基于复杂网络的突发事件VMS信息发布特性研究

发布时间：2018-08-06 15:26

【摘要】：可变信息情报板(VMS)作为城市交通动态信息发布的重要方式,对缓解城市路网中由突发事件而引起的交通拥堵,发挥着非常重要的作用。由于交通流的随机性以及可能引起的信息集中或过剩等原因,使得VMS诱导效果并不显著,甚至造成负面影响。因此,有必要对VMS信息发布策略的特征进行研究,复杂网络方法为城市路网中这一研究提供了了一个新视角。本文定义了可以反映VMS显示屏之间相关关系的概念-VMS关联度,并应用复杂网络的相关知识,构建了的VMS关联性网络模型,提出了VMS关联网络的统计分析方法。以北京市VMS信息发布为背景,针对主要包括事故和故障车事件的突发事件,分别构建了早高峰和晚高峰时段的事故VMS关联网络和故障车VMS关联网络,通过研究VMS关联网络的性质得到了北京市事故和故障车VMS信息发布的空间特征规律,从而为VMS信息发布策略的制定与优化提供支持。首先,通过对事故网络和故障车网络的拓扑结构分析发现,二种网络均表现出了明显的小世界特性,即交通流的连续性使得各VMS显示屏之间的具有明显的关联性。事故信息的发布范围会受到交通流状况的影响,而故障车事件则相关性不大。虽然故障车网络的连通性远远不如事故网络,但其集团性明显高于事故信息网络,这与统计出的故障车信息的发布范围明显小于事故信息的发布范围是一致的。其次,论文统计了网络的加权指标,包括边权、点权和单位权,得到了事故和故障车的发生在地域上的偶然性很大,早晚高峰时段事故的严重程度地域分布差别比较大,存在受事故影响较为严重的区域的结论,并且发现事故网络的关键点早高峰更加分散,而晚高峰则相对集中；故障车在网络在早高峰和晚高峰的关键点分布都较为分散。之后,论文对网络的度、单位权和点权进行相关性分析,在此基础上对于不同的VMS信息显示屏,根据度、单位权、点权和聚类系数四个属性进行了聚类分析,将VMS划分为16个类别,每个类别分别对应一种VMS信息发布特性,并给出了早晚高峰时段事故和故障车网络中D、E类的显示屏分布图。最后,论文将VMS关联网络划分为派系,进行了社团结构挖掘,挖掘出了VMS信息发布的组团性质,并利用模块度对划分结果进行评估和优化。研究发现,事故网络和早高峰时段的故障车网络的模块度均小于0,在分析原因后论文对社团结构的划分进行了人工优化,优化后的网络模块度指标更加符合VMS关联网络属于混合网络这一事实。优化后的社团结构划分结果可作为VMS信息发布的参考策略。
[Abstract]:Variable information board (VMS), as an important way to issue urban traffic dynamic information, plays a very important role in alleviating traffic congestion caused by unexpected events in urban road network. Due to the randomness of traffic flow and the possible information concentration or excess, the induction effect of VMS is not significant or even negative. Therefore, it is necessary to study the characteristics of VMS information release strategy, and the complex network method provides a new perspective for the study of urban road network. This paper defines the concept of -VMS correlation degree which can reflect the correlation relationship between VMS display screens, and applies the related knowledge of complex network to construct the VMS correlation network model, and puts forward the statistical analysis method of VMS correlation network. Based on the information release of Beijing VMS, this paper constructs the accident VMS connection network and the fault vehicle VMS associated network during the morning rush hour and the late rush hour, respectively, in view of the emergency events which mainly include the accident and the malfunction vehicle event. By studying the properties of VMS related network, the spatial characteristics of VMS information release of Beijing accident and fault vehicles are obtained, which can support the formulation and optimization of VMS information release strategy. Firstly, by analyzing the topology of the accident network and the vehicle network, it is found that the two networks have obvious small-world characteristics, that is, the continuity of the traffic flow makes the VMS display screen have obvious correlation. The range of accident information is affected by the traffic flow, but the accident is not relevant. Although the connectivity of the fault vehicle network is far less than that of the accident network, its collectivity is obviously higher than that of the accident information network, which is consistent with the statistical distribution range of the fault vehicle information is obviously smaller than that of the accident information. Secondly, the paper statistics the weighted index of the network, including side weight, point weight and unit weight, and obtains that the accident and fault vehicle occurrence in the region is very accidental, and the regional distribution of the accident severity in the morning and evening rush hour is quite different. It is found that the key points of the accident network are more dispersed in the early peak, while the late peak is relatively concentrated, and the distribution of the fault car in the critical points of the early peak and the late peak in the network is more dispersed. After that, the paper analyzes the correlation among the degree, unit weight and point weight of the network. On this basis, for different VMS information display, according to the four attributes of degree, unit weight, point weight and clustering coefficient, cluster analysis is carried out. The VMS is divided into 16 categories, each category corresponds to one kind of VMS information release characteristic, and the display screen distribution diagram of VMS E in the morning and evening rush hour accident and fault vehicle network is given. Finally, the paper divides the VMS association network into factions, mining the community structure, mining the group nature of VMS information publishing, and evaluating and optimizing the partition result by modularity. It is found that the modular degree of the accident network and the fault car network in the early rush hour is less than 0. After analyzing the reasons, the paper makes a manual optimization to the division of the community structure. The optimized network module index is more in line with the fact that the VMS network belongs to the hybrid network. The optimized community structure partition results can be used as a reference strategy for VMS information publishing.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：U492.8

【参考文献】