杨帆
入学时间：2010级
答辩时间：2015年
论文题目：基于实验与解析的城市道路网络交通流演化理研究
中文摘要
摘要
城市道路网络交通流是汽车交通流在地面道路网络中受路网拓扑结构属性、交通信号控制、交通信息诱导以及出行决策等因素的影响而衍生出的具有时空变化特性的流体。它是一种间断流，其特征体现在：在相同的交通流量水平下，车辆的行程时间因受到信号控制等因素的影响而导致其行程时间方差较之连续流情况下的行程时间方差更大。该特性会直接影响到交通出行者的路径选择以及宏观交通流的演化。城市道路网络交通流的演化则是由大量的个体交通的路径级出行活动相互交织、相互影响而形成的网络级交通现象。网络交通流演化是交通状态、特征或者某一交通参数随时间序列变化而在时空上的变化情况，具体的表征包括：交通状态的转移；特定网络区域内的交通拥堵的形成和消散；路网可靠性的时变情况等。
本文以国家高技术研究发展计划（863计划）项目专题“道路交通状态全息感知与交互处理技术和战略实验室”为依托，利用实验交通工程学和数学解析相结合的方法，以间断流行程时间作为贯穿全文微观交通数据分析到宏观交通流演化机理研究的核心数据。以实验交通工程学的方法论分析处理城市道路网络间断流行程时间数据，发现其特有的存在形式和分布特征，并利用解析的方法实现从现象到本质的转变，揭示城市道路网络间断流的内涵和外延。在宏观间断交通流演化方面，以坚实的本质特性分析为基础，利用解析方法在目前尚无法完全获取网络级间断交通流演化数据的背景下对其演化机理进行理论分析。
论文对城市道路网络交通流的概念进行了深入解析。通过对现实交通现象的观察和分析，指出城市道路网络交通流是一种间断交通流，其受到信号交叉口的控制影响，使得交通流的速度和行程时间都呈现出特有的分布特征。利用南京市的城市道路路段单元的行程时间数据分析，印证了上述结论，并明确以间断流行程时间数据作为本文的数据源，以贯穿微观到宏观的交通流机理分析，为后续研究提供了坚实的现实基础和数据来源。
在上述分析的基础上，论文对间断流行程时间进行了研究。通过基于RFID技术的城市地面道路网络路段单元间断流行程时间分布的分析，揭示城市地面道路路段单元的行程时间因受到信号控制延误的影响均呈现双峰分布特性。在本文的研究中，利用正态分布和对数正态分布可以很好的拟合这样的双峰分布。在此基础上，对间断流行程时间可靠性进行了分析，提出了适用于间断流行程时间的可靠性指标体系，并挖掘出间断流行程时间中“快速流”行程时间和“慢速流”行程时间之间的特征值——临界时间（CRS_TT），为后续的研究提供了良好的数据和理论依据。
基于间断流行程时间研究，论文对间断流条件下的交通流加载方法进行了研究。首先提出了城市道路网络中考虑信号交叉口等待时间的静态路径搜索算法。其次，通过实验交通工程学的数据分析手段，提出了城市道路网络间断流行程时间函数，将上述的静态路径算法中的等待时间扩展为宏观的延误指标，并提出相应的路径搜索算法。研究表明，信号交叉口的转向延误对出行者的路径选择产生了很大的影响，导致路径所有结果明显不同于传统网络中的路径搜索结果。最后基于间断流行程时间函数，提出了城市道路网络间断流的交通加载算法，是一种非均衡的加载方法。通过与传统交通分配算法的对比，凸显出间断交通流在网络上分配时的不均衡性和随机性，以及信号延误对于出行者的路径选择以及交通流演化的影响。
论文基于等时线的方法对上述城市道路网络交通流演化的数值结果进行分析。在上述中微观城市道路网络间断交通流研究分析的基础上，利用等时线这一新颖的研究工具，从网络层面针对单一交通源点的网络交通状态进行分析。提出了等时线的定义以及四个重要参数：等时线密度、等时线角、等时线交通平均扩散速度和等时线空间平均扩散速度。根据时间属性，提出了设计等时线、静态等时线和动态等时线三种不同类型的等时线，针对不同的分析需求对网络交通流的演化进行分析。研究结果表明，等时线能够直观地反映出交通流在时空上的分布特征，并且可以表征出连续时间段内网络交通流演化的特性。通过等时线可以从网络层面对交通流演化进行分析，为交通管理和控制提供了新的分析工具和思路。
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：城市道路网络，间断交通流，行程时间，双峰分布，等时线
英文摘要
ABSTRACT
Urban road network traffic flow is a spatial-temporal vehicle flow on road network impacted by the network topology, signal control, traffic information guidance and trip decisions. This kind of traffic flow is an interrupted flow that variance of the travel times is larger than the counterpart continuous flow. The route choice of travelers and macroscopic traffic flow evolution would be directly impacted by this feature above. Urban road network traffic flow evolution is a network-scale phenomenon combined by amount of individual route-scale trips. It is also the temporal-spatial variations of traffic status, features or one specific parameter, including traffic status transfer; the formation and dissipation of congestion in a certain region; the temporal variations of network reliability, etc.
