基于智能体的方式选择及出发时间选择模型研究

发布时间：2018-03-19 13:20

本文选题：出行行为模型　切入点：智能体模型　出处：《清华大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着经济的发展,交通拥堵日益成为制约我国经济发展的重要因素。近年来,北京、上海等大城市均出现了严重的交通拥堵,影响了城市的健康发展。为解决交通拥堵问题,政府部门采取了许多交通需求管理政策来促使出行者改变出发时间或从小汽车出行转向公共交通、非机动车出行。研究表明,建立科学、合理的居民出行行为模型(出发时间选择模型、出行方式选择模型等)来预测政策实行后出行者的反应对于评估交通系统的运行状态具有重要作用。传统的居民出行行为模型基于个人效用最大化原理,假定出行者具有完全理性和无限信息性,将会准确评估出每个选择肢的具体效用值,并选择对个人效用值最大的选择肢。但实际上,出行者具有有限理性和不完全信息性,并不能精准地知道交通系统的实际情况,也往往不是选择效用最高的选择肢,而是根据经验选择一个相对满意的选项。针对这个问题,本文展开基于智能体的出行行为模型(出发时间选择模型和交通方式选择模型)研究,放松传统模型中出行者完全理性和获取信息不需要成本的假定,建立更为符合出行者决策特点的出行行为模型。本文的研究成果主要包括以下几点:1.开展了基于智能体的出发时间选择模型和交通方式选择模型理论研究,提出模型的框架主要包括搜索效益、搜索成本、搜索规则和决策规则四个方面,并用发生式规则(if-then)来模拟出行者的搜索和决策过程。另外,本文还从理论和模型表现上对比了基于智能体的出行行为模型和传统基于个人效用最大化原理的Logti模型。2.本文深入研究了居民出行行为数据调查方法,并结合研究要求,采用JAVASCRIPT和JAVA编程,设计了动态网页版问卷和平板电脑APP版调查问卷,并收集了研究所需数据。3.利用机器学习算法,建立了基于智能体的出发时间选择和出行模式选择联合模型。利用MATLAB将模型写成仿真程序,应用在北京二环以内的路网拓扑结构上,分析了拥堵收费和需求增加对出发时间和模式选择结果的影响,结果显示该模型能够较好地预测出发时间改变和交通方式改变行为。4.将模型与开源宏微观交通仿真软件TRANSIMS相集成,搭建了动态交通规划与仿真平台,以中新天津生态城为例进行了应用案例研究。结果显示,基于智能体模型能与各种宏微观仿真软件的结合,并进行场景和政策分析。
[Abstract]:With the development of economy, traffic congestion has increasingly become an important factor restricting the economic development of our country. In recent years, Beijing, Shanghai and other big cities have appeared serious traffic congestion, which has affected the healthy development of cities. In order to solve the problem of traffic congestion, The government has adopted a number of traffic demand management policies to encourage travelers to change departure times or switch from car to public transport, non-motorized travel. Reasonable travel behavior model (departure time selection model, To predict the response of travelers after the implementation of the policy plays an important role in evaluating the operating state of the transportation system. The traditional travel behavior model of residents is based on the principle of maximizing personal utility. Assuming that the traveler has complete rationality and infinite information, the specific utility value of each selected limb will be accurately evaluated, and the selected limb with the greatest personal utility value will be selected. But in reality, the traveler has limited rationality and incomplete information. You don't know exactly what's going on in the transportation system, and often you don't choose the most effective limb, but you choose a relatively satisfactory option based on experience. In this paper, the agent-based travel behavior model (departure time selection model and traffic mode selection model) is studied to relax the assumption that the travelers are completely rational and that obtaining information requires no cost in the traditional model. The main research results of this paper are as follows: 1. The theory of departure time selection model and traffic mode selection model based on agent is studied. The framework of the model mainly includes four aspects: search efficiency, search cost, search rules and decision rules, and use the generative rules to simulate the travelers' search and decision-making process. This paper also compares the agent-based travel behavior model with the traditional Logti model based on the principle of individual utility maximization. Using JAVASCRIPT and JAVA programming, the dynamic web version questionnaire and the tablet computer APP version questionnaire are designed, and the data needed for the research are collected. 3. The machine learning algorithm is used. A joint model of departure time selection and travel mode selection based on agent is established. The model is written as a simulation program by MATLAB and applied to the road network topology within the second ring ring of Beijing. The influence of congestion charge and increasing demand on departure time and mode selection result is analyzed. The results show that the model can predict departure time change and traffic mode change behavior. 4. The model is integrated with open source macro and micro traffic simulation software TRANSIMS, and a dynamic traffic planning and simulation platform is built. The application case study of Zhongxin Tianjin Ecological City is carried out. The results show that the agent model can be combined with various macro and micro simulation software, and the scene and policy can be analyzed.
【学位授予单位】：清华大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491

【相似文献】