A Q-Learning Based Scheme for Neighbor Discovery and Power Control in Marine Opportunistic Networks

Opportunistic networks, as an emerging ad hoc networking technology, the sparse distribution of nodes poses significant challenges to data transmission. Additionally, unlike static nodes in traditional ad hoc networks that can replenish energy on demand, the inherent mobility of nodes further complicates energy management. Thus, selecting an energy-efficient neighbor discovery algorithm is critical. Passive listening conserves energy by continuously monitoring channel activity, but it fails to detect inactive neighboring nodes. Conversely, active probing discovers neighbors by broadcasting probe packets, which increases energy consumption and may lead to network congestion due to excessive probe traffic. As the primary communication nodes in the maritime environment, vessels exhibit high mobility, and networks in oceanic regions often operate as opportunistic networks. To address the challenge of limited energy in maritime opportunistic networks, this paper proposes a hybrid neighbor discovery method that combines both passive and active discovery mechanisms. The method optimizes passive listening duration and employs Q-learning for adaptive power control. Furthermore, a more suitable wireless communication model has been adopted. Simulation results demonstrate its effectiveness in enhancing neighbor discovery performance. Notably, the proposed scheme improves network throughput while achieving up to 29% energy savings at most during neighbor discovery.