Study on Obtained Advance Motion Forms of a Caterpillar Robot by using Reinforcement Learning

Abstract

A learning function is essential for human beings both to obtain and to adjust suitable motions. Especially, from a viewpoint of biological evaluation, observing process in acquisition of motion forms is a very interesting problem. However, there are few reports on acquisition of an actual robot motion form under applying learning algorithms. This paper clarifies how the actual robot obtains the motion form by use of the Reinforcement Learning method. These results show various motion form of the advance motion, the acquisition process of the motion form and the learning parameter effort on the motion form. Moreover, this paper clarifies environment effect on the motion form of the actual robot.