童梅
入学时间：2005级
答辩时间：2008届
论文题目：面向实验交通系统的建模与计算
中文摘要
摘要
随着现代科学技术特别是ITS的应用和发展，现代交通系统变得越来越复杂且具有不可还原性，传统的研究方法已经越来越不适应现代交通系统的特性变化，同时交通工程需要大量的实践或实验来对交通参与者行为进行建模与分析，对交通设施进行评价与改善，对交通管理与控制效果进行预估和评判，实验交通工程应运而生。本论文基于实验交通工程学理论体系，从实验科学的基本原理入手，建立实验交通系统的框架，提出了基于计算实验平台的主动计算实验及基于现实交通系统实时数据的被动交通实验相结合的实验交通系统的通用架构；通过与传统交通系统建模对比突出实验交通系统的建模特点：基于数据的建模，并提出了实验交通系统的四个模型层次及其相互依存关系；提出了面向实验交通系统平台的综合建模过程以及实验交通系统的数据建模框架。由于系统的主要输入是海量的实时交通数据，数据的存储和处理从规模、精度和速度均有较高的要求，因此系统以分布式并行技术为计算支撑，并引入了计算智能技术来对海量的数据进行挖掘分析，提出了实验交通系统平台级的并行计算解决方案，架构级的分析和设计解决方案，分析和决策层的计算解决方案，实验层的多智能体计算实验研究以及数据可视化的解决方案。系统的理论和方法在实际应用研究中得到验证和发展。本文以基于网格技术的交通信息服务示范系统为应用背景，开发了面向交通信息服务的实验交通系统原型，并以城市道路交通状态在线解析为例，从微观解析、中宏观解析及对未知现象的探索三个方面实践了基于实时数据解释和观察的被动交通实验过程，验证了实验交通系统框架与建模理论的科学性和有效性。
论文首先通过大量的文献阅读和资料总结，介绍了实验交通工程的起源和发展，实验科学及其在社会复杂系统中的应用，以及国内外类实验交通系统的研究进展，并就实验交通系统实施的关键技术：复杂系统理论及建模技术，数据管理技术和计算智能技术的在交通系统中的应用做了回顾，确定了论文的突破方向为实验交通系统框架研究及其面向实验交通系统的建模方法和计算技术。
接着分析了ITS环境下交通系统的研究对象的变迁以及交通系统的新特性，从实验科学的基本概念出发分析研究了实验交通及其系统的基本原理，对实验交通系统进行了需求分析研究，确定系统的用户与角色，分析研究了系统的功能需求、建模需求、计算需求以及性能需求，从而导出系统的逻辑结构、物理结构、信息流，再由系统的逻辑结构确定了系统的研究内容，并根据系统的技术特征引出对实验交通系统的关键基础问题及关键理论和方法的论述，就实验交通系统的关键的建模、实验、决策、计算和集成问题以及实现系统的关键技术展开研究，以面向交通信息服务的建模与计算作为论文的研究重点。
对于实验交通系统的建模，论文分析了传统交通系统建模的类型、目标以及对应模型的假设条件或缺陷，对比传统交通与实验交通系统的建模过程，并以行程时间的估计为例说明的实验交通系统建模的特点：基于数据的建模。从数据流转的角度将实验交通系统的建模分为四个层次：信息融合模型、实验模型、分析模型及决策模型，并描述了四者的相互依存的关系。实验交通系统需要一个物理平台来支撑。文章从软件工程的角度出发来描述实验交通系统平台的建模过程，分析了自顶向下和自底向上的两种系统建模设计方法，提出了面向实验交通系统的综合建模过程。数据建模是实验交通系统的重要部分。文章阐述了实验交通系统中与GIS系统结合的静态数据建模、交通在线解析系统的数据建模要点、交通动态计算实验系统的数据建模要点以及操作型数据管理与分析型数据管理相结合的一般做法。
对于实验交通系统的计算，论文从面向系统以及面向模型两个层面分析研究实验交通系统的计算需求。面向系统建模方面，首先分析了系统平台的计算需求，并将并行计算、分布式计算与网格计算进行了介绍，详细阐明了三者之间的关系，并提出了实验交通系统系统平台级的并行计算解决方案；其次面向系统的架构提出了实验交通系统平台中各个系统层次采用的系统计算方法：总体架构采用面向服务的计算方法，交通动态计算实验系统采用面向多智能体的计算方法，交通在线解析系统采用面向对象的计算方法，系统的模型求解采用结构化计算方法。面向模型的计算方面，首先对实验交通系统的分析和决策层，主要考虑的计算模式是联机分析处理和数据挖掘，并给出了如何实现联机分析处理及数据挖掘的解决方案；对于决策层，研究了实验层的建模需求，并介绍了基于多智能体的建模技术与平台，以及基于多智能体的交叉口控制人机交互模型实现的研究成果。数据可视化能使实验交通系统的实验和计算结果能为分析和决策者理解和应用。在实验交通系统的七个数据处理阶段，分别采用日志文件、GIS、统计报表、联机分析处理以及通过友好的人机交互界面来输出结果。
论文最后基于实验交通系统的框架，以交通信息服务示范系统为应用研究背景，介绍了面向交通信息服务的实验交通系统的建模过程，包括系统的物理模型、数据模型及实验模型，详细研究基于出租车GPS数据的路网交通状态估计实验模型，分析研究了网格计算、并行计算技术以及数据挖掘技术在实验交通系统中的应用，给出了系统实现的运行环境、物理部署、运行结果，最后以城市路网交通状态在线为例详细描述了基于被动交通实验的交通在线解析系统的三种解析方式及数据分析结果。应用研究证明，实验交通系统的建模方法比传统交通系统建模更具有科学性和有效性。
关键词： 实验交通工程，实验交通系统，实验科学，复杂系统，系统建模，计算科学，并行计算，数据挖掘
英文摘要
ABSTRACT
As modern science technology developing, especially the ITS systems applying, modern transportation systems become more complex and unreductive. The traditional study methods can not catch up with the changes of modern transportation systems. Furthermore, in order to modeling and analysis behaviors of traffic participants, evaluating and improving traffic infrastructure as well as estimating and assessing effects of the traffic management policies, massive practice or experiments are required in traffic engineering. Here comes the Experimental Transportation Enginnering (ETE). Based on theory of ETE and the basic theory of experimental science, we construct a framework of Experimental Transportation Systems (ETS) and put forward an ETS modeling method combined with data-oriented bottom-up method and task-oriented top-down method. We study about the system in function modeling and data modeling in detail. As the system inputs are massive real-time traffic data, requiring high computing performance in scale, precision and speed while storing and processing the data, we introduce parallel computing to support the computing platform, also compute intelligence to mining and analyzing the massive data. We verify and improve our theory in real application studying and apply a passive experiment using the constructed prototype of ETS.
