DDQN-Based Hybrid Routing Protocol for UWSNs With Void Repair Mechanism

To address the challenges of low delivery rate in proactive routing protocols and high forwarding delay in opportunistic routing (OR) protocols caused by network sparsity, node mobility, and localization difficulties in dynamic underwater sensor networks (UWSNs), this study proposes a hybrid routing protocol based on double deep Q network (DDQN), termed HR-DDQN. This protocol prioritizes proactive forwarding to reduce holding delay and energy consumption and employs opportunistic forwarding for path repair in case of failure. To improve the success rate of proactive forwarding, the proposed protocol uses DDQN to train the next-hop routing strategy, considering factors such as neighbor distance, speed, depth, and energy to avoid selecting unreliable forwarding targets. The protocol leverages the advantages of nodes’ active movement and the favorable hydroacoustic propagation properties of the sound channel axis to design a void repair mechanism. When even opportunistic forwarding is insufficient to repair the forwarding link, nodes actively adjust their depth for connectivity repair, thereby improving the route delivery rate in sparse networks. Simulations on NS3 demonstrate that the proposed protocol significantly optimizes packet delivery rates in sparse networks and has lower forwarding delay compared to OR protocols.