Fuzzy control is a close-to-human-reasoning process from some inputs to an output. Take band credit score as an example. Rather than thinking a score below 550 is a bad credit but 551 is neutral, fuzzy computes the output based on how close the input is to both conditions.
Fuzzy logic is a close-to-human-reasoning process from some inputs to an output. Rather than thinking of y = c when a < x < b for an input x, an output y, and some constants a, b and c; fuzzy computes the output based on how close the input is to some conditions. Consider a similar setup to the system: fuzzy is driving a car toward certain destination. The raw input is distance to destination and car velocity.
To allow tuning of fuzzy gains, it is best to remap raw input between certain constraints. Say a car is 30m away from target, and fuzzy thinks 100m away is the largest viewable distance; then car has an error of 0.3, which is relatively small but still touches the definition of medium distance. This remapping is called fuzzification.
Before fuzzy starts running, it defines a series of logic rules. For instance, if car is far from destination, move quickly; otherwise, gradually slows down until car reaches the final point. Now, consider velocity of car as the second input. Think of how a person drives the car, if car is close to target but is moving quickly, merely slowing down is not enough. That is, car should try to brake hard. The amount of moving and braking speed are calculated from rules. To simplify the process, interpret fuzzy logic as a table. This table contains multiple fuzzy outputs, each of which corresponds to one combination of fuzzy input values.
Finally, let fuzzy output goes through defuzzification. Say a fuzzy input combination of 0.3 distance and 0.1 speed gives a fuzzy output of 0.2 (car wants to move slow but not stopping). Let maximum allowable speed of car be 10m/s, then defuzzification gives an output of 2 m/s.