窦慧丽

入学时间：2007级

答辩时间：2010年

论文题目：面向预报的城市道路交通流预测关键理论研究

中文摘要

摘 要

交通预报是应用交通流变化的规律，根据当前及近期的交通流采集数据，对未来一定时期内的交通流变化情况的预先估计和预告。交通预报的主要内容是道路网络未来一段时间内交通流运行状态的定性或定量信息,具体包括交通拥堵出现的地点、时间和强度。及时准确的交通预报信息可以为交通管理者提供有力的决策依据，保证对整个城市道路交通系统进行有效的诱导、控制和管理；还可以为出行者带来更为可靠的保证，确保出行决策的正确有效；而与交通密切相关的企事业单位，可以根据所获得的交通预报信息合理安排企业活动，减少不良交通状态所带来的损失。因此，随着智能交通系统各个领域科学技术的发展，以及先进的信息采集、传输和发布等相关技术的进步，无论是交通出行者或是交通管理者都对智能交通系统提出了更高的要求。类似于“天气预报”，人们迫切想要得到“交通预报”服务。

本论文以国家高技术研究发展计划（863计划）项目专题“城市交通状态智能预测与管理支持系统”为依托，围绕面向预报的城市道路交通流预测关键理论展开研究。论文通过大量的文献阅读和资料总结，梳理了交通预报的发展趋势和相关理论研究方面的成果和不足。在此基础上，确定了本文的研究方向，即面向预报系统的城市道路交通流预测的理论问题及模型算法研究。在交通预报概念解析、面向预报的交通流预测机理解析、交通流预测模型与算法等方面取得了一系列的成果，具有重要的理论意义和实用价值。

论文首先对交通预报的概念进行了深入解析。通过交通预报和交通预测关系的分析和界定，对交通预报的概念进行了明晰。并对交通预报系统的用户主体及其信息需求层次、系统功能等进行了分析，为论文的理论模型体系指明方向。

其次，论文对交通流采集数据的质量控制问题进行了研究。针对城市道路交通系统中数据实际采样环境中存在的各种干扰，以及由此带来的错误的、不完整的、含噪声的数据大量存在于数据仓库中，论文首先对交通流数据的质量问题及成因进行了分析，在此基础上构建了针对不同问题类型数据的质量控制的模型和算法，为后续交通流预测模型的研究提供了良好的数据基础。

再次，基于交通预报系统的信息需求，论文分别从微观、中观和宏观三个层面，对城市道路交通状态预测的理论模型与算法展开研究。根据交通状态反映的空间范围的大小，交通状态可以分为微观、中观和宏观三个层面，分别满足交通出行者和交通管理者不同层次的信息需求。论文针对微观的路段层面，综合考虑了交通预报系统的需求和影响，以及交通流演变的随机性和不确定性，提出了交通状态概率预测模型，实现了交通状态的定量描述；针对中观路径层面，通过分析路径和路段行程时间的相互关系，提出了路径行程时间的滚动预测模型，通过引入动态误差修正因子，对路段行程时间进行逐期误差修正，然后进行错时累加，得到路径的行程时间，解决了起讫点距离过长时，交通预报对行程时间预测准确度的影响；针对宏观的路网区域，通过研究分析路网区域拥堵指数和区域内路段交通状态的相互关系，综合考虑交通拥堵的空间范围和持续时间，建立了路网区域拥堵指数预测模型，可以动态跟踪区域的宏观交通状态变化，反映区域状态的变化趋势。

最后，对全文进行了总结，指出了论文的创新点，对有待于进一步研究的问题进行了展望。

关键词：交通预报，质量控制，概率预测，滚动预测，拥堵指数

英文摘要

ABSTRACT

Traffic forecast is prediction and advanced notification of the running state of the traffic flow in the future, which is based on the change law of traffic flow and the historical and recent collection data. The main contents of traffic forecast are qualitative and quantitative information of the traffic flow condition of the road network in the future and the three elements of traffic forecast are time, location and intension of traffic congestion. Timely and correct traffic forecast can provide potent decision-making basis to traffic management personnel, which ensures the traffic guidance, traffic controlling and traffic management can be effectively carried out. Also, traffic forecast can provide travelers with reliable information to guarantee the correctly of the travel decision. The establishing enterprise closely related with traffic flow state can make arrangement activities according to the traffic forecast to diminish losses caused by the adverse traffic condition. So with the development of all the fields of Intelligent Traffic System and the progress of the information collection, information transmission and information publishing, both traffic management personnel and traffic travelers put forward a high requirement of Intelligent Traffic System. Similar to weather forecast, they raise a claim of traffic forecast.

This dissertation is mainly based on the research of national 863 projects “System of Intelligent Forecasting and Management Support for Urban Traffic Status”and focuses on the critical theories research of traffic flow prediction oriented to traffic forecast.According to the massive literature data, the dissertation epurates the progress trend of traffic forecast and summarize the achievement and deficiency of the related theoretic research. Based on the analysis, the research direction in determined, that is basic theory and modeling research of traffic flow prediction oriented traffic forecast of urban street. The dissertation gained a series of research achievements in some aspects such as the clarity of traffic forecast, analysis of the mechanism of traffic flow prediction oriented to traffic forecast, modeling and algorithm of traffic flow prediction, which has great theoretical significance and practical value.

