V2V Routing in VANET Based on Fuzzy Logic and Reinforcement Learning

Abstract

To ensure the transmission quality of real-time communications on the road, the research of routing protocol is crucial to improve effectiveness of data transmission in Vehicular Ad Hoc Networks (VANETs). The existing work Q-Learning based routing algorithm, QLAODV, is studied and its problems, including slow convergence speed and low accuracy, are found. Hence, we propose a new routing algorithm FLHQRP by considering the characteristics of real-time communication in VANETs in the paper. The virtual grid is introduced to divide the vehicle network into clusters. The node’s centrality and mobility, and bandwidth efficiency are processed by the Fuzzy Logic system to select the most suitable cluster head (CH) with the stable communication links in the cluster. A new heuristic function is also proposed in FLHQRP algorithm. It takes cluster as the environment state of heuristic Q-learning, by considering the delay to guide the forwarding process of the CH. This can speed up the learning convergence, and reduce the impact of node density on the convergence speed and accuracy of Q-learning. The problem of QLAODV is solved in the proposed algorithm since the experimental results show that FLHQRP has many advantages on delivery rate, end-to-end delay, and average hops in different network scenarios.