2A1-E19 強化学習を用いた自律移動ロボット群における衝突回避行動獲得

概要

論文の詳細を見る
This research proposes a collision avoidance system which is essential in a Distributed Multiple Robot System. In a real environment, it is impossible to apply a Distributed Multiple Robot System unless robots realize to avoid not only the static obstacles they recognise but also unknown obstacles and moving objects. To realize avobe, it is necessary for robots to plan pathes on demand. In this research, robots acquire their own policies in trial-and-error using machine learning for multiagent system dealing with uncertainty. In order to verify the effectiveness of the proposed system, simulator experiments are conducted in several distinctive initial settings. It is shown that some good results of the experiment are obtained up to the four robots meeting together.
一般社団法人日本機械学会の論文