Fuzzy Logic
Fuzzy logic is a form of many-valued logic in which a statement can be true to a degree between 0 and 1 rather than being strictly true or false, designed to model the graded, vague categories of everyday language such as "warm," "tall," or "fast." Instead of drawing a sharp threshold, it lets an object belong partly to a category. A room at 22 degrees Celsius might be warm to degree 0.7 and hot to degree 0.1, and rules written in ordinary words can then be combined arithmetically to produce a smooth decision.
The field was launched by Lotfi Zadeh, of the University of California, Berkeley, in his 1965 paper "Fuzzy sets" in the journal Information and Control. Zadeh generalized the classical set, where an element is either in or out, to a fuzzy set defined by a membership function mapping each element to a number in the interval from 0 to 1. His standard operators defined the complement as 1 minus the membership, the intersection as the minimum, and the union as the maximum, and later work generalized these to families of t-norms and t-conorms such as the product. In 1973 he introduced linguistic variables, whose values are words like "cold" or "very fast" tied to membership functions, and in 1978 he proposed possibility theory as a companion framework for reasoning about imprecise information.
A membership function is the modeling heart of any fuzzy system, and the common shapes are deliberately simple: triangles, trapezoids, and Gaussian bells. A temperature variable might have overlapping triangular sets for cold, warm, and hot, so that one reading belongs to two sets at once. The overlap is what produces smooth behavior, since adjacent rules blend rather than switch. Choosing these functions is partly engineering judgment and partly tuning, and in modern systems they are frequently learned from data, as in neuro-fuzzy methods such as the adaptive network-based ANFIS of Jyh-Shing Roger Jang in 1993.
Fuzzy control was the application that brought the idea out of theory and into products. In 1974 Ebrahim Mamdani and Sedrick Assilian at Queen Mary College in London used fuzzy rules to control a small steam engine, a result published in 1975 that established the Mamdani-style controller of fuzzification, rule inference, and defuzzification shown above. Takagi and Sugeno's 1985 variant made rule outputs functions of the inputs, which suits mathematical analysis. Japanese industry then adopted fuzzy control widely: the Sendai subway line that opened in 1987, with a Hitachi fuzzy controller for smooth braking and acceleration, is the best-known case, and consumer goods such as rice cookers, washing machines, and camera autofocus followed through the late 1980s and 1990s. Its attraction was that an expert's linguistic rules could be encoded directly without a precise plant model.
The most persistent misconception is that fuzzy logic is just probability in disguise. Probability measures uncertainty about whether a crisp event occurs: a coin is definitely either heads or tails, and probability expresses ignorance about which. A fuzzy membership of 0.7 expresses vagueness, partial truth in the concept itself: a person can be tall to degree 0.7 with no uncertainty about their height at all. The two can be combined, but they answer different questions. Another pitfall is that membership degrees are subjective and not automatically calibrated, and fuzzy systems have also been criticized for offering no guarantees that a well-tuned conventional or optimal controller would not provide, for which reason academic debate with statisticians was lively for decades.
In current practice fuzzy logic occupies a niche inside embedded control and decision-support software, and its ideas persist in soft computing alongside neural networks and evolutionary methods. Its remaining strength is interpretability: a rule base like "if the error is large and positive then increase output strongly" can be read, audited, and edited by an engineer, which is attractive where opaque models are hard to certify. Gradual, graded truth also echoes in the soft scores that modern classifiers produce, although those are trained probabilities and not membership degrees.
Read fuzzy logic as a way to let words like "warm" do arithmetic: Zadeh's membership functions turn vagueness into numbers between 0 and 1, and the resulting rule systems produce smooth control actions, which is a statement about degree of truth, not about chance.