Content uploaded by David Rosenwasser
Author content
All content in this area was uploaded by David Rosenwasser on Mar 13, 2019
Content may be subject to copyright.
ROBOSENSE 2.0
1 Facade mock-up of physical
prototypes, showing how user
input (beginning at the central
component) can affect pattern
and line quality of subsequent
components.
Jeremy Bilotti*
Bennett Norman*
David Rosenwasser *
*Authors contributed equally
to the research
Jingyang Leo Liu
Jenny Sabin
Sabin Design Lab, AAP, Cornell
University
Robotic Sensing and Architectural Ceramic Fabrication
1
ABSTRACT
“Robosense 2.0: Robotic Sensing and Architectural Ceramic Fabrication” demonstrates
a generative design process based on collaboration between designers, robotic tools,
advanced software, and nuanced material behavior. The project employs fabrication tools
that are typically used in highly precise and predetermined applications, but uniquely
thematizes the unpredictable aspects of these processes as applied to architectural
component design. By integrating responsive sensing systems, this paper demonstrates
real-time feedback loops that consider the spontaneous agency and intuition of the archi-
tect (or craftsperson) rather than the execution of static or predetermined designs. This
paper includes new developments in robotics software for architectural design applica-
tions, ceramic-deposition 3D printing, sensing systems, materially driven pattern design,
and techniques with roots in the arts and crafts. Considering the increasing accessibility
and advancement of 3D printing and robotic technologies, this project seeks to challenge
the erasure of materiality: when mistakes or accidents caused by inconsistencies in
natural material are avoided or intentionally hidden. Instead, the incorporation of material
and user-input data yields designs that are imbued with more nuanced traces of making.
This paper suggests the potential for architects and craftspeople to maintain a more direct
and active relationship with the production of their designs.
INTRODUCTION
Unlike robots in industry, which execute predened and
repetitive tasks in a controlled environment, robotics in
design and architecture are becoming increasingly involved
with uncertain tasks within more complex and dynamic
contexts. Manufacturing machines and robots have now
started to gain intelligence—actively communicating,
monitoring, and sensing—and with this, the ability to
react (Menges 2014). To facilitate the feedback between
design and robotic fabrication, a Python-based interface,
encapsulating communication protocols and robotic
manipulation libraries, was created in Robosense 1.0
(Moorman, Liu, and Sabin 2016). The interface seamlessly
bridges the gap between physical and digital environments
and allows for a feedback-oriented robotic fabrication
paradigm. Building upon Robosense 1.0, Robosense 2.0
steps forward and integrates the interface into design
software Rhinoceros 3D and Grasshopper.
Since 2009, the Sabin Design Lab has innovated digital
ceramics through 3D-printed ceramic bricks and
nonstandard componentry (Sabin 2010). The plastic nature
of clay offers a potent material solution to contemporary
generative design processes in architecture, which
frequently feature organic and natural forms of increasingly
complex expression and ornamentation (Sabin et al. 2014).
The use of clay to integrate the designer’s intuition and
to rationalize complex data and geometries has been
incorporated into the design process in alternate industries
such as car and boat design for decades, but professionals
in the broader eld of architecture have yet to widely explore
the potential for these materials and tools to augment
the efciency and quality of built design work. This paper
explores the transfer of information and data, including
real-time feedback via sensing technologies, through the
hybridization of crafts-based ceramic techniques with
contemporary digital design, robotic 3D printing, and
nonstandard component–based architectural assemblies.
We propose the development of a responsive feedback
system that provides the designer/maker with information
about the material, allowing for intuitive, on-the-y
modications to the design process during the course
of fabrication. Reciprocally, the designer’s choices and
changes are registered, and the software responds in real
time to create a uid workow that informs and unlocks the
potential for more a nuanced understanding of 3D-printed
clay as an architectural fabrication technique.
