1、4作者姓名：李志纯论文题目：交通行为建模与评价研究网络均衡分析法作者简介：李志纯，男，1976 年 8 月出生，2002 年 9 月师从于北京航空航天大学黄海军教授，于 2006 年 3 月获博士学位。中 文 摘 要交通行为分析是当前国际交通科学研究的一个前沿领域，也是进行交通规划与管理的基础。全面系统、深入地研究交通行为的本质是科学“诊治”城市交通“病”的关键步骤。交通系统是一个复杂的巨型系统，对交通系统中出行者的行为进行科学的分析需要使用系统科学的理论、方法与手段。当前，国际上关于交通行为的研究通常将出行者的运动行为和停车行为分离开来分别研究，这割裂了交通出行中“行”和“停”的统一，与城市
2、交通综合治理的初衷背道而驰，结果造成交通行为预测不准确、交通规划不合理、政策评价不科学。实际中，机动车总是在“行”与“停”两种状态之间转化，行车和停车是城市交通中紧密相连的两种行为，两者相辅相成、相互促进和制约，共同构成完整的交通系统。本博士论文运用系统分析方法，将城市交通中机动车的运动行为和停车行为综合起来作为一个整体，进行全面、系统的研究，分析了这两种行为之间的相互影响和相互作用。主要内容包括：（1）系统地研究了不同网络状态（静态、确定性动态和随机动态交通网络）中出行者的旅行行为和停车行为模型；（2）建立了评价随机动态交通网络性能的可靠度指标；（3）评价了不同交通信息（路径诱导信息和停车信
3、息）对不同属性出行者行为的影响。其主要目的在于通过深刻地透析交通行为的本质，使城市交通规划与管理工作更加理性、科学，更具有预见性。所得研究成果不仅有助于丰富和完善交通流理论，同时也为我国城市道路和停车设施的规划与管理、交通政策的制定与评价提供科学的理论指导。具有创新性的主要研究成果包括以下几个方面：1）静态交通网络中旅行行为和停车行为研究。现有的交通行为模型通常将出行者的运动行为和停车行为分离开来研究，而且经常忽略交通网络的能力约束，包括道路路段和停车设施的能力约束，结果可能导致预测的旅行需求大于路段和停车设施的能力。实际中，旅行需求受交通设施能力的限制，而且能力限制可能促使交通需求在不同的交
4、通方式之间转移，特别是在早晚上下班高峰期间。论文第二章首先研究了能力约束下的多方式交通网络中出行者的旅行选择和停车行为决策模型。模型同时考虑了出行者的出行路径、停车设施和交通方式选择以及道路路段和停车设施的能力约束。设计了组合增广的 Lagrangian 对偶算法和部分线性化算法来求解该模型。该研究首次发现增加停车设施可能诱发诡异现象，即在网络中增加一个新的停车设施之后，总的网络旅行费用不是减少了，反而增加了！该文发表在国际刊物 Journal of Mathematical Modelling and Algorithms（2007, Vol.6, No.2, pp.239-257）上。现有
5、的交通行为研究也经常假定出行者在一次出行过程中只使用一种交通方式，而不涉及多种交通方式之间的换乘问题。 “停车换乘”是现代城市交通需求管理的有效措施之一，5合理的换乘设施布局、停车泊位分配和停车收费，对提高城市交通系统的功效和缓解城市交通拥挤有重要的现实意义。论文第三章研究了上班高峰期进入市中心区的出行者的停车换乘行为，建立了与停车换乘均衡条件等价的变分不等式模型。使用该模型可以预测有多少出行需求选择使用换乘方式出行、每个换乘设施有多少人聚集以及交通需求在路网上如何分布等。该研究有助于加深对交通行为的理解、合理规划与布局换乘设施、协调发展多种交通方式。研究成果发表在系统科学与数学 （2006,
6、 Vol.26, No.3, pp.352-361）上，拓展的研究已经被国际刊物 Transportation Research Record 录用并正在出版。2）确定性动态交通网络中旅行行为和停车行为研究。论文第四章首次将出行者的停车决策行为结合到动态交通分配模型中，分析了不同类别出行者对出发时间、出行路径、停车设施和停车时间长度的决策行为。构造了与出行者旅行行为和停车决策行为等价的变分不等式模型，设计了求解该模型的有效算法。研究结果表明，所设计的求解算法能真正地产生动态交通网络均衡解；网络拥挤水平、停车设施的位置和容量以及停车收费方案都极大地影响人们的出行选择行为；对路外停车设施，静态交通
7、行为模型将低估平均停车时间长度和高估平均周转率、平均每小时停车设施占用率和停车利润，对路边停车设施，结论正好相反。这些成果发表在交通科学权威刊物 Transportation Research Part B（2006, Vol.40, No.5, pp.368-395）上，并在短期内被 SCI 他引 1 次（见证明材料） 。论文第五章进一步使用基于活动（Activity-Based）的方法研究了出行者的活动、旅行调度和停车行为模型。模型同时考虑了一天中出行者对活动类型、活动地点、活动时间长度、停车设施和旅行路径的选择行为，研究结果清晰地重现了出行者的链式层次行为。该论文于 2005 年 7 月
8、在本领域国际最高水平的理论性会议 International Symposium on Transportation and Traffic Theory（ISTTT）上宣读 45 分钟，同时在 TTT 丛书中发表（2005, pp.717-739， Elsevier 出版，每次限发表 35-38 篇论文） ，是 50 多年来我国大陆交通运输工程学科的博士生首次在该会议上发表论文。ISTTT 宣称“the symposium may prove to be a career-defining moment”，被 TTT 丛书收录的论文 “reflect the best work underw
9、ay anywhere in the world on the science and operation of traffic and transportation”（见网站 http:/www.isttt.org） 。TTT 丛书的封底上注明“ It is essential reading for anyone interested in researching or understanding traffic and transportation phenomena and in developing effective approaches to planning, design a
10、nd management of transport systems”。3）随机动态交通网络中旅行行为和停车行为研究。交通网络是一个复杂的动态、随机系统。衡量随机系统性能的一个重要指标是可靠度，它是指系统能够在规定的条件或时间内实现预定功能或目标的概率。随着交通系统对社会经济发展所起的支持作用越来越大以及出行者对服务水平的要求越来越高，路网的可靠度性能显得日益重要。论文第六章在利用供需相互作用原理和不动点理论系统地分析随机动态交通网络中出行者的旅行行为和停车行为的基础上，首次提出了评价随机动态交通网络稳定性能的计划可靠度和停车可靠度指标，设计了基于 Monte Carlo 模拟方法和相继平均法
11、（ Method of Successive Averages）的算法来计算可靠度。研究结果表明，提出的可靠度指标能有效地衡量一天中不同时段道路网络和停车设施的服务水平，传统的静态交通网络可靠度指标做不到这一点。研究也发现，增加道路通行能力将导致不同时段上、不同空间上旅行可靠度的相对变化程度不一致，即道路通行能力的增加将诱发时间和空间不公平性问题。这些研究成果发表在国际刊物 Networks and Spatial Economics（2007 年 1 月 22 日录用，2007 年 3 月 2 日在线发行）和论文集 Lecture Notes in 6Computer Science（200
12、5, 3828, pp.777-787）上。4）交通信息对不同属性驾驶员行为影响的评价研究。利用智能交通系统（Intelligent Transportation Systems， ITS）来引导驾驶员的出行行为，是 21 世纪交通需求管理中最有前景的措施之一。因此，对 ITS 提供的路径诱导信息和停车信息的收益、可行性和风险进行分析和评价，显得尤为重要。论文第七章研究了多用户类随机动态交通网络中路径诱导信息和停车信息对不同属性驾驶员出行行为的影响，同时讨论了如何确定一般路网均衡状态下信息系统的市场渗透率。研究发现，适度的信息质量和市场渗透率将提高驾驶员决策的理性程度和减少违章停车数；网络的不
