A Q-Learning Framework for Disaster Resilient Data Transfer in Underwater Wİreless Sensor Networks

Underwater Wireless Sensor Networks (UWSNs) play a critical role in real-time environmental monitoring during disasters such as earthquakes, tsunamis, and underwater landslides. However, highly dynamic and hazardous underwater conditions significantly reduce the effectiveness of conventional routing protocols. This study addresses data transmission challenges in disaster-prone UWSNs by proposing a Q-learning-based Adaptive Data Transfer Algorithm (QL-ADTA). The proposed approach employs a multi-objective reward function that dynamically adapts to contextual parameters including data priority, link quality, node survivability, debris density, and water current strength.The algorithm was evaluated using disaster-simulated synthetic datasets and compared with conventional routing protocols such as DSR and AODV. Experimental results demonstrate that QL-ADTA achieves an 18% improvement in packet delivery ratio, 23% reduction in latency, and 20% enhancement in energy efficiency. These findings confirm the model’s ability to adapt routing decisions in real time based on disaster severity, thereby improving communication reliability and extending network lifetime. The context-aware reward design and emergency-adaptive routing strategy distinguish the proposed method from conventional reinforcement learning approaches in UWSNs, offering a robust framework for disaster-resilient underwater communication.