The piece is called ‘Chirality’, meaning ‘handedness’, and is a large scale tensegrity sculpture that was designed and built for Burning Man 2015 in the span of four months. It was built onsite by a dedicated crew of 8, with a total volunteer base of 30.
Tensegrity structures are highly responsive, ultralight, and very strong. They also happen to be incredibly difficult to build, which is what made this such a fun engineering challenge. We set out to build a tensegrity structure that could be climbed, to allow the viewer experience the dynamics firsthand, and learn to trust the fundamental strength and compliance of tensegrity.
Chirality is seen in physics when the spin of a particle is used to define its handeness, or in chemistry when molecules cannot be superimposed onto their mirror images. ‘Handedness’ is essential to the technique of developing a tensegrity structure: when you stare down the axis, each progressive layer alternates from Left to Right, interwoven in space. When viewing Chirality, there is a moment of alignment down the axis which provides insight into the fundamental techniques of tensegrity. The form represents a loose interpretation of a molecule.
Tensegrity represents a hybrid of art and engineering, two disciplines that I’ve been passionate about. The goal was to make the technique accessible and find a way to map the forces and responsive aspect of the structure, a like a realtime FEA. The structure had LEDs triggered by accelerometers embedded within the end caps, and changed as the structure was climbed. At night it became a structural data visualization, resembling a twinkling point cloud at night.
The true beauty for me lies in the process we developed, the hard earned learning experiences, and the team that grew together. The sculpture happened almost by accident. This is the story of how it all came to life.
Task 1: Wait, What Did I Just Commit To?
Working on an art piece for Burning Man is like running a startup in your free time. It challenged me on all levels: driving concept, pioneering a new method of construction, building a team, keeping the team happy, coordinating fabrication & engineering, endless mechanical design, sweeping the shop, and taking out the trash. It took a whole new level of learning and self sacrifice to bring this thing to life.
We took a challenging construction technique and brought it from conception to fruition in an incredibly harsh environment on an accelerated timeline. Here’s what proved essential to the process:
Katherine Barton – General Counsel, Sanity Check, And Best Helper
Natalie Walsh – Fabric Design
Ken Caluawerts – Electronics Design & Software Development
Jack Kalish – Lighting Effects
Rachel Ciaverella – Electronics/Fabric/Drill Press
Atil Iscen – Software Analysis
Ekin Senturk – Structural Engineering
Many Others Including, But Not Limited To: Johann Karkheck, AJ Romine, Pamela Pascual, Vytas Sunspiral, Joe O’Connor, Jonathan McKeever, Ashley Cuppet, Jamie Trowbridge, Gordon Kirkwood, Dustin Fieder, Tory Voight, Brittany Powers, Aaron Porterfield, Bryan Hermannson, Owen Laine, Shelby Clark
Task 2: Research & Inspiration
A large chunk of time was spent researching tensegrity. I noticed that there has been quite a lot of cutting edge research and model making, but not enough large scale tensegrity in the world. We initially set out to develop a freeform tensegrity shape, but ended up developing a scalable and low cost technique for building tensegrity systems.
My interest in tensegrity started 10 years ago when I read Volume 6 of Make Magazine. I followed the steps and built a primitive four strut tower, and was pretty proud of it. I remember being frustrated with how difficult it was to manage the cable lengths with fishing line, and how tricky the eyelet connection was. I think this was a pivotal moment that set me down the path of studying engineering and art.
Some Resources & Inspiration:
The term tensegrity was coined by Buckminster Fuller as ‘tensional integrity’ and pioneered by Kenneth Snelson as a form of floating compression.
-Tomohiro Tachi has developed some really impressive freeform tensegrity generator software. This was one of the most inspiring research that I found, and what inspired one of my concepts. He did the generative tensegrity bunny rabbit above. Here’s a video of his software in action:
-Tom Flemons has done a lot of research on Biotensegrity.
