基于车载传感网的交通异常信息检测与传输关键技术研究

发布时间：2018-05-05 07:42

本文选题：车载传感器网络 + 道路拥堵检测　；参考：《北京交通大学》2014年博士论文

【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%。
[Abstract]:With the extension of wireless sensor network application in the vehicle field, vehicle sensor network, as a new generation of network technology which has attracted much attention, has a broad application prospect in urban road condition monitoring and traffic anomaly detection. In view of the traffic anomalies are easily caused by two accidents and road congestion problems, how to use the vehicle sensor network and the vehicle sensor network is very easy. When the traffic anomaly information is detected accurately, the practical information release and subscription services are provided, and the abominable impact of abnormal events on urban traffic has become the core of the construction of the intelligent city.
This paper is based on the research on the key technology of traffic anomaly information detection and transmission in vehicle sensor network. Based on the data acquisition of urban road condition, the detection content of abnormal information is supplemented and perfected through congestion information detection, and the geographic broadcast technology and vehicle based vehicle based on multi target sections are discussed in combination with the different requirements of abnormal information transmission. The main work and contributions of the cross regional information subscription for workshop communication include:
(1) there is selfishness in vehicle nodes because of minimization of energy consumption, which leads to the low reliability of existing road data acquisition protocols. Aiming at this problem, this paper proposes an energy aware social routing protocol under the framework of DTN routing and incentive cooperation in the network architecture with public vehicles as the main body. The protocol first passes the perception section. The residual energy and velocity information of the point is proportional to the limited copy token, which not only avoids the blind diffusion of copy resources, but also balances the overall energy consumption level of the nodes, and indirectly realizes the purpose of encouraging the selfish nodes to participate in the cooperation. At the same time, the protocol evaluates the social activity of the nodes according to the differences of historical nodes. The experimental results show that the probability of the protocol to deliver the data to the sink node is about 10% higher than that of the SF protocol in the simulation scene near the real condition; and for the SW protocol and the EBR protocol, the probability value can be raised to 20%..
(2) in view of the incompleteness of the content of abnormal information detection and the limitation of the single level information fusion method, from the perspective of comprehensive optimization, this paper proposes a multi level information fusion based information detection mechanism for road congestion, based on the integration of feature level information fusion and decision level information fusion technology. A clustering algorithm is used to eliminate a large number of false or redundant atomic messages. Then, a custom event probability prediction function and message reliability allocation strategy are used to screen the congestion characteristics. Based on the D-S evidence theory, an anti-jamming congestion decision method is proposed, which avoids the falsehood generated by the traffic signal waiting. The experimental results show that the efficiency of the average message aggregation of the mechanism can reach 98%. Although it is only 2% higher than the RSMA method, it can accurately extract very subtle congestion features from adjacent two lanes. The theoretical analysis also proves that the mechanism can ensure consistency and accuracy of congestion information detection.
(3) in view of the ununiqueness of the target area for abnormal information distribution, the repeatability of the geographic broadcast path and the liquidity of the broadcast object, this paper presents a kind of intelligent geographic broadcasting mechanism. This mechanism, by deploying the "Lighthouse" and "buoy" in the crossroads and the target area, and the coordinates of its coordinates. The message is encapsulated in the broadcast message packet to guide the message to execute the broadcast behavior and select the information transmission direction. In order to reduce the routing cost, the message first uses the "Lighthouse" to establish the multicast sharing path, and through the path splitting to the entrance of the multiple target areas. The experimental results show that this mechanism not only reduces the total number of message re broadcast, but also minimizes the repetition of message receiving within the range of tolerance message loss rate. Probability.
(4) due to the small and unbalanced distribution of cross region and RSU distribution between the abnormal information subscription node and the location of the target node, the subscription success rate is low and the cost is high during the abnormal information recovery process based on V2V communication. This paper proposes an opportunity routing based on the perception of vehicle community structure. The protocol first designs a quota message replication strategy based on link stability to solve the problem of limited and easy missing message copy resources. Secondly, using the sociality and regularity of vehicle mobility, a method of building a vehicle community based on access similarity is proposed, combined with "message movement trend" and "social customs". The two parameter definition designs a message forwarding algorithm based on community structure perception. The experimental results show that the protocol can obtain a successful subscription probability as high as the Epidemic protocol at a lower cost, and is about 8%. higher than the ProPHET protocol using greedy replication strategy.

【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP274.2;TN929.5

【参考文献】