【作者】 张波；

【导师】 隽志才；

【作者基本信息】 上海交通大学 ， 管理科学与工程， 2012， 博士

【摘要】 路径选择是出行行为研究领域的一个关键问题，它为许多交通问题研究提供了理论基础。传统路径选择问题研究通常假定出行者是完全理性的，其出行决策遵循期望效用理论，以出行的阻抗最小、效用或可靠性最大作为决策依据，较少考虑出行者的有限理性问题。本文围绕出行者的路径选择问题，在有限理性框架下基于前景理论进行出行行为分析和建模，研究出行者的动态路径选择行为与交通系统宏观运行状况之间的相互作用规律，探讨序贯出行决策中参照点的动态更新规律。首先，分别对路径选择问题的研究现状和前景理论在出行行为研究中的应用进行了回顾和总结，指出了现状研究存在的问题和不足，进而提出论文的研究内容和技术路线；在此基础上对前景理论在出行行为研究中的适用性进行了探讨，研究认为前景是否适用于出行行为研究，关键要看出行者所考虑的选择方案属性是否具有不确定性，这一结论为前景理论在出行行为研究中的具体应用提供了指导和依据；此外，针对前景理论自身存在的局限，对在具体研究中如何选取更为合适的理论给出了建议。第二，考虑交通网络行程时间的不确定性和出行者的感知误差，并基于行程时间可靠性和出行时间预算来设定出行路径选择的内生参照点，基于累积前景理论建立了一个随机用户均衡模型；分别在固定需求和弹性需求条件给出了等价的变分不等式，设计了求解算法并通过算例对模型及算法的有效性进行了验证；结合参数敏感性分析研究了参照点、风险偏好、损失规避和感知误差等出行者的认知和心理因素与网络均衡交通流之间的相互影响和相互作用规律。模型更加符合现实中出行者的决策行为特征，提高了交通需求预测的精确性。第三，进一步考虑出发时间选择问题进行动态交通分配建模，以早高峰工作出行为研究对象，合理设定了出发时间和出行路径同时选择问题的参照点，基于累积前景理论建立了一个随机动态用户最优模型；给出了等价的变分不等式，设计了求解算法并通过算例分析了出行者的动态路径选择行为与时变的交通流之间的关系；在此基础上建立了动态可靠性指标来评价交通网络的运行效率和服务水平，并基于算例结果展现了出发时间选择与到达可靠性之间的关系。模型较好的克服了传统出发时间选择模型基于完全理性假设，以及“时间窗”和“计划延误成本”概念的局限。第四，在不同信息提供条件下分别基于复制子动态、最优反应动态、Smith动态和BNN动态来考察出行者日常路径选择的博弈学习行为，基于累积前景理论建立了四个动态交通系统模型；分别在不同的初始状态和参照点更新规则下通过算例对交通流的动态演化与用户均衡的实现过程进行了数值模拟。均衡和演化两种分析方法的综合运用，揭示了交通流演化的内在动力机制，以及网络均衡的实现过程和条件，在有限理性框架下丰富和发展了传统的交通分配理论体系。最后，将累积前景理论与贝叶斯法则相结合，构建了一个序贯出行决策学习模型，深入分析了序贯出行决策的特征和参照点的内涵及其影响因素，设计了一个路径选择实验，基于实验数据探讨了序贯出行决策中参照点的动态更新规律。实验研究方法的引入不仅为出行者行为的实证研究提供了思路和借鉴，也为前景理论在交通领域的实际应用创造了条件。本文采用规范研究和实证分析相结合的方法，运用行为科学理论解析出行行为发生的微观机理，基于合理的行为假定和研究方法进行出行行为分析和建模，拓展了传统路径选择问题的研究思路，有关模型、算法和结论可以为交通分配、动态诱导、拥挤收费、区域限行等问题的研究提供理论依据。更多还原

