FL/S IN CONTROL

FL is also emerging as a new paradigm for nonlinear modeling, especially to represent certain kinds of uncertainty with more rigor [9,10]. FL-based systems have the following features: (1) they are based on multivalued logic as against bivalued (crisp) logic, (2) they do not have any specific architecture like neural networks, (3) they are based on certain rules that need to be a priori specified, (4) FL is a machine-intelligent approach in which desired behavior can be specified by the rules through which an expert’s (design engineer’s) experience can be captured, (5) FL/S deals with approximate reasoning in uncertain situations where truth is a matter of degree, and (6) a fuzzy system is based on the computational mechanism (algorithm) with which decisions can be inferred in spite of incomplete knowledge. This is the process of the inference engine. FL-based control is suitable for multi­variable and nonlinear processes. The measured plant variables are first fuzzified. Then the inference engine is invoked. Finally, the results are defuzzified to convert the composite membership function of the output into a single crisp value. This specifies the desired control action. Heuristic fuzzy control does not require deep knowledge of the controlled process. Heuristic knowledge of the control policy should be known a priori.

There are several ways in which FL can be used to augment the flight-control system: (1) simulating pilot response to an aircraft that is not behaving as expected due to damage or failure, (2) incorporating complex nonlinear strategies based on the pilot’s or system design engineer’s inputs within the control law, (3) performing adaptive fuzzy gain scheduling (AGS) using the fuzzy relationships between the scheduling variables and controller parameters, and (4) in FL-based adaptive tuning of the Kalman filter for adaptive estimation/control.