郝正博

入学时间：2018级

答辩时间：2024年

论文题目：考虑通行优先的城市道路应急交通路径优化方法研究

中文摘要

摘要

城市应急车辆路径优化是一个长期存在的社会热点、难点和痛点问题，同样也是典型的应用型学理问题。应急车辆（如消防救援车辆、医疗救护车辆、警用特勤车辆、工程抢险车辆和国防保障车辆等执行救援任务的专用车辆）在拯救人民生命和减少财产损失方面发挥着重要作用。与社会车辆不同，应急车辆稍有延误便可能造成严重后果。由于城市道路交通系统的高度复杂性和动态性等特征，应急车辆救援的效率性、安全性、可靠性和低社会成本性难以得到保障。

因此，秉承问题导向、需求牵引的研究原则，本文从提高应急车辆救援效率、保障其安全可靠和降低因其优先通行所造成的社会车辆负面成本等实际工作需求出发，依托车路协同环境下大数据、车联网等新兴技术快速发展优势，利用常发性较高的我国消防救援事件及其多元数据构建实证分析场景，以我国应急车辆所享有的法定优先通行权为核心切入点，探究考虑通行优先的城市道路应急交通路径优化方法。针对以往研究中偏向依赖模型或强数理假定、时空优先策略缺乏协同、社会车辆负面成本考虑较少等不足，本文开展如下四个部分的研究：

1. 基于多元数据的应急车辆救援通行特征解析。针对以往研究中偏向依赖模型或强数理假定等不足，为减少理论与应用之间的差异，增强研究成果的可应用性，该部分首先对我国应急车辆历史轨迹数据等相关多元数据进行融合分析，并提出考虑多特征约束的两阶段行程识别方法，用以构建其救援通行全过程。其后，从应急救援需求特征分析、应急救援供给特征分析、应急救援供需匹配分析、行驶速度影响因素分析、现有被动优先效果分析、主动通行优先效果模拟等多个方面解析其救援通行特征。通过特征解析，可获得大部分救援车队规模为1-3辆；92.1%的救援行程距离分布在0-4公里区间内；车道数量、路段限速等因素在99%的置信水平下显著影响应急车辆速度等诸多实用性结论。该成果有助于为后续章节中应急车辆路径规划及沿途优先方案提供研究场景与真实数据支撑。

2. 考虑时空协同优先的应急车辆路径优化方法。针对以往研究中时空优先策略缺乏协同等不足，该部分立足于为应急车辆提供无条件优先通行策略的核心思路，即为应急车辆赋予最高等级优先权，并考虑其与社会车辆在优先通行政策、路径优化目标等多方面的特征差异，提出考虑时间优先策略与空间优先策略相互协同的两阶段应急车辆路径动态优化方法。第一阶段，将用于求解静态K最短路径的Yen算法与路网时空动态性等特征相结合，增加基于层次聚类算法的路径安全风险约束与基于偏斜度指数等综合评价指标的可靠性约束，以获取备选路径集合。第二阶段，基于车道数、路段限速和拥堵延迟指数等静、动态属性特征，获得在时空协同优先条件下的最佳路径。实证研究显示，与其历史轨迹数据相比，该方法能够缩短救援时间约67.8%，并提高安全性与可靠性。此外，各备选路径在实施超过路段限速20%的超速策略后，原有次优路径的救援时间低于了原有最优路径，并随着超速策略实施强度增加而进一步扩大优势。因此，适当实施超速驾驶策略具有进一步减少救援时间，并改变最佳路径选择结果的潜能。

