Acquisition of exercise skill represented by fuzzy control rules for a rings gymnastic robot

Abstract

We have focused on rings that are one event of men's apparatus gymnastics. The ring exercises have the characteristics that the apparatus can move in all directions freely unlike horizontal bar exercises that have the constraint between hands and the horizontal bar. Realizing the robot that performs rings practically, we can expect a contribution to the coaching. Therefore we have already proposed the "rings gymnastic robot" that performs gripping the rings suspended by a rope. In this paper, our goal is to acquire an exercise skill represented by fuzzy control rules. A backward giant circle from handstand to handstand is selected as the exercise to simulate. The skill to realize the exercise is obtained by a genetic algorithm, in which an evaluation method using five subgroups with independent evaluation is introduced.