A Clear Rethinking : Competition Has U.S. Firms Warming Up to Fuzzy Technology
In the mind of Noboru Wakami exists a refrigerator that can make up a dinner menu based on what’s in cold storage, what’s losing its shelf life and what would make a nutritionally balanced meal for its owner. It can also print shopping lists as stored items are consumed.
Wakami, a Matsushita Electric Industrial Co. research manager, says his imaginary refrigerator may become real some day because of fuzzy set theory, a mathematics-based technology that has already added a primitive form of intelligence to everything from electric razors and washing machines to water-treatment plants and machine tools.
Fuzzy theory makes it possible for relatively simple machines to make the kind of rough judgments that humans routinely make by creating statistical representations of such common-sense concepts as far and close .
The American engineering community, which long criticized fuzzy logic as the product of “fuzzy thinking,” is warming to the discipline. As a result, U.S. manufacturers, including General Electric Co. and Motorola Inc., are plowing funds into fuzzy research.
The turnabout reflects the reality of competition. Japanese companies have successfully put what appeared to be half-baked theory into practice.
On some Japanese subway trains fuzzy technology is the basis for controlling brakes. Factors such as the approximate speed of the train and distance to the next station are assigned different values, which the system analyzes using fuzzy rules to make the kind of common-sense judgment that a human would make in deciding when and how hard to apply the brakes in order to stop at the right place.
The Japanese have also applied the technology to consumer products already on the market, including washing machines, cameras, vacuum cleaners, portable computers and air conditioners.
The successful application of fuzzy technology may soon prove to be the edge that Japanese companies need to crack some of the few remaining industries in which they do not have a stranglehold.
“In washing machines, refrigerators, elevators, areas that were thought to be immune (from Japan), this is a nightmare come true,” says James C. Bezdek, professor of computer science at the University of West Florida. “Even if the performance isn’t dramatically better, for an extra $50 many people will choose to buy a fully automatic washing machine.”
The conservative Institute of Electrical and Electronics Engineers, the world’s largest professional engineering organization, recently gave fuzzy technology legitimacy by agreeing to sponsor its first annual conference on the subject next month in San Diego.
In Europe too, the discipline is making headway. Hans Zimmermann, professor at the Aachen Institute of Technology, which has Germany’s largest fuzzy research project, says he attracts 750 students to his fuzzy theory lectures, five times the number he got a year ago. “We are getting fuzzy fever,” he says.
American and European scientists have been late in exploring fuzzy logic because “academics have disdain for something that is easy to build,” says Lotfi A. Zadeh, the Iranian-born professor of computer science at UC Berkeley who did the pioneering work in fuzzy logic. “They want solutions to be elegant. It is the curse of respectability.”
Meanwhile, Japanese researchers have moved on to more sophisticated fuzzy research, while at the same time searching for ways to jazz up existing products with the available technology.
“Fuzzy theory is like seasoning,” says Wakami, who is credited with developing the fuzzy-controlled washing machine that first popularized the technology. “Sometimes the seasoning simply improves the taste; sometimes it produces something dramatically different.”
The advantages sometimes appear minor, but they can be important. Drought-stricken Californians might appreciate the water-saving features of washing machines using fuzzy technology to ensure that the absolute minimum amount of water is used to clean laundry.
The fuzzy-based vacuum cleaner is not only easier to use, it offers psychic satisfaction to boot. Lights on the vacuum cleaner flicker red as the cleaner picks up dust. They turn green when a spot of the floor is clean, signaling it’s time to move on. The machine will also give a reading on how dirty the floor was, indicating that the room should be cleaned more, or less, often.
Matsushita’s fuzzy-based camcorder compensates for shaky hands while filming. Sanyo Electric’s razor has a fuzzy-based control system that adjusts the speed of the blade by measuring the thickness of the beard.
Fuji Electric is building dozens of plants that use fuzzy logic to control the flow of chemicals into its water-purification system.
“In almost every case you can build the same product without fuzzy logic,” Zadeh says, “but fuzzy is faster and cheaper.”
The rapidly decreasing cost of technology and growing competition have driven the sharp increase in the use of fuzzy technology. Virtually every consumer product today contains a microprocessor--a small computer chip. These chips, which cost about $6 each, are already built into the products, and fuzzy features can be added at almost no extra cost.
Japanese engineers say fuzzy technology has made gains at the right time.
