Q-Learning and Double Q-Learning Based Routing in Opportunistic Networks

Opportunistic Networks occur as a notable solution to address communication challenges in areas where cellular networks are overloaded and fixed infrastructure is unavailable. The critical issue in these networks lies in designing a routing protocol that can adapt to frequent network changes caused by node mobility. This research introduces two innovative routing protocols, namely “Reinforcement Learning based Opportunistic Networks (RLON)” and “Double Q-Learning based Opportunistic Networks (DQON),” which leverage Q-learning and Double Q-learning techniques, respectively, to achieve optimal selection of the next hop in Opportunistic Networks environment. Through extensive experimentations, the performance of RLON and DQON is evaluated against existing protocols such as Epidemic, PRoPHET, Spray and Wait, and RLProph. RLON and DQON exhibit significant performance gains, with improvements in Delivery Probability (4–23%), Overhead Ratio (12–35%), and Average Latency (10–32%) compared to the state-of-the-art approach, RLProph in diverse mobility scenarios.