Newsgroups: comp.robotics
Path: brunix!sgiblab!uhog.mit.edu!nntp.club.cc.cmu.edu!honeydew.srv.cs.cmu.edu!argonaut
From: argonaut+@cs.cmu.edu (Charalambos Athanassiou)
Subject: Re: Neural Networks and robots
Message-ID: <CqFBCr.7qq.3@cs.cmu.edu>
Sender: news@cs.cmu.edu (Usenet News System)
Nntp-Posting-Host: b.gp.cs.cmu.edu
Organization: School of Computer Science, Carnegie Mellon
References: <2s28d5$1hn@hermes.acs.ryerson.ca>
Date: Thu, 26 May 1994 18:58:02 GMT
Lines: 32


It depends how you define the use of neural network in controlling robots.
The CMU ALVINN networks in the CMU-NAVLAB/UGV projects are used to
directly predict the direction that the vehicle should follow. 

Neural network have also been used in robot control, i.e. control of
robotics arms to help cut down on the amount of calculation needed to do
kinematic/dynamic analysis of the arms.

Other groups use NNs as a function approximator and try to learn to predict
the best action based on current state and some perception (sensing) of the
surrounding environment. For the latter flavor, which covers most of the
Reinforcement Learning area, take a look at the WWW
Mosaic file:
http://www.cs.cmu.edu:8001/Web/Groups/reinforcement/mosaic/homepage.html

This is the homepage of the local group here in CMU and contains references
to relevant work and on-line papers.

I hope this helps

Sincerely Yours
===========================================================================
Haralabos N. ATHANASSIOU		
Robotics Doctorate student		Tel: 	(412) 268 8461 (office)
The Robotics Institute				
Carnegie Mellon University		Fax:	(412) 621 1970
Pittsburgh, PA 15213-3890					
USA
---------------------------------------------------------------------------
(InterNet): athanassiou+@cmu.edu	c.athanassiou@ieee.org
===========================================================================
