Accident; analysis and prevention最新文献_第10页

Evaluating causes of animal-vehicle collisions through the lens of driver behavior 通过驾驶员行为来评估动物与车辆碰撞的原因

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-21 DOI: 10.1016/j.aap.2025.108314

Carson J. Pakula , Olin E. Rhodes Jr. , Travis L. DeVault

Animal-vehicle collisions (AVCs) are ubiquitous in developed regions of the world and pose risks to both wildlife and humans. In the United States, collisions with deer (Odocoileus spp.) cause billions of dollars in economic losses and thousands of human injuries annually. The current AVC literature has largely focused on factors unrelated to driver behavior including AVC hotspots, wildlife movement, and damages caused by AVCs. However, despite being a component in every AVC, few studies have investigated driver behavior during animal-vehicle interactions. Here, we systematically reviewed literature databases to identify factors influencing driver behavior during these interactions and to highlight apparent gaps in the literature. We found that vehicle speed, road attributes, environmental conditions, and vehicle types show inconsistent associations with AVCs and the mechanisms by which they influence driver behavior is not well understood. Many studies focused on mitigation methods to influence driver behavior, including various warning signs; however, the effectiveness of these systems varies considerably. Other topics including wildlife attributes, roadway illumination, and inherent driver attributes directly influence driver behavior, but are understudied. Most studies relied on seemingly logical explanations for results or associations between variables to identify these influences, but few studies directly tested how specific variables influenced driver behavior and detection ability of wildlife. Given that driver behavior influences every potential AVC, future research should directly investigate the behavioral and perceptual mechanisms behind driver detection of wildlife and other factors influencing overall driver behavior during wildlife-vehicle interactions.

动物与车辆碰撞（AVCs）在世界发达地区普遍存在，对野生动物和人类都构成威胁。在美国，与鹿（Odocoileus spp.）的碰撞每年造成数十亿美元的经济损失和数千人受伤。目前的AVC文献主要集中在与驾驶员行为无关的因素上，包括AVC热点、野生动物运动和AVC造成的损害。然而，尽管这是每个AVC的组成部分，但很少有研究调查动物与车辆相互作用时驾驶员的行为。在这里，我们系统地回顾了文献数据库，以确定在这些相互作用中影响驾驶员行为的因素，并突出了文献中的明显空白。我们发现，车速、道路属性、环境条件和车辆类型与AVCs的关联并不一致，而且它们影响驾驶员行为的机制也不清楚。许多研究侧重于影响驾驶员行为的缓解方法，包括各种警告标志；然而，这些系统的有效性差别很大。其他主题包括野生动物属性、道路照明和固有驾驶员属性直接影响驾驶员行为，但尚未得到充分研究。大多数研究依赖于对结果的看似合乎逻辑的解释或变量之间的关联来确定这些影响，但很少有研究直接测试特定变量如何影响驾驶员行为和野生动物的检测能力。鉴于驾驶员行为影响每一个潜在的AVC，未来的研究应直接研究驾驶员检测野生动物背后的行为和感知机制，以及野生动物与车辆交互过程中影响驾驶员整体行为的其他因素。