First, according to massive bibliographic data, we introduce the origin and development of ETE and application of experimental science to social complex systems. We summaries the international semi-ETS systems development progress and review the key technologies in ETS, complex theory and modeling technology, data management technology and computational intelligent technology. We focus on ETS and relative technologies in modeling and computing.
Second, after analyzing of study objects under ITS background and the new characters of transportation systems, we illustrate the basic theory of experimental transportation as well as relative systems according to the concept of experimental science. Then we analyze the original requirement of ETS from aspects of system users or roles, system functions, modeling, computing and performance, consequently deducing the logical architecture and physical architecture of ETS. According to the architectures, the study contents are decided and according to the technical character of ETS, we probe into the key infrastructure, theory and method, developing the key parts of the study from system modeling, experimenting, decision making, computing, integration of the system and relative key techniques. We emphasize our study on system modeling and computing.
Third, in analyzing system modeling in chapter 4, we compare modeling process of traditional transportation systems and ETS and put forward a combined method of top-down and bottom-up. We suggest the modeling architecture of the system and analyze their relationship. We differ the two transportation modeling methods, traditional and of ETS, taking traffic flow model as an example, and put our attention on data modeling of ETS, by discussing solutions of static data models with GIS, data models of online traffic analysis system, data models of dynamical computational ETS and management of operational data and analytical data.
Fourth, in chapter 5, we induce all possible involved techniques in ETS according to its modeling system from system level, architecture level, and analytical level to data visualization. We give solutions for all levels. For the system level, we analyze the relationship among parallel computing, distributed computing and grid computing and point out our solution in ETS. For the architecture level, we apply service-oriented architecture design, multi-agent computing, object-oriented design and structure modeling to the design of system architecture of ETS. For the analytical level, what we concern about are online analysis and data mining, and we also give relative solution on it. For data visualization level, we use logs, GIS, reports, online analysis a friendly interface to show the results in seven phases of data processing.
At last, based on a case study of a grid-based traffic information system, we put the theory of ETS into practice. The system modeling processes, including business models, data models and function models are presented. A taxi-GPS-based network traffic state estimation model is described in detail. Grid computing, parallel computing as well as data mining techniques is applied into the system. The operation environments, deployment of business modules, results as well as experimental analysis are presented in the case study. We also illustrate three modes of traffic online analysis in ETS by a road network traffic state online analysis system in the case system. The application study proves that modeling methods of ETS are more effective than the traditional ones.