【Abstract】 Route choice is a key problem in travel behavior research field, which providestheoretical foundations for the research of many other traffic issues. Existing routechoice research is generally based on the assumption that travelers are entirelyrational and usually make their decisions based on minimum disutility, maximumutility or reliability of the available alternatives, and seldom considers travelers’bounded rationality in their decision-makings. Around the issue of route choice, thispaper is intended to analyze and model travel behavior based on Prospect Theorywithin the frame of bounded rationality. The interaction laws between travelers’dynamic route choice behavior and traffic system’s macroscopic operationalperformances are also investigated. Reference points’ updating rules in sequentialtravel decision scenes are explored as well.Firstly, the state-of-art of route choice research and Prospect Theory’sapplication in travel behavior research is reviewed and summarized. Severalproblems and shortages in existing research are pointed out, followed by the maincontents and technical roadmap of the dissertation. Then around the applicability ofProspect Theory in travel behavior research, it is suggested that the key criterion liein whether the alternative attribute within traveler’s consideration involvesuncertainty or not. This conclusion provides guidance and foundation for the specificapplication of Prospect Theory in travel behavior research. In addition, some adviceon how to select a more suitable theory in actual research is given to avoid of severalexisting defects of Prospect Theory.Secondly, in consideration of the uncertainty of travel time and travelers’perception errors in traffic network simultaneously, a Stochastic User Equilibriummodel based on Cumulative Prospect Theory is formulated, in which travelers’endogenous reference points are set by an integration of travel time reliability andtravel time budget. Then the equivalent variational inequalities and algorithms forthe model are given accordingly under the condition of fixed and elastic traveldemand, followed by two numerical examples respectively to provide validity of the model and algorithms. By parameter sensitivity analysis, a study of the interactionlaws between travelers’ cognitive and psychological factors, such as reference point,risk appetite, loss aversion and perception errors, and the traffic network equilibriumis then conducted. Conforming more to travelers’ decision behavior in reality, thesuggested model can improve the accuracy of traffic demand forecasting.Thirdly, in order to apply Prospect Theory to dynamic traffic assignment inconsideration of the problem of departure time choice, a Stochastic Dynamic UserOptimum model based on Cumulative Prospect Theory is formulated, whichproperly handles the reference points of departure time and route choices focusingon commute trips in morning peak. After an equivalent variational inequality andalgorithm of the model are given, the influence of travelers’ dynamic route choicebehavior on time-varying traffic flow is analyzed through a numerical example. Thena set of dynamic travel reliability indicators are established for the purpose ofevaluating traffic network’s operational performance and service efficiency. Anillustration of the relationship between departure time choice and arrival reliabilitybased on the proposed numerical example is followed. The proposed model expandsthe concept of time-window and schedule delay well and overcomes the limitation ofentire rationality assumption in existing models dealing with departure time choice.Fourthly, in scenarios of different information provision, four dynamical trafficsystem models are formulated based on Cumulative Prospect Theory, in which fourevolutionary dynamics, namely the replicator, the best response, the Smith and theBNN dynamics, are introduced to investigate travelers’ day-to-day route choicegame learning behavior respectively. Then the dynamical traffic flow evolution andachievements of User Equilibrium are presented through two numerical examplesunder different initial states and distinct reference points’ updating rules. By anintegration of equilibrium and evolution analysis methods, this paper reveals themechanisms of traffic flow evolution as well as the processes and conditions oftraffic network equilibrium, and enriches and develops traditional traffic assignmenttheories.Finally, a sequential travel decision learning model is formulated based onCumulative Prospect Theory and the Bayes Law. After a deep analysis of thecharacters of sequential travel decisions as well as the connotations and affectingfactors of reference point, a route choice experiment is designed to investigate theupdating rules of reference point. The introduction of experimental research method not only provides notion and reference for empirical study of travel behavior, butalso creates conditions for Prospect Theory’s application in practice.By an integration of normative research and empirical analysis, this paperapplies behavioral science to travel behavior analysis and modeling based on rationalbehavior assumptions and research method. The suggested models, algorithms andconclusions can provide theoretical foundation for further research of traffic issues,such as traffic assignment, dynamic route guidance, congestion pricing and regionaltraffic restriction, etc.更多还原