Dynamic Layered Clustering Routing Protocol Based on Hybrid-Optimized Neural Networks for UWSNs

To address the limitations of traditional flat routing in large-scale underwater wireless sensor networks (UWSNs), and to tackle challenges, such as long delays, low bandwidth, and high error rates encountered by sensor nodes in underwater environments, this article proposes a dynamic layered clustering routing protocol based on a hybrid optimized backpropagation neural network (PSB-NN), referred to as PSBDR. The protocol introduces nonuniform vertical layering of the network based on water depth and assigns different communication and sensing radii to nodes according to their initial layer. A trained hierarchical prediction model dynamically adjusts the virtual layer assignments of nodes throughout the network’s operation. Additionally, a method is implemented for isolated nodes to join clusters via relay nodes, and a priority mechanism, along with a dynamic weight update strategy, is employed to select the optimal next hop during routing. This protocol effectively balances the workload across nodes at varying depths and maximizes the utilization of residual energy in the nodes, contributing to an extending network lifetime. Simulation results demonstrate that PSBDR enhances network performance by improving load balancing, prolonging network longevity, reducing end-to-end delay, and increasing data collection reliability.