3. 考虑不同优先等级的应急车辆路径优化方法。针对以往研究中社会车辆负面成本考虑较少等不足，该部分立足于为应急车辆提供有条件优先通行策略的核心思路，即在为应急车辆规划路径过程中，不仅满足其救援需求，同时尽可能降低因其优先通行造成的社会车辆延误。首先，考虑应急事件自身特征、发展态势和交通及救援供给能力，构建基于多级灰度评价法的路径级优先通行等级评价模型。其后，构建以最大化应急车辆救援时间节约值和最小化社会车辆行程时间总延误值为目标的多目标优化模型，并综合考虑时变道路运行状态等参数，估算不同优先通行等级下的策略实施效果，系统优化救援路径及优先策略方案，以进一步实现应急车辆快速救援（个体最优）与社会车辆整体出行（系统最优）之间的最佳协同。实证研究显示，与推荐最佳路径相比，该方法能够缩短救援时间约50.5%。同时基于帕累托最优解集合能够获得在该时间节约比例下，社会车辆所承担的最小负面成本及其对应优先策略方案。此外，当多个（以3个为例）应急车辆形成连续车队进行救援时，能够进一步压缩单位车辆救援时间约1.7%。

4. 应急车辆救援主动交通管理即服务系统应用。针对现有应急救援实际工作中缺乏和其他部门在系统层面的协同联动等不足，该部分在总结前述章节理论模型构建和技术方法成果的基础上，面向未来城市交通高度数字化、信息化、网联化、智能化发展趋势，借鉴以往已有系统理念（如出行即服务、交通管理即服务），讨论城市应急车辆救援主动交通管理即服务系统应用，包括系统基本构思、系统需求分析、系统框架设计等内容，为未来打通业务壁垒提供系统平台思路。

本学位论文在丰富城市道路应急交通路径优化方法体系方面主要具有三点创新：1. 基于我国应急车辆多元数据赋能其救援水平提升，包括救援通行特征解析等；2. 在路径寻优算法中融入应急车辆多维度特征差异，包括综合优化救援服务水平等；3. 实施全过程主动协同型应急车辆优先通行策略集，包括构建分级分类的优先通行等级等。本文研究成果对于提升我国城市应急车辆救援出行服务水平、降低因应急车辆优先通行造成的社会车辆负面成本和提高我国城市部门联动应急处置管理水平具有较强的理论意义和应用价值。

关键词：智能交通系统，城市应急车辆，通行优先，路径优化，多元数据

英文摘要

ABSTRACT

Route optimization for urban emergency vehicles is a long-standing hotspot, challenging, and painful problem in society, and it is also a typical applied theoretical problem. Emergency vehicles (e.g., fire rescue vehicles, emergency medical vehicles, police special duty vehicles, engineering rescue vehicles, and national defense vehicles) are essential in saving people's lives and reducing property losses. Unlike social vehicles, the slight delay of emergency vehicles may cause serious consequences. Owing to the highly complex and dynamic characteristics of urban road traffic systems, it is difficult to guarantee the efficiency, safety, reliability, and low social vehicles impact of emergency vehicle rescue.

Therefore, adhering to the research principles of problem-oriented and demand-driven, this dissertation starts from the practical demand of improving the rescue efficiency of urban emergency vehicles, guaranteeing their safety and reliability, and reducing the negative social vehicles impact caused by priority strategies of emergency vehicles. Relying on the advantages of the rapid development of emerging technologies such as big data and Internet of Vehicles in vehicle infrastructure cooperative system, and utilizing the multivariate data of China's fire trucks to construct the empirical analysis scenario, and taking the statutory priority strategies of emergency vehicles as the core entry point, this dissertation explores the optimization methods for urban emergency vehicles routes considering traffic priority strategies. Aiming at the shortcomings of previous studies, such as relying on models or strong assumptions, lack of synergy between spatial and temporal priority strategies, and less consideration of the negative costs of social vehicles, this paper carries out the following four parts of the research content:

