Tracking dynamic deadlines in switched max-plus linear systems with uncontrollable workloads (Q7028068)

This article tackles the problem of tracking dynamically varying deadlines under uncontrollable workloads in systems modelled as switched max-plus-linear systems (SMPLS). Motivated by safety-critical cyber-physical applications such as medical imaging and autonomous vehicles, the authors propose an integrated control approach that combines state-feedback control with model predictive control (MPC) to minimize both deadline misses and implementation costs, such as energy consumption.\N\NThe work addresses a clear gap in the literature: most existing methods assume either continuous control spaces or fixed deadlines, assumptions that do not hold in many practical settings. By combining MPC with a finite prediction horizon and feedback gains computed via max-plus residuation theory, the proposed method enables the system to adapt dynamically to varying workloads and deadlines, while also accounting for switching costs between configurations.\N\NThe paper makes three key contributions. First, it formulates the control and scheduling problem for SMPLS with a finite set of discrete configurations and bounded model uncertainties. Second, it introduces a hybrid control framework: MPC handles workload prediction and subsystem selection, while state-feedback ensures robustness to uncertainties through just-in-time actuation policies derived from residuation theory. Third, the approach is validated on a representative image-processing pipeline, demonstrating the controller's ability to adapt to fluctuating demands while trading off energy consumption and performance.