{"title":"Evaluating causes of animal-vehicle collisions through the lens of driver behavior","authors":"Carson J. Pakula , Olin E. Rhodes Jr. , Travis L. DeVault","doi":"10.1016/j.aap.2025.108314","DOIUrl":"10.1016/j.aap.2025.108314","url":null,"abstract":"<div><div>Animal-vehicle collisions (AVCs) are ubiquitous in developed regions of the world and pose risks to both wildlife and humans. In the United States, collisions with deer (<em>Odocoileus spp</em>.) cause billions of dollars in economic losses and thousands of human injuries annually. The current AVC literature has largely focused on factors unrelated to driver behavior including AVC hotspots, wildlife movement, and damages caused by AVCs. However, despite being a component in every AVC, few studies have investigated driver behavior during animal-vehicle interactions. Here, we systematically reviewed literature databases to identify factors influencing driver behavior during these interactions and to highlight apparent gaps in the literature. We found that vehicle speed, road attributes, environmental conditions, and vehicle types show inconsistent associations with AVCs and the mechanisms by which they influence driver behavior is not well understood. Many studies focused on mitigation methods to influence driver behavior, including various warning signs; however, the effectiveness of these systems varies considerably. Other topics including wildlife attributes, roadway illumination, and inherent driver attributes directly influence driver behavior, but are understudied. Most studies relied on seemingly logical explanations for results or associations between variables to identify these influences, but few studies directly tested how specific variables influenced driver behavior and detection ability of wildlife. Given that driver behavior influences every potential AVC, future research should directly investigate the behavioral and perceptual mechanisms behind driver detection of wildlife and other factors influencing overall driver behavior during wildlife-vehicle interactions.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108314"},"PeriodicalIF":6.2,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145570589","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The hazards of skipping meals and losing sleep for ride-hailing drivers—evidence based on quasi-experimental and experience sampling methodology 不吃饭和失眠对网约车司机的危害——基于准实验和经验抽样方法的证据。

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-19 DOI: 10.1016/j.aap.2025.108304

Jingming Feng, Jian Chen, Tao Su, Huijian Fu, Ling Tan

The rapid growth of ride-hailing services, driven by big data and platform algorithms, has intensified physiological stressors among drivers, particularly during peak mealtime periods, where hunger and sleep deprivation (“skipping meals and losing sleep”) have become prevalent occupational hazards. While prior research has separately documented the adverse effects of hunger and sleep loss on self-control, their synergistic impact on driving safety remains underexplored, especially in ecologically valid settings. Guided by the self-control strength model, this study investigated how hunger and sleep deprivation interact to influence dangerous driving behaviours among ride-hailing drivers, examining the mediating role of self-control resource depletion and the moderating effects of sleep deprivation and work identity perception (contractor vs. worker).

Using a mixed-methods approach, Study 1 (N = 70 students) employed a 2 × 2 mixed quasi-experiment to establish baseline effects, revealing that hunger directly depleted self-control resources (measured via Stroop task), with sleep deprivation exacerbating this relationship. Study 2 extended these findings to 153 active ride-hailing drivers using a 2 × 2 between-subjects quasi-experiment, confirming that hungry drivers exhibited greater self-control resource depletion than fed drivers, particularly under sleep deprivation. Study 3 utilized experience sampling with 76 drivers (651 data points) to test cross-level mechanisms, demonstrating that hunger indirectly increased dangerous driving behaviors through self-control depletion, with sleep deprivation and worker identity perception amplifying these effects.

These findings highlight the compounding risks of physiological stressors in ride-hailing work, emphasizing the need for interventions targeting both hunger management and sleep hygiene. The study contributes to accident prevention research by integrating ecological validity with theoretical rigor, offering actionable insights for drivers and platform operators to mitigate fatigue-related safety risks.

在大数据和平台算法的推动下，网约车服务的快速增长加剧了司机的生理压力，尤其是在用餐高峰期，饥饿和睡眠不足（“不吃饭、失眠”）已成为普遍的职业危害。虽然先前的研究分别记录了饥饿和睡眠不足对自我控制的不利影响，但它们对驾驶安全的协同影响仍未得到充分探讨，特别是在生态有效的环境中。在自我控制强度模型的指导下，本研究探讨了饥饿和睡眠剥夺如何相互作用影响网约车司机的危险驾驶行为，考察了自我控制资源枯竭的中介作用以及睡眠剥夺和工作认同感知（承包商与工人）的调节作用。采用混合方法，研究1 （N = 70名学生）采用2 × 2混合准实验来建立基线效应，揭示饥饿直接消耗自我控制资源（通过Stroop任务测量），睡眠剥夺加剧了这种关系。研究2将这些发现扩展到153名活跃的网约车司机，使用2x2受试者之间的准实验，证实饥饿的司机比饱食的司机表现出更大的自我控制资源消耗，尤其是在睡眠不足的情况下。研究3利用76名司机（651个数点）的经验抽样来检验跨水平机制，表明饥饿通过自我控制消耗间接增加危险驾驶行为，而睡眠剥夺和工人身份感知放大了这些影响。这些发现强调了网约车工作中生理压力源的复合风险，强调了针对饥饿管理和睡眠卫生的干预措施的必要性。该研究通过将生态有效性与理论严谨性相结合，为事故预防研究做出了贡献，为司机和平台运营商提供了可操作的见解，以减轻疲劳相关的安全风险。