BACKGROUND
Robotically Fabricated Ceramics
Robosense 2.0 uses the material language of clay
deposition 3D printing, the extruded clay bead, as a medium
for developing physical case studies (Figure 1). Existing
precedents that investigate clay deposition printing tend
to use a technique borrowed from other forms of 3D
printing: predetermined extruder motions create ne
stacked layers to produce volumetric forms, similarly to
how coil pots have traditionally been constructed, and
to how low-cost plastic deposition 3D printers are able
to produce objects with high efciency. The Institute for
Advanced Architecture Catalonia (IAAC) is innovating the
use of robotically fabricated clay components for large-
scale applications with exceptional results for the purpose
of creating architectural enclosures. While their work is
innovative in scale, clay body, and precision, the production
methodology and extrusion techniques do not signicantly
vary from typical 3D-printing operations. The outcomes of
these prints have a high level of predictability because all
of the material deposited is fully supported, constraining
designs to entirely enclosed volumetric forms that lack
more advanced material intelligence (Chronis et al. 2017).
Other precedents have used the extruded clay bead to
develop more expressive architectural screens that allow
for deviation from preprogrammed behavior, but often with
difcult-to-control outcomes. As a result, the complexities
of drying, ring, and glazing these components at a large
scale of production make them difcult to envision as
applied to architectural building systems. Both Harvard
GSD’s Woven Clay project and Cornell University’s
Clay Non-Wovens use techniques inuenced by textile
manufacturing in order to develop these patterning and
screen systems, which can negotiate circumstances of light
through thin ceramic panels. Both projects are successful
in challenging ideas of patterning and demonstrating
potential for robotic fabrication in clay (deviating from the
process of printing layer on top of layer). The challenges
discovered in these projects relate to unpredictable
tolerances and warping due to the ring process.
Furthermore, joinery and connection detail in both projects
are unresolved, inhibiting either from becoming truly
scalable (Rosenwasser, Mantell, and Sabin 2017; Friedman,
Kim, and Mesa 2014).
Ron Rael and Virginia San Fratello of Emerging Objects
have worked extensively with 3D-printed ceramic material,
both powder-based using Z Corp 3D printers and with
delta bot machines (2018). Emerging Objects’ emphasis
on G-code “glitch” focuses on code manipulation of
traditional 3D-printing methods to create textures and
coded mistakes within clay material. This paper presents
an alternative to preprogrammed “glitches” through a
sensing-based process that engages the mistake, error,
and inconsistencies through human interaction and natural
material response during the process of fabrication.
Existing 6-Axis Robotics Software
When engaging in architectural fabrication using a 6-axis
robot, a designer or technician must rst create a toolpath
using computer-aided machining (CAM) and or/computer-
aided design (CAD) software for the robot to execute.
There are a handful of robotics software programs for
Rhinoceros and Grasshopper, including HAL, TACO and
KUKA|prc. HAL provides reverse kinematics solving,
simulation, and code generation. TACO is similar to HAL and
offers the ability to generate code to coordinate multiple
robots and upload RAPID code to the robot controller
directly. KUKA|prc provides a similar feature set to HAL for
KUKA robots. While these software products are excellent
for simulating and generating code, there are no existing
software interfaces readily accessible to the design and
architecture community that allow human or environmental
input to adjust and redesign the toolpath as the robot is
executing code. Thus, opportunities to understand the vast
potential of designing alongside robotic tools remain largely
inaccessible to design professionals; unincorporated into
the typical workow of an architectural design practice.
METHODS
Software for Robotic Motion Design
In order to produce a ceramic building component using
a material deposition system (a three-dimensional or 3D
printing system) on the end of a robotic arm, design intent
must be translated from the architect’s ideas into data
that controls the motion of the robotic arm. The transition
from design to code (executable by a robotic tool) is often
understood by architects and nonexperienced users on a
very rudimentary level.
Generating the Toolpath
To uidly translate design intent into robotic movement, a
script for generating bespoke warp and weft bead patterns
is developed. Instead of starting with a three-dimensional
model that is then translated into a toolpath, the designer
understands the design process beginning with the toolpath
itself, informed by the nature of the continuous bead. By
using an easily accessible visual scripting interface such
as Grasshopper, one is able to generate continuous-line
toolpaths for ceramic extrusion while simultaneously
considering the limitations and constraints of the robot’s
motions. A script is developed to function in the following
way, depicted visually in Figure 2:
• A bounding volume is assigned, either sourced from the
Rhinoceros 3D modeling workspace or generated using
the script. This volume may be modied and adapted
throughout the design process.
• The Grasshopper script produces a series of horizontal
layers within the bounding volume at an assignable
interval, each consisting of a line that weaves back and
forth across the bounding volume (like the warp and
weft of a textile). The script ensures that each layer
reaches the extents of the bounding volume.
