Adaptive Monitoring Optimization Based on Deep-Q-Network for Energy Harvesting Wireless Sensor Networks

In order to improve the energy efficiency of environmental monitoring for energy harvesting wireless sensor networks (EH-WSNs) in remote areas and achieve energy-neutral operation, an adaptive monitoring and energy management optimization method of EH-WSNs based on deep Q network (DQN) algorithm is proposed. In this paper, aiming at EH-WSNs with single-hop cluster structure, we first present a more realistic energy model established by combining different climate characteristics. Then, the optimization model of maximizing long-term monitoring utility is formulated based on harvested energy constraints. We use deep Q network (DQN) to learn random and dynamic solar energy harvesting process on solar-powered sensor nodes and optimize the monitored performance of EH-WSNs through the replay memory mechanism and freezing parameter mechanism. Finally, we present an adaptive monitoring optimization method based DQN to achieve the long-term utility. Through simulation verification and comparative analysis, in different rainy weather environments, the proposed optimization algorithm has greatly improved in terms of average monitoring reward, monitoring interruption rate and energy overflow rate. Moreover, it also indicates that the proposed algorithm has effective adaptation to the random and dynamic solar energy arrival.