Dear Robert Pease:
I enjoyed your Parts IV and V of "What's All This Fuzzy Logic Stuff, Anyhow" series (Electronic Design, Nov. 6, p. 146; Nov. 20, p. 159). Wow! What a claim from the "Fuzzy People!" I worked with Navistar International on an advanced electronically controlled diesel engine back in 1991. I guess we were just too uneducated to realize that a PID controller wasn't up to the task of speed control! (Yeah, you guys were REALLY stupid! And, ignorant too. /rap)
I'm going to go back and find out more about this book that said this was impossible. (Did you ever invent an automatic pilot for bumblebees? They can't fly either. I read it in a BOOK. /rap)
We even used the same PID controller as the engine governor. That design was implemented in a digital controller, but it included everything you described in your two articles. (Check. /rap)
We limited the integrator error input term to prevent windup and our controller worked/works very well. And this application IS on trucks! (Yeah, but you can only use it on light trucks, or heavy trucks, but not both—right? /rap)
Real-world controller designs require the engineers to not only understand the technology they're working with, but the systems that the controller is applied to. (Check. /rap) And as you noted, not just the steady-state conditions, but also transient and even error conditions! (Check! /rap)
You have to use the right tool for the job, and know how to use that tool.
Steve Zavodny
via e-mail
RAP:
I have followed your discourses on (or against) fuzzy logic over the years. I, with you, fail to see where in most control systems fuzzy logic affords any advantage over a good analog design, or even digital implementation of an analog design. (One of my major problems is that the FL guys say, "Isn't this a great solution for a simple system?" like the Mamdani steam pressure controller. They expect me to trust them when they have a whole lot of variables for a more complex problem. But when I check into it, I find that they're exaggerating all of their advantages in the simplest problem. So how am I supposed to trust them for big or complex systems? /rap)
Having said that, I must say that fuzzy logic works where the logic is fuzzy—when a person has to describe an event in terms of "sort of," "about," "kinda like," etc. I see no other conditions where it does any good. (It may apply for a really nonlinear problem. You could look up the U.S. patents of Marco Boccadoro. /rap)
I saw an advantage in sensor fusion for intelligent process control. About seven years ago, I did some R&D on using multiple sensors for controlling the stoichiometry and other qualities of tertiary compounds in edge-defined crystal growth. My company decided it didn't want to be in that business before I could cut hardware. However, as there were multiple sensors trying to control multiple parameters simultaneously, and the edges between the conditions were "fuzzy," fuzzy logic seemed the simplest implementation. Simulations worked, but that was pulling one's self up by one's boot strap.
Ralph Reinhold
via e-mail
I agree. There are many complicated systems in the world. When they're hard to model, or hard to simulate, things get messy. But think about SPICE. Electronic things are often EASY to model—and SPICE still gives stupid answers sometimes! /rap
Bob:
I have been watching your comments on electric and hybrid cars. Years ago, I started preaching to my coworkers that the "all electric" vehicle wasn't the way to go. I proposed hybrid cars, but not like the hybrids showing up now. (They're pretty good—and much better than electric cars. /rap)
Most automakers' models retain the mechanical transmission and like you said, use the electric motor only for assisting the gasoline engine. My hybrid version would use a small internal combustion engine driving a generator used to charge the batteries. The output of this pair would have to be the average power used by the vehicle. At constant load and speed, this should be a more efficient way to use the gasoline.
(Yes, that's a better way. I wonder why almost nobody does it! Maybe it's because that would be less like a conventional car. By making something and calling it a "hybrid," they can pitch that they're selling a very "green" product even though it isn't much different from ordinary gas cars. /rap)
The batteries would carry the peak load. Range would be greater than that which can be supported by batteries alone. The motor generator set could even run after the vehicle stopped to fully charge the batteries.
(Check. And if you're going on a moderate trip, such as a 30-mile commute, you can charge up the battery before and after the trip, and never even start your engine unless there are special circumstances. /rap)
Dynamic braking can also recharge the batteries. Each wheel would have an electric motor directly driving the wheel. A processor (not using Fuzzy Logic!) would control the power to each motor, resulting in all-wheel drive and no-slip traction. (I don't really care if the motor drives a gearbox or not. That's not a bad way. /rap)
Jim Jansen
Dallas, TX
via e-mail
We tend to agree on electric cars, but I'm waiting for the flywheels. /rap
All for now. / Comments invited!
RAP / Robert A. Pease / Engineer
rap@galaxy.nsc.com—or:
Mail Stop D2597A
National Semiconductor
P.O. Box 58090
Santa Clara, CA 95052-8090
Reprints
