【作者】 徐东伟；

【导师】 贾利民； 董宏辉；

【作者基本信息】 北京交通大学 ， 交通安全工程， 2014， 博士

【摘要】 智能交通系统的建设和发展是解决交通拥堵等交通问题的有效途径,而道路交通路网上的交通状态信息的实时准确获取是建设智能交通系统、提供交通信息服务、进行道路交通管理的基础,在道路交通系统运行管理控制中起着基础性和关键性的作用。道路交通状态的有效获取方法是提升道路交通系统运行和管理效率的核心问题。本论文以道路交通管理与控制的实际需求为背景,深入研究了道路交通状态多维多粒度的获取方法,基于不同层次的道路交通特性提出了不同的道路交通状态获取方法,并通过实例分析与应用,验证了提出的道路交通状态获取方法的有效性,形成了不同层次的、较为完整的道路交通状态获取方法体系。本论文的主要研究成果具体体现在以下几个方面：1)以道路交通的截面特性为基础,提出了基于交通信息模板的道路交通状态获耿方法。基于交通信息模板的道路交通状态获取方法是在有效历史数据的基础上进行。该方法首先根据不同的划分标识,将道路交通运行模态划分为不同的子模态；然后提取不同交通运行模态下的历史道路交通数据,进行相关的数据预处理,并对道路交通信息的规律性进行分析,利用道路交通信息的规律性信息获取不同交通运行模态下所对应的交通信息模板；最后判断当前时刻目标路段的交通运行模态,选取相对应的交通运行模态下的交通信息模板,利用道路交通状态信息的时间相关性的特点获得当前时刻的交通运行状态数据。2)以道路交通的路段特性为基础,提出了基于虚拟速度传感器节点的道路交通状态获取方法。基于虚拟速度传感器节点的道路交通状态获取方法是在对路段的交通状态进行软测量的基础上进行。该方法首先采用线性插值的方法在相邻的交通流传感器之间设定虚拟速度传感器节点序列；然后将交通流传感器节点时间序列的速度数据映射到空间序列上,并利用最小二乘法对权重矩阵进行训练(该权重矩阵与虚拟速度传感器节点的速度和交通流传感器检测的速度数据相关)；最后利用权重矩阵和交通流传感器节点时间序列的速度数据,计算得到虚拟速度传感器节点的速度,从而获得整个路段的交通状态的空间分布。该方法可以有效获取未布设交通检测器的路段的交通状态。3)以道路交通的区域特性为基础,提出了基于区域交通吸引子匹配的道路交通状态获取方法。基于区域交通吸引子匹配的道路交通状态获取方法是在对交通区域特性分析的基础上进行。该方法首先根据不同的划分标识,将道路交通运行模态划分为不同的子模态,并提取不同交通运行模态下的道路交通状态数据,进行相关的数据预处理,获取不同交通运行模态下的道路交通状态信息并存入道路交通运行特征参考序列中；然后引入目标路段的区域交通吸引子的概念,提取不同交通运行模态下的区域内其它各路段的道路交通状态数据,进行相关的数据预处理,获取不同交通运行模态下的目标路段的区域交通吸引子信息并存入道路交通运行特征参考序列中；最后获取当前时刻的目标路段的区域交通吸引子信息,根据一定的规则与目标路段的历史区域交通吸引子进行匹配,选取区域交通吸引子最优匹配时所对应的目标路段的道路交通状态信息,并作为初始修复数据,进而通过对初始修复数据进行卡尔曼滤波,获得最终的修复数据。该方法能够实时有效地获取当前的道路交通状态。4)以道路交通的网络特性为基础,提出了基于压缩感知的道路交通状态获耿方法。基于压缩感知的道路交通状态获取方法是在对交通路网特性分析的基础上进行的。该方法首先利用交通状态信息矩阵奇异值分解和主成分分析的方法进行道路交通状态数据的结构性和可压缩性分析;然后在道路交通状态数据可压缩性分析的基础上,利用压缩感知的理论及方法进行道路交通状态估计,并对基于压缩感知进行道路交通状态获取的过程中所涉及的主要参数进行分析和设定；最后基于实例对基于压缩感知的道路交通状态获取算法的应用进行验证。该方法可以实现对交通路网大面积道路交通状态数据的估计。更多还原

