home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.fuzzy
- Path: sparky!uunet!charon.amdahl.com!netcomsv!netcom.com!gperkins
- From: gperkins@netcom.com (Glen C. Perkins)
- Subject: Re: WHEN and WHY should I use FUZZY logic?
- Message-ID: <1993Jan21.174051.7469@netcom.com>
- Organization: Netcom - Online Communication Services (408 241-9760 guest)
- References: <C15uAt.8nH@cpqhou.se.hou.compaq.com> <johnr.727592984@pongo.kowari.cpsg.com.au>
- Date: Thu, 21 Jan 1993 17:40:51 GMT
- Lines: 41
-
- Not all devices are on/off devices. I think it's Sugeno in Japan who
- is experimenting with control of a damaged helicopter. He has a
- system (I'm told) that will regain control of a helicopter after
- part of its rotor has been shot off.
-
- There are very few on/off flight controls in a helicopter, and
- this system looks at the signal from several sensors simultaneously
- and comes to an extremely rapid conclusion regarding how and
- how much to adjust the flight controls to respond to the
- helicopter's current circumstances.
-
- Fuzzy logic seems to work best when you want fast, continuous
- response to several rapidly changing inputs and you have no
- "model" of the system (in the classical algorithmic sense.)
- In the case of a balancing rod, we have control equations, so
- it's not a very good example of advancing the frontiers of
- knowledge, but it's a good example of how I could put together
- a control system WITHOUT KNOWING those equations, and without
- needing to know them. I'd just tell the system principles such
- as "if the rod is a little bit to the left and falling slowly,
- move your hand slowly to the left" and with a half dozen or
- so of these rules I'd have a first approximation of a working
- control system. I could come up with these rules by just
- looking at the system and saying tom myself "what would I
- do in X circumstances?" and entering it as a "rule." Then
- I'd watch it work an see what happened and adjust it.
-
- It's not always as easy as I make it sound, OBVIOUSLY, but
- tweaking these rules is almost always easier than deriving
- simultaneous differential equations in multiple variables
- from first principles!! And a control system based on these
- simple fuzzy rules operates much faster than a real-time
- solution of the differential equations in general.
- --
- +-----------------------+---------------------------------------------------+
- |--- Glen C. Perkins ---| |
- | <gperkins@netcom.com> | "Don't forget, your mind |
- | Native Guide Software | only *simulates* logic." |
- | Palo Alto, California | |
- +-----------------------+---------------------------------------------------+
-
-