-NASA Tensegrity Dynamics Robotics Laboratory: Super Ball Bots
-Kulripa Bridge in Brisbane, Queensland, Australia. It cost $63 million, and is the world’s largest tensegrity
-Random photo of a tensgrity at burning man. Does anyone know who did this? To the best of my knowledge this is the only successful tensegrity structure that has been brought to Burning Man.
Ken Snelson Tensegrity (1).pdf
Ren Motro (Auth.)-Tensegrity. Structural Systems for the Future-Butterworth-Heinemann (2003) (1).pdf
Task 3: Concept V1: Climbing Structure 4 Strut Tower
The project initially started out as an idea to build a climbing platform for my friend Joe’s camp. Tensegrity seemed like a fun technique to employ for a climbing structure, as it would resemble a swaying treehouse. The concept was based on a simple four strut model, with stacked 8 ft struts bolted together. Participants would climb an axially centered rope ladder and hang out in the cargo netting.
I even played with the idea of stacking tiers, to a total anticipated height of 40 ft. I’m very glad we didn’t go this direction for a myriad of reasons, safety being the foremost.
Task 4: Initial Prototypes
It was time to start building a proof of concept. To start, Joe and I prototyped using 1” aluminum poles and bungees. It was fairly straightforward to build a four strut model after getting the bungee length, but the tension in the system wasn’t evenly distributed for a stable shape. Assembly started by weaving the struts in a single direction, and affixing the ends. We next built a tower by stacking two four strut models, with alternating CW and CCW layers. It was tricky to manage tension in the structure, and it was difficult to scale the technique. It was a total mess.
However, we made an interesting discovery. When the structure was compressed, it rotated axially and compressed. This allowed the structure to be compressed into a flat pack, and spring back to shape when released. The diagonal cables bore the load evenly and stretched. This proved critical to the final technique.
Next, we tried upgrade to steel cable. The connectivity wasn’t dialed in so it was nearly impossible to get the tension right. This was when I realized that it was going to be a challenge to build a large structure out of tensegrity. It would appear that there’s a very real reason that I haven’t seen these more often.
The poles proved time consuming, so I purchased a bulk set of Tensegritoy models off Ebay in order to explore the medium more effectively. This proved to be the most valuable purchase of the entire project.
Task 5: Initial End Cap Design & Prototypes
I wanted to come up with a system that made tensegrity more affordable and accessible, but my first concept was a long shot from that. This technique was inspired by Ken Snelson’s intricate end caps, which manages the tensions separately for every single strand. The initial concept I came up with involved a cap placed on the end of the strut. It was designed with two objectives in mind:
1. Accommodate a wire rope stop internally for individual lengths of cable
2. Grip onto existing cables to allow the modules to stack, so that it could be built upwards
I designed a CNC lathed part with slots for the rope, and a waterjet end cap that could be tightened down with a 1/4”-20 bolt. I quoted parts through MFG.com, which allowed me to post my design up for bid and allow multiple vendors to submit quotes.
At the time, I had no effective way to simulate the forces throughout the system, so it was difficult to begin analysis of the individual components. I ended up choosing 1000lbf to start, and ran a quick FEA only to determine that the poles I was going to use weren’t able to handle it.
By the time the parts had arrived, the design had evolved into a new technique. This was an expensive learning experience, since I had 45 made. It was time for a change in direction, but that’s half the fun of this type of project eh?
Task 6: Circuit & Slot Technique: Proof Of Concept Prototype
Next, I took inspiration from a technique that Rene Motro wrote, called the ‘circuit strut’, meaning a continuous circuit to define the boundary of the kites that are visible on the facets of a tensegrity. I realized this could be implemented in a very simple fashion by cutting slots into the pipes and threading it through. This technique was a winner because it reduced total labor and assembly time, as well as cut down hardware cost. Another major was was that I didn’t have to buy those expensive end caps.
A huge bonus is that it transferred load laterally through the cable vs. focusing it all on the pipe wall connection. I realized that it could manage tension across the structure if I put a turnbuckle inline.
So what better time to try this idea out than in the forest in Mendocino with a bunch of friends? I packed up the necessary supplies in a milk crate and we played for a weekend, and came up with a proof of concept.
