V2V Routing in VANET Based on Heuristic Q-Learning
Xiaoying Yang, Wanli Zhang, Hongmei Lü, Liang Zhao
Abstract
Designing efficient routing algorithms in vehicular ad hoc networks (VANETs) plays an important role in the emerging intelligent transportation systems. In this paper, a routing algorithm based on the improved Q-learning is proposed for vehicle-to-vehicle (V2V) communications in VANETs. Firstly, a link maintenance time model is established, and the maintenance time is taken as an important parameter in the design of routing algorithm to ensure the reliability of each hop link. Aiming at the low efficiency and slow convergence of Q-learning, heuristic function and evaluation function are introduced to accelerate the update of Q-value of current optimal action, reduce unnecessary exploration, accelerate the convergence speed of Q-learning process and improve learning efficiency. The learning task is dispersed in each vehicle node in the new routing algorithm and it maintains the reliable routing path by periodically exchanging beacon information with surrounding nodes, guides the node’s forwarding action by combining the delay information between nodes to improve the efficiency of data forwarding. The performance of the algorithm is evaluated by NS2 simulator. The results show that the algorithm has a good effect on the package delivery rate and end-to-end delay.