面向城市环境的车载网络瞬时拓扑与路由关键技术研究

发布时间：2018-03-09 05:03

本文选题：车载网络　切入点：节点移动性　出处：《国防科学技术大学》2014年硕士论文　论文类型：学位论文

【摘要】：车载自组织网络(Vehicular Ad Hoc Networks)简称VANET。VANET被认为在将来的智能交通系统的重要组成部分,被认为通过整合ad hoc网络、WLAN和蜂窝技术能够提升交通安全和交通效率。在车载网络中,装备有专用短程通讯设备的车辆可以直接或间接地与其他车辆通信,也可以与路边设施交换信息,基于此,车载网络可以提供很多服务,比如:车辆碰撞预警、实时交通情况报告等等。对于大多数基于车载网络的应用而言,端到端的数据通信传输的延迟时间和投递率是制约应用性能的主要瓶颈,这依赖于车辆的移动性特征,特别是车辆高速移动造成的网络拓扑动态变化,传统的对于网络拓扑的研究主要基于理论模型,但理论模型依赖于很强的前提假设,这与现实情况完全不符。随着近年来大规模车辆实验的部署,为现实条件下车载网络拓扑的研究提供了可能。本文基于旧金山和深圳的出租车轨迹数据集,对城市环境下车载网络节点移动性和网络拓扑进行分析,并提出了路由策略。我们首先对数据集和数据特征做了简单介绍,并给出了节点移动性特征的重要指标:联系间隔时间、剩余联系间隔时间及节点的度中心性。我们发现联系间隔时间的分布大致符合指数分布特征。接着我们给出了节点度的分布特征和两跳邻居节点的度的一些特征,最后给出了等待时间降低比例(WTR)和好点比例(GNR)的定义,并利检验了三种简单策略下的WTR和GNR,分析了节点的不确定性对车辆传输延迟的影响。接着我们将网络抽象成为一个带有时间戳的无向图,利用数据集来探索了车辆连通分支的时空变化特征,我们发现整个网络由大量的连通分支(Connected Component,CC)构成,绝大多数CC中所包含的节点数目非常小,但是最大的CC包含了网络中相当大一部分节点,甚至可以达到80%,而且在交通高峰和低潮时段最大CC的车辆数目变化并不明显,进一步,我们提出了连通分支稳定性的概念,来衡量节点成员的变化程度,发现当半径足够大的时候最大连通分支表现相当稳定,然后我们提出了连通分支位置独立性的概念,找到了最大连通分支的覆盖范围,为静态节点的部署提供了依据。最后我们引入了复杂网络中的经典算法k-核分解法,来找到了最大连通分支的k-核。最后我们设计了最先相遇K策略(FMK)和基于剩余相遇间隔时间的路由策略(RICT),并对路由策略的性能进行测试,同时我们得出了端到端的数据传输与S-D节点对的初始地理位置是独立的,比较发现RICT有着较好的性能。
[Abstract]:Vehicular Ad Hoc Networks (VANET.VANET) is regarded as an important component of the future intelligent transportation system. It is believed that the integration of ad hoc network and cellular technology can improve traffic safety and traffic efficiency. Vehicles equipped with special short-range communication equipment can communicate directly or indirectly with other vehicles, or can exchange information with roadside facilities. Based on this, vehicle-borne networks can provide many services, such as vehicle collision warning. For most applications based on vehicle network, the delay time and delivery rate of end-to-end data communication transmission are the main bottleneck of application performance, which depends on the mobility of vehicle. Especially the dynamic change of network topology caused by the high-speed movement of vehicle, the traditional research of network topology is mainly based on the theoretical model, but the theoretical model depends on a strong premise hypothesis. With the deployment of large scale vehicle experiments in recent years, it is possible to study the topology of vehicular network in real conditions. This paper is based on the taxi track data set in San Francisco and Shenzhen. This paper analyzes the mobility and network topology of vehicular network nodes in urban environment, and proposes a routing strategy. Firstly, we briefly introduce the data sets and data features. An important index of node mobility is given: contact interval time, We find that the distribution of the connection interval approximately accords with the exponential distribution. Then we give the distribution characteristics of the node degree and the degree of the two-hop neighbor node. At last, the author gives the definitions of waiting time reduction ratio (WTR) and good point ratio (GNR). We also test the WTR and GNR under three simple strategies and analyze the effect of node uncertainty on the vehicle transmission delay. Then we abstract the network into an undirected graph with a timestamp. Using the data set to explore the spatial and temporal characteristics of the connected branches of vehicles, we find that the whole network is composed of a large number of connected components CCs, and most of the nodes contained in CC are very small. However, the largest CC contains a large number of nodes in the network, and can even reach 80 percent. Moreover, the maximum CC number of vehicles in the rush and low tide periods does not change significantly. Further, we put forward the concept of stability of connected branches. When the radius is large enough, the maximum connected branch is quite stable. Then we put forward the concept of the location independence of the connected branch, and find the coverage of the maximum connected branch. Finally, we introduce the classical algorithm of k- kernel decomposition in complex networks. Finally, we design the first encounter K policy FMK) and the routing strategy based on the remaining encounter interval time, and test the performance of the routing policy. At the same time, we get that the end-to-end data transmission is independent of the initial geographic location of S-D node pairs, and it is found that RICT has better performance.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495;U463.67;TN929.5

【参考文献】