Automatic Generation of Fuzzy Rules for Control of Mobile Robot with Track Chassis Based on Numerical Data

In this paper, we consider the problem of generating a set of fuzzy rules for the Mamdani fuzzy inference system based on the numerical data obtained in the process of a managed system learning. The approach proposed in the article to solve this problem is based on the algorithms for clear and fuzzy clustering such as the mining clustering algorithm and the Gustafson - Kessel algorithm. It allows one to significantly simplify the process of forming a set of fuzzy rules and minimize the participation of a person in this process thus providing for an automatic selection of the number of rules as well as to determine all the necessary parameters for each of them. To implement the proposed approach, two computer programs were written. The first of them collects numeric data when a person manages a robot. Based on the collected data, this program builds a base of fuzzy rules for controlling mobile robot on a tracked chassis. This base of fuzzy rules and its computer implementation is further used in the second program for automated control of a mobile robot in the plane by varying the tractive force of each of the tracks depending on the position of the target to which the robot should approximate a given distance.