学科专业: 交通运输规划与管理
指导教师: 冉斌教授
杨东援 教授
摘要
随着道路交通管理系统智能网联化趋势的深入,道路交通管理系统对交通状态估计提出了新的任务诉求:可以刻画出针对车辆的交通状态演变机理、并能在拥堵状况下进行高精度的短时交通状态估计。此外,随着车辆自动识别、车牌照识别等新型观测技术的蓬勃发展,新型的定点车辆匹配型观测环境既提供了针对车辆的新型观测数据基础,也对现有的交通状态估计框架提出了挑战。因此,为了满足智能网联化道路管理系统对交通状态估计模块的诉求,也为了更好地适应新型的定点车辆匹配观测环境,本研究在继承现有交通状态估计框架和研究成果的基础上,以“交通状态估计”为主题,引入“拉格朗日-空间坐标系”的概念,遵循“模型建立——算法实现——算例验证”的基本主线,构建了拉格朗日-空间坐标系的交通状态估计总体框架。
首先,本研究构建了拉格朗日-空间坐标系的宏观交通流模型。根据现有的拉格朗日-空间坐标系宏观交通流模型基础,完善了拉格朗日-空间坐标系的交通状态演变模型——行程时间传输模型。根据实际数据验证了拉格朗日-空间坐标系的交通基本图规律。通过点模型、扩展项的修正思路,解决了如何在开放式路网环境构建拉格朗日-空间坐标系的宏观交通流模型的问题。
其次,本研究构建并实现了拉格朗日-空间坐标系的交通状态估计框架。在了拉格朗日-空间坐标系宏观交通流模型的基础上,考虑定点车辆匹配型观测环境的检测机理,构建了拉格朗日-空间坐标系的交通状态估计框架。基于卡尔曼滤波类的数据同化技术,设计了状态估计模型的实现算法,并给出了将拉格朗日坐标系的交通状态转换至时间坐标系的实现方法。
随后,本研究讨论了拉格朗日-空间坐标系交通状态估计在理想的定点车辆匹配观测条件下的估计性能。在对具体案例仿真建模及标定的基础上,通过仿真的轨迹数据模拟出理想的交通状态估计环境,并基于模型的参数标定,构建并实现了拉格朗日-空间坐标系的交通状态估计模型。通过与现有的交通状态估计模型的对比,评估分析了拉格朗日-空间坐标系交通状态估计针对不同空间范围在拥堵、非拥堵状况的估计性能。
最后,本研究探讨了拉格朗日-空间坐标系交通状态估计在实际观测环境的适应性。以车牌照识别技术为例,厘清了该项技术的观测机理,并分析了该项观测技术在观测设备布设、观测对象识别等方面存在的实际观测问题。通过具体的观测场景设计,本研究分别从干道观测设备损坏、干道观测区间粒度过大、匝道观测设备缺失、观测样本缺失四个方面,探讨了基于拉格朗日-空间坐标系交通状态估计模型与实际观测环境的磨合。通过论述该模型在实际观测环境中的适应情况,本研究从观测设备布设、观测技术改进等方面对未来观测环境的改善给出了有效的提升方向。
本研究立足于拉格朗日-空间坐标系,构建了既可以从车辆维度阐述交通流演变机理,又能适应定点车辆匹配型观测环境的交通状态估计框架,并通过实际算例验证了该框架具有良好的估计性能,为智能网联化道路交通管理系统的交通状态估计模块提供了理论支撑和实现途径。同时,本研究论证了拉格朗日-空间坐标系交通状态估计方法在估计精度、拥堵状况、不同道路线型等方面相对于传统估计方法的优越性,展现了该方法的潜在技术优势。此外,本研究讨论了拉格朗日-空间坐标系交通状况估计框架在实际观测环境的适应性,为如何在有限资金投入的条件下有效地改善交通信息观测环境给出了提升的方向。
关键字:交通状态估计,拉格朗日-空间坐标系,宏观交通流模型,卡尔曼滤波,定点车辆匹配型观测环境
ABSTRACT
With the high degree ofintelligence and connection of the transportation system in the future, thetraffic management system proposes a new task appeal to the traffic stateestimation, which can depict the traffic state in the vehicle dimension and canhave higher precision under the congestion condition. In addition, with therapid development of new observation technologies such as automatic vehicleidentification and license plate recognition, the new fixed-pointvehicle-matched observation environment not only provides a new observationdata base for vehicles, but also challenges the existing traffic stateestimation framework. Therefore, in order to adapt to the future intelligent andconnected traffic management system's appeal to the traffic state estimationmodule, and to better adapt to the new fixed-point vehicle matching observationenvironment, this study attempts to establish a traffic state estimationframework that can meet the new task requirements and adapt to the newobservation environment. Based on the inheritance of the existing traffic stateestimation framework and research results, this study introduces the concept of“Lagrangian-space coordinate system” with the theme of “traffic stateestimation”. Following the basic main line of "model building - algorithmimplementation - example verification", this study constructs the overallframework of traffic state estimation in Lagrangian-space coordinate system.