13、确定性水平和信息服务的质量将大大影响出行负效用及其各组成要素的大小；均衡市场渗透率水平与驾驶员的属性、出行里程以及信息服务费有关，信息服务收费水平可用来控制信息系统的市场渗透率水平。此外，研究也清楚地发现和解释了信息影响的不公平性和信息过度等新的行为现象。相关研究成果发表在系统工程理论与实践（2004, Vol.24, No.9, pp.125-130） 、 公路交通科技 （2005, Vol.22, No.2, pp. 95-99） 、 中国公路学报 （2005, Vol.18, No.3, pp. 94-98）和国际著名运筹学刊物 European Journal of Operation
14、al Research（2007, Vol.176, No.3, pp.1464-1477）上。关键词：网络均衡，旅行行为，停车行为，可靠度，交通信息7A Study on Travel Behavior Modeling and EvaluationNetwork Equilibrium Analysis ApproachLi ZhichunABSTRACTTraveler behavior analysis, as one of the frontiers of the current transportation science, is the foundation of carryin
15、g out the planning and management of urban transportation systems. A systematic and comprehensive study on traveler behavior analysis plays a pivotal role in alleviating urban traffic congestion. Urban transportation system is an enormous and complicated system, so the theories and methodologies in
16、systems science should be applied to investigate the traveler behavior in transportation science. Most existing studies on traveler behavior modeling often treat the traveling behavior and parking behavior of travelers independently. This contradicts with the ideology that urban transportation syste
17、ms should be planned and managed in an integrated manner. As a result, the existing studies on traveler behavior may generate biased prediction on traffic and parking demands, irrational planning on urban transportation systems and inaccurate evaluation on transport policies. In fact, vehicles eithe
18、r move on roads or park at some sites. Parking system and road traveling system, both as the most important components of an urban transportation network, are strongly related to each other, being improved or restricted.In this dissertation research, we investigated the joint behavior of traveling a
19、nd parking by using the systematic analysis approach. It consists of the following components: (1) modeling the traveling and parking behavior of travelers in a static context, a deterministic and dynamic context, and a stochastic and dynamic context, respectively; (2) establishing new reliability i
20、ndices for measuring the stability of stochastic and dynamic transportation networks; and (3) evaluating the impacts of route guidance and parking information services on heterogeneous drivers behavior in stochastic and dynamic networks. The main aims of the research are to develop some new methodol
21、ogies for deepening our understanding on traveler behavior such that we can more effectively and scientifically manage urban transportation systems. It is expected that the research outcomes presented in this thesis, can further enrich and improve the current theories associated with urban transport
22、ation network modeling and analyses, as well as provide fundamentals and tools for scientifically planning and managing urban transportation facilities.The main contributions of this dissertation research are as follows:1) Study on traveling and parking behavior in static transportation networks. Th
23、e most existing studies on traveler behavior modeling often separately treated the parking behavior and traveling behavior of travelers. Moreover, the existing studies also often ignored the network capacity constraints, including both link and parking capacity constraints, and their impacts on elas
24、ticity of 10travel demand, and thus may induce that the predicted travel and parking demands exceed the capacities of road links and parking facilities. In practice, the travel demands are subject to the capacity constraints of transport facilities. Moreover, these capacity constraints may cause tra
25、ffic demand shift between different transport modes in peak hour periods during a day. In order to capture the drivers responses to traffic congestion and transport facilities (road capacities and parking spaces) availability, in Chapter 2 of this thesis we developed a mathematical programming model