• Each layer is manipulable in three dimensions; the print
is not constrained to at, two-dimensional layers. In
the script developed for these tests, points on each
layer’s edges are xed to the bounding volume to ensure
proper adhesion to a plaster mold in which the clay is to
be printed; the portion of each layer in the center of the
print may be assigned a differing spacing (Z-coordinate
value), or may be translated into different positions
altogether. The potential for complex, multi-axis motion
(4, 5, 6 or more axes moving simultaneously to produce)
is not limited by the script.
• The script connects each layer to the next, regardless
of three-dimensional manipulations, ensuring that the
toolpath remains continuous. The endpoint of each layer
is necessarily adjacent to the start point of the next in
order to produce a continuous path for the robot end
effector to follow.
• The script determines orientation planes for every
target point (ordered coordinates that guide the robot’s
motion) in the toolpath, prescribing the end effector’s
orientation in space. The designer is able to access,
parse, and manipulate this data in order to design
specic orientations for the tool based on his or her
intent or knowledge of the fabrication process.
• HAL receives the continuous line toolpath from the
Grasshopper script and translates it into RAPID code,
displaying the code to the user/designer in real time.
The user may preview the motion of the robot before
beginning, using a built-in visualizer paired with HAL’s
robot simulation tools.
Sending the Toolpath
Next, RAPID code (containing joint positions and/or target
coordinates) is sent to the robot. Robosense 1.0 used the
open source software Open ABB to communicate directly
with an ABB IRB 4600 from Processing, a programming
language for visual art (Moorman, Liu, and Sabin 2016).
Open ABB provides the benet of sending individual
commands such as speed adjustments, cartesian moves,
and joint positioning directly to the robot rather than
executing an entire list of static commands produced
by HAL. The motivation of this project is to communicate
directly with the robot from design software such as
Rhino and Grasshopper. A server, written in the ABB robot
Robosense 2.0 Bilotti, Norman, Rosenwasser, Liu, Sabin
language RAPID, is loaded onto the robot controller and
a Grasshopper component creates a client in GH Python
to connect to the Open ABB server on the robot controller.
Grasshopper can directly control an ABB robot without the
use of expensive proprietary software.
Se ns ing an d Ad jus ting t he To ol path Us in g Ma te rial
Behavior
With the ability to intervene in the execution of a toolpath,
changes to the robot's motion can be made based on the
state of the fabrication environment. This is realized by
reading sensor data into Grasshopper using an Arduino
and the Grasshopper plugin Firey (Payne and Johnson
2013). A relationship is dened between the sensor value
and a toolpath adjustment. For instance, an increase in
clay body moisture content can be mapped to a decrease in
robot travel speed. A series of exercises, described in the
Exercises section, were performed to test the software and
understand the relationship between sensors, software,
and material feedback. Figure 3 is a system diagram of
Robosense 2.0.
Extruder 2.0
In order to seamlessly integrate clay extrusion into the
robotic sensing environment, a previously designed
custom extrusion system was signicantly adapted for the
purpose of engaging a variety of sensors, and allowing for
uninhibited 6-axis motion of the robot arm (Rosenwasser
et al. 2017). These revisions include a newly designed end
effector, Extruder 2.0, which is modular in its construction.
One improvement includes the addition of aluminum
mounting plates, which facilitates infrared temperature
sensors, cameras for recording extrusion, 3D scanners,
or other sensing systems (Figure 4). A hygrometer is used
to measure moisture content of clay by registering the
electrical resistance or capacitance of the material. This
device expands the sensing environment to the clay mixing
area, where one may precisely monitor and record the
moisture of the clay body during preparation.
Extruder 2.0 discharges clay using compressed air only
and does not depend on stepper motors or augers to move
material to the end effector. By continuing to use exclusively
compressed air, the extruder’s compact dimensions are
maintained, especially in comparison to alternative systems
such as piston-and-chamber.
Designing a Clay Body
To create a fully responsive design environment, material
must be considered alongside software and hardware.