It proved to still be difficult to assemble. There was no way to hold the poles in the correct position so it took many hands until the whole thing was in tension, but we battled with the ends slipping. We first built it with string, but it had no no directional reinforcement, so l I tossed it up in the air and it exploded upon landing.
Task 7: Concept V2: Freeform Human Hand
Once that concept was proven somewhat functional, I felt confident that we could do something bigger. I was inspired by Tomohiro Tachi’s generative freeform design and wanted to create a shape that referenced the musculoskeletal system. The initial concept was a scaled up human hand made of 150 struts, which turns out was a bit excessive for the scope of the project.
I quickly discovered that the sheer number of poles and cost of required hardware ruled this option out, not to mention the complexity of the design. It made more sense to go with larger struts and spend less time and money dealing with hardware, and focus more on the piece itself.
We tried to make a hand-like shape by attaching four strut & icosahedron models. This proved to be an unstable technique but presented some hilarous models.
My mentor PK advised that the human hand was a literal metaphor. There’s a well established history of scaling things up in the art world, but it wasn’t conceptually challenging. He asked me ‘What can your brick do that others can’t?’
Time for a new direction. This marked an important milestone in the project: we’d built a bunch of individual cells but lacked an overall connectivity. It was not a stable shape: I still had much learning to do regarding the fundamentals of tensegrity.
Task 8: Back To Basics: Formfinding
We now had 2.5 months. It was time to make some important decisions based on the constraints: limited time, fixed financial resources, and a small crew. And we had absolutely no idea what this thing was going to look like yet! We were starting from scratch, which turned out to be a good thing.
A notable technique involved wrapping/unwrapping, inspired by Tomohiro Tachi’s paper. I started by laying out struts in a CW/CCW arrangement, and connecting each layer’s ends in a circular fashion. The shapes looked particularly interesting when only partially wrapped.
We also tried stacking icosahedrons, stacking different polyhedra, and other techniques of integrating structures. None seemed promising. Tensegritoy models are easy enough to build, but any problems in technique scale up exponentially when you’re dealing with struts that are taller than you are. We made some pretty cool shapes but they wouldn’t translate into a stable structure.
Task 9: Fabric Exploration
Natalie Walsh spearheaded the fabric development. We explored multiple concepts of how to clad the shape in fabric to create more of a body to the piece. We chose to cover the outer kites with fabric, as they emphasized the tensegrity technique and held tension well in a parabolic shape, but this would also morph into something different.
Task 10: Final Shape: Animation & Generation
Devon Penney, our animator, was essential in finding a way to generate the tensegrity shapes. He implemented the generation technique from Tomohiro Tachi’s paper in Houdini to create shapes based on polygonal meshes. Once this system was in place, we leveraged the Houdini’s extensive procedural modeling tools to quickly try a variety of polygonal meshes (lower left image above) to get a feel for how different tensegrity shapes looked without building them. To get a feel for how these shapes responded to gravity in a physical environment, Devon used Houdini’s Grains solver, which is typically used to simulate sand, but is great for producing stable simulations of connected objects like tensegrity structures. Once we had decided on a final shape, he produced lit renders of the real world scale model. This was an essential step in allowing us to figure out what we wanted the final piece to look like, and provided a lot of insight into the process of modeling that we used as the project progressed.
The final shape was built as a 5×3 tower, 5 struts in circumference, with three layers alternating CCW/CW axially.
Here’s an initial simulation Devon built to get an idea of how the structure responded to forces:
Task 11: Electronics Initial Prototype
Ken Caluawerts from the NASA tensegrity robotics lab was our electronics lead. He showed me a video of a prototype he built with the NASA tensegrity robot with glowing lights on the ends, and it was quite beautiful. Since we wanted to have an interactive element to the piece, we decided to use his discovery and apply it on the end caps of the sculpture. We chose to use accelerometer input to trigger the lights. The more the sculpture responded to weight as people climbed it, the more it would light up.
