Error-free simplification of transparent Mamdani systems

Abstract

This paper shows that combinatorial complexity of fuzzy systems is at least in part caused by redundancy in these systems and presents the algorithm and its implementation for detection and removal of such redundancy for a special class of Mamdani systems. Performance of the simplification algorithm is demonstrated with uniformly impressive results on acknowledged benchmarks coming from different areas of engineering - truck backer-upper control, Mackey-Glass time series prediction and iris data classification.