home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.fuzzy
- Path: sparky!uunet!gatech!concert!rock!taco!eos.ncsu.edu!dlgerber
- From: dlgerber@eos.ncsu.edu (DARRELL L GERBER)
- Subject: Re: WHEN and WHY should I use FUZZY logic?
- Message-ID: <1993Jan22.212749.7020@ncsu.edu>
- Sender: news@ncsu.edu (USENET News System)
- Organization: North Carolina State University
- References: <C15uAt.8nH@cpqhou.se.hou.compaq.com>
- Date: Fri, 22 Jan 1993 21:27:49 GMT
- Lines: 60
-
- In article <C15uAt.8nH@cpqhou.se.hou.compaq.com>, pipkinsj@cpqhou.se.hou.compaq.com (Jeff Pipkins ) writes:
- |>
- |> There seem to be plenty of answers to the question of WHAT fuzzy logic
- |> is, and HOW it works. There are several good primers that provide
- |> a good introduction to those questions.
-
- (stuff deleted)
-
- |>
- |> I'm trying to keep an open mind about this whole fuzzy thing, but I just
- |> can't imagine a situation where I could benefit from it. I'm willing to
- |> assume that the problem is my ignorance... So, enlighten me!
-
- One reason for using Fuzzy Logic Controllers which hasn't been emphasised enough
- is that you don't need any kind of model to apply it. This may not seem like a
- big deal for simple systems (in fact, if compared with standard PD control of a
- simple pendulum it is much slower and more complex) but if you consider highly
- non-linear and coupled systems (such as robots) it is very fortunate to not have
- to develop a mathematical model of the system.
-
- Another important consideration is how easy (relatively) it is to make the basic
- fuzzy logic controller adaptive. In fact, I am currently working on developing
- an adaptive fuzzy logic controller which uses a second fuzzy logic controller as
- the basis for the adaptation routine. The beauty of this is that you still
- don't need any model or parameter identification routine. Unfortunately, how do
- you prove it is stable and converges?? I don't know!?!?
-
- One other thing I haven't seen mentioned is how the rule base in fuzzy can be
- chosen to resemble other types of controllers. For example, consider the simple
- PD controller. If you plot the resulting output as a function of the error in
- position and the error in velocity (e and edot) you will find it is just a plane
- whose orientation is governed by the proportional and derivative constants.
- Now, place a grid over the plane described by the e and edot axes and create
- fuzzy sets centered at the intersections between the grid and the axes. The way
- to make a fuzzy rule base to resemble PD control is create a rule for each point
- on the grid of the form:
-
- IF (applicable fuzzy set on e axis) AND (applicable fuzzy set on
- edot axis) THEN (value of PD plane directly above the intersection)
-
- This may sound pretty confusing but it is much easier to understand if you have
- a picture. This example, however, is good to get the basic idea but the
- concept is valid no matter how many degrees you have (this was 3D) and what
- type of control surface (or hypersurface) you are trying to match.
-
- You may be wondering why anyone would want to use this if they could use the
- original controller. It may not be obvious but if you are using an adaptive
- fuzzy controller you can start with a rule base found from this method and then
- let the adaptation routine "tune" it to what the best control surface for that
- particular system. Again, no models needed!!
-
- ------------------------------
-
- dlgerber@eos.ncsu.edu
- Department of Mechanical and Aerospace Engineering
- North Carolina State University
-
- -------------------------------
-
-
-