In a very short time, he developed a wifi module that had accelerometers and capacitive touch sensors built in, and would drive the lighting effects. One of the difficult parts of being on a project with such an accelerated schedule was that we didn’t have time to have a custom PCB board made. We decided to pare down to just the accelerometer effects because out in the dust, simple is better.
Task 12: 1/2 Scale Prototype: First Attempt
The first attempt didn’t go so well. We spent a lot of time dealing with attachments and hardware; I started with a copper sheath technique to protect the rope from the poles, but it gave us more trouble than it was worth. This was when I realized it would be well worth it to scale the poles up, but we had to develop a better way.
This proved to be quite a challenge. Any flaws in the assembly technique scale up exponentially when the struts are taller than you are. It was time to develop a method to get 15 struts into the right place.
The first time we tried it, it was done using a wrapping technique. All the poles were laid out in a 5×3 tower, and wrapped around. This worked well for the toy models but quickly we got lost in poles and cables during assembly. It ended in a pile of sticks and strings on the ground :/
Task 13: 1/2 Scale Prototype: Success!
The next day, Katherine Barton and I approached the jumbled pile with a fresh mindset. We tore it all apart and rebuilt it, noticing connectivity issues along the way. We started more simply, with a two layer pentagonal shape, and layered our way upwards.
We built a primitive hoist system with the shop crane to get a better idea of how it would come together. Another critical breakthrough was the decision to use wire rope clips to hold the ends in place; otherwise the ends would slip and it was nearly impossible. It was an iterative approach: once they were roughly in place, we adjusted the lengths to get a closer approximation.
The technique worked surprisingly well, but there was a lot left to iron out. In order to get accurate lengths and a fully stable structure, it was necessary to build a stable shape in CAD. Now it was time to apply the underlying geometry in CAD and build a proper tool to make assembly easier.
Task 14: How To Hold The Poles In Place? Jig Design & Underlying Geometry
Since we had just used a duct tape fixture to get the poles together for our initial assembly, it seemed reasonable to design a jig for the full size piece. I knew we needed some technique to manage all 15 struts simultaneously, so it seemed very reasonable at the time to design a jig to hold them all in space while we connected the wires.
This presented a unique design challenge of its own: the jig must be able to hold the 15lb poles, rotate in 2 axes, hold their shape when you wanted them to, and collapse down for packing. And be cheap. So i devoted a week to jig design.
The jig was designed using Tomohiro Tachi’s transformation technique: by picking the midpoint of each edge and using that as the vertex for the transformations. Each pole was rotated normal to the axis by 15 degrees, and the layers were alternating at 30, 0, -30 degrees. This provided the clockwise/counterclockwise/clockwise twist and brought the shape into a semi-spherical configuration.
Each of the three hubs had five poles aligned in a pentagonal pattern. They had thumb twist set screws to hold everything in place, which made everything easy to adjust. The hubs were nesting with schedule 40 pipe running along the axis, and set in place with dowels.
The transparent pyramid shape seen in the CAD is the underlying geometry: I was manually tweaking the jig to match my desired underlying geometry. This was a clunky process because I didn’t initially design it parametrically, it took a bit of internal assembly part editing to get what I wanted.
TENSEGRITY JIG REV01.PDF
Task 15: Structural Analysis: Strut & Wire Rope Selection
It was time to validate our design using engineering analysis software. I consider this one of the most fascinating and technically challenging parts of the whole project. It turns out traditional engineering tools aren’t equipped to solve problems of this complexity.
The closest loads we could approximate would be to simulate the structure packed full of heavy people. To start, I came up with a simple load approximation to pick the correct struts. To start, I chose 2.5” schedule 40 6061 aluminum pipe. It had a good human proportion, decent price, a weight of 15lbs, and didn’t rust, which made it attractive.
Ekin Senturk, our structural engineer working remotely from NYC, helped build out buckling simulations. Dr. Tugurul Turan from Istanbul Techical University (a friend of Ekin’s) ran a nonlinear analysis in ABAQUS, and Ekin interpreted the results and wrote the attached analysis. The loads simulated were 1,200lb compressive load (working load of the wire rope) and 500lb side load (a lot of people)
When running an analysis in ABAQUS, our structure was so responsive that it flagged an ‘excessive displacement’ error. This made it difficult to use a tool like FEA to get a read on the whole system.
