Psched: An Efficient Task Scheduling Approach Considering Atomic Tasks in Self-Powered Systems

Energy harvesting technology is a promising approach to overcome the limitations of traditional battery-powered IoT devices, enabling the devices of longer lifetime and easier maintenance. However, the unstable nature of harvested energy can compromise system performance and energy efficiency, particularly for systems tasked with scheduling large atomic tasks that are more vulnerable to energy fluctuations. In this paper, we propose a heuristic scheduling algorithm, Psched, which is based on dual-period power prediction and takes the characteristics of atomic tasks into account. The prediction model achieves an 87% accuracy rate in predicting TV-RF signals' power. Additionally, we develop an optimal scheduling algorithm (OPT) to establish a performance baseline. Experimental evaluations show that proposed Psched reaches 93% of the OPT's performance in terms of the weighted number of successful tasks, which is 12%–23% higher than that of the other algorithms. Moreover, the proposed Psched also performs better than other heuristic algorithms in terms of task success rate and energy efficiency.