Key Words：Experimental Transportation Engineering, Experimental Traffic Systems, Experimental Science, Complex System, System modeling, Computational Science, Parallel Computing, Data Ming
沈峰
入学时间：2005级
答辩时间：2008届
论文题目：城市道路交叉口群交通控制模型算法及其实现
中文摘要
摘要
缓解城市交通拥堵的问题涉及城市政策法规、城市规划建设、城市交通管理等多个层次多个方面的问题。而城市道路的信号控制技术，作为交通管理的主要手段，其智能化和现代化水平充分体现了交通管理的社会生产成本，而在此背后，强大的理论和技术支持才是其核心竞争力的体现。城市道路交叉口群是城市交通问题的微观体现，因此在追求城市交通可持续发展的大背景下，跨学科综合交叉相关领域的科研成果和技术，分析和描述城市交叉口群交通流运行的特征，为其设计适用的交通控制算法并应用先进的软件技术进行实现对解决城市交通问题具有一定的理论价值和现实意义。
本文首先对从国内外在城市交通控制相关研究的成果进行了总结和综述。从单点控制到协调控制，从定时控制到动态控制多个角度对城市交通控制模型、算法以及系统实现相关的研究成果进行了总结和分析，提出了交通控制的研究应不局限于针对局部问题进行孤立建模分析的研究思路，必须全面考虑城市道路交叉口群交通流在控制条件下的表现特征，从而引出了本文研究的主要切入点，即控制条件下的城市道路交叉口群的建模问题研究。
在第三章中，本文对控制条件下的交叉口延误模型、交叉口群交通流特征以及交叉口群控制系统的检测器布设方案进行了总结和定性化的分析，为后文的研究提供了依据和支撑。
在第四章中，本文对城市道路交叉口群的建模问题进行了研究，首先对信号控制条件下的城市道路交叉口群的组成要素进行了分析和归纳，将交叉口群分为道路路段、道路交叉口以及信号控制三个组成要素。并对现有的城市道路交通流模型进行了细致的分析和总结。提出了应用细胞传输模型(Cell Transmission Model)作为城市道路交叉口建模研究的基础，并对细胞传输模型对于描述城市道路网络间断交通流的适用性进行了分析。针对细胞传输模型无法描述城市道路间断交通流在交叉口内部受到信号控制影响而存在的通行能力变化和车流在道路路段上的离散现象，应用了时序Petri网模型(Timed Petri Net)以及对CTM模型中原有的流量传播模型进行了改进，并进行了相关的实例验证。
在第五章中，本文基于前述交叉口群的建模研究结果，在分析了交叉口群控制模型的一般化描述的基础上，提出了基于交叉口群仿真模型的多目标优化控制算法。算法将交叉口群仿真模型嵌入到多目标优化控制算法中，替代目标函数计算方案的评价指标，并采用多目标遗传算法中的非支配排序算法对问题进行求解。为了对算法进行验证，本文设计了多种实验案例对算法进行了验证，对结果进行了分析并证明了算法的有效性和适用性。
在第六章中，本文首先对动态交通控制的原则、目标、时效性、适应规模以及控制结构分类等方面进行了概述，并引出本文动态控制算法的概念：双层控制结构的动态控制策略。将控制策略分为两层：路网优化层与单点优化层。在网络优化层中完成较长时段的优化参数的计算，并根据时段内交叉口群内交通流运行指标来确定各单点优化层的优化目标及约束条件等。单点优化层则关注于单个路口在短期内的控制方案计算，单点优化层包含多个单点控制单元，每个单点控制单元能相互交换信息，进行短时交通流预测，并完成控制方案的滚动优化。在确定了控制结构以后，本文建立了动态控制策略的模型体系，在单点优化层包括：交叉口交通流短时预测模型、交叉口信号配时优化模型、交叉口交通转向估计模型，在网络优化层模型包括：参考周期的计算模型、交叉口相位差的优化模型以及车流指标影响系数模型。本文在分析了动态控制策略在时效性和方案覆盖度的特性，详细设计了模型体系中优化模型的求解算法。为了验证算法的有效性，本文在第四章交叉口群仿真建模研究的基础上设计了多种仿真应用场景，并对仿真结果进行了详细的分析。
第七章中，本文在交叉口群控制系统原型设计的层面对本文所述控制策略在系统实现层次进行了阐述。首先在系统概念、实时系统设计、系统功能分析、系统设计原则、系统总体设计以及系统软件组成等方面对控制系统原型的框架进行了设计，然后对控制系统的核心控制软件的设计与开发进行了论述。最后，对面向交通控制的数据仓库进行了初步的设想和分析。
关键词：交通控制模型，交通控制算法，交通仿真，细胞传输模型，多目标遗传算法，交叉口群控制策略，控制系统原型设计
英文摘要
ABSTRACT
Solving or relieving urban traffic congestion is a complex problem involving urban policy and law, urban planning and urban traffic management. The level of intelligence and modernization of urban traffic control which is the main method of urban traffic management represents defends on the theoretical research of traffic engineering. Urban intersections group is the micro representation of urban road network, so it is necessary to focus on the traffic flow characteristic of intersections group, then design and implement a proper traffic control algorithm combining several spheres of knowledge.
First of all, this dissertation gives an extensive literature review of traffic signal control for isolated intersections and urban network covering static and dynamic solution to the current state of the art. After analyzing the defects of the current research, the dissertation carries out a point of view which urban traffic flow characteristics should be modeled thoroughly instead of eliminating several confounding factors.
Chapter 3 of the dissertation is dedicated to introducing the intersection delay model, intersections group flow characteristics and the layout plan of detector of the control system.
In chapter 4, the dissertation focuses on the modeling problem of urban intersections group. The intersections group model is divided into three parts: road section, intersection, signal control. After analyzing and conclusion of the current popular traffic model, cell transmission model is picked as the basic model. Several improvements are made aiming at that the CTM model can not modeling the variety of traffic flow capacity because of the signal control plan and the platoon dispersion phenomenon. The improvements are verified by simulation experiments.
In chapter 5, a new traffic control algorithm is developed employing the multi objective genetic algorithm. The algorithm embeds the simulation model into the none dominated genetic algorithms to replace the object function. The algorithm is tested by several simulation experiments, and the results show the effectiveness and applicability of the algorithm.