For the cables, Ekin and I came up with a similar approach. We chose 1/4” 7×19 galvanized steel cable, as it’s rated to 1,200 lb with a 5x safety factor. Initially we held off on purchasing the rope until we had evaluated a range of other options: 3/16”, 7/32”, and even, 5/16′. In the end went with the 1/4” for the best price/strength/weight ratio.
I’m happy with the process; we used our best judgments with the tools and short time we had available. We could have spent more time analyzing and optimizing, but this would have killed us on the actual fabrication schedule. I’m looking forward to developing a new methodology of analysis for the next one.
Strut Analysis For Tensegrity Structure.pdf
Task 16: Computational Analysis & Simulation
Luckily, we had access to some cutting edge research from the NASA tensegrity robotics team (led by Vytas Sunspiral). Atil Iscen and Ken Caluawerts both helped to develop custom software to analyze tensegrity structures. This utilized some cutting edge computational design to help validate the stability of the shape.
Sadly, we didn’t have time to sort out the workflow between Devon’s animation, my CAD, and the simulator. This is something worth working on more, and adds a technical challenge for future revisions of the project.
Here’s an early simulation Atil ran:
Task 17: Jig Fabrication
I’ve been working with metal for the past decade in various capacities, so it was a great chance to share that skill with the team. It took one night of prep work (layout, grinding, measuring, cutting) and one night of welding.
The entire jig was designed to sit within a wooden spool. It went well overall, despite the minor hiccups that are inevitable from rushing so much. At one point I discovered that I had grabbed the wrong size conduit from Home Depot and it simply wouldn’t go together. I also discovered that I hadn’t checked my tolerances thoroughly enough and had chosen the wrong schedule 40 steel pipe. Luckily my friends over at Flux bailed me out that same night, and even helped me recut the pipe on the same night.
Task 18: Build #1 Attempt
After the jig was built, we began to assemble the structure. The jig worked exactly as planned, and held the poles exactly where i wanted them.
My initial design involved malleable aluminum sheaths that would protect the steel wire rope from the stress concentrations of the slots in the aluminum pipe. Unfortunately, they also crimped themselves to the wire rope as soon as they were routed and tensioned, which meant that every length had to be perfect. Unfortunately we were freestyling it the same way we had done with the prototype, so it didn’t work out so well. We used ropes at set lengths to measure the lengths from the CAD and adjusted the sleeves to those lengths, and crimped them on. This was messy and tedious; since some of the measured lengths were not perfect, giving us conflicting dimensions.
I was confident that we could get it to stand, but it was going to be messy, and not tuned to perfection. Despite having sunk two full days into the jig build, I made the difficult decision to tear it all down and start from square one.
Task 19: Recovery Via Chisel
The bent sheaths came back to haunt us: now we had about 60 aluminum sheaths that were crimped onto our wires, and they had to come off. My friend Johann came by and we spent a full Tuesday night with a hammer, chisel, and channel locks. It was messy, but it worked. It turns out if you angle the chisel correctly with a firm hand, you will glide along the wire rope without damaging any strands.
Task 20: Flat Packing: The Right And The Wrong Ways To Place A Turnbuckle
I didn’t like the jig; there was too much time spent assembling fasteners, and it was fighting itself during assembly. We could have gotten a reasonable structure, but it wasn’t elegant. My friend Gerald mentioned that he’d seen rapid deployment tensegrity domes before, so I started thinking of a better way.
I knew that it was critical to place turnbuckles strategically to allow for tension management. Each layer has to be in the exact same spot, as opposed to our initial prototype. I thought back to the compressing structure from the exploration with bungees, and how the diagonals all stretched to allow flat packing.
