【作者】 张琳娟；

【导师】 张思东；

【作者基本信息】 北京交通大学 ， 通信与信息系统， 2014， 博士

【摘要】 随着无线传感器网络应用在车辆领域中的延伸,车载传感网作为备受关注的新一代网络技术,在城市路况监测和交通异常检测等方面具有广阔的应用前景。鉴于交通异常事件极易引发二次事故和道路拥堵问题,如何利用车载传感网及时准确地检测出交通异常信息,并提供切实可行的信息发布和订阅服务,减轻异常事件对城市交通的恶劣影响成为推进智慧城市建设的核心。本文立足于车载传感网的交通异常信息检测与传输关键技术研究,以城市路况数据采集为基础,通过拥堵信息检测补充和完善异常信息的检测内容,并结合异常信息传输的不同需求,分别讨论面向多目标路段的地理广播技术和基于车车间通信的跨区域信息订阅问题,本文主要工作和贡献包括：(1)车辆节点因最小化能耗需求普遍存在自私性,导致现存路况数据采集协议可靠性不高,针对该问题,本文利用DTN路由框架和激励合作思想,在以公共车辆为主体的网络架构下提出了一种能量感知社交路由协议。该协议首先通过感知节点的剩余能量和速度信息对有限的复制令牌进行比例分配,不仅避免了拷贝资源的盲目扩散,还均衡了节点的整体能耗水平,间接实现了鼓励自私节点参与合作的目的；同时,该协议根据历史相遇节点的差异性来评估节点的社会活跃程度,并设计了基于社交关系能力的聚焦算法。实验结果表明,在接近真实条件的仿真场景下,该协议将数据成功交付给sink节点的概率要比SF协议高出约10%；而对于SW协议、EBR协议,该概率值可提升至20%。(2)针对目前异常信息检测内容的不完整性和单级信息融合方法的局限性,本文从综合优化的角度出发,通过整合特征级信息融合和决策级信息融合技术,提出了基于多级信息融合的道路拥堵信息检测机制。该机制首先设计了基于模糊分簇算法的消息聚合方法来剔除大量错误或冗余的原子消息；然后利用自定义的事件概率预测函数和消息可信度分配策略来筛选拥堵特征,并基于D-S证据理论提出了一种抗干扰拥堵判决方法,从而避免由交通信号灯等待产生的虚假拥堵特征。实验结果表明,该机制的平均消息聚合效率可达98%,虽然仅比RSMA方法高出2%,但它能够从相邻两车道间准确地提取出极其细微的拥堵特征。理论分析也证明,该机制能确保拥堵信息检测的一致性和准确性。(3)针对异常信息分发目标区域的不唯一性、地理广播路径的重复性以及广播对象的流动性问题,本文提出了一种智慧地理广播机制。该机制通过在十字路口和目标区域中分别部署”灯塔”和”浮标”等虚拟的地理标志物,并将其坐标信息封装在广播的消息包内,从而引导消息智慧地执行广播行为和选择信息传输方向。为了减少路由成本,消息先利用”灯塔”建立多播共享路径,并经过路径分裂到达多个目标区域的入口处。然后,消息在每个目标区域中基于”浮标”进行初始地理广播,并选择在距离目标路段出口最近的截面单元内,利用最优重广播时间预测方法来选拔重广播继任节点。实验结果表明,该机制不仅降低了消息重广播的总次数,还能在容忍的消息丢失率范围内最小化消息重复接收概率。(4)由于异常信息订阅节点和目标节点所在地理位置之间存在跨区域和RSU分布数量少且不均衡的特点,导致基于V2V通信的异常信息回复过程中的订阅成功率低且成本高。针对该问题,本文提出了一种基于车辆社区结构感知的机会路由协议。该协议首先设计了基于链路稳定性的配额消息复制策略,以解决消息拷贝资源有限且易丢失的问题：其次,利用车辆移动行为的社会性和规律性,提出了基于访问相似度的车辆社区构建方法,并结合”消息移动趋势”和”社交关系能力”两项参数定义设计了基于社区结构感知的消息转发算法。实验结果表明,在跨区域特点显著的场景下,该协议能够以较低的成本获得与Epidemic协议一样高的信息成功订阅概率,且比采用贪婪式复制策略的ProPHET协议要高出约8%。更多还原

