节点文献

基于生存分析的信号交叉口非机动车穿越行为研究

Crossing Behavior of Nonmotorized Vehiclesat Urban Intersections Based on Survival Analysis Method

【作者】 环梅

【导师】 贾斌; 杨小宝;

【作者基本信息】 北京交通大学 , 系统科学, 2014, 博士

【摘要】 机动车与非机动车的混合交通是当前我国城市交通的主要特征,也是造成我国城市交通拥堵和事故频发的一个重要原因。非机动车出行灵活、准时性高,是解决中短距离出行和接驳换乘的理想交通方式,作为符合我国国情并拥有广泛群众基础的代步工具,在现阶段依然具有不可替代性。以前的非机动车主要指人力自行车,近年来随着技术的发展,电动自行车的使用已越来越普遍。与人力自行车相比,电动自行车可到达的距离更远,速度更快,但伴随的交通隐患也更高,这给我国城市交通带来了一些新的问题。此外,非机动车是一种健康、环保、低能耗的出行方式。发展城市非机动车交通是预防和缓解交通拥堵、减少大气污染和能源消耗的重要途径之一,关系人民群众的生产生活和城市可持续发展。然而,城市交通中,非机动车是一个相对弱势的群体,涉及到非机动车的交通事故比重一直居高不下。非机动车交通事故中,闯红灯违规是导致事故发生的主要原因,由于较差的法律约束和人们较低的安全意识,当前非机动车的闯红灯行为在我国极为普遍。目前,研究信号交叉口非机动车闯红灯行为的文献很少,更鲜有文献运用生存分析方法来分析非机动车闯红灯行为。生存分析的优点是可以考虑删失数据,并能将事件的结果和出现此结果所经历的时间结合起来分析,非常适合用来研究信号交叉口闯红灯行为。据此,本论文结合当前我国交通的特点,以非机动车为主要研究对象,基于生存分析方法,重点分析城市干道信号交叉口非机动车的穿越行为,揭示信号交叉口非机动车骑行者的违规风险、等待忍耐时间及其关键影响因素,并对特殊的管控措施进行评价。具体来讲,本论文的主要工作如下:(1)非机动车穿越行为的实证调研。通过选取典型城市干道信号交叉口对非机动车穿越行为进行实地拍摄,获取基础数据。分析非机动车和行人穿越信号交叉口过程中的等待时间分布、等待区域空间分布、运行轨迹、运行速度、穿越间隙等,揭示信号交叉口自行车、电动车与行人的微观行为差异。结果表明电动车的违规率明显高于行人和自行车;来自两侧非机动车的违规率明显高于直行到达的;行人不等待直接违规的比例显著低于自行车和电动车;违规者中行人的等待时间明显比自行车和电动车长,电动车运行速度比自行车和行人的快,而穿越的安全界限则比自行车和行人小。(2)非机动车骑行者的等待忍耐时间分布规律。基于实证调研数据,建立信号交叉口非机动车穿越前的等待忍耐时间持续模型,对红灯期间到达的非机动车在穿越前的等待忍耐时间进行估计,探索不同等待时间下非机动车闯红灯违规率的分布。结果表明随着等待时间的增加,非机动车的违规概率逐渐增大,18.2%的骑行者几乎不等待就直接闯红灯违规;20.6%的骑行者愿意等待时间120s,甚至更长。并指出对删失数据的不当处理会导致明显高估非机动车骑行者的闯红灯违规率。(3)非机动车闯红灯行为的Cox风险模型及其影响因素分析。构建非机动车骑行者等待忍耐时间的Cox风险模型,运用调查数据对模型参数进行估计,系统地分析各个潜在因素对骑行者的违规风险和等待忍耐时间的影响。结果表明交通方式、等待位置、高峰期、从众行为和机动车流量等对非机动车闯红灯行为有显著影响。电动车骑行者比自行车的违规风险更高,愿意等待的时间更短;等待位置越靠前(靠近路)违规风险越大;平峰期的违规风险大于高峰期;正在违规的人越多,机动车流量越小,则非机动车骑行者越容易违规,等待时间越短。并指出建立的Cox风险模型可用来预测或评估这些交通运营、管理和政策的变化对非机动车闯红灯行为的影响。(4)非机动车通勤者的安全穿越可靠性建模与分析。运用可靠性思想和加速风险模型理论,结合非机动车穿越问题,建立非机动车通勤者的安全穿越可靠性模型,基于实证数据,找出最优可靠性模型的数学形式,并揭示影响通勤者安全穿越可靠性的关键因子。进一步把非机动车通勤者分为等待者和不等待者两类,分别探讨了他们的安全穿越可靠性问题。结果表明在各个潜在影响因素中,骑行者的等待位置、来自方向和从众行为等行为特征因素是影响非机动车通勤者安全穿越可靠性的最主要因素;Gompertz模型最适合用来拟合非机动车等待人群的安全穿越可靠性问题。(5)交通协管的管控效果评价。通过运用Logistic模型、方差分析、协方差分析和生存分析等方法,比较信号交叉口有无交通协管时非机动车的穿越行为特征、闯红灯违规率和等待忍耐时间的差异,据此来评价交通协管对非机动车闯红灯行为的管控效果。结果表明交通协管对闯红灯行为有显著影响。有协管时非机动车和行人的闯红灯违规率都显著低于无协管的情形,有协管时的等待忍耐时间则比无协管时更长;不同来自方向中,交通协管对直行人群的闯红灯行为具有很好的管控效果,而对来自左右两侧人群的管控效果相对较弱。