Firstly, the dissertation deeply analysis the concept of traffic forecast. Based on the analysis and boundary definition between traffic forecast and traffic prediction, the concept of traffic forecast is perspicuous. Then the dissertation carried out the requirements analysis of traffic forecast system, including user requirement, function requirement, performance requirement, based on which the theoretic research direction is designated.

Secondly, the dissertation carried out the research of the problem quality control of traffic flow data. Aimed at the existing various disturb in the practical environment of traffic flow data collection in the urban traffic system, hence lots of wrong data, lost data and polluted by noise in the database, the problem of quality control of traffic flow data is studied. The dissertation firstly analysis the types and reasons of traffic data with questions and then put forward various model and algorithm of quality control according to the types of questions. The research of this part provides a better base to the theoretic research of the subsequent traffic flow prediction.

Thirdly, the dissertation carried out the theoretic research of traffic flow prediction oriented to traffic forecast, including three level of microscopic level, medscopic level, macroscopic level, which can satisfy the different requirements of traffic manager and traffic traveler. Aimed at the microscopic road section, and comprehensively consideration of the requirements and influence of traffic forecast, and also consideration of the randomness and uncertainty of traffic flow, the probability prediction algorithm of traffic state is put forward, which can give the quantitative description of traffic state. Aimed at the medscopic level of path, and according to the analysis of the relationship of the travel time between road section and path, the rolling prediction algorithm of the travel time of path is put forward. The dynamic error correct index is introduced to correct the predicted travel time of road section and then the travel time is achieved by interleaved addition of the travel time of road section, which solved the problem of inaccuracy of travel time prediction when the distance between origin and destination. Aimed at the macroscopic traffic network area, according to the analysis of the relationship between the traffic state of network area and road section, and comprehensively consideration of the spatial scope and the length of lasting time, the prediction algorithm of traffic congestion index is put forward, which can dynamically track the changes of the traffic state and reflect the macroscopic tendency of the traffic state of network area.

Finally, the dissertation is summarized. The creative research achievements and other important research directions were pointed out.

Key Words:Traffic Forecast, Quality Control, Probability Prediction, Rolling Prediction, Congestion Index

时柏营

入学时间：2007级

答辩时间：2010年

论文题目：面向交叉口的交通流实验分析方法研究

中文摘要

摘 要

随着现代科学技术特别是智能交通系统在交通系统中的应用与发展，新的科学技术与交通系统的不断融合，交通研究领域的不断扩大，现代交通系统已经是一个复杂的大系统，传统的还原解析方法已经显现出在复杂交通现象的特性研究中的局限性。

实验方法研究不但是研究解析自然科学的利器，随着复杂系统理论研究的兴起及计算机仿真技术及其他高新技术的发展，在诸多社会科学领域也得到了广泛应用并取得了很多成果；交通系统由于其复杂性，需要在以往研究成果基础上，寻求实验方法研究的思路,当前高新技术的发展也在工具、手段上为复杂交通系统的实验研究方法研究提供了基础。基于以上背景，提出了论文研究的问题——面向交叉口的交通流实验方法研究。

.论文首先从实验相关学科发展、实验方法在交通学科研究中的应用展开论述，在此基础上，总结了实验交通工程的概念；根据交叉口交通流实验方法研究的功能需求，提出实验方法研究的概念框架并给出实验系统的构架；鉴于信息对交叉口交通流运行形式的影响，分别对常规情况下的交叉口交通流及车路协调环境包括车车通信、车路通信条件下的交叉口交通流展开了实验研究，最后针对实验系统实现的关键技术及系统集成进行了阐述。

论文首先进行文献阅读与总结，分析了实验科学在实验物理学、实验心理学、实验经济学及计算经济学的应用与发展，在此基础上，通过阐述实验方法在交通学科研究总的应用，总结了实验交通工程的概念；接着对交叉口的交通流运行方面进行了综述，确定论文的突破方向是以交叉口交通流为研究对象，进行实验方法在交通学科中应用的初步尝试以及进行实验系统集成的关键技术研究。

通过分析面向交叉口交通流实验方法的功能需求，给出了交叉口交通流实验方法研究的概念框架及系统构架，并对系统中虚拟环境建模问题进行了分析。

常规情况下的交叉口交通流运行，是以车辆传统的跟车、换道为基础，以车流离散现象为表征的间断流，根据研究的需求，建立了以微观仿真软件为基础的实验系统框架。首先对实验系统进行了校核与验证，接着进行了信号配时方案的设计及车流离散模型的对比验证分析；

高度信息水平下的车路协调系统中车—车、车—路间通过信息交互作用，影响着交通流运行的形式与交叉口运行指标；针对车路协调环境下交通流研究需求，基于复杂系统研究中多Agent技术进行了实验环境的构建，以传统的多速度差模型为基础，建立了能反映车车通信、车路通信条件下的交通流运行模型，并以此为基础，针对车车通信条件下的跟车、换道进行了实验研究；针对车路通信条件下的无信号控制交叉口建立了次路车流引导模型及算法，进行了实验分析。

