Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!news.moneng.mei.com!uwm.edu!lll-winken.llnl.gov!ames!news.dfrf.nasa.gov!neon.dfrf.nasa.gov!joep
From: joep@neon.dfrf.nasa.gov (Joe Pahle)
Subject: tuning a Fuzzy controller with a GA  
Message-ID: <1994Nov8.161626.23507@news.dfrf.nasa.gov>
Sender: news@news.dfrf.nasa.gov (Usenet news)
Reply-To: joep@neon.dfrf.nasa.gov
Organization: NASA Dryden Flight Research Center, Edwards, CA
Date: Tue, 8 Nov 1994 16:16:26 GMT
Lines: 29

I am toying with a fairly standard fuzzy controller (3 inputs, 1
output) implemented in matlab. I would like to tune the membership
functions (I'm currently using trapazoids for speed ...) with a GA I
also have in matlab.

The hard part is trying to restrict the GA to produce "proper"
membership sets, i.e. where the sum == 1 over the entire set. I could
try to put a measure of how "proper" the set was in the fitness, but
this seemed like a very indirect approach. I have read a few papers on
integrating GA and Fuzzy for online optimization, but this usually
means changing the form of the membership set, or the defuzzification,
but of which I would like to remain VERY simple for real-time
performance.

Is there a general procedure in applying a GA to a problem with 
inter-related parameters, similar to side constraints ?

Thanks in advance !!

-- 

Joseph W. Pahle 
F-18 HARV flight controls 
phone: 805 258-3185
e-mail: joep@neon.dfrf.nasa.gov 

Comment: disclaimers attract lawyers ... here boy !  
NASA Dryden Flight Research Center, Edwards,CA

