Supplementary Materials for

Toward site-specific and self-sufficient robotic fabrication on

architectural scales

Steven J. Keating, Julian C. Leland, Levi Cai, Neri Oxman*

*Corresponding author. Email: neri@mit.edu

Published 26 April 2017, Sci. Robot. 2, eaam8986 (2017)

DOI: 10.1126/scirobotics.aam8986

The PDF file includes:

Text S1. ISO 9283-1998 pose repeatability characterization.

Text S2. ISO 4287 surface roughness characterization.

Text S3. Preliminary financial analysis of PiP construction.

Text S4. Notes on estimations in construction robots comparison

table.

Table S1. ISO 4287 surface topography summary data.

Fig. S1. Laser compensation enabling maintenance of constant aboveground

height.

Fig. S2. KUKA endpoint position during programmed oscillatory movement with

laser sensor compensation vs. without laser compensation.

Fig. S3. Electrohydraulic drivetrain.

Fig. S4. Solid PV panels.

Fig. S5. PiP mold filled with conventional concrete.

Fig. S6. Foam deposition head for PiP fabrication.

Fig. S7. Foam deposition head detail.

Fig. S8. Empirical determination of optimal foam extrusion parameters.

Fig. S9. Doubly curved geometries fabricated with PiP process.

Fig. S10. Automated rebar tie inserter prototype.

Fig. S11. Adhesion failure caused by excessive moisture on the printed surface

from the dew condensation that occurred in the morning during printing.

Fig. S12. KUKA robot arm with combined milling/PiP fabrication end effector.

Fig. S13. Milled spray polyurethane foam mold used to produce concrete casting

shown in Fig. 4F.

robotics.sciencemag.org/cgi/content/full/2/5/eaam8986/DC1

Fig. S14. Range of color variation in spray polyurethane foam insulation.

Fig. S15. Completed PiP dome.

Fig. S16. The PiP dome was able to support human weight, even without any

structural material (such as concrete) cast inside the formwork.

Fig. S17. Fabrication of horizontally printed overhang feature.

Fig. S18. Fabrication using direct welding deposition technique.

Fig. S19. Concept computer renderings of future applications for DCP systems.

Fig. S20. Excavation tests with the DCP demonstrating both subtractive site

preparation and local material gathering.

Fig. S21. A rollable photovoltaic panel was used for testing with the DCP to

explore deployable large solar arrays for power self-sufficiency.

Fig. S22. Radiation scanning using a Geiger counter mounted to the DCP to

gather volumetric data around three radioactive sources.

Fig. S23. Preliminary fabrication explorations with electrosintered powdered

glass, thermally deposited ice structures, and compressed earth forms.

Fig. S24. Balluff magnetostrictive sensors.

Fig. S25. YUMO rotary encoder.

Fig. S26. Sensor board that mounts at the end of the KUKA robotic arm.

Fig. S27. DCP sensor architecture diagram.

Fig. S28. HomePrint software interface.

Fig. S29. DCP control architecture diagram.

Fig. S30. Leica retroreflector mounted to the KUKA robotic arm on the DCP.

Fig. S31. A selection of various DCP end effectors.

Fig. S32. Thermoplastic fabrication end-effector and 3D printed structure made

from acrylonitrile butadiene styrene (ABS).

Legend for movie S1

References (43–66)

Other Supplementary Material for this manuscript includes the following:

(available at robotics.sciencemag.org/cgi/content/full/2/5/eaam8986/DC1)

Movie S1 (.mov format). Short video overview of the DCP.

Supplementary Text: Text S1: ISO 9283-1998 Pose Repeatability Characterization

Analysis of ISO 9283 data is separated into two distinct segments: run segmentation and data analysis. Run Segmentation: Run segmentation is performed with a MATLAB script. This script operates by identifying locations within the dataset where both 1) the X/Y/Z position is within some tolerance of the X/Y/Z position at the start of the run, and 2) the derivative of X/Y/Z position (the velocity) is within some tolerance of zero. This analysis creates a list of “newmove” flags – locations where the script believes that a new move has begun. These flags are then filtered a second time to remove segments that are too short (do not correspond to entire move segments, but are rather created by noise in the data), and the data is finally exported as a MATLAB structure with an entry for each segment, containing measurement index, XYZ position, and time data. Data Analysis: Data analysis is performed with the script iso9283analyzer.m. This script operates by performing the following steps:

The X/Y/Z position vectors for the first segment of data are loaded, and the derivative of the position for this segment is calculated and filtered to reduce noise.

