Optimizing V2V Unicast Communication Transmission with Reinforcement Learning and Vehicle Clustering

Efficient routing algorithms based on vehicular ad hoc networks (VANETs) play an important role in emerging intelligent transportation systems. This highly dynamic topology faces a number of wireless communication service challenges. In this paper, we propose a protocol based on reinforcement learning and vehicle node clustering, the protocol is called Qucts, solve vehicle-to-fixed-destination or V2V messaging problems. Improve message delivery rates with minimal hops and latency, link stability is also taken into account. The agreement is divided into three levels, first cluster the vehicles, each cluster head broadcasts its own coordinates and speed, to get more cluster members. Also when a cluster member receives another cluster head broadcast message, the cluster head generates a list of surrounding clusters, find the best cluster to the destination as the next cluster during message passing. Second, the protocol constructs a Q-value table based on the state after clustering, used to participate in the selection of messaging clusters. Finally, we introduce parameters that express the stability of the vehicle within the cluster, for communication node selection. This protocol hierarchy makes Qucts an offline and online solution. In order to distinguish unstable nodes within a cluster, Coding of each road, will have vehicles with planned routes, For example, car hailing and public bus. Compare the overlap with other planned paths vehicles in the cluster, low overlap is labeled as unstable nodes. Vehicle path overlap rate without a planned path is set to the mean value. Comparing Qucts with existing routing protocols through simulation, Our proposed Qucts scheme provides large improvements in both data delivery rate and end-to-end delay reduction.