【作者】 李俊卫；

【导师】 林柏梁；

【作者基本信息】 北京交通大学 ， 交通运输规划与管理， 2009， 博士

【摘要】 随着国民经济的飞速发展和城市化进程的不断加快,城市的交通拥堵问题越来越引起交通研究者的关注。在过去的几十年中,许多智能交通理论和管理方法已经应用于实际的交通管理中,这对于改善路网的交通状况和提高交通网络效率具有重要意义。城市交通设施功能化是改善日益恶化的交通状况和缓解交通拥堵的一个重要措施。因而,用于承载城市内长距离运输和大交通量的快速路,已经获得了迅速的发展。动态OD估计是智能交通系统的一个重要组成部分,深入地研究快速路动态OD估计对优化城市交通控制、缓解交通拥堵和发展城市路网的OD估计等具有重要意义。本论文目的是发展有效并实用的快速路动态OD估计模型和算法,主要沿着动态OD估计的非基于交通分配方法主线对快速路动态OD估计进行深入研究。首先,总结和分析了目前静态和动态的OD估计模型及算法。其次,依据快速路的交通流特征,提出了一个宏观交通流模型参数估计的状态空间模型及其在线估计算法。再者,依据车队离散的正态分布假设,研究了城市快速路通道和城市快速路网的动态OD估计,并构建了模型,设计了相应的求解算法。最后,论文通过动态交通分配仿真,对所提出的模型进行案例评估。论文的主要创新性成果和内容如下:1.构建了宏观交通流模型参数估计的状态空间模型。为了获得更精确的动态OD间时变的车辆平均行程时间估计,建立了宏观交通流模型参数估计的状态空间模型;提出了两类估计OD间车辆平均行程时间方法,并通过数值试验进行了对比分析。2.提出了基于UKF算法的快速路通道的动态OD估计模型。通过考虑车辆期望速度的变化和交通流的时变本质,结合宏观交通流模型参数估计的状态空间模型,构建了两个快速路通道的动态OD估计模型;通过仿真试验对模型进行评估和对比分析。评估结果表明,所提出的OD估计是稳态和实用的。3.UKF算法的引入和改进。UKF算法是一种新的、适合于非线性系统的Kalman滤波算法,其优点是并不需要计算复杂的Jacobian矩阵。标准的UKF算法并不考虑存在的约束问题,这必然会给估计带来较大的误差。为了减少估计误差,论文对标准的UKF算法进行了改进,着重考虑了在约束条件下的截断问题,这将为运用UKF算法求解存在约束的其他模型提供了新的思路。4.基于行程时间的正态分布假设,构建了基于动态检测线技术的快速路网动态OD估计模型。在没有任何径路信息状况下,依据行程时间的正态分布假设和检测线技术,修改了行程时间估计和量测方程,扩展了线性动态OD估计模型为快速路网的非线性动态OD估计模型,并通过仿真数据对模型和算法进行了评估。5.提出基于交通分配的城市快速路网的动态OD估计模型。首先,考虑到路网信息获取的不完全性,论文基于交通分配模型和OD间车辆行程时间的正态分布假设发展了K.Ashok所提出的路网估计模型。其次,论文结合动态检测线技术,构建了组合的快速路网动态OD估计模型。最后,运用仿真数据对提出的模型进行了分析评估。评估结果表明,提出的快速路网动态OD估计是稳态的。更多还原

【Abstract】 With the development of national economy and accelerating urbanization process, the traffic researchers have paid more attention to the jam of city traffic. In the past several decades, many theories of intelligent traffic and management methods have been applied to real traffic management, which is important to make the traffic conditions of traffic networks better and improve traffic efficiency. Function of urban traffic facility is an important measure to make deteriorating traffic conditions better and decrease traffic jams. Urban Expressways that load heavy and long-distance traffic volume have arised. Dynamic Origin-Destination(OD) demands are essential input for on-line traffic control, dynamic traffic-assignment simulation and management systems. Hence, study on dynamic OD estimation of expressway has important significance.The purpose of the dissertation is to develop effective models and algorithms for dynamic OD estimation of urban expressway. This study mainly follows the research line of non-assignment-based methods to investigate dynamic OD estimation of expressway networks. Firstly, the thesis summarizes and analysises existing static and dynamic of OD models and algorithms. Secondly, the thesis purposes a state-space model of traffic flow model parameters and its algorithm to apply to the travel time estimation of dynamic OD matrix. Thirdly, on the basis of the assumption of normal distribution of the travel time, we study dynamic OD estimation of urban expressway corridor and urban expressway network, establish the models and design the corresponding solving algorithms. The thesis evaluates the proposed models based on numerical tests of dynamic traffic assignment.The main innovation points and contents of the study include:1. Construct a state-space model of traffic flow parameters. To obtain the travel time estimation of dynamic OD matrix, we propose the state-space model and evaluate the models of computering the travel times of OD demand.2. Propose two dynamic OD estimation models of urban expressway corridor based on Unscented Kalman Filter. The proposed models capture the speed discrepancy among drivers and integrate the state-space model of traffic flow parameters. Numeral experiments compare and analyze the two models. The results show the proposed dynamic OD estimation is stable and robust. 3. Introduce and improve Unscented Kalman Filter (UKF) algorithm. UKF algorithm is an important core algorithm for the thesis. The advantage of UKF does not need to obtain Jacobian matrix of measurement equations. Normal UKF algorithm does not consider existing constraints, which leads to larger errors. The thesis considers the problem of UKF under constraint conditions, which have important significance to solve other problems based on UKF under constraint conditions.4. Propose a dynamic OD estimation model of expressway network with screenline flows based on the assumption of normal distribution of travel times of OD pairs. By modifying measurement equations and the estimation of travel time, the thesis extends the linear dynamic OD estimation model of expressway network to a no-linear model.5. Propose a dynamic OD estimation model of expressway network based on a dynamic traffic assignment model. Considering that the information of road network is uncomplete, the thesis extends the dynamic OD estimation model proposed by K. Ashok based on a traffic assignment model and the assumpation of normal distribution of the travel times of OD pairs, which doesn’t denpend on a historic OD matrix. Subsequencely, the thesis proposes an integrated dynamic OD model for expressway network based on the technique of screenline flow and a dynamic traffic assignment model. The thesis evaluates and analyzes the proposed models by numerical tests. Numerical results show that our proposed models are stable and robust.更多还原