【作者】 向郑涛；

【导师】 熊励；

【作者基本信息】 上海大学 ， 管理科学与工程， 2013， 博士

【摘要】 随着社会生产力水平的不断提高，交通运输业在人类社会生活中的地位越来越重要，其发展直接影响到社会的文明程度和经济的可持续发展。交通运输业的快速发展在给我们的生活带来极大便利的同时，也带来了一系列社会问题，主要表现在以下三个方面：交通拥堵、交通事故和环境污染。要想从根本上解决交通运输业发展带来的交通拥堵等社会问题，不仅需要加强交通基础设施的建设，还需要对交通系统进行系统科学的研究，揭示交通系统的规律和特性，从而更好地指导交通系统的建设和管理。交通流理论是解释交通现象、分析交通问题和指导交通管理的研究基础，能够有效地指导交通拥堵等问题的解决。本文在交通流建模、交通流复杂性的定量分析以及基于信息反馈的交通流诱导策略这三个方面进行了研究，不仅有其深远的理论意义，而且具有重要的工程应用价值。(1)基于刹车灯规则的交通流元胞自动机建模研究为了研究微观意义上车辆之间的相互作用对宏观意义上交通流演化的影响，并揭示三相交通流中同步流的产生机制以及同步流与宽运动堵塞之间的相变机制，本文基于刹车灯规则，在Tian模型的基础上，考虑确定性减速对随机慢化的影响，改进驾驶行为的建模规则，提出改进刹车灯模型。改进刹车灯模型修改了加速规则和随机慢化规则，不仅使车辆之间的相互作用规则更加符合驾驶员的实际驾驶行为，而且还避免了车辆过度减速现象的发生。通过对基本图和时空图的定性分析，发现该模型不仅能够很好地模拟交通流的三种状态：自由流、同步流和宽运动堵塞，而且还能够很好地描述交通流演化的复杂性特征。(2)基于多尺度熵的交通流复杂性定量分析时空图可以定性地描述交通流演化的复杂特性，但是，仅对交通流复杂性进行定性分析还不足以深刻认识交通流的本质特征。为了定量度量交通流的复杂性，以实现不同状态下的交通流复杂性对比，并分析交通流元胞自动机模型中各个参数对交通流复杂性的影响，本文以NS模型和前述提出的改进刹车灯模型为例，采用多尺度熵方法，以车头时距时间序列为研究对象，定量分析了车头时距在不同时间尺度的复杂性。分析结果表明：对于NS模型，密度对车头时距复杂性的影响较大，随机慢化概率对车头时距复杂性的影响较小；对于改进刹车灯模型，同步流的出现会显著增加车头时距的复杂性。(3)基于信息反馈的交通流诱导策略研究为了达到理想的交通流诱导效果，需要基于实时交通信息，制定合理的交通信息反馈策略。此外，基于交通流元胞自动机模型，可以实现不同场景下的交通流模拟和动态演化，从而为验证交通信息反馈策略的有效性提供基础。本文首先从ITS的实际应用现状出发，研究车联网覆盖程度对典型信息反馈策略性能的影响，为车联网环境下信息反馈策略的选取提供参考；然后，提出了一种新的信息反馈策略――加权平均速度反馈策略，并分别基于NS模型和前述改进刹车灯模型，与典型信息反馈策略进行性能对比和分析，结果表明：该策略不仅性能更好，而且具有良好的鲁棒性，使得该策略具有很强的实用性和适用性。论文创新点在于：(1)提出了基于刹车灯规则的改进模型，揭示了同步流和宽运动堵塞之间相变的随机特性。(2)将多尺度熵方法引入到交通流分析中，实现了交通流复杂性的定量度量。(3)分析了浮动车比例对典型信息反馈策略性能的影响，为车联网环境下信息反馈策略的选取提供参考。(4)提出了加权平均速度反馈策略(WMVFS)，在不同的应用场景下都具有很好的交通流诱导性能。更多还原