This dissertation is funded by the research of national 863 projects“Road Traffic Status Information Collection, Interactive Processing Technique and Strategic Laboratory”. Both experimental traffic theory and mathematical analysis are combined in this dissertation. We take the travel times under interrupted flow as the object to analyze interrupted flow in urban road network from microscopic to macroscopic aspects. The data of interrupted traffic flow in urban road network is analyzed by experimental traffic theory, and kernel essence of urban road network traffic flow is analyzed by mathematical methods under the background of the lack of network-scale traffic flow data.
The concept of urban road network traffic flow is illustrated in this paper.The urban road network traffic flow, based on observations and analyses in real world, has been proved to be an interrupted flow impacted by signal control at signalized intersections. This conclusion was testified by real data in the city of Nanjing, and the travel time under interrupted flow was taken as the data resource in this paper for following research.
The travel time under interrupted flow is studied based on analysis above.The bimodal distributon of travel time under interrupted flow in urban road network is discussed based on the RFID data analysis. In this paper, normal and lognormal distributions have been applied for fitting bimodal distributions, and the corresponding results are excellent. The travel time reliability under interrupted flow is accordingly analyzed. The reliability indices are established and a feature point, CRS_TT, is studied in this processing in order to distinguish the travel times of fast flow from the counterpart of slow flow. This point is a critical contribution for following studies.
The traffic loading under interrupted flow is studied based on travel time study under interrtuped flow above.The static path finding algorithm is presented based on bimodal distribution. Next, the travel time function under interrupted flow in urban road network is given by experimental traffic theories and methods. In that case, the staitic path finding algorithm has been updated to a dynamic algorithm. It indicates that delay at signalized intersection plays a significant role in travelers’ rout choice and it leads to a distinguished path finding result from traditional approaches. Finally, the traffic loading under interrupted flow is studied as a non-equilibrium loading approach. Compared to traditional traffic assignemt, the distinct temporal and spatial features of network interrupted flow are highlighted. Besides, the impact of signalized delay to route choice and traffic flow evolution is also studied.
The network traffic flow evolution in urban road network is studied by traffic coutour line approach.The network traffic status origined from a single traffic centre is analyzed from network aspect based on contributions from microscopic and meseoscopic aspects above. The countour line’s definition and four critical parameters: density, angle, average propagation rate of traffic, and average spatial expansion rate of traffic are introduced. Three typical traffic countour maps: the design traffic contour map, the static traffic coutour map, and the dynamic traffic countour map are discussed for different research requirements. It is indicated that traffic contour map can directly reflect the spatial-temporal distributions of traffic flow and variations of its characteristics. It is a new analysis tool and thinking pattern for traffic management from network-scale.
Finally, the dissertation is summarized. The creative research achievements and other important research directions were pointed out.
Key Words:urban road network, interrupted traffic flow, travel time, bimodal distribution, traffic coutour line
胡盼
入学时间：2010级
答辩时间：2016年
论文题目：基于多源数据的城市道路交通质量评价方法研究
中文摘要
摘 要
交通的本质是实现人或物的移动，对人而言，交通质量体现在交通服务对出行移动需求的满足程度。而人的出行需求具有多样化、层次化及个性化等特性，其满足程度又具有主观性及客观性双重特点。随着交通信息采集技术、多源数据融合处理技术及数据挖掘技术的快速发展，促进了精细化及精准化的交通状态感知和出行行为研究，从而为出行者提供更加丰富及精准的交通信息服务成为研究趋势。本文面向出行者的交通质量评价及其对出行服务应用问题，将传统的汽车交通评价主体转变为以人为主体，围绕交通质量评价的基本需求，研究城市道路路网、路段及路径的交通质量评价方法。目标是基于数据研究更精细的交通质量评价方法及出行优化方法。
首先，针对交通质量的评价需求开展研究。运用需求层次理论构建交通质量影响因素基本体系，利用探索性因子分析法构建交通质量结构体系分析模型，利用结构方程模型模拟出行者个体特征、成本约束、出行决策及交通质量之间的量化关系。认为快捷性及可靠性为交通质量的基本要素，并以此为核心开展评价方法研究。
第二，提出基于出行时耗的路网交通质量评价方法。根据数据源的粒度差异及其数据特性，面向城市整体出行特征及个体出行特征提出交通质量评价方法，并研究了出行时耗的估算及预测技术。其中，RS-PCA-RBFNN预测算法可约简输入变量结构体系及消除变量之间的共线性关系，适用于变量维度较多且变量之间存在共线性环境；PCA-FRON预测算法适用于规模较大的复杂非线性在线预测问题。
第三，针对路段交通质量的拥堵程度及可靠性分析方法开展研究。提出基于多源数据的人均行驶速度估算方法，及其在交通拥堵评价应用及概率分布规律。具体而言，综合考虑小汽车及公交车两种出行方式，提出利用浮动车、公交IC卡、交通检测器等多源数据估算路段人均出行速度方法。利用该参数，提出了一种新型交通拥堵指数计算方法，实验表明该算法能更准确符合出行者对拥堵程度的实际感受，对合理及准确地认知及判断交通拥堵程度具有理论及实际意义。以常用的正态分布及对数正态分布为基础模型，构建四种双峰分布模型。实证结果表明人均行驶速度主要服从N_N分布及N_logN分布，且有公交专用道路段更适合N_N分布，无公交专用道路段更适合N_logN分布。
最后，提出基于广义出行成本的路径交通质量评价及出行优化方法。考虑出行者个体特征对出行成本的感知差异及认知规律，引入敏感系数调整交通质量属性在出行成本构成中的贡献率，从而将考虑感知特性的广义出行成本作为交通质量评价指标。提出利用手机GPS数据及ANFIS算法，识别交通方式及路径行程时间估算方法。考虑出行者对抵达时刻的选择偏好，假设行程时间服从正态分布条件下，对最佳出发时刻及期望出行成本进行寻优求解，结果表明它取决于期望到达时刻，行程时间的期望值及方差，出行者对早到损失及晚到延误损失时间成本的相对大小。研究成果可为制定个性化及精准化出行决策方案提供依据。
