AI learns to balance a stick

Electron · Post by **Electron** » Sun May 15, 2005 12:16 pm

Quite cool. Is the source available or is it closed?

JoeWright · Post by **JoeWright** » Sun May 15, 2005 5:53 pm

I probably won't make the source code available for this one because its a bit messy and a couple of bugs (that actually don't do any damage).

However, I will release some source code for my further projects like my walker.

Joe

katoun · Post by **katoun** » Sun May 15, 2005 7:40 pm

Good.
I'm realy forward to see THE WALKER.
Good luck.

Bot_Builder · Post by **Bot_Builder** » Sun May 15, 2005 9:46 pm

Yeah, the walker should be very entertaining. I bet if you set it up with four legs with some decent power your best walkers for a while will be frog-like jumpers using back legs for propulsion and front for balance.

Very cool implementations of walker evolution can be found in Framsticks - an awesome general purpose NN/GA/physics simulator. http://www.frams.alife.pl/

Another with a couple walking examples among some other neat AI stuff is breve - http://www.spiderland.org/breve/

Robomaniac · Post by **Robomaniac** » Mon May 16, 2005 1:08 pm

Very interesting

katoun · Post by **katoun** » Thu May 19, 2005 5:17 pm

Hi,
I have found some interesting resource:
http://www.isi.edu/isd/LOOM/LOOM-HOME.html
Try to see if it can help you(I think it can, besides you know more about AI that me so you could use it).

JoeWright · Post by **JoeWright** » Thu May 19, 2005 11:45 pm

Katoun,

Looking at that sort of reminds me of the time I spent looking at Prolog to try and solve a sport league scheduling problem. After a while I realised prolog wasn't going to give me anything special so I ditched it, learnt GA and wrote a programme in a day that solved the problem very well.

In other words, I think any half decent programmer could implement their ideas quickly enough in C++ (for example) rather than having to learn a new language.

Regards
Joe

katoun · Post by **katoun** » Fri May 20, 2005 1:14 pm

Sorry,
I have found they don't use C++ for theyr project anymore, too late(older versions where in C++ but those are no where to be found).
Sorry I thought I could help you.

JoeWright · Post by **JoeWright** » Fri May 20, 2005 1:24 pm

Katoun,

No problem, thanks for the link. Actually the AI side of things is pretty straight foward. Its the physics that's the complicated bit. I'm awaiting the next SDK release of Newton which should include the final piece I'm missing then fingers crossed...

Joe

katoun · Post by **katoun** » Fri May 20, 2005 4:44 pm

What final piece(Linux suport) or something else,you got my atention here( like you haven't with your AI )but is there somthing new going to be added to Newton

katoun · Post by **katoun** » Fri May 20, 2005 4:45 pm

What final piece(Linux suport) or something else,you got my atention here( like you haven't with your AI )but is there somthing new going to be added to Newton

JoeWright · Post by **JoeWright** » Fri May 20, 2005 7:22 pm

Julio is hopefully going to include some example code in the SDK of how to programme a motorised universal joint. He did give an example but it was wrong so he's going to have to re-write it.

Joe

katoun · Post by **katoun** » Fri May 20, 2005 9:21 pm

Uuuuuu,

cassini · Post by **cassini** » Fri May 20, 2005 9:43 pm

Actually Linux and Mac support is been done already.
http://www.physicsengine.com/downloads.html