Investigations in Robotic Craft

By Andrew Wit

This paper examines the creation of a new typology of architectural craft derived from the residual elements found through the misuse of robot manufacturing tools. The discussion of the T_Series project, and the T_Three table specifically, created by WITO* will demonstrate that through the utilization and exaggeration of mechanical imperfections, a new typology of architectural design aesthetic can be created.

As common knowledge of architectural robotics is minimal, this paper will begin with a brief introduction into architectural robotics as well as an understanding of how these projects are currently situated within the discourse of architecture.


As architectural tools and techniques progress towards more automated means of design and fabrication, architects and designers continue struggling to grasp the true design, fabrication, and workflow potentials possible through the integration of architectural robotics. Tools such as multi-axis CNC machining, 3D/4D multi-material printing, and industrial and humanoid robotics, once limited to large industries and research institutions, are now becoming more relevant within the realms of design thinking and the everyday practice of architecture. 

With their innate ability to easily adapt to various material and fabrication methodologies, flexibility in tool changing and design, as well as their ability to resolve complex formal geometries within shorter durations of time, these systems are being further integrated into everyday design and fabrication workflow. 

Unlike industrial environments, where efficiency and mechanical repeatability are key and similar operations may be repeated over the entire life of a tool, architecture tends to work in an environment of “one-off” projects. In such a setting where each client has specific needs and where internal and external environmental conditions are constantly evolving and changing over time, there is no guarantee that a single design methodology or artifact will be relevant in subsequent projects. 

As these architectural robotics and processes are currently in their infancy, the field remains fairly under documented and untested. Therefore, the designer’s ability to fully utilize the robots’ abilities within design and fabrication, as well as to realize the potential of robotic tools, remains underdeveloped. In addition to this inherent knowledge gap, the large amount of time and energy necessary for robotic programming and tool fabrication by designers tends to limit current project outcomes. Projects such as T_Three show some of the unexpected, yet beautiful outcomes possible through the utilization and misuse of these tools.



Whereas many designers and researchers currently focus on the potential for higher efficiency, precision, and form finding potentials through the implementation of advanced architectural machines and robotics, T_Series projects explore the machines/robots’ potential to directly manifest the architect’s hand as a means of rethinking architectural craft. Through a rigorous understanding of each tool’s limitations, tolerances, and methods of failure, it is possible to document, re-implement, and embrace imperfection as a strategic design aesthetic rather than as an element to be avoided. Below are the chronicles following the production of one such project, and how the residuals of imprecise machining were utilized as a new means of creating architectural craft.

Moving beyond formal operations, these projects are developed as a result of an interrelationship between the machines and series of four constraints—materials, computational tools, surface programming, and the misuse of digital fabrication. Each constraint helped to create a unique interaction between man, material, and machine. 


1. Materials

In craft, understanding your materials, their properties, and how they interact with the tools being used are extremely important. For the design of T_Three, a large volume of very beautiful and unique hard and soft woods had been donated to us from Bel-Air Woodworking in up-state New York.

Although materials were donated, finding the appropriated members was quite tedious. As many of the larger members were not planed, all lumber was first scanned and modeled for a better understanding of their sections. This created a digital library of size, species, color, and weight for the individual lumber types.

After material documentation was complete, the next task was to choose a combination of lumber with interesting visual and milling characteristics, as well as identifying those members with similar levels of distortion to allow for the smallest amount of waste.

Materials were tested and arranged for best continuity. The final chosen materials had the rough dimensions of 60 x 21 x 3 inches and consisted of five boards of four species—cherry, white oak, red maple, and walnut—of different sizes and cross sections. The material sections were left unaltered. Rather, irregular members were assembled as is, which enabled the final artifact to retain an asymmetrical form and material composition. 

Once the final material arrangement was determined, all members were laminated to form the initial massing. Weight distribution, structural loading, and member curvature determined the base geometry of this form. The largest and heaviest member (cherry at three inches thick) was placed in the center, with the lighter members on either side. After several days of curing, the base mass was rescanned and modeled, allowing for a curvature analysis of the base form. As the board of cherry had a high level of twist, distorting the whole overall form, it was necessary first to CNC plane down the top elevation, allowing for a more predictable top surface.

