Fuzzy modelling operator navigation behaviors

Abstract

The paper presents a fuzzy modelling method for static and dynamic systems and its application as a tool for learning navigation behaviors from operator's actions. The modelling process starts from a set of experimental input-output data, and automatically produces a set of fuzzy rules, whose number depends on the desired degree of model's accuracy. The method is based on least squares techniques, and can be used in association with the most common fuzzy operators and sets. The paper illustrates the application of the method to model the behaviors of a human expert operator driving a real mobile robot equipped with ultrasonic sensors. Learning the fuzzy behaviors makes possible the adaptation of the robot to the environment, and saves design time and efforts. The resulting rule bases can be used as building blocks in executing purposeful and robust autonomous navigational tasks.