To select a clay body and recipe for use with the updated
extrusion system, a range of clay recipes with multiple
viscosities was tested, including porcelain, a standard
potter’s stoneware with grog, and a higher-plasticity dark
brown high re stoneware. When used in the compressed
air extrusion system, a clay body must be thick enough
to retain its form and structural integrity once extruded,
but must also be thin enough to pass through the hopper
2 Diagram of the process of generating a toolpath.
3 Once a toolpath is generated, the Robosense plugin sends a move
command to the robot, the robot executes the command, an Arduino
reads a sensor, the toolpath is updated, and the feedback loop continues.
11
Robosense 2.0 Bilotti, Norman, Rosenwasser, Liu, Sabin
chamber and delivery tube without excessive pressure
(120 psi).
Different clay bodies exhibited a variety of unexpected
behaviors when extruded, which have the potential to
become elements of a design language informed by
feedback in the fabrication environment. Two key behaviors
were identied and tested further to examine the potential
for understanding these nuanced phenomena using
Robosense 2.0: 1) looping behavior and 2) bridging behavior.
Looping behavior is dened as what occurs when the clay
extrusion rate exceeds the typical rate for a given robot
travel speed. This overextrusion of clay produces a bead
that creates a looping pattern when the end effector nozzle
is raised (Figure 5). The bead falls approximately along the
intended path of material deposition rather than precisely
along it. Bridging behavior is dened as what occurs when
a clay bead is extruded across an unsupported span. The
bead’s plasticity allows it to remain intact when supported
on two ends. The height at which the bottom of the bridge
falls has a relationship to the length of the unsupported
span, the pressure of extrusion, and the material
composition of the clay body itself (Figure 5). A number of
tests were conducted to examine the relationships between
extrusion pressure and contextual factors (such as bridge
span distance or nozzle height), using a variety of clay
bodies (Figure 6).
Exercises
The software, tools, and materials developed in Robosense
2.0 form a new type of responsive sensing environment,
facilitating further development of techniques for the
design of ceramic building components. Testing physical
prototypes in this environment, as outlined in the
following sections, claries a design process/language
that translates sensor input into a unique materiality for
architectural ceramics.
Exercise 1:Simple Extrusion Pressure Test
The rst test is intended to demonstrate dynamic
manipulation of a toolpath based on user input from
sensor data. A bead of clay is extruded as air pressure
from the extruder is read by a Universal 150PSI Pressure
Transducer Solenoid and fed into a Grasshopper script
via an Arduino. Changes in air pressure (which controls
extrusion rate) are mapped to a resultant change in the Y
coordinate of the waypoints in a predetermined toolpath
(Figure 7 top). Line 01 is extruded at 100 psi, line 02 is
extruded with the pressure gradually increasing from 100
psi to 120 psi, and line 03 is extruded with the Y coordinate
being adjusted in the range of 0 mm to 60 mm in response
to the pressure gradually increasing from 100 psi to 120
psi. All lines are extruded 10 mm above the work surface.
4 (top) Diagram and (bottom) images of modications to the clay extrusion
system and end effector.
5 A) Low nozzle height produces a at, wide layer. B) Bead is deposited
accurately but is not compressed. C) Beginning to produce looping
behavior. D) Demonstrating signicant looping behavior. E) Bridge
behavior at a low extrusion pressure. G) Medium to low extrusion pres-
sure. G) High extrusion pressure H) Causes bridge failure.
Figure 7 (bottom) depicts a second line test in which the Z
coordinate also increases, along with the Y coordinate, as
the pressure increases. The three lines in Figure 7 (bottom)
follow the same rules as the lines in Figure 7 (top), with
the addition of the pressure being mapped to an increase
in the Z coordinate of the end effector in lines 02 and 03.
The lower bound of 100 psi is mapped to 10 mm and the
upper bound of 120 psi is mapped to 20 mm above the
work surface. Increasing the height of the end effector in
response to the pressure produces more dramatic looping
behavior.
Exercise 2: Extrusion Pressure Adjusts Toolpath
User-controlled changes in the toolpath results in a change
in material behavior: a simple, linear toolpath was extruded,
but changes in air pressure were chosen to correspond
to changes in the Z-value of each target point along the
toolpath. By dynamically changing the toolpath’s Z-value
instead of the Y-value, overextrusion (looping behavior) was
encouraged (Figure 8).