In the beginning of chapter 6, the dissertation gives a brief introduction of the dynamic traffic control including the control principle, control object, control aging effectiveness, control scale and control structure. Then a new dynamic control strategy which composed of two layers: network optimization layer and intersection optimize layer. The network optimization layer process a long period optimization due to the performance index of the network traffic flow. The intersection optimization layer focus on the short period signal plan calculation, plenty of intersection optimization units are designed in the intersection optimization layer. Each intersection optimization unit can communicate with each to process the short term traffic prediction and rolling optimization of the signal plan. The model architecture of the dynamic traffic control strategy is introduced later such as the short term traffic prediction model, signal plan optimization model. Proper algorithms of the optimal model are designed. The strategy is examined by several well designed simulation application tests.
Chapter 7 is the description of the intersections group control system prototype. The prototype concept is introduced, and several system design principle are proposed in the dissertation. Then the dissertation describes the design and deployment of the prototype software. And the traffic control oriented data warehouse is proposed at the end of the chapter.
Key Words: traffic control model, traffic control algorithm, traffic simulation, cell transmission model, multi objective genetic algorithm, grouped intersection control strategy, traffic control system prototype design
张超
入学时间：2005级
答辩时间：2008届
论文题目：面向服务的城市公共汽车交通评价方法研究
中文摘要
摘 要
小汽车交通的盛行带来了一系列的问题：交通拥堵、空气污染、能源消耗、交通事故以及社会的不公平性等等，优先发展公共交通被认为是缓解众多交通问题、节约社会资源和能源、促进交通和城市可持续发展的最主要和最有效的途径之一。在我国大部分城市，公共交通（本文特指公共汽车交通）系统规模巨大，网络复杂，在整个公交系统中占据主导地位。但纵观各城市的公交发展，公交系统的整体服务水平不高，公交出行率不断下降，公交服务水平和公交利用率之间形成了恶性循环的怪圈，因此，面向服务的城市公共交通评价方法的研究变得极具现实意义。
论文主要针对现有理论和方法的空白点和薄弱点，在大量查阅相关文献和研究成果的基础上，从我国城市公共交通服务的机理解析出发，综合运用系统工程理论、交通工程学、管理学、经济学及计算机仿真软件等对面向服务的城市公共交通评价方法展开了系统的研究，以期推动和完善公共交通评价理论，更科学合理地进行公交系统的规划和管理。
论文首先从服务科学的观点层面，对公共交通服务机理进行解析，系统地分析了公共交通服务的形式、组成和特征，确定了面向服务的公共交通的评价内容为两大块：服务水平评价和效益水平评价，论文随后分服务技术性能评价和服务质量评价两个主要部分对公共交通服务水平评价展开深入研究。
公交服务技术性能是公交服务质量实现与提高的基础。论文在前人提出的众多评价指标的基础上，根据一定的筛选方法和原则，建立了面层、线层、点层和配套设施层等四个准则层的评价指标体系，并结合仿真软件TransCAD给出了各评价指标的量化方法。
公交服务质量由两大部分组成：公交运营服务质量和公交感知服务质量，前者是从公交企业的运营角度考虑，后者是从公交乘客的感知角度出发。对公交运营服务质量的评价是从管理学大师马斯洛提出的动机理论逻辑递进层次角度，从安全、可靠、高效和经济四个准则层构建了公交运营服务质量的评价指标体系。公交感知服务质量则是借鉴服务管理学中的服务质量评价模型，结合公共交通实际对已有模型进行修正，提出了适合公共交通感知服务质量的评价模型，并对模型进行了实证研究，验证了模型和方法的有效性。
面向服务的公共交通综合评价是在公交服务水平评价和效益评价的基础上进行的，是对公共交通整体水平的综合评价。本文运用广义函数法建立起公交综合评价模型，对评价指标权重的获取是广义函数实施的难点。本文针对不同的情况分别运用层次分析法和改进的层次分析法建立各评价指标权重的初始判断矩阵，并运用相容矩阵法得出最终的权重矩阵，计算出各评价指标的权重。实例证明：该综合评价方法是有效较易实施的。
提高服务水平必须增加公交投入，要保证投入和产出间的平衡。文章的最后从经济学的角度分析了公共交通的投入和产出，给出了公共交通的短期生产函数和长期生产函数。通过对公共交通的成本和效益的分析，建立了公交企业的利润函数模型，以便公交企业平衡公交服务水平的提高和公交的投入。同时，由于公共交通准公共产品的性质，政府需对公共交通进行一定的财政支持，文中也给出了公共交通的一些政策扶持措施。
关键词：公共交通；服务水平；服务质量；评价方法
英文摘要
ABSTRACT
The popularity of transportation by car has brought about a series of problems: traffic jam, air pollution, energy consumption, traffic accidents, and social injustice, to name a few. Giving priority to the development of public transport is considered one of the most important and efficient way to relieve so many traffic problems, save social resources and energy, and promote sustainable development of transport and cities. In most cities of our country, the public transport (bus transport in this paper) system dominates the whole transport system with its large scale and complex network. However, a general survey of public transport development in all cities will show that, the overall service level of the public transport system is not high, the rate of transit trips is decreasing, and a vicious circle between service level and utilization of public transport has taken form. Therefore, a research of service-oriented evaluation method for urban public transport is of great current significance.
Focusing on the gap and weakness of existing theories and methods, and based on a research into a large amount of relevant literature and study results, the paper starts with an analysis of the service mechanism of public transport, and makes a systematic study of service-oriented evaluation method for urban public transport by employing systems engineering theory, traffic engineering, management science, economics and emulation software, with the purpose of promoting and improving public transport evaluation theories, and planning and managing public transport system in a more scientific and reasonable way.