Within this first segment of data, locations where the derivative of position is within some tolerance of zero are identified. These are used to generate a set of n positions that are defined as “waypoints,” which future runs are compared against. In the case of the ISO 9283 pose accuracy/repeatability standard, there are five waypoints for each segment, with the start and stop position sharing a waypoint.

Using these waypoints, iterate through all other segments and identify locations where the position is within some tolerance of these waypoints, and the derivative of position is within some tolerance of zero. These are the waypoint measurements for each subsequent segment. They are averages of all points measured while the system is considered to be “at” a waypoint (e.g. meeting the two criteria listed above).

From all waypoint measurements, generate an average position across all trials for each waypoint.

Calculate error between waypoint position for each segment and this average waypoint position (lj in ISO 9283)

Calculate the corrected sample standard deviation for the dataset, as per ISO 9283

Calculate final system repeatability measure.

Run data is then plotted as 3D paths as well as individual X/Y/Z trajectories. The final system repeatability for each pose is displayed on the plot as a sphere centered on the mean position of that waypoint, with radius equal to the system repeatability of that pose.

Repeatability: ISO 9283 defines system repeatability as mean position error + 3 standard deviations, giving a confidence of 99.7%. ISO 9283 defines system repeatability on a pose-by-pose basis, meaning that to report repeatability in accordance with the standard, the repeatability at each point should be reported.

For this test, conducted with the DCP, we report a worst-case ISO 9238 pose repeatability of <55 mm. The following image shows the path traversed by the DCP’s endpoint over the duration of the test, with system repeatability “spheres” for each pose superimposed over the waypoints, and mean error, standard deviation, and position repeatability for each pose reported. Because of limitations of the metrology equipment used in this test, we do not measure or report pose orientation repeatability as specified in the standard.

Text S2: ISO 4287 Surface Roughness Characterization To quantify the typical surface textural properties of Print-in-Place fabricated formworks, a 250 mm x 300 mm sample of material printed in-lab using a conventional robotic arm was analyzed in accordance with ISO 4287 (38).

Print-in-Place fabricated formwork sample

Five surface profile samples were taken using a Microscribe 3DLX digitizer, with a resolution of 0.48 mm between recording points. While this instrument is dramatically lower-resolution than typical surface topography measurement equipment, it is sufficient for this application because of the scale of surface features of interest on the PiP printed sample. Profile data was analyzed using the MountainsMap surface metrology software

package (66). Profiles were de-noised (50 m cutoff) and then segregated with an 8 mm

c filter to separate roughness and waviness components. Characteristics from all five profiles were averaged, yielding the data presented in Table S1 below.

Table S1. ISO 4287 surface topography summary data.