To create the complete panel in Figure 9 (middle), the
end effector’s height ranges from 0 to 10 mm above the
Z coordinate of the initial, unmodied toolpath. While
the looping behavior is more subtle in Figure 9 (middle)
because of the limited range of the z adjustment, this
conguration produces a wider, atter bead as more clay
is extruded. Figure 9 (right) is a component fabricated with
the end effector’s height ranging from 0 to 20 mm above the
unmodied Z coordinate of the initial toolpath.
Ex erc is e 3: E x tr u si on P res su r e In fo rms Rea l-Ti me
Design Changes
The nal test develops the idea of utilizing feedback loops as
a means of dynamically generating changes in the design
of architectural components. Adjustments in air pressure
(extrusion rate) now correspond to a change in the design
of the toolpath geometry. This allows the architect to
redesign the structure of the component during fabrication
in addition to making small local adjustments to the
toolpath. The Grasshopper script is modied so that the
values from Firey adjust an attribute of the patterning
to be translated into a toolpath and executed by the robot.
HAL generates new RAPID code from the updated toolpath.
6 Clay body experiments, which tested increases in the robot’s travel speed
and potential bridging distance of multiple clay bodies. These tests were
critical in differentiating behaviors of each clay body.
7 (Top) Two lines with change in Y coordinate in response to pressure
change. (Bottom) Two lines with change in Y and Z coordinate in response
to pressure change.
7
8 Dynamic modication of a toolpath during fabrication, depicting toolpaths
in elevation and corresponding visualizations of the resultant printed
beads. As extrusion air pressure is adjusted, the Z-coordinate value
of subsequent toolpath target points is multiplied by a factor within a
designer-determined range (in this case, 1.0 to 2.25).
6
Robosense 2.0 Bilotti, Norman, Rosenwasser, Liu, Sabin
Robosense sends the new toolpath one cartesian move at
a time until a new update to the patterning has occurred,
thus beginning a feedback loop (Figure 10).
In the case of this test, the parameter adjusted was the
distortion of each layer’s pattern. With an increase in
extrusion pressure, two changes were produced: 1) the
design of a panel’s toolpath was distorted in the X and
Y directions toward an attractor point assigned by the
designer, and 2) the clay bead exhibited an increase in
looping behavior as the Z coordinate of the toolpath was
increased in relation to the air pressure increase (Figure 11).
Agg reg at io n an d Gl oba l Pa tte rning L og ics I nf or med by
Real-Time Inputs
The design of a global facade pattern composed of
individual ceramic panels is possible by considering air
pressure changes and user input during the fabrication
process. Unique patterns are designed in which variations
in each component’s toolpath aggregate to form a global
organization. Ceramic screens and facades created
using this method have the possibility of great variation in
structural integrity, light/visual ltration, and aesthetic
novelty depending on the location of looping behavior and
pattern density.
A global pattern is generated using the following method
(Figure 12): 1) A set of relationships is designed using the
toolpath design script, which determines the sequence
of the pattern’s generation. 2) Attractor locations are
determined for each panel that can potentially distort each
toolpath in the X and Y directions, depending on user input
during fabrication. 3) Panels are fabricated in the sequence
determined in step 1. Throughout fabrication, changes
in extrusion air pressure are recorded in the Robosense
component as they dynamically modify the trajectory of
the robot’s motion. An increase in air pressure during the
fabrication of ceramic panel 01 affects subsequent panels
(numbered 02) by increasing the attractor’s distortion of
the toolpath (Figure 11).
By incorporating logics of user input and inuence not
only into the creation of individual toolpaths, but also into
the design of global patterns, the designer’s agency and
responsiveness in the process is further augmented.
RESULTS
The line tests in Exercise 1 show successful real-time
manipulation of the Y coordinates of a series of target
locations, traced by the robot’s end effector. Initially, the
Robosense 2.0 component sent cartesian moves in batches
of ten points, and a signicant delay was observed, which
resulted in a large buildup of clay between batches. This
9 (Left) Calibration of a single layer to identify the lower and upper bounds
of the air pressure and the end effector height. (Middle) A complete
component with a small Z-coordinate adjustment range. (Right) A
complete component with a large Z-coordinate adjustment range.
10 System diagram illustrating how real-time inputs allow for the creation of
bespoke global patterns during fabrication.