针对两种实验系统实现有如下两种关键技术：一是Agent模型的实现，二是平行系统的建立。论文利用Petri网及有限状态机（Finite State Machine，FSM）对多Agent系统进行了实现；接着给出了建立基于仿真软件的平行系统和基于多Agent的平行系统的实现框架及技术；最后，针对实验交通系统的未来发展及需要努力的方向进行了展望。

本文是利用实验方法对交通系统进行研究的一次尝试，力求在前人交通研究的成果基础上，利用实验分析的方法，对交叉口交通流运行进行研究，从而促进实验交通工程概念发展并建立面向交叉口交通流实验分析的平台。

关键词：实验交通，交叉口，实验方法，智能车路，仿真软件，多Agent，Petri网，有限状态机

英文摘要

Abstract

With the development of modern science and technology, especially the application of ITS, and the integration of new technology and transportation system, the research area of transportation has expanded rapidly. The modern transportation system has become so complex that the traditionalreductiveapproach has been outdated for the study.

Experimental method is not only a tool for parsing the natural sciences, with the emergence of complex systems theory and computer simulation technology and other high-tech technology development, in many areas of social sciences have also been widely used and has achieved many results. Because of Transportation system’s complexity, we needs to find the idea of experimental method based on the results of previous studies. Also at present the development of high-tech provide the basis for the research of transport systems with experimental research method in the tools and means. Based on the above background, the thesis put forward the question - the experimental method study facing the traffic flow of intersections.

This paper first started discussion from the development of related experimental subjects and the application of experiments methods in traffic and based on which summed up the concept of experiment traffic engineering. According to the functional requirements of study of intersection traffic flow by experimental method, this paper proposed the conceptual framework and gives a framework for experimental system, and based on which carried out experimental study respectively, for flow of intersection under normal circumstances and the traffic flow of the intersection under Vehicle-Infrastructure Integration environment including car-to-car and car-to- infrastructure communication. Finally the realization of the key technologies and system integration were also discussed for experimental system.

After the literature reading and summarizing this paper reviews the literature and analyzes the application and development of empirical science in the fields of experimental physics, experimental psychology, experimental economics and computational economics and points out the characteristics of empirical experimental study and reviews the development of transportation simulation system, the complicated system, the multi-Agent modeling, and computation intelligence. The paper's breakthrough direction is the study of the experimental technique and the key technologies in the intersection experiment system, taking intersection as an example.

By analyzing the functional requirements by experimental methods facing the flow of intersection traffic, gives the conceptual framework and system architecture and the modeling of virtual environment system were analyzed.

The operation of intersection traffic flow under normal circumstances is the Interrupted traffic flow for the characterization of the discrete phenomenon based on the traditional following modals and lane-changing modals. According to research needs, established a experimental system framework based on micro-simulation software; and first , carrying the Verification and Validation for the experimental and then on this basis, carrying on the design of signal timing and traffic comparison of the discrete model verification analysis.

The interaction between car-to-car and car-to-road through information under Vehicle-Infrastructure Integration environment, by which affects traffic flow and intersection operation in the form of performance indicators under the coordination of research needs of Vehicle-Infrastructure Integration environment, constructed the experimental environment based on the Multi-Agent technology for the study of complex systems. Based on Full Velocity Difference Model constructed the traffic flow operation model which can reflect the condition of car-to-car and car-to-road communication. And then carried out experimental studies.

Carrying out two kinds of experiment systems, this paper points out several key technologies: first, establishment of the multi-Agent secondly, establishment of the parallel system; Then, this paper explains how to realize these two key technologies.

Finally, this paper has an outlook into the future and the development direction of the experiment transportation system.

This paper, applying the experimental approach is a tentative study in experiment transportation system It tries to carry out empirical research in intersections under different information environments on the base of previous transportation research achievements and lays the foundation for the future development of experiment traffic engineering.

Key Words：experimental science，Experimental Traffic Engineering，intersection，Vehicle-Infrastructure Integration，traffic simulation，Multi-Agent，Petri net，finite state machine

马国胜

入学时间：2007级

答辩时间：2011年

论文题目：交叉口主动交通安全控制基础问题研究

中文摘要

摘要

道路平面交叉口是道路网络中道路与道路的交叉点，是道路的主要组成部分。城市中交通阻塞和交通流的中断主要发生在交叉口，而道路交通事故也较为集中地发生于此。本文根据交通事故发生的原因—认知、判断失误以及信号控制不当，立足于主动交通安全研究基础层面，在对主动安全理念的形成与发展、交通冲突技术、相关基础理论研究等方面的核心内容进行归纳总结的基础上，基于车路协调技术发展趋势和主动交通安全思想，围绕交通流从上游交叉口到下游交叉口经历的暴露、遭遇、避险过程，从控制的角度对交叉口主动交通安全基础问题进行了研究。