本文研究成果对于“以人文本”的交通质量分析及个性化的交通出行信息服务具有较强的理论指导意义与实用价值。
关键词：交通质量；多源数据；优化方法；拥堵指数；出行方案。
英文摘要
ABSTRACT
The essence of traffic is to realize the movement of people or objects. For a person, the traffic quality is reflected by the degree of satisfaction of traffic service to his or her travel demands. People’s travel demands have diverse, hierarchical and individualized characteristics, and their degrees of satisfaction have both subjective and objective features. With the rapid development of technology in traffic information collection, big data processing and data mining, it facilitates the research in traffic status perception and people’s travel behavior in a fine and precise way, so that there is a “people-oriented” research tendency to provide more abundant and accurate traffic information service to travelers. Aiming at the problems in travelers’ traffic quality evaluations and their applications in traffic service, this research using people instead of the traditional vehicles as research subjects for traffic quality evaluations, focused on analyzing traffic quality’s demands, working on the evaluation method of traffic quality of road net, road section and road route. The goal is to build a more refined method for evaluating the traffic operations and a more personalized method for optimizing trips.
Firstly, the evaluation demands and methods of traffic quality were investigated. Specifically, a basic factor system of traffic quality was built according to the hierarchy of needs theory, and a model analyzing traffic quality structure system was established by the exploratory factor analysis in order to verify the rationality and availability of the basic system of traffic quality. In addition, the relationships among personal features, cost constrains, trip decisions and perceptible traffic quality were simulated and quantified using the structural equation model. The calibrated results by SP data showed that the travel time and its reliability of trip time were the most important factors in traffic quality evaluations.
Secondly, the method using trip time to evaluate traffic quality of road net was studied. According to the density difference and characteristic of data source, forced on the traffic quality of urban and individual trip characteristic, the estimating and predicting method of trip time was studied. Specificly, through reducing the structure systems of the input variables and eliminating the collinear relationships among variables, the RS-PCA-RBFNN algorithm was suitable for the variables with large dimensions and showing colinearity among them. The PCA-FRON algorithm here could be applied in complicated nonlinear online predicting problems with huge scales.
Thirdly, aiming at the second basic factor of the traffic quality, the running speed, methods for its estimation, applications to the traffic congestion evaluations and the regular patterns of its probability distributions were studied. Specifically, by integrating car and bus, a method of per capita travel speed estimated by the multi-source data was proposed. A new way of computing the traffic congestion index was suggested. Test results showed that the method proposed here matched the travelers’ real perceptions of the degrees of traffic congestion more accurately, and had both theoretical and practical significances regarding perceiving and judging the degrees of traffic congestion rationally and accurately. In researching the probability distributions of per capita travel speed in interrupt flow, the normal distribution and lognormal distribution were chosen as basic forms to build four kinds of bimodal distributions. Experimental results showed that the per capita travel speed mainly behaved like the N_N bimodal distribution and the N_LogN bimodal distribution. And the N_N bimodal distribution fitted the road sections with transit lanes better, whereas the N_LogN bimodal distribution fitted the road sections without transit lanes better.
Finally, methods for computing and optimizing trip scheme by generalized travel cost were studied. Considering the travelers’ perception differences to the servicing factors of trip cost, a generalized trip cost was calibrated using sensitivity coefficients. In order to compute time of travel route, a method inferring travel modes was proposed by GPS data of smart phones and ANFIS algorithm.Based on the solution algorithm of the budget of optimal trip time and desired trip cost, in order to optimize the trip scheme, the best combination of leaving time, traffic mode and travel routes was provided. The method here might be useful for making personalized trip scheme decisions and traffic information service.
Overall, the research results might provide significant contributions in “people-oriented” traffic quality evaluations and personalizing service of traffic information both theoretically and practically.
Key words: traffic quality; multi-source data; optimization method; traffic congestion index; trip scheme.