{"title":"The hazards of skipping meals and losing sleep for ride-hailing drivers—evidence based on quasi-experimental and experience sampling methodology","authors":"Jingming Feng, Jian Chen, Tao Su, Huijian Fu, Ling Tan","doi":"10.1016/j.aap.2025.108304","DOIUrl":"10.1016/j.aap.2025.108304","url":null,"abstract":"<div><div>The rapid growth of ride-hailing services, driven by big data and platform algorithms, has intensified physiological stressors among drivers, particularly during peak mealtime periods, where hunger and sleep deprivation (“skipping meals and losing sleep”) have become prevalent occupational hazards. While prior research has separately documented the adverse effects of hunger and sleep loss on self-control, their synergistic impact on driving safety remains underexplored, especially in ecologically valid settings. Guided by the self-control strength model, this study investigated how hunger and sleep deprivation interact to influence dangerous driving behaviours among ride-hailing drivers, examining the mediating role of self-control resource depletion and the moderating effects of sleep deprivation and work identity perception (contractor vs. worker).</div><div>Using a mixed-methods approach, Study 1 (N = 70 students) employed a 2 × 2 mixed quasi-experiment to establish baseline effects, revealing that hunger directly depleted self-control resources (measured via Stroop task), with sleep deprivation exacerbating this relationship. Study 2 extended these findings to 153 active ride-hailing drivers using a 2 × 2 between-subjects quasi-experiment, confirming that hungry drivers exhibited greater self-control resource depletion than fed drivers, particularly under sleep deprivation. Study 3 utilized experience sampling with 76 drivers (651 data points) to test cross-level mechanisms, demonstrating that hunger indirectly increased dangerous driving behaviors through self-control depletion, with sleep deprivation and worker identity perception amplifying these effects.</div><div>These findings highlight the compounding risks of physiological stressors in ride-hailing work, emphasizing the need for interventions targeting both hunger management and sleep hygiene. The study contributes to accident prevention research by integrating ecological validity with theoretical rigor, offering actionable insights for drivers and platform operators to mitigate fatigue-related safety risks.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108304"},"PeriodicalIF":6.2,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145555914","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Profiling crash-associated factors and injury risk patterns among lost-in-thought (daydreaming) drivers: a combined cluster-sequence analysis approach 分析心不在焉（白日做梦）司机的撞车相关因素和伤害风险模式：一种组合聚类-序列分析方法。

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-19 DOI: 10.1016/j.aap.2025.108315

Panick Kalambay , Philip Balyagati , Angela Kitali , Emmanuel Kidando

Cognitive distraction, particularly in the form of being lost in thought or daydreaming, is a significant yet underexamined contributor to traffic crashes. This study investigates crash patterns and risk profiles associated with drivers who experienced lost-in-thought distraction at the time of a crash in Washington State. Using cluster correspondence analysis (CCA), distinct associations between crash characteristics and driver attributes were uncovered. In addition, process mining was employed to identify typical sequences of crash events. Three meaningful clusters emerged. Cluster 1 involved crashes on road segments with speed limits exceeding 40 mph, lacking traffic control, and often involving male drivers in clear weather. Cluster 2 was marked by crashes at signalized intersections under partly cloudy or foggy conditions, with a higher likelihood of injury. Cluster 3 reflected rainy or low-light crashes involving young drivers on curved, divided, high-speed roads. Across all clusters, frequent crash sequences included collisions with vehicles in transport, parked cars, and fixed objects. Cluster 1 and Cluster 2 stood out for their distinct contextual characteristics. Cluster 1 crashes often involved crossing the center line, suggestive of deep cognitive distraction. In Cluster 2, crashes frequently occurred after drivers stopped at flashing red lights or stop signs, then proceeded, indicating momentary lapses in attention. The study also highlights the limitations of current crash data and emphasizes the need for standardized reporting of distraction-related incidents. Findings support context-specific countermeasures, such as signal enhancements, curve warnings, and distraction-focused training for novice drivers, to address the multifaceted risks associated with lost-in-thought crashes.

