【作者】 刘念伯；

【导师】 曾家智；

【作者基本信息】 电子科技大学 ， 计算机系统结构， 2010， 博士

【摘要】 在现代社会,面向网络的应用和服务已经成为日常生活的一部分,传统的有线网络已经无法满足人们对随时随地自由通信的需求,各种新兴的无线网络技术应运而生并得到了长足的发展。车用自组织网络就是这样一种新型的、与人们生活息息相关的无线网络。它是由装备了无线通信设备的车辆组建而成的网络,各个车辆节点间通过自组织的方式传递数据,也可以与各种基础设施节点交互,共同为驾驶者和乘客实现安全、通信等方面的应用和服务。与普通无线网络不同的是,车用自组织网络是自组织的,车辆节点既是网络的终端也是路由器,可以实现自组织和自管理的通信。同时,与一般的无线自组织网络不同,车用自组织网络又具有节点运动速度快、节点移动受道路拓扑约束、节点数量巨大等特点,呈现出独特的网络特性,也使车用自组织网络的研究更具挑战性。节点移动性是车用自组织网络的首要问题和基本特征,也构成了网络路由、数据传输以及各种应用服务的基础。然而车用自组织网络的节点移动性研究尽管有一些初步的研究成果,却没有完善的、综合性的车辆节点移动模型和描述,这就为节点移动性研究的深入研究留下了非常广阔的空间。本文在对车用自组织网络的节点移动性研究行了全面分析和总结的基础上,深入细致地研究了车辆节点的长时间停放和规律性移动在车用自组织网络的网络特性、数据传递、交通运输等方面的影响,取得了若干创新和成果。本文的主要贡献包括:1.针对车辆节点长时间停放的特点,提出了停放车辆协助(Parked Vehicle Assistance, PVA)的思想,利用停放车辆作为静态路边节点与移动车辆通信,共同实现车用自组织网络的信息收集,递交,处理和发布。在城市范围内,处于停放状态的车辆数目众多,分布广泛,且平均停放时间较长。这些停放车辆具有无线通信设备和高容量汽车电池,可以很容易地加入到车用自组织网络通信中来,并像静态路由器一样中继和转发数据包。在各个停车场集中停放的车辆可以类聚形成不同的停车场集群(Parking Lot Clusters, PLC),作为有相当容量和处理能力的数据中心和稳定的网络骨干。通过停放车辆和运动车辆的合作,可以建立基于停车场集群的分布式数据库系统,对车辆用户和外部用户提供全面的事件感知、信息收集、数据处理、消息发布和查询响应。因此基于停放车辆协助的车用自组织网络可以在无路边设施或路边设施极少的情况下改善原有的网络应用并实现一系列的新应用。停放车辆协助发现了通信的闲置资源,自底向上地增强了车用自组织网络,也拓展了研究的深度和广度。调查和实验结果显示停放车辆协助对典型的车用自组织网络单播和广播路由均有较大的性能提高。2.针对车辆节点规律性移动的特点,提出了行程历史模型(Trip History Model, THM)的思想,利用车载设备记录车辆的移动行程,通过对行程历史的学习自动预测车辆在驾驶过程中的目的地,使车辆节点具备自身移动性知识。目前车用自组织网络中常用的随机移动模型是对大量车辆运动的简单抽象,不具备任何节点特定的移动性知识。而行程历史模型从个人和家庭日常活动中的规律性出发,对单个车辆节点自身的移动特性建立模型,根据车辆移动的时间和空间特征采用决策树等机器学习方法做出移动预测。同时,带有时间特征的行程历史数据弥补了轨迹历史数据无时间度量的缺陷,进一步提高了预测精度。调查和实验结果显示行程历史模型能较准确地预测个人车辆在驾驶过程中的目的地。3.针对交通运输的低效率问题,提出了车辆到乘客通信(Vehicle-to-Passenger communication, V2P)的思想,在移动车辆和路边乘客之间实现直接、快速和灵活的车用自组织网络通信,提高客运和货运服务的效率。装备了手持无线通信设备的乘客可以作为路边节点加入车用自组织网络,进行车辆呼叫(vehicle calling),发布其交通需求(如共乘、叫出租车等)。这些呼叫请求在网络中扩散并寻找合适的车辆节点,一旦找到,驾驶者可以决定是否回应请求并提供相应的运输服务。通过行程历史模型进行移动目的地预测,可以实现车辆移动和乘客需求之间的透明匹配。因此,车辆到乘客通信可以实现开放和高效的交通信息交互,提高交通运输效率。实验结果显示车辆呼叫可以实现高效率的临时共乘和出租车呼叫服务。更多还原

