| quotation: | | [Copy] |
| | Rui Gu1,2,3,et al.[en_title][J].Control Theory and Technology,2026,24(2):307~315.[Copy] |
|
| |
|
|
| This Paper:Browse 34 Download 0 |
码上扫一扫！ |
| Online data-driven MPC for PWA systems with unknown parameters |
| RuiGu1,2,3,AoyunMa1,2,3,4,DeweiLi1,2,3,YunwenXu1,2,3,ShaoyingHe1,2,3 |
|
|
| (School of Automation and Intelligent Sensing, Shanghai Jiao
Tong University, Shanghai, China
Key Laboratory of System Control and Information
Processing, Ministry of Education, Shanghai, China
Shanghai Key Laboratory of Perception and Control in
Industrial Network Systems, Shanghai 200240, China;State Key Laboratory of Submarine Geoscience, School of
Automation and Intelligent Sensing, Shanghai Jiao Tong
University, Shanghai 200240, China) |
|
| 摘要: |
| A data-driven model predictive control (MPC) algorithm based on the input-mapping method is proposed for piecewise affine
(PWA) systems. These systems are characterized by unknown but constant parameters and are subject to disturbances, as
well as state and input constraints. To support the control strategy, an offline algorithm is developed to compute a non-convex
robust positively invariant set that serves as the terminal set within the MPC framework tailored for PWA systems. The
online MPC algorithm directly maps the future control input and predicted state to the historical input-state data associated
with the corresponding state subregion. This mapping process leverages the more accurate relationships contained in the
historical input-state data to enhance the prediction accuracy of future states. A state-dependent weight embedded in the cost
function enables the controller to balance prediction accuracy against convergence speed, enhancing overall performance.
Moreover, conditions ensuring the recursive feasibility of the optimization problem and stability of the closed-loop system
are established. The effectiveness of the proposed algorithm is demonstrated through a numerical example, which highlights
its ability to handle complex system dynamics and constraints while maintaining robust performance. |
| 关键词: Data-driven model predictive control · Input-mapping method · Piecewise affine systems · Unknown parameters |
| DOI：https://doi.org/10.1007/s11768-025-00307-0 |
|
| 基金项目:This work was partially supported by the National Key Research and
Development Project (No. 2024YFB4105200), the National Science
Foundation of China (Nos. 62573284, 62333015, 62261160385), the
Science Foundation of Shanghai (No. 24ZR1438800) and the China
Postdoctoral Science Foundation (No. 2025M771696). |
|
| Online data-driven MPC for PWA systems with unknown parameters |
| Rui Gu1,2,3,Aoyun Ma1,2,3,4,Dewei Li1,2,3,Yunwen Xu1,2,3,Shaoying He1,2,3 |
| (School of Automation and Intelligent Sensing, Shanghai Jiao
Tong University, Shanghai, China
Key Laboratory of System Control and Information
Processing, Ministry of Education, Shanghai, China
Shanghai Key Laboratory of Perception and Control in
Industrial Network Systems, Shanghai 200240, China;State Key Laboratory of Submarine Geoscience, School of
Automation and Intelligent Sensing, Shanghai Jiao Tong
University, Shanghai 200240, China) |
| Abstract: |
| A data-driven model predictive control (MPC) algorithm based on the input-mapping method is proposed for piecewise affine
(PWA) systems. These systems are characterized by unknown but constant parameters and are subject to disturbances, as
well as state and input constraints. To support the control strategy, an offline algorithm is developed to compute a non-convex
robust positively invariant set that serves as the terminal set within the MPC framework tailored for PWA systems. The
online MPC algorithm directly maps the future control input and predicted state to the historical input-state data associated
with the corresponding state subregion. This mapping process leverages the more accurate relationships contained in the
historical input-state data to enhance the prediction accuracy of future states. A state-dependent weight embedded in the cost
function enables the controller to balance prediction accuracy against convergence speed, enhancing overall performance.
Moreover, conditions ensuring the recursive feasibility of the optimization problem and stability of the closed-loop system
are established. The effectiveness of the proposed algorithm is demonstrated through a numerical example, which highlights
its ability to handle complex system dynamics and constraints while maintaining robust performance. |
| Key words: Data-driven model predictive control · Input-mapping method · Piecewise affine systems · Unknown parameters |
|
|
|
|
|
