【作者】 刘明君；

【导师】 毛保华； 贾顺平；

【作者基本信息】 北京交通大学 ， 交通运输规划与管理， 2010， 博士

【摘要】 混合交通是中国城市交通的基本特征。在混合交通环境下,机动车、非机动车与行人的特性各异,相互干扰严重,导致道路设施利用率与交通运行质量下降,为城市道路交通管理与控制带来巨大挑战。城市道路信号交叉口是城市路网中最基本的节点,是城市道路网络中通行能力和交通安全的瓶颈,交叉口的通行能力影响着整个交通网络的运输能力。信号交叉口排队车辆车头时距是城市微观交通分析中的一个关键参数,直接反映了司机驾驶行为特性、交叉口通行能力以及服务水平。本文在综述国内外相关研究的基础上,结合混合交通调查数据,从混合交通环境下信号交叉口排队车辆队列消散特征、首车车头时距、排队车辆车头时距分析及通行能力研究等方面进行了一些有益探讨,主要研究内容及结论如下：①根据摄像采集数据,对混合交通流环境下城市道路信号交叉口排队车辆队列消散过程及车头时距进行了深入分析,提出了车辆消散的四个阶段,即初始加速阶段、致密性行驶队列阶段、队列消散阶段和自由流阶段,并指出受机非干扰等因素的影响,车流消散的四个阶段不一定全部出现。对混合交通流消散特性进行了分析,明确指出信号交叉口首车车头时距从大到小依次为专右车道、专左车道、直外车道和直内车道,并定量分析了机非干扰和车辆类型对车头时距的影响。②在适应性分析的基础上,将风险分析方法引入交通流研究领域,首次提出了基于风险分析方法的信号交叉排队车辆首车车头时距分析模型,并对偏似然函数的构建和模型参数估计方法做了详细探讨,以北京典型信号控制交叉口首车车头时距数据为对象进行了案例分析,取得了良好效果。分析结果表明北京信号交叉口排队车辆首车车头时距达到4.54s,明显大于国外同类城市；模型影响因素分析结果还揭示了启动干扰和行驶干扰对分别导致首车平均车头时距增长了37.09%和60.24%。③系统分析了既有三大车头时距模型,即Briggs确定性模型、Bonneson模型和神经网络模型的缺陷,阐述了统计学习理论和支持向量机方法的优点,首次构建了基于支持向量机方法的信号交叉口排队车辆车头时距分析模型,提出了分析框架和步骤,将本车类型、前车类型、车道功能、前车车头时距、排队位置和有无机非干扰这六个影响因素作为模型输入,排队车辆车头时距作为输出,通过相对误差、平均绝对误差和均方根绝对误差三个统计指标对模型性能进行评价。以北京市知春路口排队车辆车头时距为研究对象,应用本文提出的支持向量机模型、与确定性模型和神经网络模型进行混合交通流环境下信号交叉口排队车辆车头时距分析,结果表明,本研究提出的支持向量机模型明显优于确定性模型和神经网络模型。④通过对信号交叉口排队车辆队列消散的稳定性分析,明确指出混合交通流环境下,排队车辆队列消散存在波动性,HCM定义的饱和流率很难出现,导致HCM提出的基于饱和流率的通行能力计算模型存在不适应性。通过引入信号转换时间内平均能通过的车辆数Nvki和显示绿灯时间内平均能通过的车辆数Ngki,构建了修正通行能力模型,并利用累计曲线法对相关参数进行了标定,以北京市知春路口东进口为研究对象进行了实证分析。结果表明,本模型在准确性、简洁性和适应性方面均优于HCM模型。更多还原

【Abstract】 ABSTRACT:Most cities in China are characteristiced by heterogeneous traffic. Under the heterogeneous traffic condition, motorized vehicles, non-motorized vehicles and pedestrians all have their own characteristics, and interfere with each other, which cause the decrease of road facilities utilization and operation quality, and challenge the road transport management and control system. The intersection is the basic node of the road network in a city, and also the bottleneck of network capacity and traffic safety. Therefore, the intersection capacity has a great influence on the whole transport network.As a key parameter for traffic microscopic analysis, discharge headway of queuing vehicle in signalized intersection dircectly reflect driving behavior characteristic, as well as intersection capacity and service level. Based on the review of relavent research home and abroad, combined with heterogeneous traffic investigation data, several subjects, like discharge characteristics of queuing vehicles, the first discharge headway, successive discharge headway are analysed. Main works and conclusions of this dissertation are summarized as below:①According to the video record data, we elaborately analyses the discharging process of queuing vehicles and its discharge headway. On this foundation, the four stage of discharging are proposed, i.e. the initial speedup phase, compaction platoon phase, platoon dispersion phase, and free flow phase. However, the four stages may not appear under some situation owing to non-motorized vehicles disturbance. By analyzing the discharge characteristic of heterogeneous traffic, this dissertation expressly instructs that the discharge headways on lane type are in the following descending order:the right-turn-only-lane, the left-turn-only-lane, the off-side straight lane, and the kerb-side straight lane. And then, we also make a quantative analysis on the effects on discharge headway for non-motorized vehicles disturbance and vehicle type.②Hazard-based analysis method is introduced into traffic flow field after adaptability analysis, and for the first time first discharge headway is modeled basing on hazard-based analysis method. The partial likelihood function and parameters’ evaluation method is explored deeply next. Then a case study of typical signalized intersections at Beijing is conducted. The results provided that in Beijing city the first discharge headway arrives at 4.54s, which is obviously longer than in other equivalent foreign cities. And the factor analysis results also reveal the influence degree of the starting disturbance and the moving disturbance, which causes an increase in the average first discharge headway to 37.09 percent and 60.24 percent, respectively.③The defects of existing top-three widly-used headway models, which are Deterministic Model (DM), Bonneson Model (BM), and Neural Network Model of discharge headway (NNM), are analysed, and the principle of Statistical Learing Theory and Support Vector Machine (SVM) is represented next. In addition, a SVM-based discharge headway model at signalized intersection is constructed for the first time, and the related analytical frameworks and process are presented. In the model, subject vehicle type, preceding vehicle type, lane type, discharging headway of the preceding vehicle, queuing position, and disturbance are input, and discharge headway of the subject vehicle is output. Besides, three indexes, i.e. relative percentage error (RPE), mean absolute percentage error (MAPE), and root mean square error (RMSE), are introduced to evaluate the model perfoemance. At last, performances of three models, namely SVM-based discharge headway model (SVMM), DM, and NNM, are evaluated by case studies. The results indicate that the SVMM is better than the others.④Based on the discharge stability analysis of queuing vehicles at signalized intersections, the saturation flow rate defined in Highway Capacity Manual is pointed out that can hardly happen under the heterogeneous traffic condition at Beijing. This means that the HCM capacity model based on the saturation flow rate is not adaptive. After introduction of two variables, i.e. the expected number of vehicles that can be discharged during green interval of usable phase,, and the expected number of vehicles that can be discharged during green time of usable phase,, a modified intersection capacity model is presented and the related parameters is calibrated by cumulative curve method. Finally, the eastbound of Zhichunlukou is conducted as an empirical analysis. The results show that the modified model is better than HCM model in accuracy, simplicity and universality.更多还原