Due to the lack of fully considering dynamic traffic information, in Vehicular Ad-hoc Networks (VANETs), existing routing schemes are easily stuck in local optimum, path redundancy, and congestion problems. By combining Software Defined Network (SDN) with VANET, the emerging Software Defined Vehicular Network (SDVN) can provide a global perspective of the traffic network to bridge the gap. In this paper, we introduce an intelligent fuzzy-based routing scheme for urban SDVN. First, a large urban area is divided into multiple sub-areas, in which each area is centered on an intersection. Second, the central controller maintains a routing table that records the priorities of packets be forwarded from an area to another, and all values in the routing table are initialized using Fuzzy Logic. Finally, we propose a hierarchical greedy routing with link stability (GLS) to calculate the routing path with the highest link stability according to the routing table. Meanwhile, considering the dynamic nature of vehicles in an area, Reinforcement Learning is employed to update the routing table during the routing process. Simulation results show that the proposed routing scheme achieved a significant improvement in performance over its counterparts.

由于缺乏充分考虑动态交通信息，在车辆自组织网络（VANET）中，现有的路由方案容易陷入局部最优，路径冗余和拥塞问题。通过将软件定义的网络（SDN）与VANET结合，新兴的软件定义的汽车网络（SDVN）可以为交通网络提供一个全局视角，以弥合差距。在本文中，我们介绍了一种智能的基于模糊的城市SDVN路由方案。首先，将一个大城市区域划分为多个子区域，其中每个区域都以一个十字路口为中心。其次，中央控制器维护一个路由表，该表记录了要从一个区域转发到另一个区域的数据包的优先级，并使用模糊逻辑初始化路由表中的所有值。最后，我们提出了具有链路稳定性（GLS）的分层贪婪路由，以根据路由表计算出具有最高链路稳定性的路由路径。同时，考虑到区域内车辆的动态特性，在路由过程中采用强化学习来更新路由表。仿真结果表明，所提出的路由方案在性能上比同类方案有了很大的提高。