认知分心，尤其是陷入思考或做白日梦的形式，是导致交通事故的一个重要因素，但尚未得到充分研究。这项研究调查了在华盛顿州发生的撞车事故中，经历过分心的司机的撞车模式和风险状况。使用聚类对应分析（CCA），揭示了碰撞特征和驾驶员属性之间的明显关联。此外，还采用流程挖掘来识别典型的崩溃事件序列。三个有意义的集群出现了。第一类事故发生在限速超过每小时40英里的路段，缺乏交通控制，通常是晴天的男性司机。第二组的标志是在部分多云或有雾的天气下，在有信号的十字路口发生撞车事故，造成伤害的可能性更高。集群3反映了雨天或弱光下年轻司机在弯曲、分隔的高速公路上发生的撞车事故。在所有集群中，频繁的碰撞序列包括与运输中的车辆、停放的汽车和固定物体的碰撞。集群1和集群2因其独特的上下文特征而脱颖而出。第一类撞车事故通常涉及越过中线，暗示着严重的认知分心。在第二组中，车祸经常发生在司机在闪烁的红灯或停车标志前停车，然后继续行驶后，这表明司机的注意力有短暂的失误。该研究还强调了当前碰撞数据的局限性，并强调了对分心相关事件进行标准化报告的必要性。研究结果支持针对具体情况的对策，如信号增强、弯道警告和针对新手驾驶员的注意力分散培训，以解决与失神碰撞相关的多方面风险。

{"title":"Profiling crash-associated factors and injury risk patterns among lost-in-thought (daydreaming) drivers: a combined cluster-sequence analysis approach","authors":"Panick Kalambay , Philip Balyagati , Angela Kitali , Emmanuel Kidando","doi":"10.1016/j.aap.2025.108315","DOIUrl":"10.1016/j.aap.2025.108315","url":null,"abstract":"<div><div>Cognitive distraction, particularly in the form of being lost in thought or daydreaming, is a significant yet underexamined contributor to traffic crashes. This study investigates crash patterns and risk profiles associated with drivers who experienced lost-in-thought distraction at the time of a crash in Washington State. Using cluster correspondence analysis (CCA), distinct associations between crash characteristics and driver attributes were uncovered. In addition, process mining was employed to identify typical sequences of crash events. Three meaningful clusters emerged. Cluster 1 involved crashes on road segments with speed limits exceeding 40 mph, lacking traffic control, and often involving male drivers in clear weather. Cluster 2 was marked by crashes at signalized intersections under partly cloudy or foggy conditions, with a higher likelihood of injury. Cluster 3 reflected rainy or low-light crashes involving young drivers on curved, divided, high-speed roads. Across all clusters, frequent crash sequences included collisions with vehicles in transport, parked cars, and fixed objects. Cluster 1 and Cluster 2 stood out for their distinct contextual characteristics. Cluster 1 crashes often involved crossing the center line, suggestive of deep cognitive distraction. In Cluster 2, crashes frequently occurred after drivers stopped at flashing red lights or stop signs, then proceeded, indicating momentary lapses in attention. The study also highlights the limitations of current crash data and emphasizes the need for standardized reporting of distraction-related incidents. Findings support context-specific countermeasures, such as signal enhancements, curve warnings, and distraction-focused training for novice drivers, to address the multifaceted risks associated with lost-in-thought crashes.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108315"},"PeriodicalIF":6.2,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145562147","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Why do drivers brake later than AEB in rear-end collisions? An analysis based on drive recorder videos 在追尾事故中，为什么司机刹车晚于AEB ？基于硬盘录像机视频的分析。

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-19 DOI: 10.1016/j.aap.2025.108275

Yuqing Zhao, Yifan Lin

Rear-end collisions are among the most common crash types in Japan. However, due to their relatively low fatality and injury severity, research on them remains limited. Although previous studies have shown the effectiveness of autonomous emergency braking (AEB), further studies is needed to improve understanding of driver-related factors and enhance system performance to reduce collision risk.