由于主动交通安全是面向过程的以及主动交通安全控制实质上是对交通安全的过程即暴露阶段、遭遇阶段、避险阶段等进行控制，因此，本文应用主动交通安全思想对交叉口主动交通安全控制过程进行了分析和讨论，着重分析了交通流运行过程中的运行特征、时空关系和车路协调技术对控制参数的影响，并从现实需求出发对引起交通安全问题的关键因素—运行速度、交通冲突、信号控制进行了阐析。为弥补以往的研究在面向交通流运行过程时从信息提供或信号控制单一视角分析和解决交通安全问题的不足，本文通过对两者的整合针对交通流冲突过程中可能产生的后果采取相应的控制措施以避免事故的发生。在此基础上，构建了交叉口主动交通安全控制研究框架体系并对基础问题进行了归纳。鉴于本文着重讨论单点的控制问题，暂不涉及上游交叉口对下游交叉口的影响，因此，论文研究面向交叉口主动安全控制过程中的遭遇及避险阶段，重点研究两阶段的辅助控制和信号控制。具体来说涵盖了两个层面的研究：前者一般作为车辆安全控制的初级阶段，主要是通过信息提供的方式，使得驾驶员在接受信息后改变驾驶行为进而调整对车辆的控制策略，做出正确的决策和操作，避免碰撞保障行车安全；后者是通过信号状态的调整或信号优化控制车辆运行风险。

在交叉口范围内存在机-机、机-行人以及机-非冲突，尤其是相位衔接过程中交叉交通流间容易产生事故。本文分析了现实碰撞类型并依据事故成因对基础理论问题—信号控制参数调整与优化、冲突预测、碰撞风险估计和预警等进行了研究：第一，考虑到交叉口因信号控制不当造成的事故原因，为克服以往研究中车辆以匀速状态通过交叉口的不合理性和实时保障车辆在信号控制交叉口不发生碰撞事故，尤其是两难区的存在导致的追尾、侧面碰撞事故，提出了考虑驾驶行为的交叉口主动交通安全控制策略并对控制策略的有效性进行了仿真验证；第二，有关资料研究表明，左转车辆最容易对行人造成伤害，其事故是右转车辆与行人事故的4倍。为防止交叉口行人-车辆发生碰撞事故，并且针对单目视觉技术测距与实际存在着一定的误差，本文初步提出了基于人行横道线特征（一般来说，人行横道线会在路面上长期存在并且每一条人行横道线具有一定的宽度）的交叉口行人-车辆辅助控制方法。利用横道线特征建立了行人识别感兴趣区和人行横道线的曲线方程，这一方法可以有效地识别行人和确定行人在交叉口的位置。在此基础上预测了行人-车辆潜在冲突，结合现有的预警模型实现对驾驶员的预警以保证行人安全，最后进行了离线仿真验证；第三，在实际运行中，车辆接近交叉口时，速度通常是变化的，且具有明显的随机性。本文利用交通状态关键参数—速度这一特性克服以往研究中只考虑单一的运动状态的不足，初步建立了左转车与直行车的碰撞风险表达模型并给出了预警阈值；第四，在分析交叉口主动交通安全控制实现环境的基础上，依托现有技术，搭建了车路协同环境下的交叉口原型实验系统，并对各子系统进行性能测试。总体来说，该原型系统为测试物理设备性能、基础模型和算法适用性提供了一个实验平台；第五，分析了驾驶行为对信号交叉口绿灯间隔的影响，通过动态风险模型估计相位衔接过程中交叉车辆的碰撞风险，如果风险估计值在阈值以下则绿灯间隔期间通过交叉口是安全的，若超过阈值则应针对不同闯红灯情形采取延长绿灯时间、增加绿灯间隔或提前警示等控制策略以保证行车安全；第六，对第四章已建立的直行与左转车碰撞风险模型进行了修正，之后采用修正的风险模型分析了不同相位相序下的风险值并以风险值的大小做为设计相位相序的依据；第七，周期时长是决定单点定时信号控制交通效益的关键控制参数，同时也是确定绿灯极限延长时间的依据，周期时长过长或过短都是不利的。从安全角度分析，周期过长并且超过行人忍耐极限时间会导致行人违规过街从而引发交通事故。目前针对交通流间产生的冲突，通常都是采取一种优化的思想，在减少冲突数的同时，使通行效率、安全在某种程度上取得平衡。本文以国内常见的两相位交叉口为研究对象，推导出了基于

分布的两相位交叉口冲突点延误改进模型并验证了模型的有效性，接着进一步分析了信号控制交叉口流量、冲突和延误的关系，最后以停车线前的延误和冲突点延误最小为目标函数，采用遗传算法求解，优化了交叉口周期时长，通过优化前后冲突数的对比，实现了降低冲突次数的目标。

最后，对全文进行了总结，指出了论文的创新点，对有待于进一步研究的问题进行了展望。

关键词：主动交通安全思想，主动交通安全控制，交通安全过程，实验系统平台，控制策略，辅助控制，信号控制，碰撞风险，冲突点延误

英文摘要

ABSTRACT

Road intersection is the junction of roads and the key component of road. Traffic jam and interruption of traffic flow mainly happen at the intersection. Meanwhile, intersections lead to a relatively high proportion of accidents. According to the analysis of traffic accident causes, related research content such as the development of proactive safety thought, traffic conflict technology and related basic theory were summed up. After that, based on the trend of vehicle infrastructure integration technology andproactive traffic safety thought, the study was conducted on the basic issue of proactive traffic safety control at intersection.