【Abstract】 In modern society, network-oriented services and applications have been an essential part of our life. Traditional wired network can not satisfy the people’s requirement for establishing communication, anytime and anywhere, which launches and promotes the development of wireless network technology. Vehicular ad hoc network (VANET) is such a rising wireless network tightly related to people’s daily life. It is a network which consists of vehicle nodes that equipped with wireless communication device, and the nodes in a network of this kind can deliver packet data among each other in an ad hoc manner, can interact with various infrastructure nodes, and also can collectively provide safety and communication services and applications for drivers and passengers. Different from other kinds of wireless networks, VANET belongs to self-organizing, self-managing ad hoc network where vehicle nodes both serve as routers and hosts. Comparing to common ad hoc networks, VANET has some special characteristics, such as high-speed node movement, node movement relaying on road topologies, vast node numbers and so on, which show particular network features and make VANET researches very challenging. Node mobility is the prime problem and key feature in VANET, and constructs the basis of vehicular routing, data delivering, services and applications. Although there are some preliminary research results about vehicular node mobility, no perfect, integrative mobility model or description is found. Thus, there is a vast space for further study in vehicular node mobility researches.Based on a systematical summary of relevant mobility works on VANET, this dissertation focuses on vehicular network characteristics, data delivery, transportation and other aspects affected by long-time parking and regular movement of vehicles, and gains several achievements on some sub-topics. The major contributions of this dissertation are as below:1. Based on long-time parking of vehicles, we propose the idea of Parked Vehicle Assistance (PVA) in VANET, which exploits parked vehicles to communicate with moving vehicles as roadside infrastructure nodes for collecting, delivering, processing, and distributing vehicular information. In urban areas, parked vehicles have extensive numbers, wide distributions and long average parking time. These parked vehicles have wireless device and high volume battery, and can easily participate in vehicular communications and serve as stationary routers to store and forward data packets. Moreover, as naturally assembled communities, the parked vehicles at different parking lots can be clustered into Parking Lot Clusters (PLC) as stable backbones and data centers with considerable storage and processing abilities. After integrating the parked vehicles and the moving ones, we establish PLC-based distributed database systems to sense event, collect information, process data, disseminate message and response query for vehicle users and outside users. It makes PVA-based vehicular networks can improve current vehicular applications and afford a wide spectrum of new applications without or only with a few infrastructure nodes. PVA discovers neglected resources for vehicular communication, enhances vehicular network from down to top, and promotes vehicular research in depth and scope. Survey and simulation results indicate that PVA has good performance improvement on typical vehicular unicast and broadcast routing schemes.2. Based on regular movement of vehicles, we propose the idea of Trip History Model (THM) in VANET, which records vehicle’s trips with onboard device, predicts vehicle moving destination in driving through trip history learning automatically, and then enables vehicle to have the knowledge of self mobility. Current random mobility models in VANET come from the abstraction of substantive vehicle moving, and do not have any node-specific mobility knowledge. THM utilizes the regularities in individual or household daily activities, establishes model on single vehicle’s mobility characteristics, and make mobility prediction through machine learning methods as decision trees according to temporal and spatial features in vehicle movement. At the same time, trip history data with temporal scales remedies the limitation of lacking temporal features in trajectory history data, and greatly promotes the prediction accuracy. Survey and simulation results indicate that THM achieves accurate prediction in vehicle moving.3. Aiming at inefficiency in vehicle transportation, we propose the idea of Vehicle-to-Passenger communication (V2P), which achieves direct, fast and flexible interactions between moving vehicles and roadside passengers over VANET and promotes the efficiency in passenger and goods transportation. Passengers equipped with packet wireless device can easily join VANET as roadside nodes, launch vehicle calling queries and distribute their travel demands (e.g., taking a free-ride, calling a taxi). Once matched vehicles are found through the opportunistically disseminated queries, the drivers can decide whether to provide corresponding services, especially the carrying of passengers and goods. Through mobility prediction based on trip history model, V2P supports transparent matching between vehicles and passenger demands. Thus, V2P over VANET enables open and efficient transportation information interacting, and benefits from the relevant efficiency improvement. simulation results indicate that vehicle calling achieves good performance in casual carpooling and taxi-calling.更多还原