Research On Routing Protocol In VANET

In recent years, with the development of new technologies, for example cloud computing and big data, the concept of smart city and intelligent transportation have approached to the view of human. Because Vehicle Network has unique charm and broad market prospects, it has attracted more and more attention. Therefore the research on data routing in VANET has become a hot research field. In this paper, we analyze Q-Learning algorithm, ant colony algorithm, fuzzy logic, as well as some existing routing algorithms. We improve methods to tackle the disadvantages of those algorithms, then prove the feasibility of the improved algorithms through simulation experiments.This paper studies the application of Q-Learning and ant colony algorithm in routing algorithm in vehicle ad hoc network. We point out the shortcomings of Q-ABR algorithm and improve its updating scheme of pheromone. Q-ABR algorithm does not consider the problem of routing loop and routing error handling mechanism. We redesign the algorithm flow and remodel it which will more suitable for vehicle ad hoc network. Secondly, this paper analyzes the AODV algorithm and points out the disadvantages of using hop as metric to find shortest path. Considering the characteristics of vehicular ad hoc network, we use fuzzy logic to estimate the quality of the link to improve Q-Learning. We overcome the fault that discount rate in traditional Q-Learning cannot change adaptively according to the actual situation. We use fuzzy logic to estimate the quality of the link which is used to replace the discount rate in Q-learning. We use Q-Learning algorithm to improve AODV algorithm and remodel it. Finally, we analyze the load balancing problem in vehicle ad hoc network and the shortcomings of load balancing algorithm AD-AODV. To combat the disadvantage of the link metric in AD-AODV algorithm, we put forward a new link cost function. At the same time, the AODV algorithm is improved in order to achieve load balance strategy. Simulation experiments show that the designed algorithms achieve better performance compared with other algorithms.