This study investigated the effectiveness of AEB and explored factors contributing to braking delays by reconstructing 52 real-world taxi collisions exceeding 10 km/h using PC-Crash. Simulations were conducted with time-to-collision (TTC) and TTC_2nd-based (considering relative deceleration) AEB to evaluate collision-avoidance performance. Additionally, a decision tree was used to examine environmental, vehicular, and human factors affecting the time difference between driver-initiated braking and AEB activation.

The result indicates that AEB installed in taxis effectively reduced rear-end collisions. TTC_2nd-based AEB could avoid more collisions involving decelerating lead vehicles. Reducing AEB delay times could further enhance prevention. However, the AEB simulation showed limited effectiveness in collisions involving high speeds, wet roads, or sudden deceleration of the lead vehicle. Moreover, the primary reason drivers braked later than the simulated AEB was their failure to maintain a forward gaze before collisions. Drivers were more likely to divert their gaze from forward driving-related areas in non-critical situations, such as low-speed driving or red traffic signals ahead.

This study provides quantitative insights into human factors and AEB technology, which may inform the optimization of AEB systems and the development of driver monitoring systems, contributing to collision prevention and traffic safety.

追尾事故是日本最常见的事故类型之一。然而，由于其相对较低的致死率和损伤严重程度，对其的研究仍然有限。虽然之前的研究已经证明了自动紧急制动（AEB）的有效性，但需要进一步的研究来提高对驾驶员相关因素的理解，提高系统性能以降低碰撞风险。本研究通过PC-Crash软件重建52起超过10公里/小时的出租车碰撞事故，研究了AEB的有效性，并探讨了导致制动延迟的因素。采用碰撞时间（TTC）和基于ttc2（考虑相对减速）的AEB进行仿真，评估避碰性能。此外，还使用决策树来检查影响驾驶员主动制动和AEB启动时间差的环境、车辆和人为因素。结果表明，在出租车上安装AEB可以有效减少追尾事故。基于ttc2的AEB可以避免更多涉及减速车辆的碰撞。减少AEB延迟时间可以进一步加强预防。然而，AEB模拟显示，在涉及高速、潮湿道路或领先车辆突然减速的碰撞中，AEB的有效性有限。此外，司机刹车晚于模拟AEB的主要原因是他们在碰撞前未能保持前视。在非紧急情况下，比如低速行驶或前方有红色交通信号时，司机更有可能将目光从前方驾驶相关区域转移开。该研究为人为因素和AEB技术提供了定量的见解，可为AEB系统的优化和驾驶员监控系统的开发提供参考，有助于预防碰撞和交通安全。

{"title":"Why do drivers brake later than AEB in rear-end collisions? An analysis based on drive recorder videos","authors":"Yuqing Zhao, Yifan Lin","doi":"10.1016/j.aap.2025.108275","DOIUrl":"10.1016/j.aap.2025.108275","url":null,"abstract":"<div><div>Rear-end collisions are among the most common crash types in Japan. However, due to their relatively low fatality and injury severity, research on them remains limited. Although previous studies have shown the effectiveness of autonomous emergency braking (AEB), further studies is needed to improve understanding of driver-related factors and enhance system performance to reduce collision risk.</div><div>This study investigated the effectiveness of AEB and explored factors contributing to braking delays by reconstructing 52 real-world taxi collisions exceeding 10 km/h using PC-Crash. Simulations were conducted with time-to-collision (TTC) and TTC<sub>2nd</sub>-based (considering relative deceleration) AEB to evaluate collision-avoidance performance. Additionally, a decision tree was used to examine environmental, vehicular, and human factors affecting the time difference between driver-initiated braking and AEB activation.</div><div>The result indicates that AEB installed in taxis effectively reduced rear-end collisions. TTC<sub>2nd</sub>-based AEB could avoid more collisions involving decelerating lead vehicles. Reducing AEB delay times could further enhance prevention. However, the AEB simulation showed limited effectiveness in collisions involving high speeds, wet roads, or sudden deceleration of the lead vehicle. Moreover, the primary reason drivers braked later than the simulated AEB was their failure to maintain a forward gaze before collisions. Drivers were more likely to divert their gaze from forward driving-related areas in non-critical situations, such as low-speed driving or red traffic signals ahead.</div><div>This study provides quantitative insights into human factors and AEB technology, which may inform the optimization of AEB systems and the development of driver monitoring systems, contributing to collision prevention and traffic safety.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108275"},"PeriodicalIF":6.2,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145562149","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Have they taken effect as expected? Unpacking the black box of road safety countermeasure effects 它们是否如预期的那样生效？打开道路安全对策的黑箱效果。

