Dynamic mobile charger scheduling with partial charging strategy for WSNs using deep-Q-networks

Wireless sensor networks are a group of spatially distributed nodes deployed to sense, gather, and transmit data to the sink for further analytics. Due to continuous operations, the battery-equipped sensor nodes (SNs) drain energy rapidly, and replacing them is a hectic task. Wireless energy transfer (WET) is evolved as a promising innovation to recharge the SNs battery wirelessly to address the challenges. A WET is embedded in a vehicle called a mobile charger (MC) and traveled in the network to recharge the SNs. However, scheduling the mobile charger over the network before a sensor node dies is challenging. In this work, we introduced a partial charging strategy to avoid the long waiting time for MC because full recharging of a single node takes a long time. The partial charging strategy preempts the current charging node and moves to the newly requested node to minimize the network’s dead nodes. However, it will increase the traveling distance. Hence, adequate charging time and MC traveling path are required. In this context, this paper proposes a deep reinforcement learning-based mobile charger scheduling strategy called dynamic partial mobile charger scheduling using deep-Q-networks (DPMCS). The proposed DPMCS learns from the environment and decides each sensor’s charging duration in an identified tour. Experimental results reveal that the proposed DPMCS outperforms well compared to the existing studies, enhance the lifetime and diminish the dead nodes count.