【Abstract】 The construction and development of Intelligent Transportation System (ITS) is an effective way to solve the traffic problems such as traffic jam. The real-time and accurate acquisition of the information of road traffic states is the basis of ITS construction, traffic information service and road traffic management. It plays a fundamental and critical role in the management and control in the operation of traffic system. The acquisition approaches of the road traffic states are the key problems to improve running and management efficiency of the urban traffic system.This paper takes the actual demand of the road traffic management and control as the background. The multi-dimensional and multi-granularity acquisition of road traffic states approaches are studied in this paper. Different methods for acquisition of road traffic states are put forward based on different levels of road traffic characteristics. The effectiveness of the proposed methods for obtaining traffic states is verified through analysis and application of examples. A relatively complete road traffic states acquisition system with different levels is formed. The main results of this paper are embodied in the following aspects:1) Based on the road traffic’s section properties, the road traffic states estimation method based on the road traffic information templates is proposed.The road traffic states estimation method based on the road traffic information templates is carried on the basis of the effective historical data. Firstly, the road traffic modes are divided into several different sub-modes according to different partition identifications; then data from road traffic detectors under different traffic modes are abstracted to analyze the regularity characteristics of the road traffic information and the road traffic information templates are finally obtained based on this regularity characteristics; finally the current traffic mode of the target link is judged and the corresponding traffic information templates are selected to obtain the current traffic sates data through the road traffic states information’s temporal correlation characteristics.2) Based on the road traffic’s link properties, the road traffic states estimation method based on virtual speed sensors is put forward.The road traffic states estimation method based on virtual speed sensors is carried on the basis of the soft measurement for the link’s traffic state. First, virtual speed sensors are defined by linear interpolation between adjacent traffic flow sensors. Secondly, virtual speed sensors are designed. The weight matrix related with the speed of virtual speed sensors and the speed data from traffic flow sensors are trained by least square. The speed of virtual speed sensors are estimated with the weight matrix and the speed data from traffic flow sensors by multiple linear regressions. Thirdly, the traffic state spatial distribution of the link (the speed spatial distribution on the link) can be gained. This approach can effectively obtain the traffic state of links without any traffic detector.3) Based on the road traffic’s region properties, the road traffic states estimation method based on matching of the regional traffic attracters is put forward.The road traffic states estimation method based on matching of the regional traffic attracters is carried on the basis of the analysis of road traffic’s region properties. Firstly, the road traffic running modes are divided into several different sub-modes according to different partition identifications, and the historical road traffic data are abstracted and preprocessed to obtain the characteristic information of road traffic running states under different modes, which are stored in the reference sequences of characteristics of traffic running states. Then, the concept of the regional traffic attracters of the target link is introduced and the historical traffic data of links in the region are abstracted and preprocessed to obtain the regional traffic attracters of the target link under different traffic running modes, which are also stored in the reference sequences of characteristics of traffic running states. Finally, the data of the current regional traffic attracters of the target link are abstracted, which are matched with the historical regional traffic attracters of the target link through certain rules. The road traffic running states data of the target link corresponding to the optimal matching are selected as the initial data for recovery and the initial recovery data are processed with Kalman Filter and the final recovery data are obtained. This method can real-time and effectively reflect the current road traffic states.4) Based on the road traffic’s network properties, the road traffic states estimation method based on compressive sensing is put forward.The road traffic states estimation method based on matching of the regional traffic attracters is carried on the basis of the analysis of road traffic’s network properties. To make sure that the compressive sensing theory can be applied on the estimation of traffic states of the road traffic network, the compressibility analysis of the road traffic states data is discussed. Then the road traffic states estimation approach based on compressive sensing is presented and the parameters setting referred in this road traffic states estimation approach based on compressive sensing is discussed. Finally one typical road network in Beijing is adopted for verification of the application of this road traffic states estimation algorithm. This method can realize the estimation of large area’s road traffic states of the road network.更多还原