A Study of TSK Inference Approaches for Control Problems

Abstract

Fuzzy inference systems provide a simple yet powerful solution to complex non-linear problems, which have been widely and successfully applied in the control field. The TSK-based fuzzy inference approaches, such as the convention TSK, interval type 2 (IT2) TSK and their extensions TSK+ and IT2 TSK+ approaches, are more convenient to be employed in the control field, as they directly produce crisp outputs. This paper systematically reviews those four TSK-based inference approaches, and evaluates them empirically by applying them to a well-known cart centering control problem. The experimental results confirm the power of TSK+ and IT2 TSK+ approaches in enhancing the inference using either dense or sparse rule bases.