Newsgroups: comp.ai.fuzzy
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornell!travelers.mail.cornell.edu!news.kei.com!news.mathworks.com!uunet!in1.uu.net!olivea!charnel.ecst.csuchico.edu!csusac!csus.edu!netcom.com!thinman
From: thinman@netcom.com (Technically Sweet)
Subject: FCM: as good as it looks?
Message-ID: <thinmanD4Ds97.871@netcom.com>
Organization: International Foundation for Internal Freedom
Date: Wed, 22 Feb 1995 02:58:18 GMT
Lines: 35
Sender: thinman@netcom10.netcom.com

I'm working on a Virtual Reality scripting system and I
was intrigued by a paper on using FCMs in VR systems for
animating creatures with interesting behaviours.  It
was co-authored by Bart Kosko and one of his grad students.
(I think her name was Julie Dickinson, but am not sure).
It was in MIT Presence, a VR quarterly.

The article intrigued me because I'd been wondering how
to do just that: give non-programmers a simple technique
for design behaviours for VR creatures.  (It is rather a 
tall order!)

The article claimed that all you have to do is:
1) design an FCM network
2) design situations and desired outcomes, and
3) the theory would make you a weight set that would
cause the network to "always do the right thing".

This seems impossible.  What am I missing?  Can't you give
it contradicting situation-outcome sets? Do you have
to repetitively redesign the FCM after discovering
contradictions?

I'm also interested in finding re-distributable source code for
an FCM engine that can efficiently iterate multiple FCMs
in parallel.  Is there something good in one of the FTP sites?
Also software to generate weight sets from situation-outcome pairs.

Thanks,

-- 

Lance Norskog
thinman@netcom.com
Artisputtingtogether. Art  s th ow n  aw y.
