Newsgroups: comp.ai.fuzzy
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!uhog.mit.edu!sgiblab!rpal.rockwell.com!planets!slc
From: slc@planets.risc.rockwell.com (Steve Chiu)
Subject: RE: fuzzy control-illusion or future
Message-ID: <1995Jan5.023634.6361@planets.risc.rockwell.com>
Organization: Rockwell International Science Center
Date: Thu, 5 Jan 95 02:36:34 GMT
Lines: 44


Johan Pensar wrote "in my opinion, [fuzzy control] is an alternative
to nonlinear PID control."  This opinion is shared by most people in the
traditional control community as well as by many in the fuzzy control
community.  This is unfortunate, because it tends to narrow our conception
of how fuzzy logic can be used to solve control problems.

Traditional control methods, and the fuzzy controllers described in most
technical papers, attack the set-point regulation problem.  If you
read a paper on fuzzy control, chances are you will see that the inputs
to the fuzzy controller are some set-point error and the rate of change
of error.  When used in this way, fuzzy logic is indeed a form of nonlinear
PID control.  However, we must remind ourselves that this is only one of
myriad ways in which fuzzy logic can be used in control applications.
Having worked in the fuzzy control field for a while, I am convinced that
the real payoff is in applying fuzzy logic to high-level, or task-oriented
control.  For example, using fuzzy logic to determine what the set-point
should be, or for selecting and smoothly switching between a number of
conventional controllers.  In fact, most of the commercial applications of
fuzzy control use fuzzy logic for the task-oriented, not set-point-oriented,
part of the control problem.  For example, in the famous Sendai subway train,
fuzzy logic is used to rank the predicted results of several possible control
actions (power and brake notch settings), to determine which control action
provides the best combination of comfort, safety, on-time arrival, and
stopping-point accuracy.  In Yokogawa Electric's temperature controller,
fuzzy logic is used to determine intermediate set-points that are fed to a
conventional PID controller; the intermediate set-points lead the temperature
trajectory to the final desired value quickly without overshoot.  Fuzzy logic
can be used to classify the operating environment and determine which
conventional control law is most applicable.  It can also be used to detect
the onset of critical transitions in a process to provide a timely switch
in control strategy.  There are so many ways to use fuzzy logic in control
applications, of which the fuzzy PI or fuzzy PD controller is only one. 
Unfortunately, we have been conditioned by education to have a narrow view
of what constitutes a control problem, and hence what constitutes fuzzy
control.  Lotfi Zadeh likes to quote the aphorism "If you only have a
hammer, everthing looks like a nail".  In the case of how fuzzy control
has been perceived, I would say "if you only look at a nail, everything is
another hammer".

-Stephen Chiu
Rockwell Science Center
Thousand Oaks, California, USA
slc@risc.rockwell.com
