Reinforcement Learning-Based Technique to Restore Coverage Holes with Minimal Coverage Overlap in Wireless Sensor Networks

Coverage holes are the anomalies that can disrupt the coverage and connectivity of a wireless sensor network. It is imperative to equip the sensor nodes with energy-efficient hole detection and restoration mechanism. Existing research works either introduce a new node in the network or use the existing active nodes to recover the coverage loss. The addition of new nodes in the network, after the occurrence of a coverage hole, is not feasible if the area of interest is at a hostile location. The relocation or sensing range customization of active nodes not only results in a constantly changing network topology but also risks the generation of new coverage holes as well as increases the coverage overlapping. Current work presents three algorithms viz., minimal overlapping and zero holes coverage (MO_ZHC), predictable and non-predictable holes recovery scheme (PNP_HRS), and a game theory-based reinforcement learning (GT_RL) algorithm. During the random deployment, the nodes use MO_ZHC to achieve minimal coverage overlapping in the network. After the scheduling round, PNP_HRS utilizes the sleeping nodes to restore the coverage lost, due to the holes. The active nodes are not displaced from their location, but they learn using GT_RL, to select and wake up a sleeping node which can recover the coverage loss in the most energy-efficient manner. The proposed algorithms ensure that the mobility of the nodes is kept minimal for judicious utilization of limited energy resources. The simulation results prove the efficacy of the present approach over the previous research works.