Part 1. Rescue passage characteristics analysis of emergency vehicle based on multivariate data.Aiming at decreasing biased reliance on models or strong assumptions, to reduce the discrepancy between theory and application and enhance the applicability of research results, Part 1 initially fuses and analyzes China's emergency vehicles multivariate data (e.g., historical trajectory data) and proposes a two-phase trip identification method considering multi-feature constraints, which can be used to construct the whole process of emergency vehicle rescue. Then, Part 1 analyzes the rescue passage characteristics of China’s emergency vehicles from several aspects, such as analysis of emergency rescue demand characteristics, analysis of emergency rescue supply characteristics, analysis of emergency rescue supply-demand matching, analysis of driving speed influencing factors, analysis of the existing passive priority strategies effect, and simulation of the active priority strategies effect, and so on. By analyzing the characteristics, we can obtain the practical findings that most of the rescue fleet size ranges from 1-3 vehicles; 92.1% of the rescue travel distance is within a 0-4 km range; and the number of lanes, speed limit, and other factors significantly affect the speed of emergency vehicles at 99% confidence level. These conclusions of Part 1 provide research scenarios and real-world data support for emergency vehicle route optimization and prioritization schemes along the route in subsequent chapters.

Part 2. Route optimization method for emergency vehicle considering spatio-temporal collaborative priority.Aiming at the shortcomings, such as the lack of collaboration between temporal and spatial priority strategies in previous studies, Part 2 is based on the core idea of providing unconditional priority strategies for emergency vehicles, that is, gives the highest level of priority for emergency vehicles, and taking into account the differences (such as priority policies, route optimization objectives) between their characteristics and those of the social vehicles. Part 2 proposes a two-phase dynamic route optimization method for emergency vehicles, which takes into account the collaboration of temporal priority and spatial priority strategies. In the first stage, the Yen algorithm for solving the static K-shortest path is combined with the dynamic characteristics of the road network, and the route safety constraints based on the hierarchical clustering algorithm and reliability constraints based on the skewness index and other comprehensive evaluation indexes are added in order to obtain the set of alternative routes. In the second stage, the optimal route under spatio-temporal collaborative priority conditions is obtained based on dynamic and static attribute features such as the number of lanes, the speed limit, and the congestion delay index. Empirical analysis results show that the proposed method can reduce the time by about 67.8% and improve safety and reliability compared to its historical trajectory data. Moreover, the time of the original suboptimal route becomes lower than that of the original optimal route after the implementation of the speeding strategy that exceeds the speed limit by 20% for each alternative route, and the advantage further expands with the increase of the implementation intensity of the speeding strategy. Therefore, proper implementation of the speeding strategy has the potential to reduce the time further and change the outcome of the optimal route selection.

Part 3. Route optimization method for emergency vehicle considering different priority levels.Aiming at the shortcomings, such as less consideration of the negative impacts of social vehicles in previous studies, Part 3 is based on the core idea of providing the conditional priority strategy for emergency vehicles, that is, in the process of planning routes for emergency vehicles, not only to meet their rescue demand but also to minimize the delays for social vehicles due to priority strategies implementation as much as possible. Initially, considering the characteristics of the emergency event, the development trend, and the traffic and rescue supply capacity, Part 3 constructs a route-level priority level evaluation model based on the multi-step associated gray evaluation model. Subsequently, a multi-objective optimization model is constructed to maximize the rescue time saving value of emergency vehicles and minimize the total delay value of social vehicle travel time, and the time-varying road operation state and other parameters are considered comprehensively to estimate the strategy implementation effect under different priority levels, and to systematically optimize the rescue routes and the prioritization schemes, to realize the optimal collaboration between the rapid rescue of emergency vehicles (user optimum) and the overall travel time of social vehicles (system optimum). The empirical analysis results indicate that the proposed method can shorten the rescue time by about 50.5% compared with the recommended optimal route. Meantime, based on the set of Pareto-optimal solutions, we can obtain the minimum negative cost borne by the social vehicles and the corresponding prioritization schemes under a given time-saving ratio. In addition, when multiple emergency vehicles (three, for example) form a continuous fleet for rescue driving, the rescue time per vehicle can be further compressed by approximately 1.7%.

