anybody see this article? it sounds really cool, and I'd never thought of it before. Does anyone know more about this? I couldn't find any other reports on it. I'm guessing something like a standard neural net, but the feedback "sensors'" (it was all virtual) values were used as initial points for a chaotic function, and then _those_ results were fed back into the neural net (or maybe both pre- and post-chaotic values in parallel). Any ideas how this was done?

Quote: he is sceptical about the Japanese team’s idea that chaos plays a role in animal locomotion. “It is surprising to achieve what they call goal-directedness with a chaotic robot,” he says.

Lovely quote.

Geee, if the way it really worked involved something that wasn't surprising to you, YOU'D THINK YOU'D HAVE TRIED IT ALREADY!

lol.

you know those "biophysicist(s) at Carnegie Mellon", always commenting on Japanese roboticists. Why didn't they ask English professors at U.C. Davis while they were interviewing people?

Something similar:

http://www.imagination-engines.com/applications/robotics.htm

-dustdevil

Kind of makes me thing of Spiking neural nets, because of the recursion, they create somewhat chaotic patterns... You could possibly evolve it realtime, but that would be more like genetic algorithms than chaotic movement.

When they say chaotic algorithms, do they mean like complex functions? like z -> z^2 + c sort of deal...?

Quote: Original post by Infuscare
When they say chaotic algorithms, do they mean like complex functions? like z -> z^2 + c sort of deal...?

that's what I was thinking they meant.