Units Average Profile1 Profile2 Profile3 Profile4 Profile5

Ra Arithmeticmeandeviationoftheroughnessprofile mm 0.20 0.20 0.19 0.22 0.19 0.18

Rz Maximumheightofroughnessprofile mm 0.80 0.84 0.74 0.96 0.76 0.72

Rt Totalheightofroughnessprofile mm 2.39 2.40 3.40 2.50 1.74 1.93

RSm Meanwidthofroughnessprofileelements mm 6.64 6.42 7.38 6.49 6.53 6.37

Wa Arithmeticmeandeviationofthewavinessprofile mm 2.45 2.60 2.31 2.96 2.45 1.94

Wt Totalheightofwavinessprofile mm 11.81 12.59 10.63 15.29 10.63 9.89

WSm Meanwidthofwavinessprofileelements mm 39.70 40.45 41.48 39.61 38.74 38.22

Lam_c Lambda-Cprofilefilter mm 8 - - - - -

Ln Evaluationlength mm 241.9 241.2 241.9 241.5 242.4 242.4

Roughness

Waviness

Other

Description

Text S3: Preliminary Financial Analysis of Print-In-Place Construction Preliminary financial analysis was conducted to compare printed insulated formwork with traditional construction methods. The results show the cost of building the formwork of a house with the Print-in-Place method would be 8% less expensive than traditional wood construction methods and 31% less expensive than regular ICF construction. Most of the cost savings are from labor costs. These calculations are based around the life cycle of an average-sized one story house (perimeter = 180 ft) as seen below. Estimates from construction workers were used to calculate costs for wooden construction and ICF construction and include human labor for operating the Print-in-Place deposition system. For the purpose of the estimate, finishing costs and interior walls were not considered for all techniques. The financial estimates strongly support the fiscal feasibility for the printed formwork technique, showing that not only will it save time and increase safety, but also lower costs for even simple rectilinear buildings.

The differences in transportation volume and thermal insulation value are listed and not applied to the financial estimates. Note that the building is also a simple rectilinear building and not a complex doubly curved structure, which would have the same cost using Print-in-Place fabrication.

Text S4: Notes on Estimations in Construction Robots Comparison Table Quantitative data listed in the Construction Robotics Comparison Table are frequently challenging or impossible to find directly in the literature. In many cases, the authors have had to estimate certain parameters by extrapolating from published data, videos and other sources. This document explains estimations made in the Construction Robotics Comparison Table, in hopes of clarifying the strengths and weaknesses of this analysis, and encouraging others in the field to contribute to this effort. All estimates here represent the authors’ best efforts to capture the performance of the automated construction systems reviewed, based on publically available information in scholarly and popular literature uncovered in the authors’ literature review. Importantly, these numbers should not be viewed as absolute benchmarks for a given system’s performance. For example, many automated construction systems can fabricate objects that are larger than the value listed for Largest Fabricated Structure; or may be able to trivially expand their work volume. Category Descriptions:

- Primary/Secondary Fabrication Media: This describes the fabrication media that have been demonstrated being used with the automated construction system in publically available literature. In the event that more than one fabrication media has been used, analysis has focused on the media that is most appropriate for architectural-scale fabrication, either in terms of volumetric fabrication rate or available scale.

- Largest Fabricated Structure: This describes the approximate envelope volume of the largest structure that has been fabricated using the automated construction system in cubic meters, as demonstrated in publically available literature. This analysis focuses on envelope volume, or total enclosed volume, rather than extruded/fabricated volume because it is felt to be a more relevant benchmark for construction tasks. It is particularly important for this category to recognize that many systems are capable of fabricating larger objects than listed here: this is intended to provide a lower bounds on a system’s fabrication capability.

- Total Work Volume: This describes the system’s work volume, in cubic meters. For some systems – particularly systems with some degree of mobility – it is challenging to define a work volume. For instance, a drone-based system like the Flight Assembled Architecture system could arguably demonstrate an infinite work volume, given an appropriately defined control system. In situations like these, the authors have attempted to define a reasonable limit on the system’s performance as built.

- Typical Volumetric Fabrication Rate: This describes the typical maximum rate that a system can add volume to a fabricated object using its primary fabrication system, in cubic meters per hour. The authors’ intent in including this metric is to provide an approximation of how fast a system can fabricate a structure, which is insensitive to differences in fabrication technique or the structure being fabricated. Importantly, this analysis only considers the volumetric fabrication rate for the primary construction media, and not any secondary or finishing techniques required to complete a structure. Different fabrication techniques may yield structures with very different degrees of structural performance or “finished”-ness, and this is an important caveat to keep in mind when considering these data points. However, as all of the fabrication techniques considered here would require some degree of secondary finishing to produce a viable structure,

we feel that this is a useful starting point for comparing how fast various technique

DCP

- Primary/Secondary Fabrication Media: The Digital Construction Platform has successfully performed fabrication operations with a variety of techniques, but the two most thoroughly developed processes are print-in-place fabrication (which uses two-part spray polyurethane foam insulation as its primary fabrication media) and welded chain fabrication.