2. Computational Tools

Upon completion of assembly and mechanical planning, formal design studies were initiated. Utilization of the tensile modeling software MPanel, in conjunction with the parametric modeling plug-ins Grasshopper and Kangaroo physics, made it possible to create a series of formal operations that were reproducible through both additive and subtractive means of digital fabrication. An easy dialog between the different software allowed for the formulation of an accurate parametric representation of catenary surfaces, resulting in a simple reconfiguration based on material and aesthetic properties. 

These catenary based forms, derived from structural loading, created an overall shape which was deepest in the locations where loads were transferred (ie. legs) and thin along the edges and center of the overall artifact. This formal movement between thick and thin allowed for varying levels of computational residuals to be applied and take form along the bottom surface.

Overall Form:

Limited by the initial material sizes, the artifact takes on the simple curvature of a stretched tensile membrane. Twelve fixed control points were created along the edges of the artifact’s digital model. These points represent a theoretical jig, which defines the tables outer most edges, while also creating the focal points of the curvature. 


After the creation of the outer control points, initial non-tensioned membranes were inserted between them. Within the MPanel environment, different levels of stress, edge conditions, and material types were applied, which programmed the levels of these deformations onto the surface of the lumber. Once programmed, the surfaces were then relaxed, bringing them into their natural tensile state. 

As MPanel does not allow for continuous real-time physics calculations, this initial model was then tied into Kangaroo Physics, making possible a more one-to-one interaction. Within the software, top membranes were welded, to create a seamless enclosed curved surface. The bottom surface, on the other hand, was left un-welded, acting as a rough seam between the curved panels. This contrast in surface curvature created a series of sharp lines that move along the underside of the finished artifact. 

Material Reduction:

With an overall form constrained by the outside edges of the initial material, the artifact retained a large amount of excess mass. As a means of reducing this mass and introducing a dialog between the top and bottom of the piece, stretched holes were created along the material’s grain. These holes not only create visual effects, but also expose the true nature of the materials and construction to the user.

3. Surface Programming

Although there is a level of craft within the assembling of materials and designing of the overall form, a higher level of craft can be achieved in the programming of your tools as well as how they interact with the materials in new and innovative ways. Through the addition of texture and tool marks, the artifact begins to take on unique qualities reflective of both designer and machine.

Rather than focus on the creation of pristine artifacts that perfectly represent their digital form, T_Three envisioned an interactive system where residual surface elements from human machine interaction come together to create a new design aesthetic. Through a direct interaction between human and machine during the fabrication process, new layers of detail and complexity began to emerge on the final fabricated artifacts.

Through a series of tests, a correlation between machine location, speed, and material was realized. These variables began to control the amount of unpredicted residual elements imprinted within the material’s surface after each test had run. Upon deeper investigation of these residual imperfections, the T_Series prototypes were designed with a deeper interest residing in the mapping and harnessing of these imperfections as a means of creating architectural detail, rather then aiming for the creation of the perfect digital replica. 

The digital mapping of these unpredicted motions created a system or feedback loop through which imperfections could be mapped and then directly implemented into the fabrication process. This systematic approach allowed for the honing of imperfection in a way that added an additional layer of tactile complexity to the project. 

Upon completion of the formal geometry, the model was then introduced into the milling environment for the programming of the tool paths. To avoid complications and allow for additional testing, the project was broken into six major components: initial trimming and top planing, rough cutting of the bottom, finishing of bottom, rough cutting of bottom holes, finishing of bottom holes, and finally, finishing the top surface. Each component applied a series of unique effects to the artifact.

Top Surface:

As the artifact was to be used as a low table, the top surface was created extremely smooth to avoid long term maintenance problems. Working with the grain of the wood, the surface was milled with fine rastered increments, . This created a completely smooth top surface which required no additional finishing. Although curved because of its catenary form, the smooth finish along with the strong wood grain made this curvature almost invisible. After finishing the top surface, the perforations were added, connecting the top and bottom surfaces. 


Machining imperfections on the top surface are minimal due to its relative flatness. But as the surfaces begin to translate towards more complex curvatures, the programmed machine imperfections begin to increase. As the curvature increases, the table’s texture and residual elements slowly begin to change, transitioning to the more rough texture of the table’s bottom surface. This creates a tangible relation to the material form, perceptible as the hand moves along the surface curvature towards the bottom of the table.