“As we become an aging society, we need user-friendly products or nobody will use them,” says Wakami of Matsushita. “To add all these functions without fuzzy you would need control panels like an airplane.” While fuzzy itself is a simple concept, the difficulty is in applying the idea to build products customers want, Wakami says.
Omron Corp., a $2.6-billion Kyoto-based manufacturer of such products as ticket vending machines and health care diagnostic products, hopes to include fuzzy functions in 20% of its goods by 1994.
The company says fuzzy technology will sharply increase the identification capabilities of its automatic blood cell analyzers. And the company is working on a fuzzy system that would prevent tailgating accidents by setting off an alarm, or even gently applying brakes, when an automobile follows too closely.
Komatsu, the maker of large construction equipment, has cut the maintenance time for some of its machinery by 24% by installing a fuzzy-based “expert system.”
Information from experienced machine operators is fed into an Omron-developed software product that generates fuzzy rules. When a problem shows up--overheating of machinery, for example--the system might tell the operator that “there is a 70% chance that the oil filter is clogged.” A more accurate system might be built, but it would be time consuming, costly and more error prone, the company says.
Backers of the technology say that with further theoretical research, the technology will flourish. “I’m convinced this is just the tip of the iceberg,” says Zimmermann.
Japanese engineers are borrowing from another branch of computer research to refine their fuzzy products--neural networks. This technology, developed in the United States, has the advantage of allowing a system to “learn” by simply being exposed to different situations.
“The facts are on the wall; fuzzy is going to be worked into more and more complex systems,” Bezdek says.
He notes that there is a severe shortage of pathologists capable of reading and evaluating Pap smears designed to determine the presence of cancerous cells in women’s cervices. Fuzzy logic could be used to develop pattern-recognition systems that might distinguish a positive reading from a negative one.
“If you could do that, it would be a $25-billion industry,” Bezdek says.
To speed technological developments, Japan’s ubiquitous Ministry of International Trade and Industry has established two laboratories jointly with Japan’s largest corporations--Toshiba, Fujitsu, Nissan Motor and Hitachi--as well as with universities to conduct fuzzy research.
Toshiro Terano, head of the Laboratory for International Fuzzy Engineering Research headquartered in the port city of Yokohama, has great hopes for fuzzy as the core technology connecting various forms of artificial intelligence developed to date.
Terano says Japan has passed through the stage of simple fuzzy and is advancing to a second stage where they hope to develop fuzzy systems that can understand natural language, interpret images and guide autonomous robots.
In a recent exhibit, LIFE researchers showed how they used fuzzy technology to control the four engine blades on a model helicopter so it could fly autonomously.
“We believe that fuzzy engineering is the fusion of Oriental ambiguity and Western rationality,” Terano says. To make greater progress in this field, he suggests, Americans should be less rigidly tied to their old concepts of rationality.
FUZZY LOGIC: MAKING MACHINES REASON MORE LIKE HUMANS. Fuzzy logic is an attempt to make machines reason more like humans--taking into account a number of factors that are not precise to make a common sense judgment. When fuzzy theory is applied to control systems in machines, a number of variables are represented in mathematical terms that are analyzed simultaneously to make operational decisions. The following illustrates how fuzzy logic is used in a washing machine manufactured by Matsushita Electric Industrial Co.:
1
The optical sensor in the waste water pipe sends a beam of light through the water. The way in which this beam is distorted helps determine how much dirt or oil their is in the water as well as whether powder or liquid detergent is being used. This information is sent to the microprocessor.
2
The agitator turns the clothes in the washing machine in order to help remove dirt. The agitator motor measures the resistance it feels as it turns back and forth to help determine the size of the load and roughly what type of material is in the clothing. The larger the load and the more absorbent the cloth, for example, the greater the resistance is likely to be. The motor sends the information to the microprocessor.
3
The microprocessor receives this information in inprecise categories. For example: the optical sensor says the water is clean, somewhat dirty or very dirty and the agitator says the load is small, medium or very large. The microprocessor contains “fuzzy rules” that help it to interpret this general information and choose one of about 600 combinations. Each choice contains a set of orders telling the machine how much water to let in and how much time to allow for the wash, rinse and spin cycles (or whether to repeat any, or all of the cycles). One result after the fuzzy rules are applied might be an order for the machine to use lots of water, mutiple rinse cycles and to extend the spin time if the load is very large and the clothing very dirty.