As proactive traffic safety is oriented to process,proactive traffic safety controlcontrols the process of traffic safety in essence. Operational characteristics, time-space relations and control parameters for VII were analyzed byproactive traffic safety thought. And the key factors which influence traffic safety such as operational speed, traffic conflict and signal control were discussed proceeding from the actual demand. So as to make up the deficiency of the earlier studies on the traffic safety which focused on information provision or signal control oriented to traffic flow operating, corresponding control measures were put forward against possible conflict result through integrating info provision and signal control. After that the research framework was constructed. And the key to the study were asissit control and signal control in Encounter stage and Evasion stage. The former was that after driver was gaven information, his behavior would be affected and driving state would be changed to avoid the collision. The latter was that traffic signal controls vehicle which suffered from risk through signal state adjustment or signal optimization.

Considering the accident reasons and collision type at the intersection, this dissertation did some researchs in proactive traffic safety control model and algorithm: (1)In order to overcome the irrationality of vehicle which running through intersection at an even speed, the strategy of proactive traffic safety control for dilemma zone considering driver behavior was put forward. (2)As to provide valuable information for drivers and prevent pedestrian-to-vehicle collision and guarantee the pedestrian to be safe, a method for detectting pedestrian, acquiring the position of pedestrian and prejudging the potential pedestrian-to-vehicle conflict based on the zebra stripes feature was presented. The simulation for detecting and predicting method was performed and the method has been proved to be effective. (3)In the actual cases, vehicle speed usually changes and distributes randomly in the process of approaching intersection. The collision risk model between left-turn vehicle and straight vehicle was established to reflect this feature. (4)After the environment ofProactive traffic safety control at intersectionwhich could be realized was analyzed and discussed, the prototype experiment system in vehicle infrastructure integration environment was founded on the background of the existing technology. Then the performance of each subsystem was tested. The results show that it can carry out the control signal delivering, but positioning accuracy and communication quality should be improved further. (5)Driver behavior affecting Change interval was discussed. Collision risk was estimated in the course of phase-link-up through dynamic collision risk model. If risk estimatation value is under the threshold, vehicle will cross intersection safely during the change interval. If risk estimatation value is above the threshold, the change interval should be extended or should alarm in order to ensure vehicle safety. (6)The collision risk model between left-turn vehicle and straight vehicle was modified. Then collision risk value collision of different phase-sequence was analyzed and it provids a basis for phase-sequence design. (7)Conflict between traffic flow was resolved by optimization idea. In order to improve the deficiencies of previous research, a new conflict point delay model at two-phase intersection was put forward and the relation among volume, conflict and delay was analyzed. Then traffic signal cycle was optimized using minimum delay as target function to reduce the number of conflict.

Finally, a summary is made to generalize the work in this dissertation. The innovations are listed and suggestions are proposed for further research.

Key Words: Proactive Traffic Safety Thought, Proactive Traffic Safety Control, Traffic Safety Process, Experimental System Platform, Control Strategy, Assist Control, Signal Control, Collision Risk, Conflict Point Delay

莫汉康

入学时间：2007级

答辩时间：2010年

论文题目：基于活动的居民出行调查基础性问题研究

中文摘要

摘要

人类为满足个人事务和社会参与需要从事各种日常活动，进而产生机动化或非机动化的出行。城市交通问题也正是由于人类活动/出行的存在而变得复杂。在解决交通拥堵、环境污染、交通事故等问题的同时，提供健康可持续、具有较高运行效率及可靠性的基础设施是交通建设的发展目标之一；而交通基础数据，特别是与居民出行相关的数据无疑是实现这一目标的重要基础。因此，自20世纪80年代旅日和旅美学者率先引进居民出行调查及其方法以来，我国居民出行调查得到快速发展，并在交通政策制定、规划方案编制及决策支持等方面扮演着越来越重要的角色。然而，总结我国居民出行调查30年来的实践，其却存在着调查方法欠妥、调查内容烦杂、问卷类型落后、样本量大、抽样及数据扩算方法不统一、调查实施难、调查质量评估标准欠缺等问题。为此，论文在总结国内外大量居民出行调查案例的基础上，借鉴其它调查领域专业调查知识，采用理论与实证相结合的分析方法，对居民出行调查各阶段、各环节涉及的基本概念、原理和方法等基础性问题进行系统研究。

调查方法选择是居民出行调查实践中最为关键的步骤，因为后续工作均围绕具体方法开展，因此论文首先进行这方面的研究，主要是在详细分析询问调查法、观测调查法和组合调查法等各种常用方法优缺点及适用范围基础上，建立了居民出行调查方法综合评估指标体系，指出两次入户面访调查是适合我国国情的基于活动的居民出行调查首选方法。