【Abstract】 With the improvement of the level of social productive forces, transportationindustry is more and more important in human society because it influents thecivilization degree of society and the sustainable development of economy. However,the rapid development of transportation industry brings great convenience to our life,as well as several social problems, such as traffic congestion, traffic accidents andenvironmental pollution. Efforts on the transportation infrastructure are not sufficientto solve the transportation problems. The systematic and scientific research of trafficsystem is also very important for the reveal of transportation system for the laws andcharacteristics and the guidance of development and management of transportationsystem. Traffic flow theory provides theoretical basis to explain the trafficphenomena, analyze the traffic problems and guide the traffic management, whichcan help to solve the social problems, such as traffic congestion. In this dissertation,traffic flow modeling, quantitative analysis of traffic flow complexity and trafficflow guidance strategy based on information feedback are focused. The research inthe dissertation has academic significance and promising application.(1) Traffic flow cellular automaton model based on brake light rulesTo investigate the interaction of vehicles from micro aspect and the influence ofthe interaction on traffic flow evolution from macro aspect, an improved brake lightmodel is proposed based on Tian model considering the influence of deterministicdeceleration on randomization and improving of modeling rules for drivingbehaviors, with which the generation mechanism of synchronized flow inthree-phase traffic theory and the mechanism of phase transitions betweensynchronized flow and wide moving jam are explored. The improved brake lightmodel modifies the acceleration and randomization rules. The modifications makethe interaction between vehicles more realistic and avoid the phenomenon of over-deceleration. The qualitative analyses of the fundamental diagram andspatial-temporal diagrams show that the new model can reproduce the three trafficphases: free flow, synchronized flow and wide moving jam. In addition, the newmodel can well describe the complexity of traffic flow evolution.(2) Quantitative complexity analysis of traffic flow based on the multi-scaleentropy methodThe spatial-temporal diagrams can describe the complexity of traffic flowevolution qualitatively. However, qualitative analysis is not enough for thecomplexity analysis of traffic flow. To describe the complexity of traffic flowquantitatively for further investigation, such as the comparison of complexity oftraffic flow between different phases and the influence of parameters of traffic flowmodel on the complexity of traffic flow, multi-scale entropy method is used toquantitatively analyze the complexity of time headway at different time scales. Thetime series of time headway are generated based on NS model and our improvedbrake light model, respectively. Analysis results show that for NS model, the vehicledensity has larger influence on the complexity of time headway than randomizationprobability does; for our improved brake light model, the emergence of synchronizedflow will increase the complexity of time headway greatly.(3) Traffic flow guidance strategy based on information feedbackTo fulfill successful traffic flow guidance, reasonable information feedbackstrategies based on real-time traffic information are needed. In addition, based on thetraffic flow automaton model, traffic flow simulation and evolution with differentscenarios can be achieved to provide basis for the validation of traffic informationfeedback strategies. In this dissertation, the application of Intelligent TransportationSystem (ITS) is discussed and the influence of the coverage of Internet of Vehicleson the performance of typical information feedback strategies is analyzed, which canprovide a reference for the selection of information feedback strategies with differentcoverage of Internet of Vehicles. And a new information feedback strategy is proposed, which is the Weighted Mean Velocity Feedback Strategy (WMVFS).Based on the NS model and our improved brake light model, the performances ofWMVFS and typical information feedback strategies are compared and thecharacters of WMVFS are analyzed. Results show that our new strategy has betterperformance and better robustness, which means WMVFS has better practicabilityand applicability.The main contributions are as follows:(1) The improved brake light model isproposed, which can reveal the random character of phase transitions betweensynchronized flow and wide moving jam.(2) The multi-scale entropy method is usedin the analysis of traffic flow, which can quantitatively measure the complexityanalysis of traffic flow.(3) The influence of the proportion of float cars on theperformance of typical information feedback strategies is analyzed, which canprovide a reference for the selection of information feedback strategies in scenarioswith Internet of Vehicles.(4) The Weighted Mean Velocity Feedback Strategy isproposed, which has good performance for traffic flow guidance in differentapplication scenarios.更多还原