Toolpath planning methodology for multi-gantry fused filament
[65] G. Pritschow, M. Dalacker, J. Kurz, Configurable Control System of a Mobile fabrication 3D printing, Aug. 2019. Grad. Theses Diss.
Robot for On-Site Construction of Masonry May, 1993, https://doi.org/ [92] L.E. Parker, Multiple mobile robot systems, in: Springer Handb. Robot., 2008,
10.22260/ISARC1993/0012. pp. 921–941, https://doi.org/10.1007/978-3-540-30301-5_41.
[66] D.K. Ballast, Handbook of Construction Tolerances, Wiley, 2007. [93] A. Costa, B. Jenett, I. Kostitsyna, A. Abdel-Rahman, N. Gershenfeld, K. Cheung,
[67] S. Szyminski, Allgemeintoleranzen, in: Toler. und Passungen, 1993, pp. 111–116, Algorithmic approaches to reconfigurable assembly systems, in: IEEE Aerosp.
https://doi.org/10.1007/978-3-663-05801-4_9. Conf. Proc 2019-March, Aug. 2020, https://doi.org/10.1109/
[68] S.G. Tzafestas, Introduction to Mobile robot control, Introd. to Mob. Robot aero.2019.8741572.
Control (2013) 1–691. [94] A. Stroupe, T. Huntsberger, A. Okon, H. Aghazarian, M. Robinson, Behavior-
[69] N.L. Doh, H. Choset, W.K. Chung, Relative localization using path odometry based multi-robot collaboration for autonomous construction tasks, in: 2005
information, Auton. Robots 21 (2) (2006) 143–154, https://doi.org/10.1007/ IEEE/RSJ Int. Conf. Intell. Robot. Syst, IROS, 2005, pp. 1495–1500, https://doi.
s10514-006-6474-8. org/10.1109/IROS.2005.1545269.
[70] M.B. Alatise, G.P. Hancke, A review on challenges of autonomous Mobile robot [95] C.A.C. Parker, H. Zhang, Collective robotic site preparation, Adapt. Behav. 14 (1)
and sensor fusion methods, IEEE Access 8 (2020) 39830–39846, https://doi.org/ (2006) 5–19, https://doi.org/10.1177/105971230601400101.
10.1109/ACCESS.2020.2975643. [96] T. Recker, M. Heinrich, A. Raatz, A comparison of different approaches for
[71] S.M. Malagon-Soldara, M. Toledano-Ayala, G. Soto-Zarazua, R.V. Carrillo- formation control of nonholonomic mobile robots regarding object transport,
Serrano, E.A. Rivas-Araiza, Mobile robot localization: a review of probabilistic Procedia CIRP 96 (Jan. 2021) 248–253, https://doi.org/10.1016/J.
map- based techniques [Online]. Available, in: IAES Int. J. Robot. Autom 4, 2015, PROCIR.2021.01.082.
pp. 73–81, http://www.iaescore.com/journals/index.php/IJRA/article/view [97] S. Parascho, I. Han, S. Walker, A. Beghini, E. Bruun, S. Adriaenssens, Robotic
/968/pdf. vault: a cooperative robotic assembly method for brick vault construction, Constr.
[72] Z. Wang, H. Li, X. Yang, Vision-based robotic system for on-site construction and Robot. 4 (2020), https://doi.org/10.1007/s41693-020-00041-w.
demolition waste sorting and recycling, J. Build. Eng. 32 (2020), 101769, https:// [98] “OATD: Parascho, Stefana - Cooperative Robotic Assembly: Computational
doi.org/10.1016/j.jobe.2020.101769. Design and Robotic Fabrication of Spatial Metal Structures.” n.d.
[73] M. Drumheller, Mobile robot localization using sonar, IEEE Trans. Pattern Anal. [99] S. Neudecker, et al., A new robotic spray Technology for Generative
Mach. Intell. vol. PAMI-9, no. 2 (1987) 325–332, https://doi.org/10.1109/ Manufacturing of complex concrete structures without formwork, Procedia CIRP
TPAMI.1987.4767907. 43 (Jan. 2016) 333–338, https://doi.org/10.1016/J.PROCIR.2016.02.107.
[74] S. Ercan, S. Meier, F. Gramazio, M. Kohler, Automated localization of a mobile [100] J. Mclurkin D. Yamins n.d. “Dynamic Task Assignment in Robot Swarms.”.
construction robot with an external measurement device, in: Proc. 36th Int. [101] N. Melenbrink, J. Werfel, A. Menges, On-site autonomous construction robots:
Symp. Autom. Robot. Constr. ISARC 2019, Isarc, 2019, pp. 929–936, https://doi. towards unsupervised building, Autom. Constr. 119 (2020), 103312, https://doi.
org/10.22260/isarc2019/0124. org/10.1016/j.autcon.2020.103312.
[75] V. Rosas-Cervantes, S.-G. Lee, 3D localization of a mobile robot by using Monte [102] “ Baubot | Mobile Construction Robots.” n.d. https://www.baubot.com/
Carlo algorithm and 2D features of 3D point cloud, Int. J. Control. Autom. Syst. (accessed Aug. 24, 2021).
18 (11) (2020) 2955–2965, https://doi.org/10.1007/s12555-019-0313-0. [103] G. Dielemans, K. Dörfler, Mobile additive manufacturing: a robotic system for
[76] P. Kim, J. Chen, J. Kim, Y.K. Cho, SLAM-driven intelligent autonomous mobile cooperative on-site construction, in: IROS 2021 Workshop - Robotic Fabrication:
robot navigation for construction applications, in: Lect. Notes Comput. Sci. Sensing in Additive Construction, 2021.
(including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics) 10863,

13

You might also like