At first, the paper analyzes the service mechanism of public transport from the perspective of service science. By systematically analyzing the form, composition and features of public transport service, it defines two major components of the service-oriented evaluation of public transport system, namely evaluation of service level and evaluation of benefits. Then, the paper goes into service level evaluation of public transport system from two essential aspects, evaluation of technical performance of public transport service and evaluation of public transport service quality.
The technical performance of public transport service is the basis for realizing and improving public transport quality. On the basis of many existing evaluation indicators, and in accordance with some screening method and principle, the paper establishes a system of evaluation indicators at four criterion layers (surface, line, spot and supporting facilities), and provides methods to quantify all indicators with the help of TransCAD.
Public transport service quality comprises public transport operation service quality and public transport perceived service quality. The former starts from the angle of public transport enterprises’ operations, while the latter from the perception of passengers. In respect of public transport operation service quality, a system of evaluation indicators is set up based on Maslow’s motivation theory and from four criterion layers of safety, reliability, efficiency and economy. As to public transport perceived service quality, a suitable evaluation model is proposed by referring to service quality evaluation models in management science and amending existing models in view of the realities of public transport system. Meanwhile, an empirical study of the model is made to verify effectiveness of the model and method.
The service-oriented evaluation of urban public transport is carried out based on the evaluation of public transport service level and the evaluation of benefits. It is a comprehensive evaluation of the general level of public transport. The paper builds the comprehensive evaluation model by the generalized function method, with which the difficulty is to acquire the weights of evaluation indicators. This paper generates the primary judgment matrix of each evaluation indicator’s weight by means of analytic hierarchy process and improved analytic hierarchy process respectively according to different situations, and arrives at the final weight matrix with the tolerance matrix method. From the successful applications, it is proved that the comprehensive evaluation method is effective and easy to apply.
Enhancement of service level requires increasing input in public transport, and balancing input and output. In the end, the paper analyzes the input and output of public transport from the economic perspective, and provides the short-term and long-term production functions of public transport. With an analysis of the cost and benefits of public transport, a profit function model is established to help public transport enterprises balance improvement of service level and input in public transport. In addition, as public transport is a quasi public product, the government shall provide certain financial support. Therefore, the paper also provides some policy support measures to public transport.
Key words: public transport; service level; service quality; evaluation method
丁威
入学时间：2005级
答辩时间：2008年
论文题目：基于活动的城市居民出行前选择行为研究
中文摘要
摘 要
城市是经济、文化、政治活动的中心，城市交通系统是承载这些活动的基本构件。活动亦称行为，指某主体在连续时间段内实现某种目的的过程。居民在城市的不同场所完成的上班、家务、娱乐、购物等活动以及场所间的空间移动构成了庞杂的城市活动系统的主体，其中在场所间的单向空间移动过程即为出行行为。出行行为作为居民日常活动的派生性行为，是社会经济系统对交通需求的具体表现。完整的出行行为过程包含出行前选择（包括出行目的地选择、出发时间选择、交通方式选择）、出行中选择（包括出行路径选择、换乘选择）等一系列选择行为，每一选择行为都和引起出行行为的本源——活动本身有着直接的联系。基于活动的研究方法，是从活动入手解析出行行为机理的有效途径之一。
本论文应用非集计理论及效用理论，对我国城市居民活动日志调查数据进行归纳分析，基于活动剖析居民出行前行为的机理和规律，构建了居民出行前离散选择模型。本论文研究的重点在三个方面：一是以城市居民个体或家庭为研究单位，解析城市居民活动与出行行为的关系，通过调查数据分析城市居民出行链特征和出行前选择行为特征；二是建立基于出行链的出行前选择行为模型，并对模型的参量进行标定和适用性评价；三是对交通信息提供条件下居民的出行前选择行为特征、影响机理、交通信息效用模型进行分析，为合理诱导居民出行提供相关决策依据。
论文首先对城市居民日常活动与出行行为的关系进行解析。通过居民出行日志调查，从性别、年龄、收入、家庭规模、交通方式、出行时耗等出行链相关属性进行统计分析，从通勤出行和非通勤出行两个方面总结出我国城市居民出行链的基本特征，分析了居民日常活动与出行前选择行为的内在作用机制。
其次，构建居民出行前选择行为分析的框架体系。应用决策理论及效用理论，分析居民活动和出行决策过程，建立居民日活动模式和出行前行为选择集。
再次，建立基于出行链的居民出行前选择模型。主要从出行时间选择、出行方式选择、出行目的地选择层面建立离散选择模型。引入巢式logit模型，建立出行前综合选择概率模型，形成完整的模型体系。并对模型数据搜集和处理、模型的标定以及适应性进行分析。
最后，信息对居民出行行为的影响。先进的交通信息系统（ATIS）条件下，居民出行行为的特征同非信息提供条件下相比出现了较大的变化。分析了信息化条件下居民出行前选择行为的机理，建立交通信息效用模型。研究表明，模型能定量地描述交通信息在减少出行选择不确定性的作用，合理地反映交通信息对于出行决策的影响，为研究提供交通信息条件下的出行奠定了理论基础。
本论文的研究成果应用基于活动的交通需求方法，建立了居民出行前离散选择行为模型。为进一步深入研究提供交通信息条件下的出行前选择行为奠定了理论基础，同时也为交通需求预测的优化利用提供了理论和指导方法。
关键词：基于活动，出行链，出行选择，效用理论，交通信息
英文摘要
ABSTRACT
City is an activity center of economy, culture and polity, and traffic system is the crucial carriers of these activities. Activity means the process of residents realize their some kind of purpose in a continuous time slice. Travel behavior as an activity-derived demand, it is an incorporate of economy demand on transportation system. There is direct touch between all kinds of activity daily space-time distribution and travel choice behavior of residents, such as destination choice, go-off choice, mode choice, and route choice. Analyzing inhabitant’s travel behavior based on travel demand forecasting, finding the key of the transportation problem. Study on travel demand, especially on people’s travel choice behavior, is an important aspect in TDM. The activity-based approach (ABA) is one of the effective ways to analyze people’s travel behavior.
