2-203 弾性ネットワークと運動制御

概要

論文の詳細を見る
This paper describes the research on movement control of an elastic network. An elastic network is a robot which has arranged the spring of a shape memory alloy (SMA) in the shape of a network in the shape of a lattice. reinforcement learning is used for control of movement in this research. reinforcement learning of a discrete value, and continuous value reinforcement learning is compared and each feature is clarified. As a result, although based also on the degree of a sampling, although re-learning is quick, optimal movement control cannot be performed in reinforcement learning of a discrete value. However, in reinforcement learning of a continuous value, although it was easy to obtain the optimal solution compared with reinforcement learning of a discrete value, it turns out that re-learning is late
一般社団法人日本機械学会の論文
2002-11-14