I built two different maquettes that would test the turnbuckle placement. Each was diametrically opposite from its location on the kite. I chose a piece of tape to represent a turnbuckle, and disconnected that section. Both methods packed flat and seemed to work fairly well, but there were noticeable differences
The first technique (Method #1) I chose to prototype was the one where the pole layout is obvious: it seemed a good candidate for assembly because it would allow us to easily place the poles and see their pattern.
However, the elastic lengths of the model weren’t able to give me a real feel for the technique. This was more of a peeling than a flat packing. It proved to be very difficult to assemble this as there was a facet on each shape that didn’t have a turbuckle attached; it was fighting itself during assembly. So we tore it down and tried Method #2.
Method #2 was based on learning the hard way that each facet of the shape (triangle, kite, or quad) would have a turbuckle attached to allow for a systematic adjustment of the structure. By placing them in the correct spot, there is a continuous line of turnbuckles that can be traced along the structure. These turned out to be the diagonals from the bungee example. It also allowed us an easy way to conceal wiring for the electronics, as we had to cover the turnbuckles as well. With the turnbuckles falling on each diagonal (CW/CCW/CW) it allowed the structure to compress and pack flat along the axis. Success!
This may be the most critical innovation in the entire project. This technique is scalable, and will be used for the much more complex and challenging piece that we’ll be doing next year.
Task 21: Making New Models
It was time to build a model that would simulate the actual construction technique, as opposed to elastic lengths. I took some cord and tied knots on either side of the vertex. Each layer (A,B,C) had a different length pattern, as the turnbuckles fell on different lengths.
The first structure I built was sloppy and didn’t hold its shape. All my lengths were off because I didn’t account for the slots in the toy models, which reduced the vertex lengths by 5%. Katherine was nervous about this, as any error would propagate, so I rebuilt the model. The difference was easily apparent; the technique was solid. Full steam ahead!
Task 22: Prep Work & Assembly Line
Now that we had our plan of attack, it was time to spend a full night messing with hardware. After swaging all the necessary ends, we laid them out in the warehouse and got to work.
Note the new sheaths: I found some ultra-flexible steel conduit from McMaster Carr and slid it on there. This proved to not do much in the long run, as the flex tube unraveled under the pressure of the rope. I had a feeling it wasn’t going to work out but I figured it couldn’t hurt to have them on there.
We also installed rope clips. This was a precaution to temporarily hold all vertices in place to test the placements. The sheer amount of time spent assembling all this harware was staggering. There were hundreds of fasteners to deal with…
Just for fun, something my friend Dustin Feider showed us :
Task 23: Build #2: How To Grow A Tensegrity From The Ground?
Since I wanted to try it without the jig, we used ratchets to tension each layer incrementally, which caused it to deploy quite beautifully. This was our first run, so we spent a lot of time figuring out the hardware, fittings, etc. We didn’t tighten the wire rope clips enough so some points slid, causing some lengths to really fight each other. It was a difficult build but it worked! This gave me the confidence to proceed and refine the tactic.
Task 24: End Cap Design
The end cap was a fun mini-project in its own. How to pack all of those electronics into an end cap?
The first design was a very simple pentagonal pyramid, referencing the underlying geometry of the tensegrity shape. I initially had a sliding cap design, but it wasn’t easily manufacturable. I ended up choosing on a shape that could be bolted down onto the outer pipe, and something that could be easily manufactured. After getting quotes from multiple vendors I realized any custom manufacturing was out of the budget, so we went with DIY vacuum forming as it was a cheap and quick way to make 30 caps.
The second and final end cap was designed by my friend Johann Karkheck, which took the same pentagonal draft and lofted it with a twist. This was a beautiful touch and reflected the generative structure of the tensegrity.
Task 25: End Caps: Mold Design, Mold Creation, And Mold Plugs
Vacuum forming the caps meant that it was time to learn moldmaking! This was an enjoyable mini project, and fun to learn. I designed a two part mold that would be cast with two-part silicone (Castaldo putty from McMaster). The flexible mold allowed for it to be peeled off the plaster plugs so that I could reuse the molds. First, I 3D printed a two part mold, and cast silicone molds. These were then filled by plaster of paris plugs to be used for the final vacuum forming.