Cloud-Edge Cooperative MPC With Event-Triggered Strategy for Large-Scale Complex Systems

Leveraging cloud computing for solving nonlinear Model Predictive Control (NMPC) can address issues with slow computational efficiency and effectively handle the complexity of large-scale complex systems (LSS), characterized by numerous variables, nonlinearities, and constraints. However, current studies often overlook critical aspects such as reliability, feasibility, stability, and resource efficiency in cloud-based NMPC, potentially limiting its application in LSS. To address these challenges, this paper explores a cloud-edge cooperative MPC architecture with an event-triggered strategy. The proposed architecture comprises high-fidelity Cloud NMPC, tube-based Edge LMPC, and a Switch Module with an event-triggered strategy, seamlessly combining abundant cloud computing resources with reliable edge computing. In case of Cloud NMPC failure, the tube-based LMPC at the edge layer promptly takes over control, ensuring both the reliability of Cloud NMPC and recursive feasibility under arbitrary switching sequences. By leveraging Lyapunov functions, the minimum stability modedependent dwell time (MDT) is pre-determined offline to guarantee exponential asymptotic stability. To strike a balance between control performance and resource efficiency, an eventtriggered strategy for Cloud NMPC is devised to prevent wastage of resources due to unnecessary communication and computation. Simulations on plug-in hybrid electric vehicles (PHEVs) validate the effectiveness of the theoretical results. The superiority of the proposed scheme is underscored through a comparison with four other MPC schemes.