Part 4. Application of the Emergency Vehicle Rescue Active Traffic Management as a Service System.Aiming at the shortcomings, such as lack of collaboration with other departments in the system level in the existing emergency rescue practice, Part 4 builds on the theoretical model construction and technical method research outcomes of the preceding chapters and faces the future trends of highly digitalized, informatized, networked, and intelligent urban traffic development, and draws on previous concepts of systems (e.g. Mobility as a Service, Traffic Management as a Service) to discuss the application of the Urban Emergency Vehicle Rescue Active Traffic Management as a Service System. This includes the basic concept of the system, system requirements analysis, and system framework design, among other aspects, for the future to open up the collaboration barriers to provide a system platform ideas.

This dissertation primarily contributes three innovations to the enhancement of urban road emergency traffic route optimization method system: 1. Leveraging multivariate data of China's emergency vehicles to enhance their rescue performance, including the analysis of rescue passage characteristics; 2. Integrating multi-dimensional characteristics of emergency vehicles into route optimization algorithms, including incorporating comprehensive optimization of rescue service levels; and 3. Implementing a full-process proactive collaboration of priority strategies for emergency vehicles, including the establishment of hierarchical classifications for priority passage levels. The research findings of this dissertation are of significant theoretical and practical value in improving the performance of rescue services for emergency vehicles in China's cities, reducing the social vehicles impact caused by the priority strategies and enhancing the management and collaboration among departments in China's cities to cope with emergencies.

Key Words:Intelligent Transportation System, Urban Emergency Vehicle, Traffic Priority,Route Optimization,Multimodal Data

孙伟

入学时间：2018级

答辩时间：2025年

论文题目：信号控制交叉口交通安全综合评价研究

中文摘要

摘要

最新统计显示，全球每年约110万人死于道路交通事故，其中约40%的事故集中在交叉口区域，凸显了交叉口在道路交通安全体系中的关键地位。交叉口安全领域已取得诸多成果，但受限于高质量交通事故数据的稀缺性，以往的信号控制交叉口安全评价方法多基于单一因果链的分析框架。此外，随着智能技术的快速发展以及城市精细精准治理的推进，信号控制交叉口交通安全评价的研究基础和高度结构化的时空要素数据与高精度交通运行数据的深度融合，为开展精细化、精准化、系统性、本质性、多维度的交通安全分析提供了充分条件。鉴于此，本文以信号控制交叉口交通安全全过程分析为基础，以“相对稳定的时空要素评价、动态的交通冲突评价以及动静关联的暴露风险评价”为研究主线，构建面向信号控制交叉口的交通安全综合评价方法。

全球每年道路交通事故导致约110万人死亡，其中交叉口事故占比高达40%，凸显其安全问题的严峻性。传统安全评价方法受数据限制，多基于单一因果分析，难以揭示多因素耦合风险的内在机制。随着多源数据与智能技术的发展，深度融合时空要素与高精度交通运行为系统性安全评价提供了可能。为此，本文以信号控制交叉口为研究对象，构建集时空要素评价、交通冲突评价与暴露风险评价于一体的交通安全综合评价体系，旨在实现更精细化、多维度的安全分析，为交叉口安全治理提供理论依据与实践支撑。

本论文首先基于各类交通事故视频数据，从时空维度解析了事故过程与成因，将事故类型划分为常规、非常规轨迹、非常规时间和“双非”交通事故四类，并构建了事故与交叉口时空要素的内在关联。结合事故致因理论、驾驶员感知决策理论及交通风险4E过程理论，本文将信号控制交叉口“交通安全全过程”（规划设计→建设管理→投入使用→产生风险→事故发生）划分为规划设计建设、时空要素优化、出行主体使用和交通风险过程四个阶段，并分别构建“合规度”、“完整度”、“暴露度”、“冲突度”指标评价各阶段安全水平。