针对调查问卷设计，论文从基于活动的交通需求预测模型体系数据输入需求角度出发，结合国内外实践经验，提出我国基于活动的居民出行调查的基本调查项和备选调查项；并系统研究如何将推荐调查内容转化成问题及备选答案、以及如何进行问卷结构设计、问卷类型选择等；指出系列问题类型的调查问卷有利于改善居民出行调查结果质量，建议推广使用。

围绕调查抽样设计，论文首先从居民出行调查总体概念出发，介绍如何界定居民出行调查对象、调查范围及时间。其次，系统研究居民出行调查抽样框，指出区域框是我国目前最常用的抽样框。然后，基于抽样理论，系统分析各类概率和非概率抽样方法的特性及适用条件。最后，在分析调查误差和抽样估计基础上，提出基于活动的居民出行调查样本量确定方法，并针对我国居民出行调查样本量确定的特点，给出实用抽样率模型。

在调查实施研究方面，论文首先对居民出行调查活动的实施主体及其职责进行分析；接着详细探讨调查人员和数据处理人员队伍组建、配备规模、素质要求等问题；其次，系统分析如何有效组织实施居民出行调查培训、开展调查宣传等；最后，对居民出行调查质量控制原则和方法、试实验和预调查的样本量确定及结果评估等问题进行重点研究。

对于数据处理问题，论文着重探讨了数据编码、数据审核与插补、数据扩样等的基本原理和方法，旨在为数据处理人员提供技术支持和现场操作指导。

调查质量评估是居民出行调查的最后一个环节。本文建立了由回答率、数据缺失率、数据修正率、问卷代填率和抽样框误差率等组成的居民出行调查质量评估指标体系，并给出各指标的具体计算方法。

最后，论文通过一个案例研究，分析说明与基于出行的居民出行调查相比，基于活动的居民出行调查更能显著提高调查的数据质量。

上述所有研究内容，均以提高居民出行调查实施效率、改善数据可靠性和准确性为基本出发点，研究成果弥补了我国在基于活动的居民出行调查领域的欠缺，可为缺乏居民出行调查实践经验的城市和地区提供技术指导，亦可为未来我国居民出行调查的规范化和实施指南编制提供借鉴与参考。

关键词：基于活动的居民出行调查；调查方法；问卷设计；抽样方法；调查实施；数据处理；调查质量评估

英文摘要

ABSTRACT

To meet individuals’ affairs and daily social communication needs, people make short distance or long distance, motorized or non-motorized travel. Urban transportation problems are becoming complicated nowadays due to human activities / travels. Providing sustainable development, high reliability and low cost of travel is one of the transport infrastructure construction and development goals in solving traffic congestion, environmental pollution, traffic accidents and other problems. In order to meet this goal, ‘Personal Travel Data’ is the key foundation data on all of the basic traffic support data. Therefore, follow by the academics from Japan and American start to use personal travel survey from 1980’s, this survey methods has been rapid growth and development in China, and it also play a important role in transport policy formulation, transportation planning and other aspects. However, to sum up our country’s last 30 years of practices in personal travel survey, we find out so many serious problems, such as, improper survey methodology, complicate survey content, unenlightened questionnaire type, large sample size, non-uniform sampling and data expansion calculation methods, difficult to implement survey, lack of criteria of survey quality assessment, etc. Therefore, this paper is to complete a systematic study by concluding a large number of personal travel survey cases, referring to professional knowledge of the survey in other areas, and based on the knowledge of statistics, sociology, psychology, traffic engineering, using a combination of theoretical and empirical methods of analysis, survey implementation at all stages, involving various aspects of the basic concepts, principles and methods.

Survey method is the most critical step in personal travel survey, due to the follow-up survey work are carried out around specific method. Paper analyzed the advantages, disadvantages and application scope of inquiry survey method, observation survey method and their combinations. Paper also established comprehensive evaluation index system for travel survey methods selection, Survey shows two-stage in-home (TSIH) survey is suitable and preferred method for China's activity-based personal travel survey.

In the chapter of questionnaire design, start off the traffic demand forecasting model system data input requirements, combined with practical experience in the investigation at home and abroad, research recommended the basic survey items and alternative items of activity-based personal travel survey investigating contents. Systematically study about how to convert the recommended investigation content into possible questions and answers, questionnaire structure, questionnaire type selection and other issues. In order to point out series type of survey questionnaire to improve the quality of personal travel survey, then promote to use.

In the chapter of sampling design, start off from the concept of personal travel survey, introduced how to define the object of travel survey, survey scope and survey time. Secondly, study of the structure of sampling to point out a mostly suitable sampling structure. Then, systematically analysis various types of probability and non-probability sampling methods and its applicable conditions based on the sampling theory. Finally, based on the analysis of survey errors and sampling estimates, proposed activity-based personal travel survey to determine the method of sample size, and determine the practical sampling rate model according to the characteristics of personal travel sample size determination in our country.