26、 that simultaneously considered the choices of drivers on mode, route and parking location, and the road and parking capacity constraints. A heuristic solution algorithm that combined the augmented Lagrangian dual algorithm and the partial linearization approach was designed to solve the proposed mo
27、del. The results showed that parking capacity expansion may cause the paradoxical phenomenon, i.e., addition of a new parking facility in the network may incur an increase of total network travel disutility, which is contrary to what expected. The results of this research have been published in Jour
28、nal of Mathematical Modelling and Algorithms (2007, Vol.6, No.2, pp.239-257).In addition, most of the previous studies on the traveler behavior modeling generally assumed that a trip from an origin to a destination was completed by using a single transport mode such as the auto mode or the public tr
29、ansit mode. Little attention has been paid to modeling the park-and-ride (P&R) behavior of travelers, which involves the modal transfer between different modes. The P&R schemes have been widely recognized as an effective means for managing traffic demand in the densely developed urban areas. A ratio
30、nal and efficient P&R scheme will play an important role in improving the function and efficiency of urban transportation systems and alleviating urban traffic congestion. In Chapter 3, the P&R behavior of travelers in a general transportation network was investigated. The network equilibrium condit
31、ions on mode, route and transfer location choices were formulated as an equivalent variational inequality formulation. This proposed model can be employed to effectively predict the P&R demand in the transportation system, the traffic volume at each of P&R facilities and the traffic flow distributio
32、n over the network. This research is helpful to further understand the traveler behavior, design and layout the P&R system and to harmonize the use of various transport modes in the network. This part of research work has been published in Journal of Systems Science and Mathematical Sciences (2006,
33、Vol.26, No.3, pp.352-361). The extended study has also been accepted for publication in Journal of Transportation Research Record (for the acceptance notification, see the attached proof).2) Study on traveling and parking behavior in deterministic and dynamic transportation networks. In Chapter 4, t
34、he parking behavior of drivers was first incorporated into the dynamic transportation network modeling framework. A dynamic network equilibrium model was proposed that simultaneously considered a travelers choices of departure time, route, parking location and parking duration in road networks with
35、multiple user classes and multiple parking facilities. The model was formulated as a variational inequality problem, and was solved by a heuristic solution algorithm. The results indicated that the proposed model can indeed generate a solution that satisfies the dynamic network equilibrium condition
36、s. The traveling and parking behavior of drivers are significantly affected by travel demand, walking distance, parking capacity, and parking charge. For the off-street parking locations, the static model underestimates the average parking duration 4and overestimates the average turnover, average ho
37、urly occupancy of parking location, and parking revenue compared to the time-dependent model, but that the opposite is the case for the on-street parking locations. The main results have been published in Transportation Research Part B (2006, Vol.40, No.5, pp.368-395), and are cited by other scholar
