基于交通冲突极值统计的安全分析模型研究

发布时间：2018-06-06 23:24

本文选题：交通冲突 + 极值理论　；参考：《哈尔滨工业大学》2014年博士论文

【摘要】：道路交通安全问题一直是影响世界可持续发展的重要问题之一，每年都会导致大量的人员伤亡和巨额的财产损失。现阶段为改善道路交通安全状况所进行的研究与实践主要是以交通事故数据的统计和分析为基础。然而，由于交通事故发生的随机性和稀少性、交通事故记录的不完备性以及交通事故统计分析的集计性和事后性等缺点，探索基于非事故数据的交通安全分析方法也一直是广大研究人员和工程实践人员的努力方向。其中，交通冲突技术是目前发展较快但在理论体系和应用方法方面仍相对薄弱的一个研究方向，具有较大的开发空间和广阔的应用前景。通过交通冲突技术与极值理论的结合，本文研究了基于交通冲突极值统计的交通安全分析模型的构建、解析和扩展应用。首先从交通冲突技术和极值理论两个方面对国内外研究现状进行了综述，重点分析了交通冲突技术研究的相关文献，总结了交通冲突界定、交通冲突数据采集和交通冲突预测有效性等方面的研究成果，分析了交通冲突技术的优势和存在的主要问题，，提出了未来可能的研究方向。以广东省京珠高速公路粤北段、开阳高速公路和粤赣高速公路上的交通运行状况视频和交通事故记录为数据基础，分析了三条高速公路上的交通流量特征和车道变换行为特性，探讨了车道变换过程中后侵入时间的测定方法，并初步分析了车道变换冲突与交通事故之间的关系。根据交通冲突和极值理论都是以较为常见事件的统计分布特性外推极为少见的极值事件分布规律的共通本质，提出了将交通冲突技术与极值理论相结合进行交通安全分析的方法，分别构建了交通冲突区组极值模型和超阈值极值模型，并基于两种极值模型的参数标定结果提出了估计事故数和事故重现水平两个安全分析指标；进而，以模型标定和安全分析指标计算结果为基础，对区组极值模型、超阈值极值模型和基于事故数据的回归分析模型进行了对比分析。以消除超阈值极值模型中阈值选取的主观性和不确定性为出发点，构建了移位Gamma-GPD(Generalized Pareto Distribution)模型，探讨了模型参数的标定方法，并对模型所蕴含的安全信息进行了进一步的解析。本文将区组极值理论和超阈值理论分别应用到交通冲突极值统计中，并对两种极值理论在交通安全分析方面的应用进行了对比分析，结果表明在观测时间较短或样本量较少的情况下，超阈值极值模型在数据利用效率、模型预测可靠性、模型预测精度等方面要明显优于区组极值模型。结合极值理论中常用的“重现水平”概念，探讨性地提出了“事故重现水平”安全分析指标，并通过将超阈值极值模型与传统的基于事故数据的负二项回归模型的对比，说明了该指标的可靠性以及其与交通事故之间存在较强的相关性，进而也从侧面进一步证实了交通冲突技术的有效性。为了消除超阈值极值模型中阈值选取的主观性并减少不确定性，建立并标定了移位Gamma-GPD模型，该模型同时也实现了交通事件安全连续体的严重程度分级模型的参数化表达，有助于推动安全连续体模型应用向定量化方向发展以及交通安全分析向非事故数据方向扩展。研究结果一方面说明了极值统计在交通安全研究领域应用的可行性和优势，进一步验证了交通冲突技术的预测有效性，为道路交通安全分析与评价研究提供了新的思路和方法；另一方面，所构建的事故重现水平等安全分析指标也可以直接应用到工程实践中，为工程实践人员进行道路交通安全分析与评价提供一定的借鉴和参考。
[Abstract]:The problem of road traffic safety has been one of the most important issues affecting the sustainable development of the world. It will lead to a large number of casualties and huge loss of property every year. At this stage, the research and practice of improving road traffic safety mainly mainly based on the analysis and analysis of traffic accident data. However, traffic accidents are due to traffic accidents. The randomness and sparsity of the occurrence, the incompleteness of the traffic accident records, the collection and the hindsight of the traffic accident statistical analysis, and the exploration of the traffic safety analysis method based on the non accident data have always been the efforts of the researchers and engineering practitioners. A relatively weak research direction in theory system and application method has larger development space and broad application prospects.
Through the combination of traffic conflict technology and extreme value theory, this paper studies the construction of traffic safety analysis model based on the extreme value statistics of traffic conflict, and analyzes and extends its application.
From two aspects of traffic conflict technology and extreme value theory, the current research status at home and abroad is reviewed, the related literature of traffic conflict technology research is analyzed, and the research results of traffic conflict definition, traffic conflict data collection and traffic conflict prediction effectiveness are summarized, and the advantages and storage of traffic conflict technology are analyzed. The main problems in the future are put forward.
Based on the traffic operation status video and traffic accident records on the northern section of Guangdong Jingzhu Expressway, the Kaiyang Expressway and the Guangdong Jiangxi expressway, the traffic flow characteristics and the lane changing behavior characteristics on the three freeways are analyzed. The method for determining the post invasion time in the lane transformation process is discussed, and the preliminary classification is also discussed. The relationship between lane change conflict and traffic accidents is analyzed.
According to the common nature of the traffic conflict and extreme value theory, which are rarely extrapolated from the statistical distribution characteristics of more common events, the traffic conflict technology and extreme value theory are combined to carry out the traffic safety analysis, and the extreme value model and the extreme value model of the traffic conflict zone group are constructed. And based on the parameters calibration results of two extreme value models, two safety analysis indexes are proposed to estimate the number of accidents and the level of accident recurrence. Then, based on the model calibration and the results of the safety analysis index, the extreme value model of the region group, the ultra threshold extreme value model and the regression analysis model based on the accident number are compared and analyzed.
In order to eliminate the subjectivity and uncertainty of the threshold selection in the ultra threshold extreme value model, the shift Gamma-GPD (Generalized Pareto Distribution) model is constructed, and the calibration method of the model parameters is discussed, and the security information contained in the model is further analyzed.
In this paper, the extreme value theory and the super threshold theory are applied to the extreme value statistics of traffic conflict respectively, and the application of the two extreme value theories in traffic safety analysis is compared. The results show that the ultra threshold extreme value model is used in the data utilization efficiency and the model is reliable when the observation time is short or the sample size is less. The accuracy of model prediction is obviously superior to that of the area group extremum model. Combined with the concept of "recurrence level" commonly used in extreme value theory, the safety analysis index of "accident recurrence level" is proposed, and the index is illustrated by comparing the ultra threshold extreme value model with the traditional negative two term regression model based on the accident data. There is a strong correlation between reliability and traffic accidents, and further confirms the effectiveness of traffic conflict technology. In order to eliminate the subjectivity of threshold selection and reduce uncertainty in the ultra threshold extreme value model, the shift Gamma-GPD model is established and calibrated. The model also realizes traffic incident safety at the same time. The parameterized expression of the seriousness classification model of continuum helps to promote the development of the application of the security continuum model to the quantitative direction and to expand the traffic safety analysis to the non accident data.
On the one hand, the results show the feasibility and advantages of extreme statistics in the field of traffic safety research, further verify the effectiveness of the traffic conflict technology, and provide new ideas and methods for the road traffic safety analysis and evaluation research. On the other hand, the safety analysis indexes such as the level of accident recurrence are also available. It can be used directly in engineering practice to provide reference and reference for engineering practice personnel to carry out road traffic safety analysis and evaluation.
【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：U492.8;U491

【参考文献】