Realization and validation of task switched model predictive control

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K04533
• 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: Nagoya University
• Principal Investigator: OKUDA HIROYUKI 名古屋大学, 工学研究科, 准教授 (90456690)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: モデル予測制御 / 自動運転 / ハイブリッドダイナミカルシステム

Outline of Final Research Achievements

We proposed and verified a system that automatically generates a wide variety of tasks in real time and performs multitasking in parallel and in series using a model predictive control framework. The system defines multiple MPC primitives, which are subproblems of the optimization problem that represent simple and small tasks. The complex tasks consist of combination of these primitives, and are synthesized in real time. Furthermore, by automatically generating intermediate MPCs that smoothly connect MPCs for composite tasks, it is possible to switch between multiple complex tasks while maintaining smoothness and continuous feasibility. A fast nonlinear MPC solver was also applied for the realization of real-time computability for practical use.

Free Research Field

制御応用

Academic Significance and Societal Importance of the Research Achievements

提案した手法は，自動運転の分野のみならず，複数の達成目標や制約条件が実時間に並列または直列に変化するような複雑なマルチタスクシステム一般を対象としており，ロボットやプラント制御など広い分野での応用が期待される．このような問題に対し，マルチタスク内で考慮すべき共通の要素を抽出，MPCプリミティブと定義することで，これらの組み合わせによる膨大な種類の自動制御を自動で設計・実行できるため，大幅にシステムの設計時間を短縮することができる．MPCの産業応用は幅広く，制約を考慮できる等，システムの安全・安心の向上にも寄与できるため，多様な分野で実用化が進めば大きな経済効果が期待できると考える．