【作者】 周元峰；

【导师】 吴建平；

【作者基本信息】 北京交通大学 ， 系统工程， 2007， 博士

【摘要】 城市交通网络中，交通信息对交通流的影响是一个非常复杂的动态过程。交通流的实质是个人出行决策的集合，用户的交通行为在很大程度上决定了ATIS这项新技术能否被有效地利用。对于动态诱导系统来说，最具挑战性的任务就是真实的捕获信息对交通流的动态影响，以及驾驶员的反应，制定能够使出行用户的利益以及系统管理需求都得到满足的诱导策略。在对国内外相关研究综述的基础上，本文基于多方面的调查和实测数据，探讨北京出行者路径选择行为特性以及合理利用信息进行突发交通事件下的动态诱导的方法，并结合实际数据对提出的模型和算法进行了验证。本研究的主要工作内容包括：1．设计并利用SP问卷调查获得了大量北京市出租驾驶员路径选择行为数据；建立了路径转换概率预测的logistic分析模型；总结了影响北京驾驶员路径选择行为的关键因素及其作用：引入ROC Curve过程和曲线诊断方法进行模型的择优、诊断，并设计了路径转换概率预测诊断点的优化迭代算法。2．结合微波检测的宏观交通流数据和VMS信息分析出行者路径选择行为。提出了内部检测数据预处理的编码修正办法，路段流量转移状态识别模型和算法。考虑到信息条件下路段交通参数的不稳定性，本文提出了相应的流量转播时间算法、被测车流完全通过检测断面的周期系数迭代算法，并在这些关键参数算法基础上构造了转移流量累积模型以及基于实时流量对比和基于概率对比的转移概率估算算法。通过实测数据对不同交通信息条件下驾驶员路径改变概率进行估算以及行为时机和地点的差异性比较。3．通过浮动车的累积GPS／GIS数据分析和轨迹分析，初步分析了北京出租出行用户在路径选择方面的偏好，包括偏好使用高等级道路，避免拥堵和路径复杂性，追求时间性和出行流畅性，舒适性等。讨论了突发交通事件条件下，出行者感知路段／路径效益将如何变化并影响其路径选择行为。并提出动态绕行效率的概念和计算方法，为科学地、定量地评价出行者路径选择行为提供评判参考指标。4．基于博弈论分析了出行者在路径选择过程中的个体理性博弈行为(包括个体一个体博弈和个体-群体博弈)，结果暗示城市智能交通管理需要从个体选择与系统诱导的博弈协调方面来进行优化和改善。针对动态流量分配的特点，本文提出了考虑路段流量分段流出和FIFO条件的改进一般化离散流量传播约束，使动态流量约束及交通流分配模型更加符合实际交通流传播的连续性特点。作者利用双层规划的建模方法构造了协调系统整体与出行者个体利益的基本模型，提出了两个突发交通事件下的流量疏导方案，以及基于滚动时段方法的模型求解算法：变换信息供给策略的交通协调诱导模型和部分付费用户协调诱导模型。仿真实验证明，基于此双层规划模型的诱导策略能够使以用户最优为原则的出行者在选择自身感知利益最大路径的同时，整体用户选择趋向管理者期望的系统最优方向，减短事件影响和持续时间，缓解交通事件下的交通流振荡现象，改善路网使用效益。该信息策略方法具有可实施、操作性强的优点。更多还原

【Abstract】 In urban traffic network, how traffic information influences the change of traffic flow is a very complex dynamic process. Traffic flow is formed of individual drivers/travelers. Travelers’ behavior will greatly influence the effect of ATIS. One of the challenging tasks of dynamic route guidance system is to truly capture the drivers’ response to information and its dynamic influence on traffic flow, and then make an effective guidance strategy which can meet the requirements of both travelers and traffic management department. Based on the review of domestic and abroad research, and on the basis of survey and real traffic data from detectors in Beijing, the author constructed the framework and analytic methods for the research of driver route choice behavior in the presence of traffic information. The developed models and algorithms were validated by the real traffic data. The contents of the paper are as follows:1. The author designed a SP survey questionnaire for taxi drivers’ route choice behavior study and collected data in Beijing through the SP investigation. A Logistic model for predicting the probability of driver route diversion was developed; The key factors that influence drivers’ diversion behavior were found based on the surveyed data; A ROC Curve diagnose method was introduced for model selection and calibration, and a relative algorithm of prediction diagnosis was designed.2. Drivers’ route choice behavior model in the presence of VMS has been validated with the macro traffic flow data collected by microwave detectors in Beijing. A coding modification method for data recognition of link traffic flow diversion was developed. Considering the dynamic variety of traffic parameters, the author put forward a algorithm to estimate the vehicles’ travel time and the time for whole group of the traffic passing through downstream detectors. On the basis of this, a cumulative flow diversion model as well as the estimation algorithms based on real-time traffic volume comparison and probability comparison were developed, which can provide a estimation of driver diversion probability in the presence of dynamic VMS information. Based on detecting data, a comparison of drivers’ diversion probability, time and place in different information conditions was given.3. Through the analysis of cumulative GPS/GIS data and route tracks of floating vehicles in Beijing, taxi travelers’ route choice preference was further analyzed. Revealed preference includes preference of using high-grade roads, congestion avoidance, route complexity avoidance, minimal travel time, travel fluency and comfort. The author discussed how taxi traveler’s perceptive travel profit changes in the situation of traffic incident and how this will affect his route choice behavior. The concept and algorithm of the index of dynamic routing efficiency was put forward, which can provide a reference for quantitative evaluation of route choice behavior for travelers.4. Based on the "Game analysis" which includes drivers’ individual-individual game behavior and individual-group game behavior, it was suggested that the urban traffic management needs to be optimized from the coordination of individual choice and system guidance. Considering the characteristics of dynamic traffic assignment model, the paper developed a modified general discrete flow propagation model to make traffic assignment model closer to reality. The author developed a basic model through the approach of bilevel programming to deal with the profit conflict of users and system optimization, and two flow guidance strategies: i) the coordination guidance model with changing information provision, which is suitable for normal congestion guidance; And ii) the coordination guidance model for proportional paid travelers who will strictly abide to the guidance. The algorithms based on rolling horizon approach were also developed. The simulation results suggested that the models and information strategies can meet the demands of both users and traffic managers while travelers’ route choice is inclined to the direction of manager’s expectation, and the phenomenon of Oscillation, the influence of the incident can be reduced and the efficiency of network will be improved. The guidance strategies are of strong maneuverability, and could be applied easily in factual traffic management.更多还原