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-19 DOI: 10.1016/j.aap.2025.108309

Yingheng Zhang , Haojie Li

Countermeasure evaluation is an important subfield of road traffic safety research. Numerous works have been carried out to establish the cause–effect relationship between countermeasures and crash-related outcomes. However, a single estimate of the countermeasure effect may not always reflect their actual road safety benefits since the underlying causal mechanisms are neglected. To be specific, conventional evaluation frameworks cannot provide a comprehensive view of how a countermeasure influences defined outcomes of interest, e.g., via influencing a specific intermediate variable (mediator). This paper aims to unpack the black box by exploiting the causal inference framework for mediation analysis, which is able to decompose the total causal effect into the indirect effect that relays through the putative mediator and the direct effect. In particular, this framework builds on the potential outcome paradigm, which offers general definitions for the causal estimands of interest and clarifies the core assumptions for nonparametric identification. As a major superiority over traditional mediation analysis, evaluators are permitted to choose from a wide range of statistical methods to estimate the component models (e.g., the mediator model and the outcome model). Simulation experiments are conducted to show the applicability of the causal mediation method to count data analysis with negative binomial models. Also, the method is applied to a speed enforcement camera case study to examine the presence of crash migration phenomena by modelling traffic volume as mediator. Our empirical analysis has found a null indirect effect on crash frequency at enforced road sites, implying that crash migration due to drivers choosing alternative routes to avoid speed cameras is unlikely in this particular case. In summary, by properly applying the causal mediation method to countermeasure evaluation, we could gain a deeper insight into systematic causality.

对策评价是道路交通安全研究的一个重要分支。已经开展了大量的工作来建立对策与碰撞相关结果之间的因果关系。然而，由于忽视了潜在的因果机制，对对策效果的单一估计可能并不总是反映其实际的道路安全效益。具体而言，传统的评估框架无法全面了解对策如何影响已定义的利益结果，例如通过影响特定的中间变量（中介）。本文旨在通过中介分析的因果推理框架来打开黑箱，该框架能够将总因果效应分解为通过假定中介传递的间接效应和直接效应。特别是，该框架建立在潜在结果范式的基础上，该范式为感兴趣的因果估计提供了一般定义，并澄清了非参数识别的核心假设。作为传统中介分析的主要优势，评估者可以从广泛的统计方法中进行选择，以估计成分模型（例如，中介模型和结果模型）。通过仿真实验验证了因果中介方法在负二项模型计数数据分析中的适用性。此外，该方法被应用于一个超速执法摄像机的案例研究，通过模拟交通量作为中介来检查碰撞迁移现象的存在。我们的实证分析发现，在强制道路上，事故发生频率的间接影响为零，这意味着在这种特殊情况下，由于驾驶员选择替代路线以避开超速摄像头而导致的事故迁移不太可能发生。综上所述，适当地将因果中介方法应用于对策评价，可以更深入地了解系统因果关系。