Firstly, this study constructs amacroscopic traffic flow model of the Lagrangian-space coordinate system.According to the existing macroscopic traffic flow model of Lagrangian-spacecoordinate system, the traffic state evolution model of Lagrangian-spacecoordinate system-travel time transmission model is proposed. According to the actualdata, the basic traffic fundamental diagram of Lagrangian-space coordinatesystem is verified. Through the point model correction and extension termcorrection, the macroscopic traffic flow model of Lagrangian-space coordinatesystem is constructed to fit the open road network environment.
Secondly, this study constructsand implements a traffic state estimation framework for the Lagrangian-spacecoordinate system. Based on the macroscopic traffic flow model ofLagrangian-space coordinate system, considering the observation mechanism ofmatching observation environment, the traffic state estimation framework ofLagrangian-space coordinate system is constructed. Based on the dataassimilation technique of Kalman filter, the implementation algorithm of stateestimation model is designed, and the realization method of transforming thetraffic state of the Lagrangiancoordinate system to time coordinate system is given.
Subsequently, this studydiscusses the performance of the traffic state estimation of theLagrangian-space coordinate system in an ideal observation environment. Basedon the simulation and calibration of specific cases, the ideal traffic stateestimation environment is simulated by the simulated trajectory data. Then, thetraffic state estimation of Lagrangian-space coordinate system is constructedand realized. By comparing with the existing traffic state estimation model,the performance of traffic state estimation in the Lagrangian-space coordinatesystem is evaluated and analyzed in terms of different cells and differenttraffic conditions.
Finally, this study explores theadaptability of the traffic state estimation of the Lagrangian-space coordinatesystem in the actual observation environment. Taking the license platerecognition technology as an example, the observation mechanism of the technologyis clarified, and the actual observation problems of the observation technologyin the arrangement of observation equipment and the identification ofobservation objects are analyzed. Through the design of the specificobservation scene, this study analyzes the Lagrangian-spatial traffic stateestimation model’s performance from four aspects: damage of the main roadobservation equipment, excessive grainsize of the main road observation section, lack of the ramp observationequipment, and lack of the observation samples. By exploring the adaptation ofthe proposed traffic state estimation model in the actual observationenvironment, and the improvement direction of the future observationenvironment from the aspects of observation equipment layout and observationtechnology is proposed.
Based on the Lagrangian-spacecoordinate system, this study constructs a traffic state estimation frameworkthat can explain the evolution mechanism of traffic flow from the vehicledimension and can adapt to the new matching observation environment. Byverifying the adaptability of the estimation framework in the ideal and actualobservation environment, this study provides an effective theoretical supportand implementation approach for the traffic state estimation module of theintelligent traffic management system. At the same time, by showing thesuperiority in terms of estimation accuracy, congestion status estimation, andadaptability to different road line types, this study demonstrates thepotential technical advantages of the of the Lagrange-spatial traffic stateestimation method. In addition, by discussing the adaptability of theLagrangian-spatial traffic state estimation framework in the actual observationenvironment, and this study gives an improvement direction for how toeffectively improve the traffic information observation environment under thecondition of limited capital investment.
Keywords: traffic state estimation, Lagrangian-space coordinatesystem, macroscopic traffic flow model, Kalman filter, fixed-point vehicle-matchedobservation environment