Hierarchical Modular Reinforcement Learning
(HMRL) consists of 2 layered learning where Profit-Sharing works to plan a target position in the higher layer and Q-learning trains the state-action pair to the target in
the lower layer. The method can divide a complex task into subtasks, and it reduces to state dimension and improves learning capability. In order to solve this problem, we propose the learning method based on Restricted Boltzmann Machine (RBM) with subspace divided by Profit Sharing. In this paper, to verify the effectiveness of the proposed method, the assignment problem of taxies was investigated.

開催日：平成27年7月18日
会場：広島市立大学