In the chapter of survey implementation, paper start off from analysis of the survey main body and its duty, discus the issues of team setting up, size, quality requirements and other issues for investigators and data processing. In the mean time, analyze about how to effectively organize the training and publication for survey implementation. Finally, focus on the study of quality control standards and methods, pre-trial investigation, sample size determination and its results assessment.

In the chapter of data processing, paper focus on discussion of the basic concept and analyze methodology of data coding, review, data expansion, in order to provide technical support and field manual for data processing technicians.

In the last chapter of data quality assessment, paper established the quality evaluation system of response rate, data missing rate, data correction rate, proxy reporting rate and sampling frame error rate, then point out the detailed calculation method.

Finally, paper gave a case study and point out the improvement on the results by using activity-based personal travel survey rather than travel-based personal travel survey.

Summaries, researches carried in the paper are to improve the personal survey design and implementation efficiency, it also improves the reliability and accuracy of data collection. The result is to provide the technical support for some cities which lack of experience for personal travel survey. It also can be used for the personal travel survey standardizing and provide a guideline for future surveys implementation with a strong theoretical and practical value.

Key Words:activity-based personal travel survey; survey method; survey instrument design; sample method; survey implementation; data processing; survey quality assessment

姚佼

入学时间：2007级

答辩时间：2011年

论文题目：基于车辆行驶数据的交叉口交通控制机制研究

中文摘要：

摘要

交通控制是调控交通流，改善其阻塞，提高安全性乃至节能减排的极为重要的措施之一，其进步和发展始终与信息、通信、计算机技术以及系统工程同步。由于以往交通信息采集、处理和提供技术等的局限，交通控制无法真正实现高效、稳定地对交通流及其状态变化的实时响应与控制，导致城市道路交叉口高峰时段排队溢出、绿灯信号得不到充分利用、交通安全隐患重重。车路协调系统的出现，使得单车实时行驶数据的获得成为可能，交通信息真正实现了从时间维的线采集到时间、空间两个维度上的全天候、全方位无缝覆盖的全时空采集，可以很大程度上克服传统交通控制的弊端，真正意义上实现城市道路交叉口的自适应控制。资深学者Michael G. H. Bell在发表于Transport Research Part A的文章“Future Directions in Traffic Signal Control”中指出，仅仅依靠交通信号本身已经不能够解决日益严重的交通拥堵问题，与实时车辆行驶数据的集成及整合将是交通控制发展的必然趋势。作为车路协调环境下交通控制研究的第一步，研究基于车辆行驶数据的城市道路交叉口交通控制机制，对于拓展交通信号控制新领域具有十分重要的理论意义，同时对于进一步提高交叉口运行的安全性和效率性，以及预防与快速排解交通拥堵等也具有重要的理论应用价值。

论文首先综述了城市道路交叉口控制理论研究的现实需求背景和理论研究背景，并在总结分析国内外现有相关研究成果的基础上，指出了在车路协调环境下其未来发展的方向和定位。

进一步，从车辆行驶数据的解析入手，通过建立交叉口车路协调实验平台，获取实际的车辆运行数据，从准确性、时间特性和空间特性三个角度进行分析，抓住本质问题，分析并提炼出车辆行驶数据生成频率和车载设备市场普及率分别作为车辆行驶数据时间特性和空间特性的关键因素。其次，通过对交叉口交通控制机制的解析，分析了车辆行驶数据对其的内在影响，并基于此，提出利用指数加权平均控制的方法对车辆在交叉口处的转向和变道行为进行识别，在其中提出通过扩展卡尔曼滤波法消除由于车载设备存在的绝对误差，通过线性规划模型消除车辆转向带来的地图匹配相对误差；此外，分三种场景对交通控制信息对驾驶员在交叉口停车线前的反应行为进行分析，为后续章节的研究奠定了坚实的基础。另一方面，亦从交叉口交通控制的实际功能出发，提出了不同功能模块对车辆行驶数据的具体需求。

再次，关于基于车辆行驶数据的城市道路交叉口控制，离线方面，通过对存档车辆行驶数据的分析，挖掘出传统离线控制的输入，进而提出通过层次聚类分析的方法，确定交叉口离线TOD（Time-Of-Day）控制的最优方案数以及最佳时间切换点，进一步通过分类回归树法及时发现控制方案的老化，对其进行及时更新；在线方面，分场景提出了不同的自适应控制策略，在非饱和条件下，从相位切换顺序、车队聚类、相位切换判断三个角度给出了交叉口实时自适应控制的逻辑算法；在过饱和条件下，提出了绿灯提前切断、绿灯延迟开启、相交道路协调三种控制逻辑及这三者的五种组合算法，并通过案例分析给出了推荐的逻辑算法。在控制策略的制定中，考虑到车辆行驶数据生成频率、车载设备普及率这两个关键因素对控制策略效果的影响，对其进行了敏感度分析，给出满足交叉口两种策略性能要求的建议车辆行驶数据生成频率和车载设备最小市场普及率，为下一步车路协调系统的逐步普及和推广奠定坚实的基础。

