Learn to produce fuzzy control rules with UCL

Abstract

Bart Kosko (1990) has proposed a method to produce fuzzy control rules using unsupervised competitive learning (UCL). But experiments show that some problems are waiting for further study. This paper discusses two of those problems. (1) How to partition the universe of discourse of the fuzzy variables in the rules. (2) How to determine the rules according to the clustering results. This paper presents two new methods to solve these two problems. Their effectiveness was proved by the simulation results from an inverted pendulum system.<>