- Largest Fabricated Structure: The largest structure fabricated with the DCP system is a 7.32 m radius, 3.66 m height, hemi-ellipsoidal dome section described in (37). The enclosed volume of this structure may be approximated as

a spherical segment, with 𝑉 =1

3𝜋ℎ(3𝑟2 − ℎ2).

- Total Work Volume: The work volume of the DCP can be approximated by a spherical cap, with the center of the sphere 1.866 m above the ground plane. The radial reach of the DCP at this height is 10.149 m. The volume of this

spherical cap may be calculated as 1

3𝜋ℎ2(3𝑅 − ℎ), where h is the height from the

ground to the top of the cap, equal to 10.149 + 1.866 = 2,786 m3. It is important to note that variations in the pose of the DCP – such as changes in the J2 joint position, or the use of different end effectors – can impact this value, although this is considered generally representative of the DCP’s work volume.

- Typical Volumetric Fabrication Rate: The typical volumetric fabrication rate is calculated using parameters used during the fabrication of the hemi-ellipsoidal dome section. The DCP extruded PU foam at a feed rate of 150 mm/s, and produced a 40 mm x 80 mm trace.

In-Situ Fabricator

- Primary/Secondary Fabrication Media: The In-Situ Fabricator has been demonstrated using both brick assembly techniques and the Mesh Mould free-space wire assembly technique in (24).

- Largest Fabricated Structure: The volume listed here is based on the brickwork wall described in (24). The wall is described as measuring 6.5 m long and 2 m high. We have estimated its width as 0.75 m, which we believe to be a high estimate.

- Total Work Volume: The In-Situ Fabricator is based around an ABB IRB 4600 robot arm, which has a 2.55 m maximum radial reach and approximately spherical kinematics. The total work volume of the In-Situ Fabricator is estimated

as a hemisphere, win the total work volume calculated as 𝑉 =2

3𝜋𝑟3. Importantly,

we also disregard the In-Situ Fabricator’s mobile base and navigation/localization capability – which is substantial - in calculating the system’s work volume. We feel that while the In-Situ Fabricator is certainly capable of building structures that are larger laterally than what is described here, the maximum height of a given structure is still fundamentally limited by the reach of the robot arm. Consequently, while work volumes cited for the In-Situ Fabricator are lower-bound estimates, we believe they are still useful for understanding what scale of structure the system can produce.

- Typical Volumetric Fabrication Rate: This rate is based on the fabrication of the brickwork wall. We assume a standard German brick’s dimensions (240 mm x 115 mm x 71 mm), placed every 40 seconds as described in (43).

Apis Cor - Primary/Secondary Fabrication Media: The Apis Cor automated construction

system uses a direct concrete deposition technique to fabricate structures. To our knowledge, it has not been used to demonstrate other fabrication techniques.

- Largest Fabricated Structure: The largest structure found that the Apis Cor system has built is the circular house described in a February 2017 press release (44). It is cited as having a floor area of 132 m^2 (although this exceeds the company’s citation for the system’s maximum lateral reach). Using videos and images of the house for reference, we estimate its height to be 3 meters tall, standard for a single-story structure in the EU.

- Total Work Volume: The total work volume of the Apis Cor system is a cylinder measuring 3.3 m high with a base measuring 131.4 m^2 (21).

- Typical Volumetric Fabrication Rate: The Apis Cor system’s standard extruded filament measures 25 mm x 25 mm, and the system prints at a rate of 10 m/min (21).

Hadrian 105

- Primary/Secondary Fabrication Media: The Hadrian 105 automated construction system is purpose-built to replicate traditional brick masonry construction techniques. To our knowledge, it has not been used to demonstrate other fabrication techniques.

- Largest Fabricated Structure: The largest structure found that the Hadrian 105 system has built is the brick structure demonstrated in a time-lapse video produced by Fastbrick Robotics (45). Fastbrick cites this structure as measuring 1.6 x 2.4 x 2.4 m (46)

- Total Work Volume: The Hadrian system’s work volume is estimated based on a 28 meter lateral reach, as cited by the system’s inventor in (20).

