A Study on Behavior Acquisition by Deep Reinforcement Learning in Multi-Robot Environment

カテゴリ: 部門大会

論文No: SS3-3

グループ名: 【C】2019年電気学会電子・情報・システム部門大会プログラム

発行日: 2019/08/28

タイトル(英語): A Study on Behavior Acquisition by Deep Reinforcement Learning in Multi-Robot Environment

著者名: Watanuki Ryoma(National Institute of Technology, Matsue College),Horiuchi Tadashi(National Institute of Technology, Matsue College),Aodai Toshiyuki(National Institute of Technology, Matsue College)

著者名(英語): Ryoma Watanuki|Tadashi Horiuchi|Toshiyuki Aodai

要約(日本語): Deep Q-network (DQN) is one of the most famous methods of deep reinforcement learning. DQN approximates the action-value function using convolutional neural network (CNN) and updates the weights of CNN using Q-learning framework. We have already realized that a single mobile robot acquired behavior to avoid walls and obstacles by using DQN. In this research, we apply deep reinforcement learning to multi-robot environment and we aim to realize that multiple mobile robots acquire behavior to avoid walls and other robots using distance sensors and images of camera mounted on them. This task in multi-robot environment is more difficult than the task in a single robot environment, due to the increase of the state space and the influence by the actions of other robots.

PDFファイルサイズ: 459 Kバイト