Based on the summarization of activity-based traffic demand theories and methods at home and abroad, this thesis studies the travel behavior characteristics and the choice characteristics found on the characteristic analysis in the unit of individuals or household. The thesis analyzes the residents’ activity log survey and data processing are favor of holding residents trip rule impersonally and puts forward the travel choice behavior model. At last, the thesis analyzes the characters and mechanism of residents’ travel choice behavior considering the traffic information provided, and the traffic information utility model is put forward. Through these researches, we can well understand travel demand deriving from activity demand, people facing the constraint of the time and space, the life style influencing people’s decisions, and the travel decisions influencing each other dynamically and mutually under variable conditions.
Firstly, the thesis parses the mechanism between activity and travel behavior. It describes the travel distribution rule with the resident’s trip behavior log investigation. Characteristic of typical patterns of daily trip-chaining according difference trip intention are developed separately based on work and non-work behavior, then the gender, age, income, household size, travel mode, travel time and other correlative attributes about trip–chaining are analyzed.
Secondly, the thesis introduces the activity-based travel choice theories. The concept and hard core of trip chaining are presented, and resident’s travel requirement, choice contents and travel are analyzed based on microcosmic economy theory and travel utility theory. Then the paper puts forward activity mode and travel choice aggregate for succedent study.
Thirdly, travel choice behavior model based on trip chaining, destination choice model, go-off choice model, mode choice model included, are put forward. We introduceMNL（Multinomial Logit）model into the study, and establishes non-linear nested Logit integrated probability model. And then it analyzes the data processing, calibrates the model by using the Logistic regression method, and evaluates model applicability.
Lastly, the thesis analyzes resident’s travel characteristics under traffic information provided (ATIS) condition. People’s travel behavior changes greatly in ATIS correspondingly by using experience knowledge. Based on reveal mechanism of travel choice behavior in ATIS, we establish traffic information utility models, and explain the effects of each factor on residential activities by doing some related influencing analysis. At last, the sensitivity of traffic information utility affected by the quality of traffic information, the influence of traffic information and experience of travelers are analyzed in this model.
In conclusion, this thesis studies residents’ travel choice rule and model from the view of disaggregated theory and decision theory, and proposed a mathematics expression method for the model. This study not only develops foundation for further studying the travel choice theory in ATIS, but also provides a theory and method foundation for optimizing and utilizing the traffic demand management (TDM).
Key Words: activity-based, trip chain, travel choice behavior, utility theory, traffic information
刘好德
入学时间：2005级
答辩时间：2008年
论文题目：公交线网优化设计理论及实现方法研究
中文摘要
摘要
优先发展城市公共交通系统是解决大、中城市交通问题的最佳途径。在国家大力发展公共交通和提倡公交优先的背景下，本文以国家“十五”科技攻关项目“公交专用道信号优先控制策略与技术研究”和国家自然科学基金重点项目“城市交通网络优化与管理”为依托，围绕城市公共汽车交通网络设计的关键理论与实现技术展开研究，以学术意义和实用价值并重为原则，在公交网络设计方法研究、公交线网优化设计模型与算法研究，公交线路规模与资源配置优化研究，公交线网优化设计辅助决策研究等方面取得了一系列成果。
首先，论文从方法构筑层面，提出了理想条件下的公交网络及枢纽布局模式，并依据不同时期与条件的网络设计需求特点，将公交线网优化设计分为方案改进型和方案生成型两类，并详细分析了每种设计方法的输入和处理流程；其次，论文从网络优化层面，以用户出行时间和未满足出行需求量（无有效公交出行路径出行量）两者费用最小为目标构筑了公交线网优化设计模型，基于候选线路集生成方法的研究，结合公交出行路径搜索与客流分配方法，提出了一种基于路线优选的公交线网优化设计方法，并引入模拟退火拉伸思想，改进了遗传算法的遗传选择操作，实现了一种基于改进遗传算法对公交线网优化模型的求解方法；再次，论文从线路优化层面分别提出了公交线路规模优化与资源配置优化问题，并研究了线路布设、站点布局、车辆配置、运营组织等四个方面的优化问题，重点研究了基于乘客出行距离分布规律，乘客平均出行时间最小化的公交站距优化模型及算法。随后，论文从实现层面研究了公交线网优化设计辅助决策系统的功能需求、总体设计以及开发方法，并基于组件式GIS实现了原型系统的开发。最后，论文对全文进行了总结，指出了论文的创新点，并对有待于进一步研究的问题进行了展望。
本论文的研究成果在理论层面有助于公交线网优化设计、公交线路站点布局优化方案的实施；应用层面为开发公交线网优化设计辅助决策系统提供了方法指导，为进一步开发线网优化决策支持系统奠定了基础，是未来公交网络设计领域研究的重要方向。
关键词：公共汽车交通网络，网络设计，优化，辅助决策系统
英文摘要
ABSTRACT
It is the common sense of all spheres that urban public transit priority is an important method to resolve urban traffic problems. Under the background of our country putting emphasis on developing transit and propose transit prioriy, the dissertation is mainly based on the research achievements of two national projects. One isResearch on the Strategy and Technology of Transit Signal Priority along Bus Exclusive Lanesupported by the Key Science-Technology Project of the National “Tenth Five-Year-Plan”, and the other one isUrban Transportation Network Optimization and Managementsupported by National Natural Science Foundation Committee (NSFC), and it focuses on the critical theories and implementation technologies of Transit Routes Network Design (TRND). A series of research achievements have been fulfilled in some aspects such as TRND method architecture and procedure, transit network optimization models and solution algorithm, scale and resource of transit route optimization and computer aided design for TRND. The main contents are as follows.
