Advanced Inference Filter Defuzzification

Abstract

In the field of modeling, fuzzy models are one efficient approach for representing technical systems or human control strategies. Fuzzy models have the advantage of supplying a transparent and interpretable model. Conventional fuzzy models are based on fuzzification, inference and defuzzification. The fuzzification and inference operations are theoretically well-established in the framework of fuzzy logic. In contrast, conventional defuzzification methods are essentially empirically motivated. First, we recapitulate the inference filter concept, which supplies a new understanding of the defuzzification process and a theoretical framework. Second, we extend this approach to the advanced inference filter concept, which leads to a defuzzification method that is better suited to imitate the behavior of a human expert.