Evolving Motion and Neurosystems in Soft Robots

Continuing the discussion from Comments on Gpuccio: Functional Information Methodology:

Really interesting reference by @Timothy_Horton!

See the full story at the lab web page here:

The related papers here:

Unshackling evolution: evolving soft robots with multiple materials and a powerful generative encoding
N Cheney, R MacCurdy, J Clune, H Lipson - Proceedings of the 15th annual conference on Genetic …, 2013-18814.
Evolved electrophysiological soft robots
N Cheney, J Clune, H Lipson ALIFE 14, 222-229
Evolving Soft Robots in Tight Spaces
N Cheney, J Bongard, H Lipson Proceedings of the 2015 Annual Conference on Genetic and Evolutionary
Topological evolution for embodied cellular automata
N Cheney, H Lipson Theoretical Computer Science 633, 19-27 …

The simulator they used is freely available: VoxCad - Creative Machines Lab - Columbia University

The video itself is really worth watching. @Jordan and @cwhenderson this would be a great basis for a class.

@AndyWalsh and @NLENTS this would also make for a really interesting article! I hope you take a look.


I’m still waiting for someone from the Evolution Is Impossible camp to tell us how the new FI in these evolving robots was somehow “smuggled in”, and how much FI was smuggled. :slightly_smiling_face:


Well, a designer programed the software, you know. :crazy_face:

I’ve gone over examples like this with them in the past. They privately acknowledge there is no “smuggling,” but they go silent about it publicly. That is why you are not getting a response, here, I imagine.

For the recored (observers in mind), the researchers did program the software, and they pre-specified the function they wanted (movement), but they did not provide the evolutionary algorithm with any answers at all. The researchers themselves did not know the answers the software would arrive at!

I suspect many marriages will be negatively impacted by the development a ‘soft robots’. But if it keeps incels occupied and off the streets, could that be a net positive? :crazy_face:

Feel free to delete this…

Here is some code:

1 Like

They gave the soft robots everything they needed to do what they do. They were given muscles, tissue, bone and a means to stimulate the muscles. This is a great example of evolution by design.

While I can understand your point, I think it’s missing the point of the research a bit. The point was to use these simulations as a striped-down experiment to understand better how evolutionary mechanism work.

What could be interesting from a design perspective is that the “designer” in this case does not predict or control the outcome (there is no targeted shape), only the optimization conditions (go faster).

So who’s the designer? Is it the algorithm? Is it the researcher? In the immediate sense it looks like randomness but we understand that there is an ultimate cause (the researcher) behind this proximate cause (optimization of random changes).

So a relevant question might be, how does the methodology used in these experiments change the role of the researcher and in what way does that affect notions of “design” in this case?