2A2-D20 強化学習による超冗長ロボットの障害物回避軌道の探索

概要

論文の詳細を見る
It is necessary for a robot manipulator to search a path, which can adapt to unknown and changeable environment. A method of path searching by using reinforcement learning is proposed. A method of multi-stage learning for obstacle avoidance is proposed, the first stage of which searches a path using 1-4th joints from the base of the robot, while other joints are fixed, on and after the second stage of which searches a path by adding movable joints two by two. The second search is carried out in the configuration space (C-space) near the first path, which reduces searched space and convergence time. A method is also proposed, in which many of paths corresponding to many kinds of obstacle layout are learned, and registered as the referential paths. The searched space is further narrowed by adjusting the path to the referential path by using a dynamic programming (DP) matching.
一般社団法人日本機械学会の論文
2009-05-25