其次，针对现有交叉口时空要素评价通常依赖主观“专家法”且缺乏客观量化依据的不足，本文系统地构建了基于现行规范标准的“合规度”评价体系，对时空要素进行法定“基准安全”量化。该体系以我国交叉口交通安全相关的7类70项规范条文为基础，筛选刚性与规范性条款构建指标库，并采用直接判定、阈值比较、差异分析和条件匹配四种量化方法计算合规度。相比以往方法，合规度评价体现法律认可的最低安全基准，可按地域灵活引入地方规范，并随着标准更新自动迭代。在现行规范对交叉口时空要素约束仍显不足的背景下，本文借鉴“完整街道”、“完整社区”等理念，首创性地提出交叉口安全“完整度”概念，衡量全时空要素对各类出行者安全需求的满足程度。以系统构成完整、时空要素完整、功能与性能完整三个维度构建指标体系，经主成分分析，从32项要素中遴选出管理设施、绿化高度、相交路数、中央驻足区、横断面形式、路幅宽度、积水(雪)状况和公交站点8个关键指标，并用优序图法赋权。针对量化标准不统一，采用分类分项赋分法获取单项得分并加权求取完整度。该指标与“合规度”互为补充，共同形成时空要素综合评价。为便于工程应用，开发了合规-完整度智能评估系统（含框架、数据库与原型系统），并在两个交叉口改造前后实测验证，评价结果与问卷反馈高度一致，验证了方法的有效性与可操作性。

再次，针对现有时空要素致因评价与事故/冲突结果评价间缺乏有机联系、难以深入揭示安全本质的问题，本文以交叉口交通事故发生的必要条件“暴露”为切入点，创新性地提出微观层面的“交叉口暴露”理论。区别于宏观层面的暴露概念，交叉口暴露理论解析了交叉口范围内的时空要素引导出行行为进而影响动态风险的关联关系，并构建“暴露度”和“冲突度”指标评估动态风险过程。其中，暴露度指标表征各类出行者暴露于特定交叉口时空要素组合下的风险水平，从空间暴露（暴露面积比、隔离设施比）、时间暴露（信号相位、暴露时长、绿间隔）及行为暴露（行为系数）三个层面构建了包含六项分指标的暴露度评价体系，弥补了现有方法难以反映时空要素对行为及安全结果影响的缺陷。冲突度评价方法中，基于交通事故致因链将交通冲突分为常规与非常规两类，在常规轨迹冲突中，结合四种典型事故过程及交叉口时空特征，细化为机-机、机-非、非-非、机-人、非-人5大71小类，并量化各小类冲突严重性，综合信号配时与流量饱和度等因素建立冲突度模型，实现动态风险度量。以同一交叉口在不同交通组织方案下的实测数据验证暴露度与冲突度指标，与传统TTC指标结果高度一致，证实暴露度与冲突度评价指标的科学性与有效性。

最后，在分析现有综合评价方法适用性的基础上，围绕多阶段、多维度及多指标重构的需求，提出兼具非结构数据量化、时空风险耦合、各阶段独立评价的信号控制交叉口交通安全综合评价体系。通过层次分析法将合规度(Compliance)、完整度(Completeness)、暴露度(Exposure)和冲突度(Conflict)四项核心指标加权融合，选取各指标英文首字母形成CCEC评价方法。以高德数据为基础，选取8个城市的21处信号控制交叉口计算其CCEC指，并按8:2划分为训练集与测试集，采用多层感知机（MLP）构建评价指标与事故频次的映射模型；针对事件样本稀缺，引入SMOTE过采样和L1正则化提升模型可靠性，并进行指标贡献度及事故概率敏感性分析。结果表明，CCEC评价结果与事故量显著相关，能有效表征交叉口的安全水平，证明了CCEC综合评价方法的有效性。

综上所述，本文以“相对稳定的时空要素评价、动态的交通冲突评价以及动静关联的暴露风险评价”为研究主线，从信号控制交叉口交通安全本质出发，系统梳理了其交通安全全过程，提出了合规度、完整度、暴露度、冲突度四阶段评价指标，构建CCEC综合评价体系，为信号控制交叉口交通安全综合评价提供了理论支撑与方法工具，具有重要的应用价值与理论意义。