11 Matrix of local pattern possibilities. The numeric ranges indicated refer
to distortion strength of the attractor from the bottom layer to the top
layer of each toolpath, with a value of –1 representing the strongest
repelling inuence, and a value of 1 representing the strongest attracting
inuence.
12 (Top) Diagrams of pattern-generation sequence (1), attractor positioning
(2), and resultant pattern distortion in printed ceramic panels (3). (Bottom)
Diagram of attractor distortion strength as modied during the fabrica-
tion of ceramic panels.
delay inhibits an intuitive relationship between the architect
and the material behavior during fabrication. The delay was
removed prior to performing exercises 2 and 3 by sending
cartesian moves individually instead of in batches, allowing
for immediate adjustment of the toolpath in response to
an architect’s input. The extrusion system’s pressure
regulator, paired with an air pressure sensor, acted as a
controller of the robot and a design tool for the user.
The extrusion test described in 3.6.2 was the rst test to
take advantage of the material behavior control Robosense
2.0 provides. Looping behavior is produced at a high
pressure and Z-coordinate value, while a more typical,
controlled bead is produced at a low height and pressure
value. In all instances, an understanding of bridging
behavior allows for the possibility of atypically porous,
self-ventilating ceramic components (Figure 13). Drying
times and failure rates of these components were found
to be lower than is typical for ceramics of this size, with
greater than 90% success rate. By increasing the upper
bound of the Z-coordinate adjustment, the designer has a
larger range of extrusion behaviors to work with during
fabrication.
Reflection
Through testing a range of clay bodies, the high-plasticity
dark brown stoneware is identied as most suitable given
the intent to explore both bridging and looping behavior
(Figure 14). The design of a new end effector with entirely
modular construction allows for the attachment of sensors,
scanners, and documentation equipment. The clay hopper
tube can be easily removed from the new mounting system
and is held securely away from the work. By using only
compressed air without electronically controlled motors,
the system is a model for an easy-to-construct clay
deposition system that is accessible for a low cost (Table 1).
By using an interface and toolpath-generation script
that begins with the design of a toolpath itself (and not
necessarily with a predetermined 3D model), users
understand how their changes directly affect the code
produced, which controls the motion of the robotic arm;
the motion of the arm itself is designed, rather than an
object. This unique design process, paired with the ability
to make changes in real time (during fabrication), allows for
a uid and dynamic workow in which the architect has a
more direct relationship with the tools being used. These
feedback loops not only allow for more precise control
and understanding of fabrication processes, but they also
suggest the possibility for design to take advantage of more
idiosyncratic material behaviors otherwise understood as
mistakes or accidents.
Improvements
Robosense 2.0 can be made more user friendly by
encapsulating the feedback loops described in Figures
3 and 10. Firey and code-generation components can
be included within the Robosense 2.0 component so
the architect only has to design a toolpath and dene
relationships between sensors and robot movement. Users
could then save and reload these relationships to easily
recreate complex fabrication setups.
13 Porous ceramic tiles composed of bridged and looping clay beads.
Tab le 1 Porous ceramic tiles composed of bridged and looping clay beads.
14 (Left) Component geometry and local patterning tests in dark browstoneware.
(Right) Detail of unique material behavior in clay.
15 Sectional representation of multi-axis clay deposition within a tapered
plaster mold. As moisture is wicked from the clay, it adheres to the
tapered mold, supporting the ends of peripheral bridges.
Robosense 2.0 Bilotti, Norman, Rosenwasser, Liu, Sabin
Robosense can also facilitate an expanded breadth of
fabrication-based inputs, such as 3D scanning to reference
the existing material in real time. By scanning the physical
space, the robot could design future formwork to better
complement the existing imprecise formwork. Monitoring
the physical space and material more closely will be useful
for controlling new material behaviors introduced when
using more than 3 axes for clay deposition (Figure 15).
CONCLUSION
Robosense 2.0 integrates informed material feedback
into the design and production of ceramic architectural
componentry. The conceptual framework and software
can also be applied to other materials and architectural
assembly logics. This paper implements three experiments
and case studies in clay, which showcase the potential
for responsive sensing in an architectural fabrication
environment. Building upon research from Robosense
1.0, the project moves its software into Rhinoceros/
Grasshopper with help from the software Open ABB and
HAL, thus liberating architects to engage responsive
feedback loops at a more accessible level. By leveraging
extruded ceramic 3D printing research in Robosense 2.0
case studies, resultant components suggest a new age of
digital craft within our built environment.