【Abstract】 With the in-depth promotion of wireless sensor networks in the field of vehicles, ve-hicular sensor networks(VSNs) has attracted more attention as a new generation of network technology. Its application into the urban road condition monitoring, traffic abnormal event detection and other aspects would be hopefully promising. To solve the problem of traffic jam and secondary accident caused by the abnormal traffic event, it is essential to know the road condition accurately and timely, and provide the feasible traffic information release ser-vice and subscription service to mitigate the serious impact caused by the abnormal event on urban traffic through VSNs. This is very crucial in the construction of a smart city.The dissertation researches on the perspective of traffic abnormal event detection and information transmission protocols. Under the premise of discovering road congestion, it discusses the issues of multi-target geocasting and deals with the inter-region information subscription problem based on the collected data of urban road condition. The main contri-butions of this dissertation can be summarized as follows:(1) Due to the drivers’energy-saving need and the unfair energy consumption among nodes, the existing data collection mechanisms are unreliable. To address this issue, this dissertation proposes an energy-aware socially-based routing protocol based on the DTN routing and stimulation ideas, and designs a new data collection framework by taking the public transport vehicles as the main part. Firstly, this protocol prorates the limited replication tokens by sensing the residual energy and speed of vehicle nodes. It not only can avoid blind data spraying, but also equalize the whole energy consumption level. So, it indirectly achieves the purpose of encouraging selfish nodes to participate the data forwarding. Meanwhile, this protocol evaluate the social rela-tionship ability of nodes according to the difference of their history encounters, and designs a socially-based focus algorithm. Amount of simulation experiment results show that, under scenarios closing to the real environment, the probability of data suc-cessful delivery to the sink nodes of this protocol is about10%higher than SF, and20%higher than SW and EBR.(2) To deal with the incompleteness of abnormal information detection and the limitation of single-level information fusion, from the perspective of comprehensive optimiza-tion, this dissertation proposes a multi-level information fusion mechanism to detect the road congestion information, by combining the feature level information fusion with the decision-level information fusion. Firstly, this mechanism implements the fuzzy-clustering-based message aggregation method to remove the inaccurate and re-dundant atomic messages. Then, it selects congestion feature information utilizing the custom event prediction function and message credibility assignment strategy, and proposes an anti-jamming congestion decision method based on the Dempster-Shafer evidence theory, so as to avoid the false congestion evidences generated by the long-time traffic lights. The experimental results show that the average message aggregation ratio of this mechanism can reach to98%, which is only2%higher than RSMA, but it can extract more subtle congestion feature information from the neighbor lanes ac-curately. Also, the theoretical analysis shows that it can ensure the consistency and accuracy of event detection.(3) To address the non-uniqueness of the target regions where disseminating abnormal information, the repeatability of message transmission paths, and the mobility of mes-sage receiving nodes in each target region, this dissertation proposes a smart geo-casting protocol. It guides the message broadcast behaviors and controls the mes-sage transmission direction, by deploying some virtual landmarks like lighthouses and buoys at cross-road center and in each target region, and embedding their coor-dinates information in each broadcast message. To reduce the routing overhead, the message firstly builds the shared multicast path using the lighthouse. After the path splitting, the message arrives at the entrance of the target region. Then, the initial message broadcast is completed based on buoys in each target region, and the mes-sage rebroadcast maintenance mechanism is implemented by cutting a small unit area near the exit of the target region and predicting the next optimal rebroadcast time. The experimental results show that, this protocol can greatly reduce the total message rebroadcast times, and minimize the message repeatedly receiving probability while limiting the message missing probability in a tolerant range.(4) Due to the cross-regional feature between the subscribing node and the publishing node, the less and uneven distribution of RSUs, the abnormal information reply based on the vehicle-to-vehicle communication has the problems of a high transmission cost and low successful subscribing probability. To this, this dissertation proposes an op-portunistic routing protocol. It designs the quota-style message replication procedure based on the link stability, to avoid the limited number of message copies are easily aborted. Meanwhile, with the sociality and regularity of vehicles’mobile behaviors, this protocol proposes a method to construct the vehicle community structure based on the visiting similarity degree, and designs a community-aware message forward-ing algorithm by integrating "the message moving tendency" with "the social rela-tionship ability". The amount experiment results show that, in obvious cross-region scenario, the proposed protocol can reach the same high successful subscribing prob-ability like Epidemic with lower transmission cost, and its value is about8%higher than ProPHET.更多还原