State-Aware Dynamic Frequency Selection Scheme for Energy-Harvesting Real-Time Systems

With the increasing deployment of battery-powered embedded systems such as sensor nodes in extreme environments, harvesting renewable energy from ambient environments to achieve near perpetual operation of a system has attracted considerable research efforts in the recent past. In this paper, the authors propose a dynamic frequency selection scheme for energy-harvesting real-time systems. The proposed scheme characterizes the state of a system from the perspectives of system utilization and harvested energy with respect to a certain period of time. A portion of the battery energy is allocated to a group of tasks in the period of time by jointly considering the system utilization and energy state, and the operating frequency is selected based on the allocated energy. The derived operating frequency is fine tuned to further enhance energy efficiency when overflow occurs. Simulation results demonstrate the effectiveness of the proposed scheme. Compared with the state-of-the-art scheme that decouples the energy and timing design constraints, the proposed scheme achieves comparable deadline miss rate when the battery capacity is lower than 5000 J and achieves about 11.5% lower deadline miss rate when the battery capacity is greater than 30 000 J. The proposed scheme also outperforms the benchmarking scheme in energy efficiency. When the battery is near a full charge or overflow occurs, the proposed scheme incurs less energy waste when compared with the benchmarking algorithm, which is favorable for autonomous operation of the system. Furthermore, the time complexity of the proposed scheme is one order of magnitude lower than that of the benchmarking scheme, which makes the proposed scheme well suited for dynamic scheduling.