本研究通过对车辆行驶数据特性的研究，对其进行挖掘利用，制定相关的交叉口离线TOD和在线实时自适应控制策略，不仅可以提高交叉口交通的安全与效率，提高交通管理与控制的主动性，为发展车路协调环境下的下一代道路交通控制系统提供坚实的基础，而且对于研究车路协调系统环境下交通网络协调控制与诱导管理整合等问题亦具有广泛的借鉴作用。受研究时间以及实际客观条件所限，研究尚有一定缺陷和不足，如车载设备不同的市场普及率阶段下不同车辆行驶数据生成频率对于控制策略的共同作用结果，基于车辆行驶数据的交叉口交通控制策略实时评价等都将是后续研究的重要内容。

关键词：车辆行驶数据特性，车路协调系统，离线TOD控制，在线实时自适应控制，车载设备市场普及率

英文摘要

ABSTRACT

Traffic control is one of important measures to adjust traffic flow, mitigate congestion, improve safety and reduce emission at intersection. Its improvement always keeps pace of the development of information，communication, computer technology and system engineering. Limited by the disadvantages of traditional traffic information collection, process, traffic control now still can’t adjust the traffic flow and its status efficiently and stably, not to mention the real time response and optimization, all which result tophenomena such as spillback, block, green starvation, and numerous potential safety hazards at intersection in urban area. With the emerging of VII(Vehicle Infrastructure Intergration), we can get probe data, which means we can get the taffic inform from the temporal demension dimension only to all day, all location seamless covered. In this way, a large part of drawbacs in traditional traffic control will be overcomed, which means the real adaptive control at isolated intersection in urban area will come ture. As the senior scholar Michael G. H. Bell stated in his paper “Future directions in traffic signal control” which published in Transport Research Part A, it’s impossible to solve more and more serious congestion problems only depending on signal control, and it’s the inevitable trend to integrate probe data with traffic control. So in this background, as the 1^ststep of traffic control study in VII (Vehicle Infrastructure Integration) , research on probe data based traffic control mechanism at isolated intersectionis meaningful to extend the new area of traffic control, and also valuable to avoid and mitigate traffic congestion.

To begin with, we first summarized the research requirement and background of traffic control at isolated intersection, and finished a overall literature review of relative reseach achievements home and abroad, based on which we pointed the its opportunity and direction in VII in the future.

Moreover, vehicle-infrastructure integration system experimental platform was first established to get the vehicle probe data. And then detailed discussion on its accuracy,temporal and spatial characteristics was done, andvehicle probe data generation frequencyandonboard equipment market penetration ratioare refined as the key factora to represent its temporal and spatial characteristics. And then afteranalysing theconcept of traffic control mechanism at isolated intersection, we studied the influence of vehicle probe data on traffic control, in which Exponentially Weighted Moving Average was used to identify vehicle turning movement and lane change at intersection, in which Extended Kalman Filter was uesd to eliminate absolute error caused by GPS equipment on vehicle, and linear programming model was established to relative error of map matching when make turning at intersection. In addition, influence of traffic control information on driver behavior was studied at stop line in 3 scenarios, all these made firm foudation for the research in following chapters. On the other side, detailed requirement of vehicle probe data was also brought forward from the view of practical functions of traffic control at isolated intersection.

Furthermore, regard to the control strategies at isolated intersection in VII, about off-line ones, probe data from database were mined to input the traditional off-line optimization software, such as SYNCHRO, and then we determined the optimal TOD (Time-Of-Day) plan numbers and the switch points among them with the method ofHierarchical Clustering, and aging state of existing plans was identified with the method of CART（Classification and Regression Trees）and updated in time. About on-line ones, with the scenario that the degree of saturation smaller than 1, real time adaptive control logic at isolated intersection was given, which take into account following factors, phase switch sequence, platoon classification and phase switch judgment logic, with the scenario of over saturation, three control logic: GCA(Green Cut in Advance), GSD ( Green Start Delay) and COR(Coordination) and five combination control logic from them were brought forward. And next through case study, we gave the recommended one. In addition, considering the factors during the popularization of VII mentioned above, we did some sensitivity analysis and relative effect study of different probe data generation frequencies, onboard equipment market penetration ratios on control strategies proposed above, based on which the suggested probe data generation frequencies and minimum onboard equipment market penetration ratio which can satisfy the requirement of control strategies were given, all which made the firm foundation for VII popularization in the future.

In one word, all achievements in this dissertation will not only improve the safety and efficiency of traffic operation at isolated intersection, enhance the initiative of traffic management and control, and make a firm foundation for the development of next generation traffic control system, but also can used for reference to some problems, such as combination of traffic network coordination and guidance management. However, with the limitation of research duration and objective conditions, there are also some drawbacks in this research, such as combination impact of differernt peneration ratios of onboard equipment in vehicles and generation strategies of probe data on traffic control strategies, probe data based real time MOE(Measurement Of Effectiveness) of traffic control strategy, etc., all which we should focus on in the future.

Key Words:Vehicle Probe data, VII(Vehicle Infrastructure Integration),off-line TOD control, on-line real time adaptive control, onboard equipment market penetration ratio