Firstly, the dissertation proposed an idealism transit and hubs network with perfect condition and designed two scenarios according different period and condition of transit network design demand: Scheme-Modified and Scheme-Created, whose input data and procedure are discussed in detail. Secondly, a transit routes network optimization design model was established based onminimum user travel time and unsatisfied demand. After research Initial Candidate Routes Generation and Transit Routes Optimal Choice which is related to Transit Assignment. An Improved Genetic Algorithm which imported the idea of Simulated Annealing Extending was proposed which modified Selection Operator of Genetic Algorithm and was used in Transit Routes Network Design. Then, 4 aspects optimization problem of transit network are discussed generally which are route distribution, stop layout, vehicle allocation and operation organization. Especially research focused on stop spacing optimization model based on probability distribution of trip distance for the minimum average trip time. For practice, the dissertation studied the function request, scheme design and developing method of computer aided TRND decision support system, and a prototypesystem Component Geographic Information System oriented was developed. Finally, the summarization of the dissertation is discussed, as the same time, the creative research achievements and other important research orientations was pointed out.
The research results of the dissertation have two main significances. theoretically, it is helpful to Transit Routes Network Design, transit routes and stops layout scheme implementation at the theoretical; practically, it provides method guidance for developing computer aided TRND Decision Support System and proposed future research direction for Transit Routes Network Design Problem – Transit Routes Network Design Decision Support System.
Keyword：Bus Transit Network, network design, optimization, Aided Decision support system
吴震
入学时间：2005级
答辩时间：2008年
论文题目：面向事件的城市道路交通控制理论与方法研究
中文摘要
摘要
城市交通阻塞问题已成为城市化发展过程中最为突出的问题之一，交通阻塞不仅带来了时间、经济、燃料的损失，还污染了环境，并给城市生活带来了巨大的心理压力。交通事件在城市道路交通中时常发生，如车辆碰撞、车辆抛锚、危险物品泄漏、路面维护等等，事件性交通阻塞已成为交通阻塞的重要组成部分，每年都造成巨大的经济耗费。事件发生以后，如果交通控制系统缺乏对应策略来响应事件的发生，就容易造成交通事件影响的肆意蔓延，形成超常排队，最终引发溢流和路网的死锁。本文从阻塞的形成与演变规律、应用于控制的交通状态信息估计与预测，控制模型、算法和控制策略等几个方面进行面向事件的信号控制理论和方法的研究。
针对事件性交通阻塞现象，本文运用交通波及激波理论，深入分析了信号交叉口运行的多种状态：非阻塞运行状态、轻度阻塞运行状态和重度阻塞运行状态，并构建了信号交叉口的阻塞传播与阻塞消散特性模型，以及事件发生情况下事件点（段）阻塞演变模型。通过对交叉口重度阻塞运行状态分析，求得进口道排队系统中“可压缩段”向排队尾部的移动规律。根据交叉口之间的流量与信号控制的关系，分析了上游交叉口与下游交叉口之间及事件点（段）与交叉口之间阻塞演变影响，阐述了阻塞演变与交通控制策略的紧密关系。
交通控制自身的时滞特性需要由交通状态信息预测予以弥补，从而实现控制的实时性。面向事件的动态交通控制策略以交通状态信息为基础，准确地获取控制策略所需的状态信息能够提高在线控制效益。本文尝试剖析交通控制的实质，分析交通状态信息与交通控制的关系，总结了事件发生条件下的交通状态信息体系、特征及相互之间的关系。在交通事件发生的情况下，实时准确地跟踪实际转向比例和转向流量有利于实施灵活的控制策略，响应交通事件的发生。本文分别在粗粒度、细粒度条件下，构建了交叉口转向比例与转向流量的实时估计模型，实现了对应的实时估计算法。
在模型的构建中，针对中国典型特征的交叉口模式，分别构建了相位相序、出口道控制流、车流相位差、车队离散以及车流的延误、排队长度、停车次数等的计算模型。在交叉口基本运行指标的构建中以车流为基本计算单位，区别考虑含事件车流与不含事件车流；构建了以各时段、各类路段、各类车流的加权延误总和最小为目标函数，考虑了过饱和情况下的车流溢流约束的控制算法，并以实数编码的遗传算法（RGA）为主的组合算法对各控制算法进行实现。由于交通事件发生和处理的过程是一个变化的过程，本文考虑了交通事件管理其他措施与控制的关联性，对控制参数进行了多阶段的优化，采用多阶段准动态化优化来响应实际通行条件的变化。
本文结合交通事件发生的特殊条件与一般控制条件，考虑控制参数的优化功能和控制方案的评价功能，对面向事件的交通控制模块进行研究，形成了控制模型和算法。通过含有六个交叉口的路网进行验证和分析，采用的评价指标（MOE）为整个路网的总延误，结果表明本文采用的方法对于事件性阻塞具有较好改善效果。
结合动态控制自身特性与面向事件的控制要求，本文提出了基于规则和优化的混合动态控制框架和流程。面向事件的动态控制流程包含了事件阶段的整体控制优化和阶段内的调整控制优化。在事件发生的各阶段以预案控制库方式采用经过整体优化的初值；而在各阶段过程内，则使用滑动时间窗策略的CTM（Cell Transmission Model）模型进行微量调整优化。
本文的研究成果是城市交通控制系统理论的组成部分，同时也期待能为开发新一代的交通事件管理系统提供理论基础和技术支持。
关键词：交通事件，事件性阻塞，阻塞演变，交通状态估计，交通控制模型，遗传算法，交通控制结构
英文摘要
ABSTRACT
Urban traffic congestion has become one of the most prominent questions in the urbanization process, which not only has brought time, money, and fuel loss, and has polluted the environment, but also has brought huge mental pressure to urban live. Traffic incidents, such as vehicle collision, vehicle malfunction, dangerous good leaking, road maintenance, and so on, often occur in the urban transportation system. Incident-induced congestion has become a big part of the urban traffic congestion and lead to huge economical consumption every year. When an incident occurs in a traffic network, if we have no proper control strategy to respond it, serious congestion, even gridlocks, will be leaded. This dissertation studied the theory and methodology of urban traffic signal control under incident conditions by analyzing congestion forming and transmitting, traffic state information estimating and forecasting for the applies of control, control model, control algorithm and the control strategy.
