A practical and useful self-learning fuzzy controller

Abstract

This paper describes the implementation and application tests of a simple and useful self-learning fuzzy controller. The proposed controller is able to perform a model-free control of time-variable plants with just a few prior information. Based on a simple and fast learning algorithm that emulates the human learning process, the fuzzy controller self-adapts its control surface at a rate of 150 updates per second. The present implementation was based on the inexpensive and standard 8-bit micro-controller Intel 8031, which makes the self-learning fuzzy controller very cost effective, even to small-scale embedded applications, like home appliances, for instance . Practical experiments using a servo-mechanical plant show the good performance of the intelligent fuzzy controller to regulate the plant output and to reject severe time-variable disturbances.