An adaptive charging scheme for large-scale wireless rechargeable sensor networks inspired by deep Q-network

Nowadays, Wireless Rechargeable Sensor Networks utilize a Mobile Charger (MC) to prevent node failure by replenishing the sensor node’s energy. Existing studies primarily focus on small-size networks using a single-node charging method, lacking scalability for large-scale networks with diverse energy consumption rates. Additionally, previous charging algorithms are often sensitive to a pre-established charging request threshold, reducing flexibility in charging decisions. This study addresses the challenges by proposing an adaptive charging scheme for large-scale networks. First, we exploit the “multi-node charging” strategy, where the MC charges multiple sensors simultaneously to maximize charging utilities. We then model the charging problem as a Markov Decision Process and devise a Graph Neural Network-based representation method to reduce the state space’s dimension. Subsequently, the Deep Q-Network algorithm will determine the MC’s optimal charging policy, which automatically selects the next charging location in each round. Extensive experiments demonstrate our proposal’s efficiency, reducing approximately 51% of node failures compared to the most related works.