Layered Design for a Full-Body Tensegrity Model

Model vertebrate anatomy using two layers. The inner layer is essentially a stick figure that tensegrally joins a torso, spine, pelvis and limbs. The outer layer is an enveloping mesh that wraps the body in the form of a fascial surface. Outer layer meshes can be created using Tachi’s work. The form would naturally move like a living being does.

Context: Tom Flemons Archive and Tensegrity Biomimicry of an Entire Organism

Tensegrity torso, spine, pelvis, limbs

(Feb 22, 2014) A more accurate modelling of the body has to start with tensegrity components linked in tensegrity arrays. I have suggested in my paper The Geometry of Anatomy a good place to start would be to model the torso as an expanded octahedral tensegrity linked by means of the spine to a pelvis which is best modelled as an octet truss tensegrity. Of course the spine has to be represented as a tensegrity column and depending on the level of resolution aimed for, each of the bones must be modelled as a tensegrity structure also. But for a low res first approximation I think a torso, spine, pelvis model would suffice. The long bones of the legs and arms (and subsequent bifurcations into appendages) can be represented simply as successions of compression members floating inside of tensegrity joint capsules. I have attempted modelling most if not all components necessary to the task. I am not claiming any final solution to the forms – just a first approximation that has some plausibility. As for actuation of the resulting marionette, some kind of universal mesh that represents the fascia, wraps the structure and continuous actuators ‘muscles’ thread through the mesh and attach to the bones. The actuator system has to affect the fascia system and vice versa.

(Feb 22, 2014) As for linear joints in the arms, legs and appendages, much depends on how the joint is modelled. I propose that there would have to be continuous membrane-wrapped actuator bundles that are not fixed in any one way or place, that crossed multiple joints, and merge imperceptibly into the fascial network that surrounds it. While it is understood that in a spherical tensegrity structure local perturbations propagate equally in all directions almost instantaneously and spread the resultant load evenly through the entire structure, it is somewhat different in asymmetrical structures with elongated components attached linearly and differentially to a central torso. Forces do generally resolve successfully (except at high velocities) but perhaps not quite as fast and not symmetrically.

Model vertebrate anatomy using two layers

Tensegrity stick figure enveloped by a tensegrity mesh

(May 25, 2015) We also talked about the application of tensegrity models to improve CGI rendering in programs such as Maya. I explained to Perry about my theory that the best way to model vertebrate anatomy i.e. quadrupeds and bipeds is to build bodies with two layers of tensegrity. The inner layer is essentially a stick figure very similar to my models of the torso and the complete body. The second layer is an enveloping mesh that wraps the body in the form of a fascial surface. These meshes can be modelled as asymmetrical tensegrity masts (or their equivalent – a braided cable for example) composed of rows of double helices.

The form would naturally move like a living being does

(May 25, 2015) I pointed out that such meshes enveloping a stick model would create a fairly life like rendering of a body in motion without a lot of computation needed – the form would naturally move like a living being does by nature of its tensegral geometry. Perry felt that applying this principle to a modeling engine for CGI would prove to be revolutionary. I have no idea whether it is possible to build a tensegrity rig which would speed up rendering and save studios millions in the cost of making movies and in the gaming world would allow revolutionary life-like creations of avatars who moved with the fluidity of living beings but he must. If he thinks it’s possible then it’s probably worth investigating.


(May 29, 2015) The dragon is a good example of what I’m talking about. This sketchup jpeg shows how the double layer tensegrity might work:

Outer layer meshes can be created using Tachi’s work

(May 27, 2015) Yes I am familiar with that paper [Tomohiro Tachi (2012) Interactive Freeform Design of Tensegrity] and what he proposes in it is what I had in mind when discussing the idea of creating tensegrity meshes that could describe the boundaries of any object. If such a generated form could be imported into the NTRT software (or equivalent) and then actuated we would be well on our way to reproducing functional plausible kinematics. However, I strongly believe that an inner skeleton must be added to form to give it vertebrate characteristics. What a skin mesh replicates is an invertebrate like an octopus – without the inner skeletal form it would look and act like a bowl of jello.

(May 25, 2015. This earlier discussion provides context for the preceding paragraph) The way the Superball moves is, it essentially deflates part itself and as it rolls towards the deflation inflates behind itself to keep the motion going. This is undoubtably how we as vertebrates move ourselves. Our muscles and fascia modulate the prestress in our bodies to allow movement to take place. There is no top down external marionette control system helping us move. It is distributed force generation that is decentralized.

What I’m proposing is a kind of simplification of that process. Tensegrity components can model anything from a torso to a spine or even the individual bones of the leg or arm. Coupling separate pre-stressed (and therefore stable) components into a series of linkages or joints which are controlled by a separate set of lines is it kind of a workaround solution. It’s what we have available with current technology and materials. Again this is in aid of solving real-world problems  to creating actual tensegrity robots or tensegrity prostheses.

However these limitations don’t exist inside a computer simulation so the situation is slightly different. I like your idea you proposed a while ago to make as many connections as possible between compression elements and then see which ones get selected for as the models are placed under some evolutionary pressure. Might it be possible to code for some tensegrity blanket grid which could be mapped like a mesh on to a 3-D scanned object for example, cause it to move, and from that generate a minimal functioning tensegrity structure?