Columnist

Let's grab this bull by the horns

David Brubaker, contributing editor

In a recent letter to the editor in another electronics trade publication, a reader, identifying himself only as a "Senior Engineer," slammed the publication for devoting too many pages to pushing "the aging rope of fuzzy logic." He said that, because fuzzy-logic devotees made exaggerated claims of the technology's applicability when it first appeared, it had lost credibility. He stated there probably are niche markets for fuzzy-logic products, but resented the technology being presented, "in an insultingly arrogant and unbalanced manner," as a panacea for design woes. He finished by saying he was as sick of reading this "many-times warmed-over BS."

  In one of my earlier columns in EDN, I stated that I avoid involvement in pro-fuzzy/anti-fuzzy debates. But several issues raised in this reader's letter call for a response:

Dear Senior Engineer,

  You've been duped! And you don't even realize it. But don't feel bad—many others have been, too.

  A friend of mine, Bob Pease, who is also a senior engineer and a columnist in another electronics trade magazine, holds similar views and is continually sending me mail he has received from engineers who are also happy to expound on the frivolity of fuzzy logic.

  But, as someone who has actually designed fuzzy systems, I assure you that you have been hoodwinked. It probably went something like this (you fill in the details):

  Years ago, your first reaction to the term "fuzzy logic" was that academicians were playing name games and had created a cutesy title to increase the chances of getting papers published. With no need to find out about something as childish-sounding as "fuzzy," you assumed it would quickly disappear.

  But it didn't, and, in fact, companies, most often Japanese, claimed that using fuzzy logic resulted in better products and reduced development times. Because fuzzy logic wasn't going away, you owed it to yourself and your company to find out a little about the technology.

  With an ever-skeptical eye, you searched out articles, in both academic journals and in a number of the electronics trade magazines. What you read confirmed your suspicions. Described in the academic papers were "toy" systems, in no way resembling the real engineering problems you faced daily. Worse still, the trade-magazine articles contained large doses of hype, written by (or under the direction of) marketing directors of small, underfunded, start-up companies, anxious to sell their fuzzy system development tools. Moreover, data that appeared in articles from both types of journals was almost always simulated, not taken from actual operational systems.

  In addition, comparisons were made in the articles between fuzzy rulebased systems and "traditional" design techniques, with the fuzzy systems always providing superior performance and always being easier to design. What chafed you the most (and this would be clear to even a junior engineer) was that comparisons were made to simple-minded examples of the traditional technologies—straw men that anyone could shoot down. You concluded (and, for the most part, correctly) that these authors, although possibly fuzzy-logic experts, certainly did not understand the traditional technologies they were shooting down. The comparisons were therefore grossly unfair.

  As a final and purely technical point, you realized that solving complex systems problems with a rulebase sounded like merely using lookup tables. You and every other system engineer knows lookup tables are viable only for the very simplest of systems.

  But, again, fuzzy logic didn't self-destruct. More articles and papers appeared, including a column or two in trade magazines (fancy that). Sampling these, the previous pattern again emerged, although now less often: hype, poor engineering, and unfair comparisons. Your conclusion remained intact. Now it was time to be more vocal, and when a major trade publication devoted an entire section to fuzzy-logic techniques, you felt obligated to respond. Based on all you had read and your experience as a senior engineer, you declared fuzzy logic itself to be a straw man, supported by smoke and mirrors and needing to be exposed to the engineering community.

  With all due respect, there is an error in your logic. Hype, poor engineering, and unfair comparisons do not necessarily imply weakness of the basic technology. They are merely examples of hype, poor engineering, and unfair comparisons.

  There is a saying that the receiver of bad news shouldn't shoot the messenger, with the implication that it is OK to shoot the message originator. A corollary might be, upon receipt of a hyped message, to go and shoot both the messenger and the message originator, but not necessarily the message. Fuzzy logic lost its credibility when you found flaws in articles intended to support its value. If you were to read a poorly written article about a technology you know, use, and trust, would that technology also lose credibility? Or only the writer? Poorly designed circuits that use operational amplifiers are not an indictment of operational amplifiers, but rather of the circuit designers.

  But why, you ask, are so many articles replete with hype, poor engineering, and unfair comparisons? The hype, of course, is an attempt to sell a product. As to the poor engineering demonstrated in academic journals, despite advanced degrees in engineering, most professors and researchers are rather inexperienced at applied engineering. They concentrate on concept and theory, and are rarely, if ever, involved in getting actual products to the market. Furthermore, those who are good engineers and who try to write informative papers for academic journals will most often have them rejected in the review process by theory-oriented "peers." I know—I have tried. The bottom line is: If you read the academic and research-oriented journals, plan on learning theory and concept, not how to implement a working system.

  If fuzzy logic is valuable, why aren't hype-free articles employing good engineering and legitimate comparisons being published? They are and they aren't. I hope this column qualifies as the former. On the other hand, as someone who has designed numerous fuzzy rulebased systems for numerous clients, I have seen a distinct tendency of companies to hide from their competition (or their customers) the fact that they're using fuzzy logic. It's easy for me to see why successful applications of fuzzy rulebased systems become proprietary information, not to be published for the world to see.

  As a fellow senior engineer, I assure you that behind the hype, poor engineering, and unfair comparisons, fuzzy logic—both in rulebased and other structures—is a powerful design technique in more than just a few niche markets. Fuzzy logic is not a cure-all, nor should anyone claim it to be. Also, the use of a fuzzy rulebase does not guarantee a good design—garbage in, garbage out still applies.

Sincerely,

David Brubaker
Senior Engineer

PS. I infer from your letter (although it may have been edited out) that you have never actually designed a fuzzy system yourself; that your criticism is based entirely on what you have read. If you have the time and inclination, try one or two. Buy CubiCalc from HyperLogic ($495, Escondido, CA, (619) 746-2765) and work through some of the examples to get a feel for the technique. Then select a system you have first-hand experience with and try implementing it using a fuzzy rulebase. If you have trouble, give me a call and I may be able to help. Even if you don't need help, I would be interested in hearing about your fuzzy design experience.

PPS.  Or were you just hollering?

| EDN Access | feedback | subscribe to EDN! |
| Table of Contents |