A Saturation based Interpolation Method for Fuzzy Systems

Authors:	Navarro Jose Luis, Universidad Politecnica de Valencia, Spain Ariño Carlos, Universidad Politecnica de Valencia, Spain Sala Antonio, Universidad Politecnica de Valencia, Spain Diez Jose Luis, Universidad Politecnica de Valencia, Spain
Topic:	3.2 Cognition and Control ( AI, Fuzzy, Neuro, Evolut.Comp.)
Session:	Intelligent Modelling and Identification II
Keywords:	interpolation, fuzzy modelling, linearization.

Abstract

This paper presents an alternative inference-defuzzification algorithm forTakagi-Sugeno fuzzy systems that preserves local-model interpretation and convexityproperties. The linear model in the rule consequent is saturated outside the core set of theantecedent membership functions. This allows the interpretation of the consequents offuzzy rules as a local linearization of the model restricted to the subset where it is valid.The setting has readability advantages over Takagi-Sugeno frameworks, and it is simplerthat other interpolation proposals. Some examples illustrate the approach