Model Predictive Control under Intermittent Measurements due to Computational Constraints: Feasibility, Stability, and Robustness

We propose a continuous-time model predictive control (MPC) strategy in the presence of intermittent sampling due to limited computational power. Using hybrid systems tools, the proposed scheme explicitly models the (not necessarily periodic) computation events associated with prediction and optimization. When the terminal cost is a control Lyapunov function and the implicit MPC control law in the proposed setting is continuous, the closed-loop system can tolerate disturbances, unmodeled dynamics, and measurement noise, as well as errors due to asynchronous actuation and sporadic data losses. These findings apply to a wide range of linear systems, and are particularly important for the target application area of cyber-physical systems, where real-time safety constraints require robustness in the presence of computational limitations.