Rule base reduction on a self-learning fuzzy controller
Ramón Ferreiro García, Xoán Carlos Pardo Martínez, J. Vidal Paz · European Society for Fuzzy Logic and Technology Conference · 1999
In this paper a methodology to develop a fuzzy rule-based controller is described. The rulebase is obtained by learning from a virtual feedback (PID) controller. The designed controller is supposed to support typical control requirements (i.e. fast-following, zero steady state error, overshoot suppression, etc.) using only a small number of rules. The virtual controller is designed by manual adjusting with one of two alternative methods: simulation of a feedback controlled system using a estimated process model or by means of an on-line training phase.