【Abstract】 As a developing country, China has its own traffic characteristics. A mix of non-motorized and motorized vehicles is an important traffic type in China. It is also one of the most important reasons which cause urban traffic congestion and frequent accidents. Because of its high flexibility and punctuality, a non-motor vehicle is the ideal mode of transportation for short/middle distance travel and transfer to public transport. Non-motorized vehicle is irreplaceable at this stage because it is suitable to our national conditions and has widely basis of the masses. In the past, Non-motorized vehicle only refers to human bicycle. In recent years, with the development of technology, electric bike is widely used. Compared with human bicycle, riding an electric bike can reach faster and farther, but more unsafely. It is a new problem of urban traffic in China. In addition, riding a non-motorized vehicle is healthy, non-polluting, and energy-efficient. Development of urban non-motorized traffic can prevent and mitigate traffic congestion, reduce air pollution and energy consumption. It is very beneficial to people’s living and urban sustainable development.However, non-motor vehicle is a relatively weak group in urban traffic. There are always a high proportion of traffic accidents involving non-motor vehicles. One typical type of rule violation behavior is red-light running. Because of the poor law enforcement and peoples’low safety awareness, red-light running is rather prevalent in China. The literature review suggests that very little has been done on the red-light running of non-motor vehicles, much less on this study based on survival analysis method. Survival analysis has the advantage that can consider censored data, and can combine to study the result of the event and the time that the result experienced. This method is very suitable for studying red-light crossing behavior at signalized intersections. Therefore, according to our national traffic characteristic, based on survival analysis methods, this dissertation focused on the study of crossing behavior and waiting endurance times of non-motorized vehicles and their influence factors. Then, a special management measure was evaluated. Specifically, the contents of this dissertation are as follows:(1) Crossing behavior of non-motorized vehicles was analyzed by the empirical research. First, typical intersections on main urban roads were chosen, basic data about crossing behavior was collected by field observation. Then, several variables which described crossing behavior at intersection were coded and were used to reveal microscopic behavior differences among cyclists, electric bike riders and pedestrians. These variables included waiting time, waiting position, moving trajectory, travel speed and safety margin, etc. The results show that red-light crossing rates of electric bike riders are significantly higher than those of cyclists and pedestrians. The riders coming from both sides are more likely to run against a red light than the ones of straight arrival. Pedestrians are less likely to not wait to cross the red light than bicycles and electric vehicles. Generally, they are likely to wait longer times than cyclists and electric bike riders. Among three modes of transportation, an electric bike has the fastest speed and the smallest safety margin.(2) Waiting endurance time distributions of non-motor vehicles were explored. Based on the empirical research, a duration model of waiting times for non-motor vehicles crossing an intersection was proposed. Their waiting times were estimated. The red-light crossing rates of non-motor vehicles were explored. The results show that the red-light crossing rates increase with the increasing waiting time. About18.2%of riders are at high risk of violation and low waiting time to cross the intersections. About20.6%of all the riders are generally non-risk takers who can obey the traffic rules after waiting for120seconds. In addition, it is noted that the improper handle of censored data would overestimate the red-light crossing rates of non-motor vehicles.(3) Cox hazard-based models and the influence factors of red-light running behavior were investigated. Cox hazard-based models of riders’ waiting endurance times were proposed. Based on the surveyed data, the model parameters were estimated. The effects of various potential factors on riders’ violation risk and waiting time were analyzed systematically. The results show that traffic mode, waiting position, rush hour, conformity behavior and motorized vehicle volume have significantly impact on red-light running behavior. Electric bike riders have higher risks and shorter waiting times than cyclists. Nearer to main roads waiting position is, more likely to run against a red light riders are. Riders in off-peak hours are more likely to run against a red light than those in peak hours. Riders have higher risk and shorter waiting times with less volume of motorized vehicles, as well as the bigger number of other riders that are crossing against the red light when arrives. The Cox hazard model formulated in this chapter can be applied to forecast temporal shifts in waiting duration times of non-motorized vehciles due to changes in traffic operation, management and control.(4) Safety crossing reliability of commuter riders was modeled and analyzed. Crossing reliability of commuter riders at intersections was proposed by using reliability theory and accelerated hazard models method. Based on the empirical data, the optimal mathematical model of safety crossing reliability was chosen. The key factors affecting the crossing reliability were investigated. Furthermore, commuter riders are divided into two categories:wait and don’t wait. Their safety crossing reliabilities were discussed respectively. The results show that some potential variables including waiting position, moving direction and conformity behavior have significantly impacts on safety crossing reliability of commuter riders. Gompertz distribution model is very appropriate for fitting of safety crossing reliability for waiting commuter riders.(5) Control effects of traffic wardens on red-light running behavior were evaluated. By using Logistic model, analysis of variance, covariance analysis and survival analysis methods, riders’behavior characteristics, red-light running rates and waiting endurance times were analyzed with and without traffic wardens. According to the comparison, the control effect of traffic wardens on red-light crossing behavior was evaluated. The results show that traffic wardens have a significantly impact on red-light crossing behavior. Red-light crossing rates of riders and pedestrians with traffic wardens are lower than those without traffic wardens. Waiting endurance times of riders and pedestrians with traffic wardens are longer than those without traffic wardens. Traffic wardens have good control effects on the straight-arrival groups, but no significant effects on groups coming from the left and right sides.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络