Scheduling for Cyber-Physical Systems with Heterogeneous Processing Units under Real-World Constraints

Justin McGowen, Ismet Dagli, Neil T. Dantam, Mehmet E. Belviranli

Abstract: Cyber-physical systems (CPS) such as robots and self-driving cars pose strict physical requirements to avoid failure. The scheduling choices impact these requirements. This presents a challenge: How do we find efficient schedules for CPS with heterogeneous processing units, such that the schedules are resource-bounded to meet the physical requirements? For example, tasks that require significant computation time in a self-driving car can delay reaction, decreasing available braking time. Heterogeneous computing systems — containing CPUs, GPUs, and other types of domain-specific accelerators — offer effective capabilities to reduce computation time or energy consumption and expand such operating conditions. However, doing so under physical requirements presents several challenges that existing scheduling solutions fail to address. We propose the creation of a structured system, the Constrained Autonomous Workload Scheduler (CAuWS). This structured and system-agnostic approach determines scheduling decisions with direct relations to the environment and differs from current ad hoc approaches which either lack heterogeneity, system generality, or this consideration of the physical world. By using a representation language (AuWL), timed Petri nets, and mixed-integer linear programming, CAuWS offers novel capabilities to represent and schedule many types of CPS workloads, real-world constraints, and optimization criteria, creating optimal assignment of heterogeneous processing units to tasks. We demonstrate the utility of CAuWS with a drone simulation under multiple physical constraints. The autonomous computation for the drone is made up of commonly used workloads (i.e., SLAM, and vision networks) and is run on a popular heterogeneous system-on-chip, NVIDIA Xavier AGX.