TD3法によるスライディングモード制御のパラメータ決定

カテゴリ: 論文誌(論文単位)

グループ名: 【C】電子・情報・システム部門

発行日: 2024/05/01

タイトル(英語): Determination of Parameters for Sliding Mode Control using TD3

著者名: 松木　俊貴（防衛大学校），赤峰　諒（大分県立大分工業高等学校），原　正佳（大分大学），高橋　将徳（大分大学）

著者名(英語): Toshitaka Matsuki (National Defense Academy of Japan), Makoto Akamine (Oita Industrial High School), Masayoshi Hara (Oita University), Masanori Takahashi (Oita University)

キーワード: スライディングモード制御，深層強化学習，TD3，ボール＆ビーム系　　sliding mode control，deep reinforcement learning，TD3，ball and beam systems

要約(英語): Reinforcement learning (RL) has been studied as an effective method for tasks without explicit training data. In recent years, it has been shown that deep RL (DRL), which introduces deep learning algorithms into RL, can be applied to more complex tasks. Recent studies have introduced DRL into sliding mode control (SMC), yielding important research results. However, the black box in the computational process of a neural network (NN) is a troublesome problem, especially when NNs treat control engineering. In this paper, we propose a novel approach to determine parameters for SMC using DRL. In this method, a policy model constructed to be equivalent to the equation of the SMC input is trained by a DRL method that can treat continuous action spaces. The results show that the proposed method can determine the SMC parameters that successfully control the ball beam system through numerical experiments.

本誌: 電気学会論文誌C（電子・情報・システム部門誌） Vol.144 No.5 （2024） 特集：医用・生体工学関連技術

本誌掲載ページ: 504-511 p

原稿種別: 論文／日本語

電子版へのリンク: https://www.jstage.jst.go.jp/article/ieejeiss/144/5/144_504/_article/-char/ja/