Practical Study of Nonlinear System Identification in the Frequency Domain

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K04112
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 21040:Control and system engineering-related
• Research Institution: Keio University
• Principal Investigator: Adachi Shuichi 慶應義塾大学, 理工学部(矢上), 教授 (40222624)
• Co-Investigator(Kenkyū-buntansha): 丸田 一郎 京都大学, 工学研究科, 准教授 (20625511) ; 川口 貴弘 群馬大学, 大学院理工学府, 助教 (00869844)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 非線形システム同定 / 深層学習 / 周波数 / 制御 / モデル縮約

Outline of Final Research Achievements

There are nonlinear dynamic systems where the application of machine learning in AI has not been adequately explored. This study aims to propose a new model reduction method for the identification problem of the nonlinear dynamic systems. In this research, a method was proposed to construct deep neural networks (DNNs) capable of switching computational loads in a single learning process, and its effectiveness was confirmed through numerical examples. Additionally, new insights were gained by interpreting problems studied in machine learning within the framework of control theory.

Free Research Field

制御工学

Academic Significance and Societal Importance of the Research Achievements

現在活発に研究されているAIの分野の機械学習は，制御理論の分野では非線形システム同定に対応する．二つの分野の共通点が多いにも関わらず，それらの融合研究は進んでいない．本研究では，制御理論の視点から機械学習を考察することにより，さまざまな知見を得ることができた．また，申請者が長年研究を進めてきた，本研究に関連するシステム同定の著書をまとめており，その社会的意義は大きいと思われる.