38、s from SCI in a short time (see the attached proof).In Chapter 5 of this thesis, we further investigated the activity-travel scheduling and parking behavior in the framework of the activity-based travel analysis. The proposed model simultaneously considered the travelers choices on activity type, ac
39、tivity location, activity duration, parking location, and traffic route over the time of day. The results showed that the proposed model can clearly reproduce the travelers trip-chaining behavior during a day. The research work was presented at the 16th International Symposium on Transportation and
40、Traffic Theory (ISTTT) in July 2005, and was published in the TTT series (2005, pp.717-739). It should be mentioned that each issue of the TTT series at most includes 35 to 38 papers and is published by the famous Elsevier Publishing Company. To our knowledge, it is the first time in the past fifty
41、years that the research paper of a PhD student from mainland China is published in this prestigious international symposium. The ISTTT claimed “the symposium may prove to be a career-defining moment”, and the papers included in the TTT series “reflect the best work underway anywhere in the world on
42、the science and operation of traffic and transportation” (for the details, see the website: http:/www.isttt.org). And, the cover page of the TTT series states “It is essential reading for anyone interested in researching or understanding traffic and transportation phenomena and in developing effecti
43、ve approaches to planning, design and management of transport systems”.3) Study on traveling and parking behavior in stochastic and dynamic transportation networks. Transportation network is a complicated dynamic and stochastic system. Measure of reliability has been widely recognized as an importan
44、t index for measuring the stability performance of stochastic systems. The measure of reliability is generally defined as the probability that a system can perform the pre-specified tasks or objectives under a given condition or within a given time threshold. As the value of time increases, people d
45、esire a reliable transportation network with a high level of service. In Chapter 6, a fixed-point model was proposed to investigate the traveling and parking behaviors of travelers in stochastic and dynamic networks in terms of the supply-demand equilibrium relationship and fixed-point theory. Sched
46、ule reliability and parking reliability, as new network performance indices, were first introduced to measure the stability of performance of road networks. A heuristic solution algorithm that used a combination of the Monte Carlo simulation approach with the method of successive averages was propos
47、ed to estimate these two reliability measures. The results showed that the proposed measures of reliability can reflect the differences of the networks service levels during different time intervals of a day. This cannot be achieved in the static paradigm that only studies the issues of one interval
48、. The results also showed that the change of scheduled reliability for different routes at different times of a day as a result of the link capacity expansion is very different. This means that road capacity expansion would cause the spatial equity issue as well as temporal equity issue. The main re
49、search results are published in Networks and Spatial Economics (accepted on January 22, 2007 and available online on March 2, 2007, see the 4attached proof) and Lecture Notes in Computer Science (2005, 3828, pp.777-787).4) Study on evaluating the impacts of information services on heterogeneous drivers behavior. Intelligent Transportation Systems (ITS) has been recognized as the most promising means for managing urban traffic demand in the 21st century. Consequently, it is specially important to explore the effects of the route guidance information and parking information
