Newsgroups: comp.ai.neural-nets
Path: utzoo!utgpu!watserv1!watmath!mwtilden
From: mwtilden@watmath.waterloo.edu (Mark W. Tilden)
Subject: Re: Walking References Wanted
Message-ID: <1991Mar18.152636.17955@watmath.waterloo.edu>
Organization: University of Waterloo
References: <1991Mar17.103753.25500@qut.edu.au> <39252@netnews.upenn.edu>
Date: Mon, 18 Mar 1991 15:26:36 GMT
Lines: 27


In article <39252@netnews.upenn.edu> peretz@grad1.cis.upenn.edu (Samuel R. Peretz) writes:
>>sammut@qut.edu.au (David Sammut) writes:
>>
>>I am currently looking into implementing a simulation of a two-legged stickman
>>to see if I could make it "learn" to walk either through a neural network
>>of fuzzy logic (or possible a little of both).
>
>I believe that Rod Brooks at MIT (AI Lab) does this sort of stuff.  He
>uses connectionist models for "insect" robots, and tries to teach
>these 6 legged creatures to walk.

Marc Raibert at MIT would probably be a better contact for this particular
problem.  He's been working on dynamic balance machines for quite a few
years now and has had some real success.  Brooks' designs always use a
static balance assumption (ie: always at least three legs on the floor).
Also, as Brooks deliberately does not do computer simulations of his
creatures, there is no database to evolve a stickman from.

Is all.


-- 
Mark Tilden: _-_-_-__--__--_      /(glitch!)  M.F.C.F Hardware Design Lab.
-_-___       |              \  /\/            U of Waterloo. Ont. Can, N2L-3G1
     |__-_-_-|               \/               (519) - 885 - 1211 ext.2454,
"MY OPINIONS, YOU HEAR!? MINE! MINE! MINE! MINE! MINE! AH HAHAHAHAHAHAHAHAHA!!"