Back in the 1980s I was getting excited about the application of fuzzy logic to control applications. But it seems to have not made such a big impact, at least on hardware. But perhaps fuzzy logic's time has come at last.
I remember back in the 1980s getting quite excited about the application of fuzzy logic to control applications. I think there was a startup company but I certainly remember speaking on the telephone with Lotfi Zadeh for an article I was writing.
Lotfali Zadeh, is the father of fuzzy logic having done most of the mathematical groundwork. He is also a classic example of western migration to Silicon Valley. He was born in Azerbaijan of Russian and Iranian parents in 1921 and went on to be an outstanding mathematician, electrical engineer, computer scientist, artificial intelligence researcher and professor emeritus of computer science at the University of California Berkeley.
I also remember how in the article I was writing I tried to explain how a double inverted pendulum could be kept upright by applying a relatively simple fuzzy logic control algorithm to the movement of the trolley on which it was balanced. Latency meant other more conventional DSP control solutions on the hardware of the day were not up to the job.
Fuzzy logic is described as many-valued logic or probabilistic logic and it deals with reasoning that is approximate rather than fixed and exact.
So rather than having an air-conditioning system that works with temperatures sensed and calibrated to the nearest degree with lots of detailed math, fuzzy logic would just work with terms such as "cold" chilly, warm and hot.
I remember that programming was usually based on IF-THEN-ELSE rules-based statements along the lines of "IF room hot THEN start fan."
But I also remember what I saw as a potential weakness of fuzzy logic, the use of membership functions to define the fuzzy variables for the given application. These were usually in the form of pseudo-Gaussian distributions, usually simplified to overlapping triangular or trapezoidal distributions.
thanks for the laugh, I was an FAE for Motorola back then and I remember when we had to promote this to customers, give out dev boards - or sell them. I dont recall any customer for micros ever caring about it at all. These days in the semi industry we would still be tracking the status that "design win" and your development board all the way to the landfill.
Examples of consumer applications of fuzzy logic often seemed to be better served with linear or analog solutions (in the winter, the cooler the house is, the harder you run the furnace and the air blowers). Deciding between low, medium, and high seemed artificial. Some examples do survive (dishwashers offer light, normal, and heavy wash options). Certainly the observations about computing power are also quite relevant - the fuzzy logic decisions typically are trivial compared with the computing power that is running idle.
In grad school at Berkeley in the mid '80s, I was in Prof. Zadeh's Expert and Knowledge-Based Systems seminar for a few years, and he never once mentioned FL unless someone brought it up. In private, he would actively promote its' application to IR (i.e., search), even if it didn't quite fit. At one time (I'm too busy to check), there were at least two proper academic journals dedicated to FL, and now decades later there is at least an institutional echo of the movement in the Berkeley Initiative in Soft Computing (http://www-bisc.cs.berkeley.edu/BISCProgram/default.htm), which seems a bit like the Hair Club for Persians. Prof. Zadeh was a great guy, in the cosmopolitan/humanitarian/uber-nerd tradition.
My understanding of fuzzy logic is that it was a way to have a controller that did not have enough horse power, do the job. Since almost all of the devices today have more than enough power to do the job, fuzzy logic doesn't have much of a use.