Bottom Surface:

The bottom surface takes on a much more textured appearance through both material and machine usage. Moving from the short edges into the center, the materials change from the very smooth white oak and mahogany into the more grainy cherry. Working with this natural material transition, the tool’s path accentuates these embedded properties through a variation in milled texture. 


Utilizing the half-inch ball bit for the entirety of the bottom surface, a high level of variation is achieved. Beginning on the edges, the bit’s overlap begins at 1/16 inch steps parallel to the grain. These overlaps increase in distance reaching one half inch when arriving at the central cherry surface. This gradient creates a transition from an almost smooth surface on the surface wood, to a rough center. The shift also highlights the variations in the wood’s natural grain spacing.

Nearing the perforations, the milled texture is quickly transformed back into a smoother one. This transition allows for the seamless flow from bottom to top surfaces. 

4. Digital Fabrication & its Misuse

Upon completion of the surface programming, the code is written and transferred to the machine for processing. The milling of the T_Three took place in a series of six steps that required ten passes and four flips of the material. Realignment of the material was critical as the edges from top to bottom were continuous and the perforations between the sides needed to be aligned perfectly. As there were no flat surfaces on the unfinished base form, temporary jigs were created to adhere the material in a level position during the milling process.

The final piece retains an interesting blend of both machine and human qualities. The top surface is extremely clean in execution, yet is still reminiscent of handcraft. The simplicity of form contrasted against the wood’s grain and coloration give the piece a soft, warm feel. Changing with age, the colors and scent, in combination with the materials imperfections, tell a story of a handcrafted artifact.


In contrast, the bottom surface tells a different story, one of complexity in the retention of a machine-like quality. The striations made by the tool bring out the colors and grain of the wood. As the tool paths continue from one material to the next, subtle changes in color, texture, and quality appear. Imperfections that would be extremely obvious on the top surface disappear within the field of lines.

As the radius of the tool path begins to tighten in following the grain of the cherry wood, the quality of cut begins to degrade. The sharpness of the curves caused un-programmed reverberations in the machine, which in turn imprinted this story onto the material. These imperfections from the machine do not take away from the craft of the artifact, but rather add a new layer of depth. The further honing of these imperfections based on the material curvature creates residuals that spark the curiosity of the user, and draw their hands underneath the table.

The perforations in the wood tell yet another story. Cutting deep into the holes expose the residuals of construction in the display of material depths, grain, and color. Puncturing of the top surface also allows for light to pass, slowly aging the wood’s color at a different rate between the top and bottom surfaces.


In a time where the artifacts we interact with on a daily basis retain minimal human trace, these programmed and non-programmed movements of the machine add distinctive residual elements to the piece and its process. These elements of imperfection, completely unique to this given machine and set of tools, would be difficult to replicate on any other machine or using any other method of fabrication. The bringing together of these flawed materials with misused machines has created a new typology of materiality, one which reintroduces craft and encourages curiosity.

Investigations in Robotic Craft continues with an interview entitled Making Machines Uncomfortable with Andrew Wit.

Andrew Wit is currently the International Practitioner in Residence within the College of Architecture and Planning at Ball State University. He is also co-founder of the interdisciplinary design office WITO*. Wit earned his Bachelors in Science in Architecture from The University of Texas at San Antonio, and his Masters in Architecture from the Massachusetts Institute of Technology. 

Wit has practiced, taught, and researched in the U.S., Japan, China, Taiwan and Hong Kong, including offices of Atelier Bow Wow, Tsushima Design Studio, and Toyo Ito Associates in Tokyo, and Poteet Architects in San Antonio. His collaborative works have been highly published and won several awards such as the 2007 AIA Best of Practice award for UTenSAils, 2007 IFAI Outstanding Achievement Award, and the 2013 Guangzhou Vanke Project of the Year for Guangzhou One.

Wit’s current research focuses on the relationship between robotics, digital fabrication, adaptive environments and their relationship to architectural craft. Most recently his research has taken form in a collaboration with The Boeing Company, exploring the potential for integrating aircraft technologies into the realm of architecture.

For more information on Andrew Wit's work, visit