Design of a simplified neuro-fuzzy system based on CMAC Architectures

Abstract

This work presents the design of a fuzzy controller with simplified architecture that use an artificial neural network working as the aggregation operator for several active fuzzy rules. The simplified architecture of the fuzzy controller is used to minimize the time processing used in the closed loop system operation, the basic procedures of fuzzification are simplified to maximum while all the inference procedures are computed in a private way. As consequence, this simplified architecture allows a fast and easy configuration of the simplified fuzzy controller. The structuring of the fuzzy rules that define the control actions is previously computed using an artificial neural network based on CMAC Cerebellar Model Articulation Controller. The operational limits are standardized and all the control actions are previously calculated and stored in memory. For applications, results and conclusions several configurations of this fuzzy controller are considered.