关键词：交叉口，交通安全过程，评价方法，合规性，完整性，交通暴露

英文摘要

ABSTRACT

The latest statistics show that about 1.1 million people die in road traffic accidents worldwide each year,accounting for one third of all types of accidental deaths.In China, the average annual number of road accidents in the past five years has exceeded 200,000, resulting in more than 60,000 deaths. Among them, approximately 40% of the accidents were concentrated in intersection areas, highlighting the crucial position of intersections in the road traffic safety system. Intersection safety issues, as a long-standing research focus, have achieved many results. However, due to the scarcity of high-quality traffic accident data, previous safety evaluation methods for signal-controlled intersections were mostly based on a single causal chain analysis framework. By constructing correlation models between risk events such as accidents, conflicts, and defects in spatio-temporal elements and driving behaviors or traffic flow characteristics, Then quantify the safety indicators and conduct evaluations. Given the highly complex and multi-factor coupled driving nature of traffic safety, these methods have limitations in terms of systematicness and essential insufficiency in revealing the internal mechanism and dynamic evolution of safety risks at signal-controlled intersections. Furthermore, with the increasing demands for urban refinement and precise traffic governance, establishing a scientific, systematic and operational safety evaluation system for signal-controlled intersections has become a key to enhancing facility functions and ensuring travel safety. Thus, conducting this research in depth not only holds significant theoretical value but also provides direct support for engineering practice.

In recent years, the rapid development of multi-source data perception and intelligent technologies in road traffic has profoundly reshaped the research foundation of traffic safety evaluation at signal-controlled intersections.The deep integration of highly structured spatio-temporal element data and high-precision traffic operation data provides sufficient conditions for conducting refined,precise,systematic,essential and multi-dimensional traffic safety analysis.In view of this, this paper takes the full-process analysis of traffic safety at signal-controlled intersections as the basis, and takes "relatively stable spatio-temporal element evaluation, dynamic traffic conflict evaluation, and exposure risk evaluation" as the research main line to construct a comprehensive traffic safety evaluation method for signal-controlled intersections.

This paper first analyzes the accident process and causes from the spatiotemporal dimension based on various types of traffic accident video data, classifies the accident types into four categories: conventional, unconventional trajectory, unconventional time, and "dual-non-" traffic accidents, and constructs the intrinsic correlation between accidents and the spatiotemporal elements of intersections. Combining the theory of accident causes,the theory of driver perception and decision-making,and the 4E process theory of traffic risks,his paper divides the"entire process of traffic safety"at signal-controlled intersections(planning and design→construction and management→commissioning→risk generation→accident occurrence)into four stages:planning, design and construction, optimization of spatio-temporal elements, use by travel entities, and the traffic risk process.And respectively construct the indicators of "compliance degree", "completeness degree", "exposure degree" and "conflict degree" to evaluate the security level at each stage.

Secondly, in view of the deficiency that the evaluation of spatio-temporal elements at existing intersections usually relies on subjective"expert methods"and lacks objective quantitative basis,this paper systematically constructs a "compliance" evaluation system based on current norms and standards, and quantifies spatio-temporal elements into legal "benchmark safety". This system is based on 70 normative provisions of 7 categories related to intersection traffic safety.It screens rigid and normative provisions to build an index library and uses four quantitative methods: direct determination, threshold comparison, difference analysis and condition matching to calculate the compliance degree. Compared with previous methods, compliance evaluation reflects the minimum safety benchmark recognized by law, can flexibly introduce local norms by region, and automatically iterate along with the update of standards. Under the background that the current norms still impose insufficient constraints on the spatio-temporal elements of intersections, this paper draws on concepts such as "complete streets" and "complete communities", and innovatively proposes the concept of "completeness" of intersection safety to measure the degree to which all spatio-temporal elements meet the safety needs of various travelers. An index system was constructed from three dimensions: complete system composition, complete spatio-temporal elements, and complete function and performance. Through principal component analysis, eight key indicators, namely management facilities, green height, number of intersecting roads, central stop area, cross-sectional form, road width, water (snow) accumulation conditions, and bus stops, were selected from 32 elements and weighted using the order graph method. In view of the non-uniform quantitative standards, the classification and item-by-item scoring method is adopted to obtain the individual item scores and then weighted to calculate the completeness. This indicator and "compliance" complement each other, jointly forming a comprehensive evaluation of spatio-temporal elements.To facilitate engineering applications,a compliance-completeness intelligent assessment system(including framework, database and prototype system)was developed and verified by actual measurement before and after the renovation of two intersections.The evaluation results were highly consistent with the questionnaire feedback,verifying the effectiveness and operability of the method.

