[ODE] Community page

[Ian Munro]

Mattias Wrote:
> 
> My current implementation evolve NNs unsing the NEAT method. Nothing 
> online yet, though.

I've thought about NNs but i'm not really upto speed on them. I should
really investigate the area more.
 
> 
> That looks incredibly cool! I'd love to see him in action. 
> Let us know 
> when you have something to look at!

Sure. Of course you can't see from a still frame the crash that followed 1
second later ;-)


> > I'd previously used the sort of control systems described 
> by Railbert,
> 
> Railbert? Never heard of him, nor could I find any online 
> papers - any 
> links you could send?

Marc Raibert (sorry, mispelt his name) wrote a book called "legged robots
that balance" and was responsible for the major work at the MIT leg lab.

http://www.ai.mit.edu/projects/leglab/
(lots of interesting stuff at the site)

His main contribution was demonstrating that dynamically unstable robots
could be controlled with fairly simple control alogorithms rather than a
complete mathematical analysis of the robots dynamics.

Previous simulations I've done have been using working model 2d, inspired by
the work of John Nagle
http://www.animats.com/papers/leggedrun/leggedrun.html

But working model 2d was incredibly slow to simulate and communicate with.
ODE has been a real eye opener as to how fast simulations can run. It's made
life a lot easier.

My interests actually lie in making physical robots rather than simulations,
but they're a great "what if?" tool.

Cheers
Ian