Aiming at the phenomenon of incident-induced congestion, the dissertation has thoroughly analyzed some running status of signalized intersection such as non-congestion running status, middle-congestion running status and high-congestion running status by using traffic wave and shock wave theory. And then the dissertation constructed the congestion transmission and dissipation model for signalized intersection and for incident location if an incident occurred. By analyzing the high-congestion running status of intersection, a general rule has be modeled that a compressible section in the queue system keeps moving from the stop line to the back of queue. According to the relationship between the flows in intersections and signal control, the influence of congestion both between upstream intersection and downstream intersection and between incident location and adjacent intersections have been analyzed. The dissertation elaborated the close relation in the congestion evolution and traffic control strategy.
The time-lag characteristic of traffic control needs to be remedied by good estimation and prediction of traffic information status, so that real time control can carry out. Dynamic control strategy under incident condition based traffic state information, and accurate state information for control strategy could improve the benefit of online traffic control. The dissertation attempted to anatomize the essence of traffic control and the relationship of traffic state information and traffic control, summarized the traffic state information system and the relation among them. With the condition of traffic incident, it is helpful to flexible control strategy responding incident that turning proportion and turning flows are tracked accurately in real time. The dissertation respectively developed the real time estimation model and algorithm of turning proportion and turning flows in bigtime granularity and in small time granularity.
In the model construction, the dissertation analyzed requirement and characteristics of the intersection group division under traffic incident conditions. According to the types of the Chinese typical intersections, models of the phase sequence, control flows at exit lanes, offsets, vehicle platoon dispersion, vehicle delay, queue length, and vehicle stops are constructed respectively where movement, including incident movement and non-incident movement, is the basic unit in calculation of performance index. The sum of weighted delay for all kinds of movements in every time intervals was minimized in the objective function. And the overflow constraints under over-saturated conditions are considered in the model. Then a real-code genetic algorithm was presented to implement the optimal control under incident condition, in which we used a penalty function to ensure constraints. For the incident process is a dynamic changing process, the dissertation considered the relationship of traffic control and incident management, and carried out a multi-stage optimization which could respond the changing traffic condition semi-dynamically.
This dissertation studied the traffic control module which could run under both incident condition and general condition, considered the ability of both control parameters optimization and control plan evaluation. After that, a road network covering six intersections is used for model testing, where the measure of effectiveness is the whole network delay. The result shows a good improvement in alleviating incident-induced traffic congestion for urban traffic network.
Considering dynamic control characteristic and control requirement under incident condition, a mix control frame based on both rule and optimization was proposed. The control frame under incident condition included the whole optimization and inner stage adjusting optimization. Whole optimization occurs at the beginning of each stage by using predetermined control plan storehouse, and adjusting optimization occurs inside stage by using cell transmission model with time window gliding strategy.
Effort of this dissertation belongs to urban traffic control theory, and is hoped to provide technical support for development of new incident management systems.
Key Words: traffic incident, congestion transmission, traffic state estimation, traffic control model, genetic algorithm, structure of traffic control
