🚀 go-pugleaf

In article <1992Sep12.190738.892@csus.edu>, tching@target.water.ca.gov
(Tracy Ching <SysAdmin>) writes:
|> In article <68037@hydra.gatech.EDU> gm26@prism.gatech.EDU
|> (MCMURRAY,GARY V) writes:
|> > I am having trouble understanding how the membership function
|> >is created in order to create linguistic control rules and the
|> defuzzification
|> >rules. 
|> >Internet: gm26@prism.gatech.edu
|> 
|> 	What techniques (if any) are available to come up with the
|> membership
|> functions.
|> tching@water.ca.gov

There are some, but most depend on having an existing controller that
you
are trying to duplicate.  In practice, that's a fairly rare situation.
We
at TIL are currently working on genetic algorithm techniques to
automatically tune a fuzzy control system in place.  This seems to be
a
promising approach.

The general technique for doing the job by hand is to start off with
an
arbitrary set of membership functions (I like to use 50% crossover
points
between triangular membership functions, with equally spaced peaks
ranging
between the extreme values of the associated variables), and to tune
the
results in place, or in simulation.  The technique works reasonably
well,
but requires some experience, and takes quite a bit of time.
-- 
Erik Horstkotte, Manager, Software Development, Togai InfraLogic,
Inc.
erik@til.til.com, uunet!til!erik - (714) 975-8522
Standard disclaimer: the preceding are my opinions, not TIL's.