1A1-O04 シェイピングを適用した協調箱押しタスクにおける強化学習ロボット群の協調行動の解析(進化・学習とロボティクス)

概要

論文の詳細を見る
Reinforcement Learning (RL) is one of promising approaches for controlling an autonomous robot. However, its performance is quite sensitive to the segmentation of state and action spaces. To deal with this problem, we have proposed one RL technique, named BRL. BRL has the feature of segmenting state and action spaces autonomously. In this paper, we investigate the performance of BRL through cooperative Box-Pushing task using four mobile robots. Moreover, for treating that tasks, we adopt Shaping which is the notion extensively applied in various fields. Shaping make robot learn efficiently while raising difficulty step by step. It is considered effective to deal with task, because robots can search with previous learning. In addition, the cooperative behavior which is acquired by using Shaping is analyzed.
一般社団法人日本機械学会の論文
2011-05-26