Tuesday, September 18, 2007

Fuzzy Logic

Business India, November 30-December 13, 1998

Closer to real life

Fuzzy-logic-based consumer goods may not be worth their premium prices, but in complex systems, and where safety is involved, fuzzy logic scores high.

Shivanand Kanavi

What's the first thing that comes to mind when you think of a machine? That it is mechanical. Which means that it does exactly what it is instructed to do at the press of a button or turn of a knob. The machine sees everything in terms of discrete numbers with no choices in between. That means that you, the user, have a limited choice - you cannot get the benefit of the in-between values, be they temperature for an air-conditioner or intermedi­ate distances for an auto focus camera.

That is changing, thanks to what is called fuzzy logic. This logic, used to programme newer, more sophisti­cated machines, works differently. Unlike conventional machines, which act on simple yes-no instruc­tions, fuzzy logic machines can oper­ate in more complex conditions. In that sense they behave more like humans, whose thought processes are complex.

Let us take the example of a family deciding to shift house. Many consid­erations are weighed before a deci­sion is taken. For example, the new house is bigger, but is further from the husband's office, though it is closer to the wife's dispensary and daughter's school. The cost per square foot of the built-up area is higher, but the location is cleaner and quieter. And so on.

In short, the new house has a number of pluses and minuses. The family's decision will ultimately be either yes or no, but it will have been arrived at as a result of a complex process in which the factors involved are given varying degrees of impor­tance or weights. That is what statis­ticians would call a weighted average. To put it simply, that is how fuzzy­ logic-based machines work.

Fuzzy logic has found numerous applications in the control systems of complex machinery. In the 1990s Japanese and Korean companies have launched a large number of consumer goods with fuzzy controls. For exam­ple, a fuzzy-logic washing machine uses sensors to measure the size of the wash load and the turbidity in the wash water (which will indicate the amount of dirt in the wash). A few fuzzy rules then turn these signals into patterns of water agitation for different lengths of time and different amounts of detergent to be released by the dispenser. Accurate and inexpensive sensors became widely available in the late 1980s, as did fuzzy chips, and thus consumer goods with fuzzy controls became a reality (see table).

The shopper's guide to fuzzy logic
Product Manufacturers The fuzzy advantage
Air-conditioner Hitachi, Mitsubishi, Sharp, Matsushita (Videocon) Consumes less power
Auto engine Nissan/NOK Controls fuel injection
Camcorder Matsushita Cancels hand-held Jitter and adjusts auto focus
Photocopier Canon Adjusts drum voltage based on picture density, temperature, and humidity
Dishwasher Matsushita
Adjusts cleaning cycle and rinse and wash Strategies
Refrigerator Sharp, Daewoo(India) Sets defrosting and cooling times based on
Rice cooker Matsushita, Sanyo Sets cooking time according to amounts of rice and water
Television Sanyo(BPL), LG, Samsung, Sony Adjusts screen and texture for each frame
Video Camera Canon, Sanyo Adjusts auto-focus and lighting
Washing machine Daewoo(India), Matsushita(Videocon), Sanyo(BPL), LG, Hitachi, Samsung
Adjusts washing according to dirt level, fabric type, load, and water level.

Some of these products have reached the Indian market recently. For example, Videocon and BPL have introduced fuzzy-logic washing machines based on Matsushita and Sanyo technology. Daewoo has intro­duced its own fuzzy washing machines and refrigerators. Videocon has a fuzzy air-conditioner, BPL a colour TV, and so on. All these machines are priced 10-20 percent higher than the non-fuzzy models. The companies claim that the payoff is in ease of use and better perfor­mance. But will a fuzzy washing machine save Rs.3,000 worth of power and detergent in its design life of, say, 5-7 years? Not very likely. Besides, if something happens to the fuzzy circuitry, the repair charges are steep as the companies keep the design proprietary.

Doubtful value for money
Today fuzzy control systems have further evolved into even more advanced adaptive fuzzy. These systems change their fuzzy rules as the environment changes or as the machine undergoes wear and tear. Now we have refrigerators with adap­tive fuzzy logic which change their compressor cycles on the basis of how the consumer uses the fridge. Is the door opened very often in the morn­ing and evening and not during the rest of the day and most of the night (as a working couple with no children might do)? In a house with many children, the door might be opened often, except when they are in school or sleeping. The pattern might change again during the summer and winter vacations, and so on. The adaptive fuzzy chip learns the pattern of usage, records it in an internal clock, and triggers off the compressor accordingly. Consumer goods with adaptive fuzzy logic control are even more expensive and doubtful value for money.

“Fuzzy is wrong, wrong, and pernicious"

“So for as the laws of mathematics refer to reality, they are not certain. And so far as they are certain, they do not refer to reality.
– Albert Einstein (Geometrie und Erfahrung)

Fuzzy logic and its application suffered from official neglect and even ridicule in the US. A distinguished electrical engi­neer once said, Fuzzy theory is wrong, wrong, and pernicious. Fuzzy logic is the cocaine of science.”

Another traditionalist added: "Fuzzification is a kind of scientific permissiveness. It tends to result in socially appealing slogans unaccompanied by hard scientific work." Such strong opin­ions were a product of intolerance and fundamentalism that no doubt exist in many influential members of the scientific establishment. It was also provoked by the fact that initial advocates of fuzzy thinking gave only "hand-waving arguments" and no "hard science". Today the conserva­tives have had to eat crow. IEEE, the most prestigious body of electrical engineers worldwide, has a separate journal for research in fuzzy logic. Many Japanese and Korean companies have also turned these ideas into commercial success.

Lotfi Zadeh, an Iranian born in Azerbai­jan, developed fuzzy logic while teaching electrical engineering at the University of California, Berkeley, in the mid- 1960s. He used his prestige as a brilliant systems engineer to encourage people to work in fuzzy logic, but he faced constant ridicule. Today, he has been vindicated after a hard struggle. However hard nuts among the traditionalists have tried rationalise the fuzzy logic phenomenon by calling it some sort of Oriental mysticism (and hence Asian companies were the pioneers). However, today fuzzy is part of the arsenal of any expert in artificial intelligence.

Interestingly there was a school of Jain logicians in ancient India who had devel­oped a six-valued logic called shyadvad, instead of the Aris­totelian yes-no type of binary logic.

However, adaptive fuzzy logic is a must in more complex systems like a steel rolling mill, an aircraft, or a high-speed train. For example, if the control system of a helicopter can adjust itself to wear and tear, and changes in the outside temperature and dusty conditions, it can fly safely even in severe conditions. The absence of such adaptive controls led to failure of the commando operation launched by Jimmy Carter during the hostage crisis in Iran. Many of the US choppers crashed in the hot and dusty deserts of Iran before they could get anywhere near the hostages!

Today, adaptive fuzzy logic is being used in a large number of non-mechanical applications as well, such as evaluating takeover targets, modelling econometric changes, simulating test marketing, project management, and so on.

Fuzzy logic tries to accommodate the greyness of life as against the black and white of Aristotelian logic and is thus an advance of theory. Control systems or simulation programmes based on it are a step closer to the complexities of real life and play an important role where the cost of a mistake can be frightful. However, applications where the controls are not critical, as in a wash­ing machine or air-conditioner or fridge, are of doubtful value to the consumer.

No comments: