Keep Forwarding Path Freshest in VANET via Applying Reinforcement Learning

In Vehicular Ad Hoc NETworks (VANET), dynamic topology changes of network and inconstant bandwidth make it hard to maintain an end-to-end path to complete long-time stable data transmission. Facing this challenge, researchers have proposed the hybrid routing approach, which tries to combine both the advantages of recalculating route when topology changes and looking up routing table as long as the network topology is relatively stable. However, the existing hybrid routing algorithms can easily cause the blind path problem, meaning a route entry is kept in the routing table without expiration according to the timeout mechanism but it is actually invalid, because the next hop is already unavailable. To address this issue, we propose a Reinforcement learning based Hybrid Routing algorithm (RHR) that can online track the available paths with their status and use packet-carry-on information as real-time feedback to guide routing. RHR keeps the forwarding path always the freshest and thus improves the system performance. Simulation results show that RHR achieves better performance in packet delivery ratio (PDR), roundtrip time (RTT) and overhead than other peers under different scenarios of network scale, request frequency and vehicle velocity.