Ultralight Situation-Aware Hypervisor with Adaptive Starvation-Free Feedback Control for Embedded System

In recent years, microcontrollers are increasingly using hypervisors to achieve efficient and adaptable real-time performance in resource-constrained environments, enabling better resource management and system isolation. However, traditional hypervisor scheduling approaches often rely on static scheduling or require all scheduling decisions to be determined at compile time, limiting their ability to adapt to dynamic workload changes and real-time demands. To address these issues, this study introduces a scheduling technique that combines round-robin and priority scheduling. Rather than simply merging the two methods—which would also combine their disadvantages—the proposed approach resolves the fixed time-quantum issue of round-robin by dynamically adjusting the time-quantum based on priority for each round. In addition, to address the starvation problem in priority scheduling, a feedback control system is designed to maintain the average waiting time at an acceptable level. To mitigate the temporal overhead of determining time quanta, the design incorporates memory and functional isolation, significantly reducing context switching overhead. Our experimental results show a 54% reduction in context switching overhead compared to FreeRTOS, requiring only 44us for switching operations. In dynamic priority scenarios using computationally intensive dummy tasks to simulate real workloads, the system reduced the execution time by up to 38% for high priority conditions while appropriately deferring lower priority tasks. By improving resource utilization and responsiveness to external environmental changes, this approach enables robust and adaptive software execution, providing an effective solution for real-time embedded systems.