強化学習による二次元移動ロボットの行動獲得とその学習知識の操作(機械力学,計測,自動制御)

概要

論文の詳細を見る
For motion leanings field, how to get a motion form defined by a repeated motion patterns is an exciting research issue. Although there have been so many studies reported in relation to evolutionary learning of robotic motion, there are few studies concerning with motion form. We reported that reinforcement learning method has ability to emerge the motion form of robots using a one-dimensional caterpillar-type mobile robot. This paper clarifies transit motion forms under learning process of a starfish-type robot with two-dimensional mobile function. Moreover, the learning results show both unexpected motion results that are essential function for the reinforcement learning and two-dimensional learning knowledge manipulation in the viewpoint of learning knowledge generality.
社団法人日本機械学会の論文
2009-01-25