Efficient Opportunistic Routing in UASNs Using Dual-Time-Scale Learning: Fast LNB-2 and Slow Q-Learning

In underwater acoustic sensor networks (UASNs), traditional greedy forwarding is prone to routing void regions (RVRs) under time-varying acoustic links, intermittent connectivity, and unfavorable local topology, which degrades reliability and causes ineffective forwarding. To address this issue, this paper proposes LQ-T2OR, an efficient opportunistic routing protocol based on dual-time-scale learning. On the fast time scale, LNB-2 estimates the non-RVR forwarding capability of candidate relays using two-hop common-neighbor evidence. On the slow time scale, Q-learning optimizes long-term relay selection through a reward function that incorporates energy consumption, energy balance, forwarding distance, delay, and RVR risk. A distributed holding-time mechanism is further adopted to suppress redundant forwarding. Simulation results show that LQ-T2OR improves delivery reliability and routing robustness while maintaining competitive delay and energy-related performance.