- Typical Volumetric Fabrication Rate: We estimate volumetric fabrication rate for the Hadrian system as the volume of a placed brick (standard Australian brick, measuring 230 mm x 110 mm x 76 mm) multiplied by the feed rate in bricks/hour. There are a wide range of stated feed rates for the Hadrian 105 system: most articles claim 250 bricks/hour, but the inventors claim “…~200 bricks/hour” in (46). We assume a rate of 225 bricks/hour.

Concrete Printing

- Primary/Secondary Fabrication Media: The Concrete Printing system is focused on the use of its namesake fabrication process. No evidence of the Concrete Printing system being used with substantially different material/fabrication systems has been found.

- Largest Fabricated Structure: The bench described in (10) is the largest artifact that we have been able to identify which the Concrete Printing system has successfully fabricated. The authors state that its footprint measures 2 m x 0.9 m x 0.8 m. We estimate that the actual printed volume of the bench occupies roughly 60% of this volume to derive our estimate. A structure with a larger projected area is reported in (47), but its enclosed volume is smaller.

- Total Work Volume: The gantry used for Concrete Printing described in (10) has a work volume measuring 5.4 m x 4.4 m x 5.4 m.

- Typical Volumetric Fabrication Rate: In (47), the authors report a maximum filament size of 25 mm x 15 mm, and report a maximum used feedrate of 4 m/min.

Contour Crafting - Primary/Secondary Fabrication Media: Like Concrete Printing, the Contour

Crafting system is primarily focused on use of the Contour Crafting process, and not on application of new material/fabrication techniques to the system.

- Largest Fabricated Structure: The largest structure we have been able to identify that has been fabricated with Contour Crafting, which been documented in the literature, is a wall section described in (48). This wall section measures 1.52 m x 0.61 m x 0.15 m. There is evidence that the Contour Crafting project has produced substantially larger structures with a different gantry (49, 50), but no description of this gantry, or of components built with it, have been found in the literature. It is also important to note that there is some contention whether the additive concrete fabrication process used by the WinSun Corporation (51) is functionally identical to the Contour Crafting process. The WinSun Corporation has fabricated structures that are substantially larger than the specimen described above. However, it is extremely difficult to find reliable information about the WinSun printer’s performance, and it has consequently been excluded from this analysis, as described previously.

- Total Work Volume: The work volume of the system used in (48) appears to be the same as the volume of the structure fabricated. However, images of a larger gantry system visible in (49, 50) suggest that the Contour Crafting process has been implemented using substantially larger devices. We estimate the height of the person shown standing in front of the gantry in (50) to be roughly 1.65 m. Using this as a reference, and assuming that the length of the gantry is the same as the width of the gantry, we estimate the gantry to measure approximately 4.5 m long x 4.5 m wide x 2.75 m tall.

- Typical Volumetric Fabrication Rate: The filament size used in (48) is cited as 19 mm x 13 mm, and the feedrate is set at 20 mm/s.

3DCP

- Primary/Secondary Fabrication Media: The 3DCP development team states that they “…adopt the Contour Crafting approach…” and use a specialized, pumped concrete for their printing process (12).

- Largest Fabricated Structure: We have not found specific measurements of the largest structures fabricated with the 3DCP system, but images from the 3DCP team (52) and from a 2016 demonstration of a pavilion printed using the system (53) provide some reference for the scale of objects that have been fabricated with 3DCP. Based on these images, we estimate the largest to measure 2 m x 1.5 m x 0.5 m.

- Total Work Volume: In (12), the work volume of the 3DCP system is specified to be 9 m x 4.5 m x 2.8 m.

- Typical Volumetric Fabrication Rate: In (12), the authors claim that the optimal nozzle geometry has been found to be a 40 mm x 10 mm nozzle, operated at a feed rate of 100 mm/s.

BAAM

- Primary/Secondary Fabrication Media: The Big Area Additive Manufacturing (BAAM) system has been purpose-built for fused-filament fabrication using a variety of thermoplastics at very large filament sizes. While there are a range of different thermoplastic mixtures that have been experimented with using BAAM, they are all fundamentally part of the same material group.

