A learning-based probabilistic routing protocol for vehicular delay tolerant networks

Existing routing protocols for vehicular delay tolerant networks (VDTNs) do not adequately address the multi-hop delivery probability in the forwarder node selection. We propose a probabilistic routing protocol for VDTNs. The protocol takes into account the vehicle velocity, node centrality, and node buffer size using a fuzzy logic-based approach. Multi-hop forwarding probability is also considered for the next hop node selection by employing a Q-learning algorithm which discounts the encounter probability with the increase of hops. We conduct extensive simulations to evaluate the proposed protocol in various scenarios and show the advantage of the protocol over existing well-known approaches.