Control rule acquisition for an arm wrestling robot

摘要

Describes knowledge acquisition for an arm-wrestling virtual reality (VR) system. The ultimate goal of this research is to obtain control rules which give the impression that the operator is arm wrestling against a real person and to analyze what factors enhance this. Acquiring the rules to win games consists of two steps (the control rules are obtained in two stages because it is difficult to obtain rules directly for VR). Firstly, the rules necessary to win arm-wrestling games are obtained, and then they are modified into ones which enhance VR. This paper focuses on the former stage. We are evaluating if the obtained fuzzy rules are equivalent to the original 20,000 precise rules through a subjective test.