Reinforcement Learning-Based Routing Protocol for Opportunistic Networks

This paper proposes a novel routing protocol for opportunistic networks called Fuzzy logic-based Q-Learning Routing Protocol (FQLRP), which uses fuzzy based Qlearning for efficient routing. The proposed protocol predicts the next optimal forwarder of a message based on a reward mechanism that considers the node's energy, movement, and buffer space as parameters. Throughout the routing process, the residual energy of each node and the energy distribution of a group of nodes, are both considered in determining a reward function, which in turn helps in deciding the most suitable forwarders of the message towards its destination. Simulation results show that the proposed FQLRP scheme outperforms the Q-Learning based routing and the Epidemic routing protocols, chosen as benchmarks, in terms of delivery rate, average delay and overhead ratio.