This is a response to the challenge thrown out by Bob Pease in the May 13 issue for a practical example of Fuzzy Logic. Bob, look no further than yourself. In case that sounds like an insult, it might be better to say that we should "look no further than ourselves." Most of our reactions to everyday events are fuzzy. We are neither entirely deterministic or stochastic beings; we apply variable judgment to virtually all of our inputs, and act accordingly.
In the quest for more intelligent behavior from our hardware, we are recognizing the need for electronic systems to be more adaptable. What one calls this approach to design is not the point. I agree that the term "Fuzzy Logic" is nowadays used like "transistorized" was a couple of decades ago, but surely Bob, you understood back then it was just a way of saying "tres moderne." There was substance to the claim (that is, transistors did provide a more forward-looking solution than vacuum tubes, wouldn't you say?) but it was unduly stressed.
Fuzzy Logic is simply that branch of electronic system design that seeks to cope with uncertainty more effectively than a rigidly programmed logical structure could. As one who scorns the predictability of even a scientific calculator, you ought to be a fan: Instead of being bound to the confines of a certain "yes" or "no" mediated through one-bit variables, Fuzzy Logic admits to the possibility that "the truth" (or at least the best estimate) may lie somewhere in between. Digital radio channels routinely make use of Fuzzy Logic within their DSP to continuously determine the optimal conditions for minimum bit error rate.
I believe we are at a juncture where the development of highly adaptive, richly interconnected, quasi-analog, nonlinear, signal-processing structures (though only loosely related to contemporary Fuzzy Logic implementations) will be essential to the emergence of more "intelligent" non-organic companions. Is that important? You would be singularly oblivious of long-term trends not to recognize the potential impact of such on society.
- Barrie Gilbert, Fellow, ADI, IEEE, Analog Devices, Beaverton, Ore.
Barrie, you have my permission to use F.L., or adaptive controllers, or human thought, or any other not-necessarily-linear control system. I'm just cautioning about Fuzzy people who brag and exaggerate unconscionably. Look in IEEE Control Systems magazine, June 1993, p. 5-7 for cases where other people are objecting to overzealous F.L. claims. And thanks for the EXAMPLE of F.L. within digital radio channels.—RAP
Dear Mr. Pease:
Three quick points about your column on Fuzzy Logic.
Fuzzy Logic is Threshold Logic: This is easy. Just look at how it is implemented. Minima and Maxima of Sums is Threshold Logic.
Threshold Logic Computes Probabilities: Instead of True/False bits in and True/False bits out, what we have is probabilities in, probabilities out.
Synthesis of Threshold Logic Circuits is Efficient: Thanks to the Binomial Polytopes, polynomial-time Threshold-Logic circuit synthesis is a reality. Hence, if instead of computing bits we wish to compute probabilities, Threshold Logic, a.k.a. Fuzzy Logic, is the way to go. It is also clear that in most cases, the computation of probabilities is NOT what is required. Many times, however, it is the only way to go.
- Silvio Ursic, Ursic Computing, Madison, Wisc.
Isn't it funny that F.L. people say "probability" when they mean proportionality!!—RAP
Sometimes I agree with you, and sometimes I don't. But I must say that you're right on target regarding Fuzzy Logic. I have been studying the literature on Fuzzy Logic since I first heard the name a few years ago. Our library has considerable information on the subject (mostly in trade journals), and I have yet to read of a single application in which a properly designed and configured digital controller would not provide equal or better performance.
I will concede that Fuzzy Logic is interesting conceptually, and there is a certain curious familiarity with it (probably because the human mind is a bit fuzzy itself). Your comment that you have no documented evidence, either in the form of a workable circuit/controller or a legitimate comparison between performances appears to be a valid one. I, too, have seen much brouhaha and little factual information regarding performance. I look forward to the proof, as pointed in your column. Here's to keeping them honest...
- J.J. Bradshaw, Electrical Engineer, University of Missouri-Rolla
I am really finding some special (nonlinear) cases where F.L. shows advantages. But for general-purpose linear controllers, most claims of F.L. superiority are just hype.—RAP
Your article on Fuzzy Logic in the May 13 issue was right on the mark! After studying the hype on this technique, I came to the following conclusion: Fuzzy Logic is to control design what PLC ladders are to computer programming.
If a designer is limited to a choice of a "bang-bang" design versus a Fuzzy Logic black box, he can probably do better with the Fuzzy. However, if he can apply analysis and a bit more sophisticated hardware (and perhaps some software) he can beat or at least equal the Fuzzy design.
If he needed transition functions and rules for his design to be competitive, then Fuzzy Logic is forcing him to do some analysis that he should have done in the first place.
- Robert M. Stanley, Los Gatos, Calif.
I'm still looking for real examples.—RAP
I have read your column for some time now and find myself in almost total agreement with most of what you say. However, I think you have ignored a major problem with most electronic designers these days. Most of the younger technicians and engineers have never smelled ozone! I have been in the field of high-power broadcasting for 17 years now, and like Tim Taylor of "Home Improvement," I believe the answer to most problems is more power. Most of these young whipper-snappers have never experienced the thrill of having a 25,000-V, 10-A plate transformer arc over right next to their head. Shucks, most of these guys don't even know what an ARC is!
Modern electronics don't even smell right. Most of us broadcast engineers freely admit that we got into the field strictly for olfactory reasons....we love the smell of hot insulation and PCBs! (By PCBs I mean Polychlorinated biphenyls, not printed-circuit boards!)
I was born and semi-raised in Silicon Valley back when it was Vacuum Tube Valley. We have forgotten that wonderful heritage.
Silicon is wimpy stuff compared with the heavy iron and copper associated with transmitting tubes. (I suppose a more politically correct term would be thermionic devices.)
Mind you, I have no objection to modern digital circuits; we use scads of them in the broadcast field. But I think some of the younger folks could benefit from an internship in traditional high-power RF technology.
I guess I'd better close now. I'm feeling a little weak and need to go back to the shop for another snort of ozone....Ahhhh.
- Eric P. Nichols, Director of Engineering, KJNP AM FM TV, North Pole, AK
More power can be fun and educational, but I'd hate to be designing an electric car with a 25,000-V battery and vacuum-tube controllers.—RAP