既存制御の操作量活用による強化学習のミニモデル検証

カテゴリ: 部門大会

論文No: TC13-1

グループ名: 【C】2023年電気学会電子・情報・システム部門大会

発行日: 2023/08/23

タイトル(英語): Mini Model Verification of Reinforcement Learning by Utilizing the Operation of Existing Controller

著者名: 秦 洋（東芝インフラシステムズ）,高野 俊也（東芝インフラシステムズ）

著者名(英語): Yang Qin (TOSHIBA Infrastructure Systems & Solutions Corporation),Toshiya Takano (TOSHIBA Infrastructure Systems & Solutions Corporation)

キーワード: 強化学習|PI制御|協調制御|モーター制御|自動走行|reinforcement learning|PI controller|collaborative control|motor control|autonomous driving

要約(日本語): Reinforcement learning (RL) methods have achieved impressive results on a wide range of control problems, like autonomous driving, industry automation. In previous work, we proposed a novel collaborative control method that combines the RL and the existing controller, e.g., Proportional Integral (PI) controller, to improve the training efficiency of RL. In this work, we aim to evaluate the proposal for improving high response motor control for a mini autonomous driving system. Evaluation of a simulated driving environment demonstrates that the RL achieves less training time and higher stability by utilizing the operation of PI controller. Furthermore, experiment on an actual robot driving system proves that the proposed method achieves a less control error compared with the PI controller.