【作者】 于尧；

【导师】 杨兆升；

【作者基本信息】 吉林大学 ， 交通信息工程及控制， 2014， 博士

【摘要】 近年来，随着我国城镇建设的迅猛发展，短短三十年便完成了西方发达国家近百年的城镇发展进程。城市的大踏步前进释放了巨大的交通需求，交通供需矛盾日益突出。传统解决交通问题的方法即增加交通供给，但如新建道路等传统手段已被证明是无法从根本上解决交通问题的。新建的道路往往会诱发新的交通出行需求。供-需之间的平衡关系总是需求大于供给。所以单纯的新增道路不仅不科学，往往还会加剧土地资源矛盾，得不偿失。在土地面积有限的前提下，通过智能交通手段对出行需求进行管理和分配才是缓解当前交通矛盾突出，提高路网运行效率，改善道路交通环境的有效途径。根据《国家中长期科学与技术发展纲要（2006-2020）》中将智能交通管理系统的建设与发展列为优先发展主题的系列要求，以智能交通系统理论方法为基础，大力发展智能交通诱导技术在城市路网中的应用将是今后交通领域研究的热点。本文依托“国家科技支撑计划”课题以及“国家自然基金”项目，对基于出行者行为的动态交通分配相关模型、方法展开深入研究，为实现智能化交通管理和改善道路交通环境、提高路网运行效率提供理论支撑和技术支持。本文在对现有动态交通分配模型研究的基础上，对当前分布式动态交通分配以及中心式动态交通分配的各自优劣进行分析总结，并结合出行者行为约束，旨在提出满足实际用户出行需求，实时性强，策略更新快的闭环中心式动态交通分配模型。该方式可有效结合分布式的实时更新性强等优势，形成可交互反馈的“分布-中心”新型诱导机制，进而弥补当前单一诱导形式存在的开环、求解难等不足。首先，通过对网络动态均衡理论的分析，在已有理论思想的指导之上，引入动态交通分配过程中的博弈机制，为今后更加人性化、以人为本的智能化交通管控策略提供理论依据；紧接着，设计全新的路径选择偏好调查方法，选取出行者在出行过程中感受最为明显的数据作为调查指标，经过统计分析后的数据不仅可为动态交通分配策略的制定提供数据支持，也为未来交通出行偏好调查提供一种新的调查方向。然后，提出了区域自治的分布式动态交通分配模型，并对车载路径算法进行了深入的研究；最后考虑到诱导分配的不同用户群体，结合分布式与中心式二者诱导模式，研究了面向不同用户级别的中心式动态交通分配，设计了中心式动态交通诱导模型框架，并构建了基于博弈行为的模型及其求解方法。通过本文的研究，所提出的分布式动态交通分配模型、面向多用户等级的中心式动态交通分配模型等模型方法为提高智能交通诱导效果，改善路网交通负载情况提供了有力的技术方法保障。主要研究工作如下：（1）基于动态需求的网络均衡建模方法论文在广泛阅读网络均衡理论研究的基础上，归纳既有交通分配建模方法，以交通供需矛盾与应用智能交通诱导技术的必要性为切入点，结合动态交通流的时变特性，重点分析了动态系统最优交通分配模型和用户最优交通分配模型，并提出采用Stackelberg博弈行为对出行者与交通管理者二者间的动态博弈行为过程进行研究。（2）基于期望效益的路径选择行为建模考虑到出行者的主观路径选择偏好，在分析期望效用理论和非期望效用理论在出行行为研究中的适用性基础上，结合行程时间的可靠性与不确定性，提出基于期望效益的路径选择行为模型，并对不确定条件下路径选择过程中的“风险条件”和“模糊条件”情境下的选择行为进行了深入研究，得到延误时间概率与选择偏好概率之间的函数关系。实验结果可以较好的反映现实中道路出行者的选择行为，这对出行者对路径未有任何认知的前提下，做出准确路径偏好选择估计具有积极的推动作用。（3）多兴趣点连续搜索方法在详细分析车载最短路径算法的基础上，结合实际出行需求，提出了一种可满足出行者一次出行中访问多个兴趣点的连续路径搜索方法。提出的方法在路网搜索结构和数据访问机制等方面皆进行了优化，通过兴趣点近邻区域的时空关联推理，计算最优出行路径。并对提出的路径搜索方法进行了实验验证，结果表明，与一般最近邻算法相比，可有效提高计算性能，避免不合理路径的出行，满足出行者不同规则下的访问需求。（4）基于边际行程时间的分布式动态交通分配模型论文针对传统交通分配方法在较大规模路网下难以实施，且策略更新慢等不足，以效用理论为依托，提出了基于边际行程时间的分布式动态交通分配模型，并给出了模型求解算法。提出的分布式结构提供了一种基于局部有限信息下的实时自适应分配策略，这种基于局部控制范围的策略可以有效响应路网交通流的实时变化，计算量低，更新速度快。此外针对模型参数选取以及路网（节点）覆盖程度等影响因素分别进行了数值验证。提出的分布式动态交通分配方法不但可以有效满足系统运行需要，亦为ITS中诱导系统的建设提供了一个新的思路和理论支撑；（5）面向多级用户的中心式诱导策略对于中心式交通诱导问题，论文从博弈理论的角度出发，结合分布式动态交通分配策略优势，在分析分布端与中心端的二者间的Stackelberg博弈行为，面向不同诱导等级用户，提出一个新的交互式中心式诱导策略，并构建了基于Stackelberg的多用户诱导模型及其求解算法。仿真结果表明所提中心式诱导策略可以提高系统运行效率并有效保障系统运行的稳定性，完善了现有动态交通分配方法，并为城市智能交通管理技术的实现提供了理论支撑。更多还原