Thirdly,in view of the lack of organic connection between the evaluation of the causes of existing spatio-temporal elements and the evaluation of accident/conflict results, and the difficulty in deeply revealing the essence of safety, this paper takes the necessary condition "exposure" for the occurrence of traffic accidents at intersections as the entry point and innovatively proposes the "intersection exposure" theory at the micro level. Different from the concept of macro exposure,the intersection exposure theory analyzes the correlation between spatio-temporal elements within the intersection range guiding travel behavior and thereby influencing dynamic risks, and constructs "exposure degree" and "conflict degree" indicators to evaluate the dynamic risk process.The exposure index characterizes the risk level of various travelers exposed to the combination of spatio-temporal elements at a specific intersection. An exposure evaluation system containing six sub-indicators was constructed from three aspects: spatial exposure (exposure area ratio, isolation facility ratio), temporal exposure (signal phase, exposure duration, green interval), and behavioral exposure (behavioral coefficient). It makes up for the defect that the existing methods are difficult to reflect the influence of spatio-temporal elements on behavior and safety outcomes. In the conflict degree evaluation method, traffic conflicts are classified into two categories based on the cause chain of traffic accidents: conventional and unconventional. In the conventional trajectory conflicts, combined with four typical accident processes and the spatio-temporal characteristics of intersections, they are refined into five major and 71 minor categories: machine-machine, machine-non-machine, non-non-machine, machine-human, and non-human, and the severity of conflicts in each minor category is quantified. A conflict degree model is established by integrating factors such as signal timing and flow saturation to achieve dynamic risk measurement. The exposure and conflict degree indicators were verified by the measured data of the same intersection under different traffic organization schemes, which were highly consistent with the results of the traditional TTC indicators, confirming the scientificity and effectiveness of the evaluation indicators of exposure and conflict degree.

Finally, based on the analysis of the applicability of the existing comprehensive evaluation methods, and in response to the requirements of multi-stage, multi-dimensional and multi-index reconstruction, a comprehensive evaluation system for traffic safety at signal-controlled intersections that integrates unstructured data quantification, spatio-temporal risk coupling, and independent evaluation at each stage is proposed. The four core indicators of Compliance, Completeness, Exposure and Conflict are weighted and fused through the Analytic Hierarchy Process (AHP), and the first letters of each indicator in English are selected to form the CCEC evaluation method. Based on Gaode data, 21 signal-controlled intersections in 8 cities were selected to calculate their CCEC indicators, and they were divided into the training set and the test set in an 8:2 ratio. A mapping model between evaluation indicators and accident frequencies was constructed using a multi-layer perceptron (MLP). In view of the scarcity of event samples, SMOTE oversampling and L1 regularization are introduced to enhance the reliability of the model, and the contribution of indicators and the sensitivity of accident probability are analyzed. The results show that the CCEC evaluation results are significantly correlated with the number of accidents and can effectively characterize the safety level of intersections, which proves the effectiveness of the CCEC comprehensive evaluation method.

In summary, this paper takes "relatively stable spatio-temporal element evaluation, dynamic traffic conflict evaluation, and exposure risk evaluation related to movement and stillness" as the research main line. Starting from the essence of traffic safety at signal-controlled intersections, it systematically sorts out the entire process of traffic safety, proposes four-stage evaluation indicators of compliance, completeness, exposure, and conflict, and constructs a CCEC comprehensive evaluation system. It provides theoretical support and methodological tools for the comprehensive evaluation of traffic safety at signal-controlled intersections, and has significant application value and theoretical significance..

Key Words:Intersection, traffic safety process, evaluation method, compliance, integrity, traffic exposuremethod