Model Identification for Fuzzy Dynamic Systems

This paper presents methods for identifying the structure of a fussy model and determination of a fussy ruleset which maps input data to output data of a dynamic system. The structure identification is based on the evaluation of `conflicting rules´ in fussy associative memory (FAM) cells, and the primary rule-set is obtained by using the adaptive fussy associative memory (AFAM) method. A neural-network-based error compensator is developed to improve the accuracy of the fussy model.