【作者】 赵凛；

【导师】 张星臣；

【作者基本信息】 北京交通大学 ， 交通运输规划与管理， 2007， 博士

【摘要】 信息技术的发展为城市交通问题的缓解带来了契机。但信息技术能否发挥其作用取决于出行者在出行中对信息的反应。因此，对出行者的出行决策行为建模具有重要的理论意义和实践价值。传统的对出行决策行为建模是在“期望效用理论”的框架下进行的，但该理论与现实存在着某些不适应的方面。本文将基于“前景理论”，对出行者的出行决策行为进行理论建模并利用该模型来研究ATIS实施后对出行者的出行决策行为以及城市交通系统演化的影响。 出行者的出行过程按照时间维度可分为一日内出行和多日出行。论文首先在“前景理论”的框架下，考虑出行者在面临风险时的态度以及出行经验等各种个体特征，建立了出行者在一日内出行的路径选择行为模型，并给出了出行者决策“参照点”的定义以及连续型路径属性离散化的方法。 结合认知心理学的相关理论，对出行者多日出行中利用经验和ATIS信息进行学习的过程进行了分析，对原有贝叶斯更新模型进行了修正，考虑了出行者记忆容量的限制，并引入出行者对ATIS信任度的更新机制。在此基础上，论文将多日出行中出行者的出发时刻选择视为一个学习调整的过程，给出了出行者出发时刻调整的启发式“延误补偿”算法，定义了出发时刻学习调整阈值。基于该算法的仿真实验结果接近于现实中的出发时刻选择，表明该算法在一定程度上能够有效的描述出行者的出发时刻选择行为。 在复杂适应系统理论的指导下，根据上述出行者出行决策的理论模型，利用Repast平台建立了基于Agent的出行者决策仿真实验系统——ABTSS。 论文从理论推算、实证调查、仿真实验各个方面对比了“前景理论”与“期望效用理论”对出行者路径选择行为的刻画，结果证明“前景理论”与“期望效用理论”相比，能够更加准确地描述出行者在不确定性条件下的出行决策行为。 利用ABTSS，对不同路网负荷下ATIS的诱导效果进行了多次仿真实验，考察了不同ATIS市场渗透率下出行者出行决策的行为特征以及城市交通系统演化规律。更多还原

【Abstract】 The development of information technology brings turnaround for the traffic problem.But how the information technology will influence travelers’ behavior and the evolution pattern of urban transportation system is sill not clear. Therefore, the research of travelers’ trip decision has an important theoretical and practical value. The existing models for travelers’ trip decision are mostly based on the Expect Utility Theory, which seems to have some devia(?)ons from the real world. So the paper will construct the model according to the Prospect Theory as well as analysing travelers’ trip decisi(?)on and the evolution of urban transportation system by making use of the model.Travelers’ trip decision can be divided into two kinds of processes from the time dimension: within-day and day-to-day. Under the theoretical framework of Prospect Theory, considering a variety of travelers’ individual characters such as risk attitude and trip experience etc., the paper builds the theoretical model for travelers’ route choice behavior within a day. The definition of reference point for travelers’ trip decision and the method of discretization of continuous route attributes are also given.According to the Cognitive Psychology, the paper uses the Bayesian updating model to analyze the learning of travelers from their experiences and the ATIS information during the day-to-day trip process.The paper studies the threshold for learning triggering and the travelers’ memory capacity; also incorporates the updating mechanism of travelers’ confidence to ATIS, which relates the travelers’ subjective error with their confidence to ATIS. At the same time, travelers’ depart time choice behavior during day-to-day trip is viewed as learning and adjusting process in the paper. Then the delay compensation algorithms and learning threshold for depart time choice is given. The results of simulation based on the algorithms are similar to the travelers’ behavior in real world, which shows the effectiveness of the algorithms in modeling travelers’ depart time choice behavior .Referring to the Complex Adaptive System theory, the paper establishes the Agent Based Trip Simulation System (ABTSS) under the environment of REPAST.The paper contrast the Prospect Theory with the Expect Utility Theory in their ability of modeling travelers’ route choice behavior from such aspects as theoretical computation, SP surveys and simulation experiments respectively. The results prove that the Prospect Theory could give a more precise description of travelers’ route choice更多还原