avr-chat
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [avr-chat] fuzzy questions


From: Matthew MacClary
Subject: Re: [avr-chat] fuzzy questions
Date: Wed, 20 Jul 2005 11:38:30 -0700
User-agent: Mutt/1.4.1i

On Wed, Jul 20, 2005 at 07:03:11PM +0200, Marcin Brodnicki wrote:
> I'm looking for some components, libraries, code exaples, etc.

    The last time this question came up on this list, I posted an
example controller along with the fuzzy library I use:
http://web.engr.oregonstate.edu/~macclary/fuzz_steer/
http://web.engr.oregonstate.edu/~macclary/fuzz_steer/data/fuzz_steer_example_0.1.tgz

    The library is the two files fuzzy.[ch] and the example is in
fuzz_steer_test.c The library provides fuzzification, fuzzy AND, fuzzy
OR, and fuzzy NOT. There are analogous functions for both integers and
floats.
    One thing that isn't present though is a selection of
defuzzification options, since each controller has its own defuzz
requirements, I just write it into the controller.

> I don't want to write too much code and focus only on my fuzzy logic
> controller,

    You sound like me! Be careful though, when I try to ignore an
implementation detail it usually comes back and bites me.

> I`m also interested in some tutorials concernig fuzzy PID controller
> implementation. Thanks for help in advance.

    I have never understood why people want to make a PID using fuzzy
logic, maybe this is just a semantic problem though.  Are you using
the term "PID controller" to mean the same thing as just
"controller"?
    I would say if you have a PID controller that will work just drop
it in and tune it up.  If you need or want fuzzy control then throw
the PID away completely along with the differential term noise spikes,
integrator wind-up and temperamental co-dependent gain tuning! Smarter
people than me frequently write papers about fuzzy PID controllers
though so you should find lots to read on the topic.
    Fuzzy controllers are user-friendly because you can work with real
physical quantities like angular rate, or temperature instead of
playing with abstract gain numbers.  If you feel like describing your
control problem maybe I or someone else could outline how to attack
it?

-Matt




reply via email to

[Prev in Thread] Current Thread [Next in Thread]