【Abstract】 In recent years, urban developmentprocess only cost a short span of30yearscompare with western developed countriesnearly a centurywith the rapid development ofurban construction in China. A strong traffic demand was released by a big strideforwardon the development of cites, which leads to prominent contradiction between supply anddemand increasingly. The traditional method to solve the traffic problem has been provedto be useless as to purely increase supply by the new roads. New roads tend to inducenew travel demand, and the balance always lay the demand rather than the supply. Hence,the single method of increasing new road is not reasonable. Eventually, it can alsoexacerbate the contradiction of land resources. Consequently, the effective measure toalleviate the prominent contradiction and improve the network running efficiency is byusing intelligent traffic management under the background of limited area of land.According to “National medium and long-term science and technology developmentplan (2006-2020)”, the topic of “intelligent traffic management system construction anddevelopment” is listed as a priority developing theme. Striving to develop the technologyof urban traffic guidance will be a hotspot in the future. This dissertation studied dynamictraffic assignment combined with travelers’ constraint based on "National Science andTechnology Support Plan" project and "National Natural Fund” Project, in order torealize the true intelligent traffic management and improve the road environment. Atfinal, these studies can be the foundation and support for improving the network runningefficiency.Based on the existing research on the basis of dynamic traffic assignment model,this dissertation analyzed the advantage and disadvantage of distributed dynamic trafficassignment and centralized dynamic traffic assignment respectively, aiming to proposea real-time and strategy update fast closed-loop centralized dynamic traffic assignmentmodel. The proposed strategy can effectively combine the advantage of the distributeddynamic traffic assignment strategy which is hereby form a new mechanism withinteractive feedback. Furthermore, it is to make up current open-loop of single inducedpatternand the dilemma of hard to solve. First of all, game mechanism of dynamic trafficassignment was introduced through the analysis of the network dynamic equilibriumbased on the research of existing theories. And then, the author designed a creative routechoice preference investigation method by selecting the significant factors which thetravelers feel the most obvious. These investigated data not only can be support for thedynamic traffic management but also can provide a brand new angle of the travel preference investigation. Afterwards, regional autonomy distributed dynamic trafficassignment model was proposed as well as in-vehicle route calculate solutions werediscussed. At the final, this dissertation constructed a centralized guidance framework tostudy the centralized dynamic traffic assignment confronted different user levelscombined the induce pattern with distributed and centralized by considering thedistribution of user groups. According to above studies, the results of proposed modelscan improve the intelligent traffic guidance efficiency meanwhile provide powerfultechnical method guarantee of developing network loading. The main research asfollows:（1）Network equilibrium modeling method based on the dynamic demandThis paper summarizes the existing modeling methods of traffic distribution first onthe basis of extensively reading network equilibrium theory researches. Then the paperanalyzes optimum traffic assignment model of dynamic system and user optimum trafficassignment model specially and proposes a method to study the process of dynamic gamebehavior using Stackelberg game behavior theory whose breakthrough point is thecontradiction between supply and demand of traffic as well as the necessity of theapplication of intelligent traffic guidance technology, combining with time-variantcharacteristics of dynamic traffic flow.（2）Route choice behavior modeling based on the expected benefitsConsidering the travelers’ subjective route choice preference and combining withreliability and uncertainty of travel time, this paper puts forward aroute choice behaviormodel based on the expected benefits which is on the basis of the applicability researchof expected utility theory and the expected utility theory in travel behavior. Besides, thispaper researches the behavior choice deeply under the condition of "risk" and "fuzzy"situation in the route choosing process of uncertainty condition, getting the functionrelationship between delay time probability and the probability of choice preference.Experimental results can well reflect choice behavior of road travelers in reality, whichhas a positive promoting role in making an accurate path preference estimates whentravelers are under the premise of no cognition on the path.（3）Continuous search method with multi-interest pointsThis paper proposes a continuous path search method which can meet travelers toaccess to more than one points of interest during a trip on the basis of detailed analysis ofvehicle shortest path algorithm and combination with the actual travel demand. Themethod is optimized in road network structure and data access mechanism and calculatesthe optimal travel paththrough the temporal and spatial relation reasoning of the adjacentarea of the interest points. The proposed path search method is verified byexperimentsand the results show that the new path search methodcan effectively improvethe computing performancecompared with the average nearest neighbor algorithm, to avoid unreasonable path of travel and meet the access requirements under different rulesof traveler.（4）The distributed dynamic traffic assignment model based on marginal traveltimeThis paper puts forward a distributed dynamic traffic assignment model based onmarginal travel time and gives the model solving algorithmin view ofthedifficultimplement of traditional traffic assignment methodin large-scale road networkand slow update of strategy, which is based on utility theory. The proposed distributedstructure providesa local real-time adaptive allocation strategy under limited information.This proposed method which is based on local control strategy can effectively responseto real-time network traffic flow changes, whose amount of calculation is less and updatespeed is fast. In addition, choosing model parameters and influence factors such asnetwork coverage (nodes) are verified numericallyrespectively.The proposed distributeddynamic traffic assignment method not only can effectively meet the needs of the systemrunning, but also provides a new train of thought and theory support for the constructionof guidance system in ITS.（5）Multiple users oriented center-basedguidance strategyThis paper analyzes the Stackelberg game behavior between distribution side andcenter from the perspective of game theory and combination with the distributed dynamictraffic assignment strategy advantage for central type induced traffic problems and putsforward a new interactive center type induction strategy and builds a multi-user inductionmodel based on Stackelberg model whose solution algorithm is given, geared to theneeds of different level users. The simulation results show that the proposed center typeinduction strategy can improve the system efficiency and effectively guarantee thestability of the system, perfectthe existing dynamic traffic assignment methodandprovides a theoretical support for the realization of the urban intelligent trafficmanagement technology.更多还原