Aperiodic Data-Driven Model Predictive Control With Feasibility and Stability Guarantees

This paper proposes an event-triggered data-driven model predictive control method with terminal ingredients for unknown linear time-invariant (LTI) systems under the input constraints. Specifically, the proposed method utilizes an implicit model description based on behavioral systems theory as well as input-output measurements, without requiring prior system identification steps. We explicitly design terminal ingredients: the terminal data-driven controller, terminal weight matrix and terminal region. To reduce computational costs, we develop an event-triggered scheme to activate the optimization problem only when necessary, rather than periodically. Moreover, a terminal data-driven controller is employed in the terminal region to avoid solving the optimization problem, further reducing computational requirements. We prove that the proposed method guarantees recursive feasibility and closed-loop stability under sufficiently small channel noise. Finally, the effectiveness of our method is verified by simulations.