Energy-Efficient Routing in Wireless Sensor Networks Using Deep Belief Networks and LSTM for Mobile Sink Path Optimization and Cluster Head Selection

Wireless Sensor Networks (WSNs) are critical in numerous applications due to their ability to sense and transmit data. However, energy limitations of sensor nodes, powered by finite batteries, significantly impact network longevity. Traditional routing methods involving multi-hop transmissions and cluster formation can result in substantial energy consumption, particularly by Cluster Heads (CHs) involved in data aggregation and transmission. This research addresses the problem by optimizing energy-efficient routing using a Deep Belief Network (DBN) with Long Short-Term Memory (LSTM) for routing and CH selection. A mobile sink moving in a linear path minimizes energy consumption by reducing cluster formation and promoting single-hop transmissions. The proposed method utilizes LSTM-based CH selection to ensure that nodes with the highest residual energy are chosen, enhancing network lifetime. Experimental results demonstrate that the proposed method reduces energy consumption by up to 25% compared to circular path sink movement and multi-hop data transmission, resulting in a 40% increase in network lifetime. Performance was evaluated on a 100-node network with varying sink velocities, achieving an energy efficiency of 15% over traditional models.