Input-Mapping-Based Data-Driven Model Predictive Control for Unknown Linear Systems via Online Learning
Abstract
This paper develops a multistep input-mapping data-driven model predictive control scheme for unknown constrained linear systems. Instead of identifying the model online or relying on costly offline experiments, the method maps current and future inputs to previously measured online input-state trajectories and expresses future states as combinations of past states. A moving data window continuously refreshes the information used by the controller. Two computationally efficient robust MPC formulations are derived with feasibility and stability guarantees, and simulations demonstrate improved performance and computational efficiency.
Lingyi Yang
PhD
Dewei LI
Professor | Lab Leader
Currently serves as the Director of the Technical Committee on Predictive Control and Intelligent Decision of Chinese Association of Automation, Member of the Control Theory and Applications Education Working Group, and Editorial Member of Control Engineering Practice. He has led over 20 national-level projects, including Key Projects, General Projects and International Cooperation Projects from NSFC, Key Research and Development Programs from MOST. He has also undertaken more than 20 commissioned projects from key enterprises.
Aoyun MA
Assistant Professor