By integrating the design process with the fabrication
process, Robosense 2.0 allows each piece of an
architectural component system (Figure 16) to better
leverage the nuances of a material’s behavior, as well
as the designer’s input, imbuing traces of making into
the creation of crafted components. The designer is able
to use real-time fabrication data to generatively inform
subsequent parts of a fabrication project (whether the
implications are functional, structural, or aesthetic),
demonstrating the potential for an unprecedented level of
involvement that architects and designers may have in the
realization of their designs.
REFERENCES
Brugnaro, Giulio, and Sean Hanna. 2017. “Adaptive Robotic
Training Methods for Subtractive Manufacturing.” In Disciplines
& Disruption: Proceedings of the 37th Annual Conference of the
Association for Computer Aided Design in Architecture, edited by T.
Nagakura, S. Tibbits, M. Ibanez, and C. Mueller, 164–69. Cambridge,
MA: ACADIA.
Cederberg, Per, Magnus Olsson, and Gunnar Bolmsjo. 2002.
“Remote control of a standard ABB Robot System In Real Time Using
The Robot Application Protocol (RAP).” In Proceedings of the 33rd
International Symposium on Robotics. Seoul, Korea: ISR.
Chronis, Angelos, Alexandre Dubor, Edouard Cabay, and Mostapha
Sadeghipour Roudsari. 2017. “Integration of CFD in Computational
Design: An Evaluation of the Current State of the Art.” In ShoCK!:
Sharing Computational Knowledge!—Proceedings of the 35th
eCAADe Conference, vol. 1, edited by A. Fioravanti, S. Cursi, S.
Elahmar, S. Gargaro, G. Loffreda, G. Novembri, and A. Trento,
601–10. Rome: eCAADe.
Elashry, Khaled, and Ruairi Glynn. 2014. “An Approach to
Automated Construction Using Adaptive Programing.” In Robotic
Fabrication in Architecture, Art and Design 2014, edited by Wes
McGee and Monica Ponce de Leon, 51–66. Cham, Switzerland:
Springer.
Feng, Chen, Xiao, Yong, Willette, Aaron, McGee, Wes and Vineet
R. Kamat. 2014. “Towards Autonomous Robotic In-situ Assembly
on Unstructured Construction Sites using Monocular Vision.” In
Proceedings of the 31st International Symposium on Automation
and Robotics in Construction and Mining, 163–70. Sydney,
Australia: ISARC.
Friedman, Jared, Heamin Kim, and Olga Mesa. 2014. “Experiments
in Additive Clay Depositions-Woven Clay.” In Robotic Fabrication
in Architecture, Art and Design 2014, edited by Wes McGee and
Monica Ponce de Leon, 261–72. Cham, Switzerland: Springer.
16 Detail of facade mock-up.
Giftthaler, Markus, Timothy Sandy, Kathrin Dörer, Ian Brooks,
Mark Buckingham, Gonzalo Rey, Matthias Kohler, Fabio Gramazio,
and Jonas Buchli. 2017. “Mobile Robotic Fabrication at 1:1 scale:
The In situ Fabricator.” arXiv:1701.03573.
Johns, Ryan Luke, Kilian, Axel and Nicholas Foley. 2014. “Design
Approaches Through Augmented Materiality and Embodied
Computation.” In Robotic Fabrication in Architecture, Art and
Design 2014, edited by Wes McGee and Monica Ponce de Leon,
319–32. Cham, Switzerland: Springer.
Lind, Morten, Lars Tingelstad, and Johannes Schrimpf. 2012.
“Real Time Robot Trajectory Generation With Python” In IEEE/
RSJ International Conference on Intelligent Robots and Systems,
Workshop on Robot Motion Planning: Online, Reactive, and in Real-
time. Lausanne, Switzerland: IROS.
Meagher, Mark, David Van der Maas, Christian Abegg, and Jeffrey
Huang. 2013. “Dynamic ornament: An Investigation of Responsive
Thermochromic Surfaces in Architecture.” International Journal of
Architectural Computing 11 (3): 301–18.
