Closing the gap between stability and schedulability: A new task model for cyber-physical systems

A cyber-physical system (CPS) usually contains multiple control loops, each responsible for controlling different physical subprocesses, that run simultaneously upon a shared platform. The foremost design goal for CPSes is to guarantee system stability and control quality with limited cyber resources. We show, via an in-depth case study, that two inter-related design parameters - sampling period and consecutive control update misses - play a key role in determining stability and control performance. However, most CPS designs, such as control-schedule co-design and fault-tolerant scheduling, focus on either sampling period or control update misses alone, but not both. To remedy this problem, we propose a new CPS task model that captures both system stability and control performance in terms of sampling period and maximum allowable number of consecutive control update misses. To demonstrate the utility and power of this model, we develop two new scheduling mechanisms, offline parameter assignment and online state-aware scheduling. The former determines the sampling period and the maximum allowable number of consecutive job deadline misses for each task while preserving system stability. The latter then generates a schedule by exploiting the state of each physical subprocess to manage job deadline misses so as to improve the overall system performance without compromising system stability. Our in-depth evaluation results demonstrate that the proposed task model and the corresponding scheduling algorithm not only enable the efficient use of computing resource, but also significantly improve control performance without compromising system stability.