Dynamic Adaptation of Resource Scheduling for Real-time Tasks in Flexible Manufacturing Systems

Flexible manufacturing systems increasingly face the need to execute both real-time tasks with hard deadline constraints and human-interactive tasks that arrive unpredictably at runtime. Such dynamic workloads challenge conventional static scheduling techniques, particularly in systems that must also achieve energy efficiency. This paper introduces a dynamic adaptive resource scheduling scheme (DARS) that jointly optimizes CPU clock frequency, DRAM power states, and task migration decisions. The proposed scheme employs evolutionary optimization to generate a set of resource configurations in advance, enabling immediate adaptation to changes in the task set. To support unpredictable human interactions, the proposed scheme reserves runtime slots during offline scheduling, ensuring responsiveness without violating real-time guarantees. Simulation results under diverse workload scenarios demonstrate that DARS achieves significant energy savings over existing scheduling methods, while consistently maintaining full deadline compliance and high CPU utilization. These findings suggest that DARS offers an effective and scalable solution for energy-aware real-time scheduling in dynamic manufacturing environments.