Adaptive learning defuzzification techniques and applications

Abstract

In this paper, adaptive learning defuzzification techniques are studied under the consideration of system performance indices. By treating defuzzification processes as continuous mappings from space [0,1] n to the real line, the concept of the optimal defuzzification mapping can be developed. Since all the continuous defuzzification mappings considered in this paper form a Banach space, approximation to the optimal mapping with some known functions can be expressed as the parametric optimization problem. To find the optimal parameters, adaptive learning of the optimal defuzzification mapping is investigated. Learning laws for the parameters in the defuzzification mapping are derived in one case. Numerical results indicate that the adaptive learning defuzzification method can give superior defuzzification results to some popular defuzzification methods.