Automatic Generation of Fuzzy Rules for the Control of a Mobile Robot

Abstract

This paper is aimed at looking into the generation of a fuzzy rule base in order to control a mobile robot in a navigation space. The procedure is based on a partitioning of the state space into subsets called cell groups. Membership functions are assigned for the state and controls. The transition from one conditional subspace to another is accomplished via a center point mapping of the cell groups, under the applied action of each IF-THEN rule. Two methods generating fuzzy rules are used. First, a manual method generating the rule base, in the case of rough quantification of variables, is proposed. This proposition is very complex and nonapplicable in the case of fine quantification. Next, we used an automatic method for generating the rule base: (i) the way transitional elements are clustered on the basis of an explicitly defined performance measure, (ii) optimum transitions are selected using the dynamic programming procedure, (iii) the fuzzy rules are generated automatically for transitioning from any initial cell to the target cell. Three improvements are proposed for the automatic method: the use of a fine quantification, the use of the polar coordinates, and the use of a first-order Sugeno fuzzy controller.