Reinforcement Learning-Based Connectivity Restoration in an Ocean Network of Fishing Vessels

An Offshore Communication Network (OCN) is a network of fishing vessels at sea aimed at providing wireless Internet access over the ocean. The connectivity of fishing vessels is essential to disseminate messages, monitor emergency management, and provide information services. The impact of extreme weather conditions on wireless signals, the inability to deploy additional infrastructure, the movements induced by sea waves, the expanded mobility freedom at sea, and the non-uniform density of nodes create connectivity holes in an OCN. This paper proposes a reinforcement learning-based connectivity restoration scheme for OCNs. During the planning phase, a node examines the history of its contacts with other nodes and estimates their mobility vector to determine the expected location. Mobile nodes discover the best spots to re-establish connectivity and the most appropriate path to reach these spots via a trial-and-error strategy. In a reinforcement learning framework, we simulate actions to move toward the expected contact locations and learn the optimal movement directions without guessing the contact's actual position. During the control phase, these learned policies are utilized to relocate isolated OCN nodes and restore high-quality connectivity. Simulation results show that our scheme improves the nodes' connectivity probability.