Theo Jansen style robot leg: prototypes #8 and #9, I suppose
February 16, 2009
My Theo Jansen Leg is finally starting to become something reproducible. Every step in in this arduous journey has been aimed at reproduction. No robot needs just one leg. And also, I wanted to share my design with anyone who fancies a walking Strandbeest robot.
Update 25 feb 2009: New video demonstrating the first powered test. Maybe that should read overpowered test...
(The ugly green screens are an editing artifact. Imagine beatiful fades.)
As explained in the video (?), I decided against the hybrid design of my previous model (with the vertical parts in the bones). It was too bulky and introducing pivots or hinges through a piece of 3.6 mm plywood on end just was not practical. Pivots are a headache as it is.
So the decision was made to produce the entire leg from layers of plywood on top of each other. I needed four layers (not counting layer zero: the robots flank). I always stuck with the ideal of pretty design. Nice curves and blunt points. A little bit like skeletal bones and a little bit like wooden hour pieces. Here is a step by step build up of prototype #7 (which still functions as the set of mother templates; I never glued them together).
The dimensions that really matter to the functionality, Theo Jansen's "magic numbers", were maintained throughout the process. That's why I can say this latest instalment is equal to prototype #5. Here is an image of my drawing with the numbers.
The designations are mine except for letters A-M. Those are by the original artist. The numbering of different prototypes has become quite blurry with several late night mistakes and all the different redoes and mods to individual parts. At one point I had recreated the entire "fibula" part because the holes were drilled askew. Then I continued testing with the old, crooked, part. Several frustrating days later, I rediscovered the remade part and an awful lot of swearing was nullified all of a sudden.
So I retake control over my project by simply announcing (more to myself than any one else) that this one (the one in the video anyway) is number eight. Still need to update my templates though...
I discovered very late in the process that resistance in the joints does matter a lot. I had reasoned that each link (bone) provides sufficient leverage to overcome a little friction between pieces of plywood pressed together by nut and bolt. I were wrong!
When the crank is in this position (which I call 180 degrees), all the narrow bones in the middle are folded together. The projected lever is at a minimum now. The friction in pivots X and Y is very significant in this phase of the cycle. When I loosened the nuts almost, to the point where they were no longer pressing any wood, the clunky action finally disappeared. Almost completely.
I also discovered how sensitive the design is for little inaccuracies in fabrication. When a pivot hole were drilled half a millimetre out of its proper position, the parts would touch each other when in motion. Sometimes to the point where they could no longer complete the whole revolution. Especially at 180 degrees, as shown here. The reason for this is also the very function of the four bar linkage (which there are two of in this leg). It serves to amplify the motion of the crank.
The crank has a diameter of 60 mm. The leg has a horizontal reach of 120 mm. That is an amplification by factor 2.
The next version will hopefully benefit from all these lessons. And then, I will truly have a set of golden templates from which I can cut and assemble a full set of legs. Four or six, that remains to be seen. And I might publish my templates as a CNC design on Ponoko.com. To have cut by CNC laser just makes it so much cooler. And more expensive... Oh well, I deserve it.