- Largest Fabricated Structure: Of the range of structures fabricated using BAAM, the largest – and most relevant to this analysis – appear to be the panels used in the AMIE House. From references describing the house (54, 55), we have determined that the AMIE House is constructed from 36 separate panels, each measuring approximately 3.96 m x 1.83 m x 0.61 m.

- Total Work Volume: Cincinnati specifies the BAAM’s work volume as 6.1 m x 2.9 m x 1.83 m (56).

- Typical Volumetric Fabrication Rate: In (57), the authors describe an extrusion rate for a prototype BAAM system of 40 lb/hr, which they claim is equivalent (for their particular material density) to 1000 in^3/hr. However, from (56), it appears that the extrusion rate has been improved to 80 lb/hr.

Minibuilders

- Primary/Secondary Fabrication Media: The Minibuilders system was designed around a specific extruded material, comprised of “…40% Axon Easymax two component polymer, together with 60% marble powder.” (58)

- Largest Fabricated Structure: The largest structure fabricated using the Minibuilders system measured 1 m x 1 m x 1.5 m. (59)

- Total Work Volume: Because the Minibuilders system is nominally mobile, it is extremely challenging to specify a work volume for the system, as it can in theory move infinitely in the lateral plane. However, as with the In-Situ Fabricator, above, its fabrication capacity is still fundamentally limited by the height that the system can reach. We consequently limit the rectilinear and total work volume for the system to the largest object that it has been used to fabricate. Like with the In-Situ Fabricator, this is clearly a lower bound on the system’s performance: it is likely to be able to fabricate substantially larger structures trivially.

- Typical Volumetric Fabrication Rate: The typical volumetric fabrication rate achieved by the project was 15 in^3/min. (59)

Flight Assembled Architecture

- Primary/Secondary Fabrication Media: The Flight Assembled Architecture system was developed around the concept of brick assembly using UAVs. Other work has been done by the same team using UAVs to construct tensile rope structures, but this is not part of the Flight Assembled Architecture project, in our understanding.

- Largest Fabricated Structure: The largest structure fabricated by the Flight Assembled Architecture system was the cylindrical tower described in (13). The tower is described as being 6 m tall, and composed of 1,500 foam “bricks”. Based on extrapolation from this and figures presented by the authors, we estimate the diameter of the tower to be roughly 3.5 m.

- Total Work Volume: The Flight Assembled Architecture system is also nominally a mobile system, like the In-Situ Fabricator and the Minibuilders system. However, unlike these systems, it relies on a static optical tracking system to function. The system used for the Flight Assembled Architecture system is the Flying Machine Arena at ETH Zurich, which measures 10 m x 10 m x 10 m (60). This provides a useful maximum bound on the total work volume of the system. However, it is important to consider how other factors – such as occlusion of UAVs inside or behind the structure – may become more important as the structure’s volume increases.

- Typical Volumetric Fabrication Rate: In (13), the authors claim that it took 18 hours to assemble 1,500 bricks, each measuring 300 mm x 150 mm x 100 mm.

Guedel Gantry Robot

- Primary/Secondary Fabrication Media: The Guedel Gantry Robot system has been designed specifically for the automated fabrication of large timber structures. It is capable of a wide variety of sub-tasks involved in timber fabrication, including sawing, positioning and fastening, plus subtractive machining of timber and steel (61). However, to our knowledge no substantially different material/fabrication systems – such as fused-filament fabrication systems – have been demonstrated with the Guedel Gantry Robot system.

- Largest Fabricated Structure: The Guedel Gantry Robot was used to fabricate timber trusses as part of the Sequential Roof project at ETH Zurich. While the final, finished roof is enormous, the largest individual truss section is described as measuring 15 m long x 1.15 m wide (62). From images of the finished trusses, we estimate that the largest truss measures 3 m from its lowest to highest point.

- Total Work Volume: The rectilinear work volume of the Guedel Gantry Robot system is cited in (63) as 48 m x 6.1 m x 1.9 m.

