【作者】 韩凌辉；

【导师】 孙会君；

【作者基本信息】 北京交通大学 ， 系统科学， 2014， 博士

【摘要】 随着经济的发展和人口的增长,机动车保有量急剧增加。许多城市因此面临严重的交通拥堵。因为出行者的出行选择行为决定了交通网络中的流量分布,同时出行者具有自主性,交通管理措施只能对出行者的出行选择进行诱导。只有充分理解和把握出行者的出行选择行为,才能制定有效的交通管理措施。在不同交通网络条件下,出行者的出行选择行为具有不同特征。本论文主要研究四种交通网络条件下出行者的出行选择行为,建立相应的逐日动态配流模型,设计交通管理措施。这四种交通网络条件分别为：有出行信息诱导的交通网络、多模式交通网络、城市轨道交通网络和随机交通网络。本论文的主要研究内容如下：(1)结合现有不动点分别与用户均衡和随机用户均衡对应的两个逐日动态配流模型,本论文分别给出了准确出行诱导信息条件下和非准确出行诱导信息条件下的逐日动态配流模型。基于这两个逐日动态配流模型,研究了准确出行诱导信息和非准确出行诱导信息条件下交通系统的稳定性。通过研究发现,无论出行诱导信息是否准确,先进出行者信息系统的市场占有率、交通需求量和出行者感知误差对交通系统的稳定性都有很大的影响。(2)本论文将一个基于用户约束下系统最优的出行诱导信息与连续网络设计相结合,并将该问题用一个双层规划模型进行描述,其下层问题的可行域随着上层问题的决策变量的变化而变化。本论文对现有求解连续网络设计问题的模拟退火算法进行改进,用于求解该双层规划问题,避免了因温度过低而出现迭代步长不能计算的问题。并设计了基于路径的交通配流算法,使其能够求解带有用户约束的系统最优问题。通过数值试验发现,与只实施连续网络设计相比,本论文提出的出行信息诱导与连续网络设计相结合的交通管理措施能够更好地减少系统总出行时间,进而缓解交通拥堵。(3)本论文基于有限理性设计了城市轨道交通网络的逐日动态配流模型。在本论文中,若出行者的期望感知出行费用与实际出行费用的差值在可接受的范围内,则出行者保持当前的路径选择；否则,出行者按照随机用户最优原则重新选择路径。为了较为真实地反映出行者乘坐城市轨道交通的出行费用,本论文将出行者在车厢内受到的拥挤费用计算在出行者的出行成本内。在建立基于有限理性的城市轨道交通逐日动态配流模型后,本论文给出了该模型不动点的充分必要条件。利用该城市轨道交通逐日动态配流模型,通过数值试验发现,出行者对期望感知出费用与实际出行费用间差距的容忍域值、交通需求和出行者感知误差对轨道交通系统流量演化都有较大影响。(4)本论文分析了混合交通条件和设置有公交车专用道条件下,只有公交车和私家车两种交通方式的双模式交通系统的稳定性。同时也考虑了出行者的换乘行为对该交通系统稳定性的影响。为了反映公交车与私家车间不对称的相互影响,本论文设计了混合交通条件下和设置有公交车专用道条件下出行者使用私家车、公交车及停车换乘的出行费用。将这些出行费用与一个现有的不动点与随机用户均衡相对应的逐日动态配流模型相结合,得到了一个针对公交车和私家车的双模式交通网络的逐日动态配流模型。通过研究发现,无论是否考虑停车换乘,与混合交通条件相比,设置有公交车专用道的交通系统在保证系统的稳定性的同时还能加载更大的交通需求量。(5)现有随机交通网络条件下的逐日动态配流模型多假设交通需求是确定的,本论文假设交通需求由通勤出行者和非通勤出行者组成。通勤出行者具有确定的交通需求量,而非通勤出行者的交通需求量是一个有界的离散随机变量。本论文首先说明了在这种随机交通需求条件下,交通状态的有限性,之后给出了交通状态的演化过程满足马尔可夫链的一个充分条件,该充分条件同时保证了该马尔可夫链平稳分布的唯一性。在这个充分条件下和有关通勤出行者的学习机制、更新机制假设的基础上,本论文建立了一个马尔可夫配流模型,并给出实现该模型的算法。基于该算法,本论文通过数值试验分析了通勤出行者学习机制中的记忆长度参数、交通需求、出行者的感知误差对系统平稳分布的影响。通过研究发现,通勤出行者的记忆长度对系统平稳分布的影响较小,而交通需求和出行者的感知误差则对系统平稳分布的影响较大。更多还原

【Abstract】 With the growth of population and urbanization, traffic congestion has become one of the most serious problems in most cities all over the world. It really affects the healthy development of a city and social economy. To develop the effective measure alleviating traffic congestion, it is needed to perfectly understand the behavior of travelers’travel choice. This dissertation studies four traffic conditions:traffic network with Advanced Traveler Information System (ATIS), multi-modal traffic network, the urban railway network, and the stochastic traffic network. The corresponding day-to-day traffic assignment model would be presented, and some traffic management measures are developed based on the analysis of travel choice in these traffic coditions. The specific contents of the dissertation are as follows:(1) The travel information of ATIS is classfied into two classes:accurate travel information and inaccurate travel information. Two day-to-day dynamic assignment models are respectively developed to study the travel choice of day-to-day and the stability of traffic system for the both cases. Based on the numerical results, it is found that the market penetration of ATIS, traffic demand, the perceived error all have effect on the stability of traffic system in both cases of the accuracy of travel information.(2) A traffic management measure is presented by combining the route guidance of ATIS and the continue network design (CNDP). With the measure, the travelers are recommended to choose the shortest path with user constraints. The problem is described by a bi-level program, and the feasible domain of its lower level is determined by the deterministic variable of upper level. The Simulated Annealing algorithm is improved to solve the bi-level problem. And the path-based traffic algorithm is developed to calculate the traffic assignment problem under the route guidance. Compared to the present results of CNDP, the measure presented in this dissertation can better alleviate the traffic congestion.(3) A bounded rationality based day-to-day urban railway dynamic assignment model is developed to study the day-to-day evolution of urban railway passenger flow. The passengers’bounded rationality is described as follows:If the difference between their expect pericevied travel cost and their expericed travel cost is more than the tolerance of passengers on a day, they would reconsider their travel choice in the next day; otherwise, they would use the same route in the next day. The travel cost in the model considers the cost casued by the congestion in carriage of urban railway. It is founded that the parameter of travelers’ tolerance and traffic demand can obviously affect the convergency of the evolution of urban railway system. The travel cost has less effect on the system’s convergnecy, and it has influence on the total travel cost significantly.(4) Based on a day-to-day traffic dynamic assignment model, the stability of bi-modal (car and bus) traffic system is studied both in mixed traffic conditions and traffic network with dedicated bus lane. Meanwhile, the travelers’ transferring behavior is considered in the study. It is found that, both considering and without transferring behavior, traffic system with dedicated bus lane can maintain stable with more traffic demand than that in mixed traffic conditions.(5) In stochastic traffic network, a Markovian assignment model with stochastic traffic demand is presented. The traffic demand is grouped into two parts:commuters or regular travelers with fixed traffic demand and irregular travelers with bounded discrete random demand. With some wild conditions, it is proved that our Markovian assignment model is ergodic and has unique stable distribution. An algorithm is also developed to describe the Markovian model. Based on a simple tested example, it is found that the length of commuters’ memory has less effect on the stable distribution, but the traffic demand and the commuters’ perceived error has remarkable influence on the stable distribution.更多还原