【作者】 刘玉麟；

【导师】 徐瑞华；

【作者基本信息】 同济大学 ， 交通运输规划与管理， 2008， 硕士

【摘要】 随着城市轨道交通的迅速发展,基于网络化的系统复杂化趋势是发达的城市轨道交通系统的普遍特征,如何有效地管理复杂的网络系统已成为紧迫的问题。其中,乘客在网络中的出行行为是研究的关键之一。基于独立线路的传统运营组织方法不能反映网络化条件下的乘客出行行为,而研究乘客在多路径之间的选择行为是必须的。乘客的出行行为和网络系统的运营组织之间形成了相互影响的互动关系;此外,网络中拥挤的非线性因素又使乘客路径选择行为的研究难度加大。因此,如何描述该问题的动态性和复杂性是研究的重点和难点。本文正是基于对该问题复杂性本质的认识,从复杂系统演化的角度,研究乘客在网络系统中的出行行为,试图利用多主体系统仿真的方法模拟系统运行过程。文章首先从乘客个体行为与网络客流分配之间辨证关系的角度,系统地分析了问题的研究对象,指出本问题和道路网中交通分配的联系和区别;其次,本文反思了传统的基于静态和还原方法论的路网配流理论之局限性。文章根据复杂系统理论分析了本问题的复杂性特征,指出均衡理论的缺陷和乘客路径选择行为的适应性,并且提出了基于多主体建模系统仿真的定量研究思路;之后,根据城市轨道交通网的特点,本文构建了三层结构的网络系统表示模型——分别反映路网的规划、基础设施和行车组织,并设计了相应的多路径搜索算法和路径信息获取算法;本文建立了基于路径效用函数的乘客主体路径选择决策模型、基于强化学习的乘客主体适应性学习模型和基于多种类主体之间交互的出行过程模型;最后,作者以计算机程序实现了系统仿真,并进行了仿真实验和结果分析。计算机仿真实验的结果显示,所构建的系统仿真模型有效地模拟了系统的运行,乘客个体的适应性行为和网络客流分配状态的演化过程得到了良好的表现。更多还原

【Abstract】 Along with the rapid development of urban mass transit (UMT), the complexity based on the networking is universal among all the advanced UMT systems. How to cope with the complex system in an effective way has become a tough issue, and the study on traveler behavior in UMT network is one of the key points. Conventional methods of train operation management are oriented to the traveler flow volume on individual lines. In order to explore the traveler behavior in network system, the studies on traveler’s path-choose mechanism is necessary. Traveler behavior and network operation management influence each other; moreover, crowd effect adds the nonlinear factors in path-choose behavior. Therefore, it is very difficult to describe the dynamic and complex mechanism of the target issue. Based on the understanding of the complex nature of the study issue, this thesis tries to research the traveler behavior in network from the angle of the evolution of complex systems and to simulate the system through multi-agent modeling.This thesis systematically analyses the target issue from the perspectives of both individual traveler and traveler flow respectively and in a dialectical way, indicating the similarities and differences between this issue and the traffic assignment on road network. After reflecting on the limitations of the conventional network assignment theories which are based on static and reductive methodologies, this thesis reveals the complex nature and describes the characteristics of this academic issue. The limitation of common equilibrium theories and the adaptation of traveler behavior are also analyzed, meanwhile, the multi-agent simulation method is chosen as the quantity research approach. With regard to the special attributes of UMT network, a three-layer structure network model is built, depicting the network system from the layers of planning, infrastructure and train operation in details. The multi-path-search and path information algorithms are designed in line with that network model as well. A path-chose model of individual traveler is built on the basis of utility function theory. The adaptation and travel process of traveler agent are modeled using enforced learning method and interaction rules among model objects respectively. At last, the computer program of the system simulation model is developed, and the simulation experiment is carried out using a case of a supposed UMT network system. The data resulting from the experiment illustrates that the adaptation of traveler individuals and the evolution process of traveler flow assignment in the network system are well expressed through the multi-agent model.更多还原