- Typical Volumetric Fabrication Rate: To determine the volumetric fabrication rate of the Guedel Gantry Robot system, we attempt to estimate the amount of time required to install a typical volume of wood. From videos of the system in operation, we estimate the time to select, cut and install a single plank is ~1 min (64). From (65), we estimate that the average volume of a plank used in the Sequential Roof is 50 mm x 145 mm x 1.65 m.

Supplementary Figures:

Fig S1. Laser compensation enabling maintenance of constant aboveground height.

Fig S2. KUKA endpoint position during programmed oscillatory movement with laser

sensor compensation vs. without laser compensation.

Fig. S3. Electrohydraulic drivetrain.

3a: Electrohydraulic pump.

3b: Battery pack, motor controller, and battery management controller.

Fig. S4. Solid PV panels.

Fig. S5. PiP mold filled with conventional concrete.

Fig. S6. Foam deposition head for PiP fabrication.

Fig. S7. Foam deposition head detail.

Fig. S8. Empirical determination of optimal foam extrusion parameters.

Fig. S9. Doubly curved geometries fabricated with PiP process.

Fig. S10. Automated rebar tie inserter prototype.

Fig. S11. Adhesion failure caused by excessive moisture on printed surface from the

dew condensation that occurred in the morning during printing.

Fig. S12. KUKA robot arm with combined milling/PiP fabrication end effector.

Fig. S13. Milled spray polyurethane foam mold used to produce concrete casting shown

in Fig. 4F.

Fig. S14. Range of color variation in spray polyurethane foam insulation.

Fig. S15. Completed PiP dome.

Fig. S16. The PiP dome was able to support human weight, even without any structural

material (such as concrete) cast inside the formwork.

Fig. S17. Fabrication of horizontally printed overhang feature.

Fig. S18. Fabrication using direct welding deposition technique.

Fig. S19. Concept computer renderings of future applications for DCP systems.

19a: Automated ice structure fabrication in polar environment with power sourced through rollable photovoltaic panels and materials gathered locally.

19b: Fabrication with local sand to create fractal structures for future immersion in the ocean to support coral reef regrowth. Power sourced via deployable rollable

photovoltaics.

Fig. S20. Excavation tests with the DCP demonstrating both subtractive site preparation and local material gathering.

20a: View 1.

20b: View 2.

Fig. S21. A rollable photovoltaic panel was used for testing with the DCP to explore

deployable large solar arrays for power self-sufficiency.

Fig. S22. Radiation scanning using a Geiger counter mounted to the DCP to gather volumetric data around three radioactive sources. On-site abilities to gather and use

real-time environmental data in construction allows new functionalities, such as future applications in repairing radioactive disaster sites that are hazardous for humans.

22a: Radioactive sample (natural uranium ore) on sawhorse, and DCP set up to scan

over sawhorses.

22b: Long-exposure photograph of scan over radioactive sample. Color indicates count intensity detected by Geiger counter, with redder areas corresponding to higher counts.

Fig. S23. Preliminary fabrication explorations with electrosintered powdered glass (top

row), thermally deposited ice structures (middle row), and compressed earth forms made of locally available gravel and hay fibers (bottom row).

Fig. S24. Balluff magnetostrictive sensors.

24a: Balluff magnetostrictivesensor mounted to J4 prismatic joint, receiver end.

24b: Balluff sensor mounted to J4 prismatic joint, distal attachment.

Fig. S25. YUMO rotary encoder.

Fig. S26. Sensor board that mounts at the end of the KUKA robotic arm.

Fig. S27: DCP sensor architecture diagram.

Fig. S28. HomePrint software interface.

Fig. S29. DCP control architecture diagram.

Fig. S30. Leica retroreflector mounted to the KUKA robotic arm on the DCP.

Fig. S31. A selection of various DCP end effectors including a rotary milling tool, a

thermoplastic extruder, spray extruder, inkjet deposition head, and holders for sensors, controlled lights and welding end effector attachments.

Fig. S32. Thermoplastic fabrication end-effector and 3D printed structure made from

acrylonitrile butadiene styrene (ABS).

Supplementary Movies: Movie S1. Short video overview of the DCP.