Making a Flexible Strandbeest

Thu Sep 22 2022

The Strandbeests are a family of kinetic sculptures developed by Theo Jansen. At their centre is a linkage which is responsible for their organic gait. The linkage is driven by a crank, and its end effector is a foot. It can be driven both fowards and backwards and at different speeds, making it analagous to a wheel. If several of them are arranged in tandem, and the cycles are syncopated using a crankshaft so that the feet do not leave the ground all at once, you have a sort of walking trolley that can either be powered at the crank, pulled along, or made to descend an incline under its own weight.

I thought it would be possible to make the linkages out of a single piece of plastic, making use of flexures instead of joints. This comes with several advantages:

The difficulty of simulating large deflections

Unfortuantely, flexures are significantly more mathematically complex than joints. I used Larry Howell’s Compliant Mechanisms almost exclusively, it is comprehensive and very readable.

I won’t bore you with the theory, if you are interested I’ve detailed my adventures in simulating compliant mechanisms with multiple flexures here (spoiler: I didn’t get very far).

Fortunately, you can make a pretty good approximation of a flexure with linear springs and links, called the Pseudo-Rigid-Body Model (PRBM). They are explained in Chapter 5 of Compliant Mechanisms. The most obvious PRBM is to model a small flexure as a single torsional spring in the middle.

Fig. 1

In Fig. 1 we see three simple flexures operating in tandem with their PRBM equivalent. A has a very short flexure, so the PRBM is accurate through a large range of motion, B and C have much longer flexures, so the PRBMs are only accurate through a range of about 90 degrees. The value of a, used to determine the position of the pivot in B and C, is a empirically derived number.

Using FreeCAD to Design Linkages

I tried several different rigid body solvers, but found them clumsy to use. You would have to define your model using code, or import it from some CAD package. I wanted something I could work on in realtime. I started using FreeCAD to sketch out ideas, using the contraint solver in the Sketcher workbench as a poor man’s rigid body solver. I found that FreeCAD worked fine for my use case (I wasn’t worryed about kinematics or dynamics at this point, I was looking for a mapping between crank angles and end effector position). FreeCAD has a python interface, so I wrote a script that would periodically change the angle of the input link.

Fig. 2

I could then change the lengths of links as I watched the linkage cycle, observing how the charactaristics of the path of the foot changed. Once I had something that more or less worked I wrote a second script that recorded the path of the foot as a set of coordinates, then changed the geometry of the foot, and repeated. When it had finished running I had a set of a hundred different paths, and I just had to select the best one. Now that I had my finished PRBM, I set about turning it into the compliant linkage it was supposed to be representing. I replaced all the joints with flexures and filled the spaces between with rigid bodies.

Fig. 3

I attempted to run a FEM analyis using CalculiX in Freecad, applying different displacements to try and replicate the effect of the crank turning. The results aren’t good, I suspect because the internal solver is linear. Notice how the pivot hole expands and contracts oddly.


The Printed Linkage

I mirrored the linkage to make a part with two feet with one input and sent it off to a 3D printing service. I did have to reassure them that I wanted the flexures to be that thin, and that I would take the of risk them getting damaged in the post. Here is the final linkage in action:


As you can see the parallel flexures buckle due to the torque on the foot from the link to the pivot. This is a limitation of using a PRBM, if I had found a better way of modelling the mechanism I might have been able to mitigate it. That being said it still works reasonably well. I never made a full Strandbeest using my mechanism, at this point I was completely sick of the project.