Menges, Achim. 2015. “The New Cyber Physical Making in
Architecture: Computational Construction.” Architectural Design 85
(5): 28–33.
Moorman, Andrew, Jingyang Liu, and Jenny Sabin. 2016.
“RoboSense: Context-Dependent Robotic Design Protocols and
Tools.” In Posthuman Frontiers: Data, Designers, and Cognitive
Machines, Proceedings of the 36th Annual Conference of the
Association for Computer Aided Design in Architecture, edited by
Kathy Velikov, Sean Ahlquist, Matias del Campo, and Geoffrey Thün,
174–83. Ann Arbor: ACADIA.
Nicholas, Paul, Mateusz Zwierzycki, Esben Clausen Nørgaard,
David Stasiuk, Christopher Hutchinson, and Mette Thomsen.
2017. “Adaptive Robotic Fabrication on for Conditions of Material
Inconsistency: Increasing the Geometric Accuracy of Incrementally
Formed Metal Panels.” In Fabricate 2017: Rethinking Design and
Construction, edited by Achim Menges, Bob Sheil, Ruairi Glynn, and
Marilena Skavara, 114–21. London: UCL Press.
Payne, Andrew O. and Jason Kelly Johnson. 2013. “Firey:
Interactive Prototypes for Architectural Design.” Architectural
Design 83 (2): 144–147.
Rael, Ronald, and Virginia San Fratello. 2018. Printing Architecture:
Materials and Methods for 3D Printing. Hudson, NY: Princeton
Architectural Press.
Rosenwasser, David, Sonya Mantell, and Jenny Sabin. 2017.
“Clay Non-Wovens: Robotic Fabrication and Digital Ceramics.”
In Disciplines & Disruption: Proceedings of the 37th Annual
Conference of the Association for Computer Aided Design in
Architecture, edited by T. Nagakura, S. Tibbits, M. Ibanez, and C.
Mueller, 502–11. Cambridge, MA: ACADIA.
Sabin, Jenny, Martin Miller, Nicholas Cassab, and Andrew Lucia.
2014. “PolyBrick: Variegated Additive Ceramic Component
Manufacturing (ACCM).” 3D Printing and Additive Manufacturing 1
(2).
Schwartz, Thibault. 2013. “HAL: Extension of a Visual Programming
Language to Support Teaching and Research on Robotics Applied
to Construction.” In Rob|Arch 2012: Robotic Fabrication in
Architecture, Art and Design, edited by Sigrid Brell-Cokcan and
Johannes Braumann, 92–101. Vienna: Springer-Verlag.
Vasey, Lauren, Iain Maxwell, and Dave Pigram. 2014. “Adaptive Part
Variation A Near Real-Time Approach to Construction Tolerances.”
In Robotic Fabrication in Architecture, Art and Design 2014,
edited by Wes McGee and Monica Ponce de Leon, 67–81. Cham,
Switzerland: Springer.
Jeremy Bilotti is a research associate in the Sabin Design Lab and
a faculty teaching associate in Cornell AAP's department of archi-
tecture. Bilotti is a recent graduate of Cornell University’s B.Arch
program. He is also a Hunter R. Rawlings III Presidential Research
Scholar.
Bennett Norman is a graduate student on Cornell Tech’s new campus
in the Master of Computer Science program. Norman is a recent
graduate of Cornell University’s Computer Science program. He was
formerly a research associate in the Sabin Design Lab.
David Rosenwasser is Senior Personnel in the Sabin Design Lab
at Cornell AAP and a recent graduate of Cornell University’s B.Arch
program. He is also a Hunter R. Rawlings III Presidential Research
Scholar.
Jingyang Liu is a PhD candidate in Carnegie Mellon’s Department of
Architecture. He was formerly Senior Personnel in the Sabin Design
Lab. Liu graduated in 2018 as the rst recipient of Cornell’s Master of
Matter, Design, Computation degree. Liu holds an M.Arch from Cornell.
Jenny Sabin’s work is at the forefront of a new direction for 21st
century architectural practice—one that investigates the inter-
sections of architecture and science. Sabin is the Wiesenberger
Professor in the area of Design and Emerging Technologies and the
Director of Graduate Studies in the Department of Architecture at
Cornell University. She is principal of Jenny Sabin Studio and director
of the Sabin Design Lab at Cornell AAP.
