Task-Degradation Aware Adaptive Dynamic Scheduling for Priority-Based Automotive Cyber-Physical Systems

The advancement of autonomous driving technology presents significant challenges to automotive cyber-physical systems (ACPS) due to the dynamic and random nature of task releases. These systems classify tasks into different priority levels based on real-time and security requirements, necessitating heterogeneous computing platforms to meet diverse computational needs. Static scheduling methods often fall short in addressing these complexities effectively. A dynamic scheduling approach is provided for multi-priority ACPS, abstracting different tasks as directed acyclic graphs (DAGs). The proposed method establishes a mixed-integer linear programming (MILP) model to handle tasks with multi-priority levels, providing optimal scheduling solutions for dynamically released tasks. The MILP optimization results are then analyzed to identify near-optimal scheduling rules, which are used to design an adaptive dynamic scheduling approach suitable for dynamically and randomly released tasks. Simulation experiments with randomly generated task sets validate the effectiveness of the proposed task-degradation aware adaptive dynamic scheduling (TD_ADS) algorithm, which significantly reduces the Deadline Miss Rate (DMR) for multi-priority tasks. The TD_ADS algorithm demonstrates superior performance in minimizing DMR for high-priority tasks, while maintaining a lower overall DMR and achieving greater time efficiency across varying workloads and dynamic task releases.