1P1-A04 Q-learningによる把持戦略の獲得(ロボットハンドの機構と把持戦略)

概要

論文の詳細を見る
We have been trying to develop an Assistive MObile robot System (AMOS). The aim of AMOS is to replace only simple assistive work that is to transport the daily use objects in a designated indoor location to reduce the burden of the care taker. In the system, at first an operator gives a command to the system what to bring, and then mobile unit approaches to an object, pick up and relocate the object semi autonomously. In this paper, we proposed a method using Reinforcement Learning to acquire grasp planning to pick up a daily life object. We assumed that pose and Voxel data of the object are estimated previously by vision components. And then state space is compressed to reduce calculation time, and the calculation is performed in off-line. Afterwards we performed on-line experiments.
一般社団法人日本機械学会の論文
2008-06-06