{"title":"Have they taken effect as expected? Unpacking the black box of road safety countermeasure effects","authors":"Yingheng Zhang , Haojie Li","doi":"10.1016/j.aap.2025.108309","DOIUrl":"10.1016/j.aap.2025.108309","url":null,"abstract":"<div><div>Countermeasure evaluation is an important subfield of road traffic safety research. Numerous works have been carried out to establish the cause–effect relationship between countermeasures and crash-related outcomes. However, a single estimate of the countermeasure effect may not always reflect their actual road safety benefits since the underlying causal mechanisms are neglected. To be specific, conventional evaluation frameworks cannot provide a comprehensive view of how a countermeasure influences defined outcomes of interest, e.g., via influencing a specific intermediate variable (mediator). This paper aims to unpack the black box by exploiting the causal inference framework for mediation analysis, which is able to decompose the total causal effect into the indirect effect that relays through the putative mediator and the direct effect. In particular, this framework builds on the potential outcome paradigm, which offers general definitions for the causal estimands of interest and clarifies the core assumptions for nonparametric identification. As a major superiority over traditional mediation analysis, evaluators are permitted to choose from a wide range of statistical methods to estimate the component models (e.g., the mediator model and the outcome model). Simulation experiments are conducted to show the applicability of the causal mediation method to count data analysis with negative binomial models. Also, the method is applied to a speed enforcement camera case study to examine the presence of crash migration phenomena by modelling traffic volume as mediator. Our empirical analysis has found a null indirect effect on crash frequency at enforced road sites, implying that crash migration due to drivers choosing alternative routes to avoid speed cameras is unlikely in this particular case. In summary, by properly applying the causal mediation method to countermeasure evaluation, we could gain a deeper insight into systematic causality.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108309"},"PeriodicalIF":6.2,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145562224","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The effect of license plate number-based vehicle restrictions on crash frequency 车牌限行对碰撞频率的影响。

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-19 DOI: 10.1016/j.aap.2025.108326

Maria C. Valencia-Cardenas , John Jairo Posada-Henao , Jhan Kevin Gil-Marin , Carlos A. Gonzalez-Calderon , Alexander Paz

This study investigates the effects of license plate-based vehicular circulation restrictions, known as “peak-and-plate” policies, on urban traffic crashes in Medellín, Colombia. Although such policies are widely implemented to reduce congestion and emissions, limited research has examined their unintended safety impacts on crash frequency before and during the restriction periods. Considering high-resolution crash data from 2008 to 2022, this study evaluates the temporal effects of these restrictions on crash frequency across different hours, months, and weekdays. Statistical preprocessing and hierarchical clustering techniques were applied to discover crash temporal patterns, and negative binomial, pooled, fixed effects, and random effects models were tested to account for overdispersion and panel structure in the data. Results reveal that crash frequencies increased consistently in the 30 min preceding the enforcement of circulation restrictions, particularly during the morning hours. Surprisingly, crash rates remained elevated during the restriction period, contrary to expectations that reduced vehicle volume would lower crash occurrences. These findings suggest the presence of confounding behavioral, spatial, or enforcement-related factors. Additionally, monthly and weekly crash patterns correlate with seasonal variations, holidays, and changes in traffic volume, emphasizing the importance of context-aware traffic policies. Statistical evidence suggests that current vehicle restriction policies may inadvertently concentrate risk in specific time windows. These insights highlight the need for policymakers to refine urban mobility strategies by considering crash and traffic data simultaneously, implementing targeted public awareness campaigns, targeting speed enforcement, and incorporating enforcement dynamics into the design of vehicle restriction programs.

本研究调查了以牌照为基础的车辆流通限制，即所谓的“高峰和牌照”政策，对哥伦比亚Medellín城市交通事故的影响。尽管这些政策被广泛实施以减少拥堵和排放，但有限的研究已经检查了它们在限制期之前和期间对碰撞频率的意外安全影响。考虑到2008年至2022年的高分辨率碰撞数据，本研究评估了这些限制在不同时间、月份和工作日对碰撞频率的时间影响。应用统计预处理和分层聚类技术发现崩溃时间模式，并测试负二项、池效应、固定效应和随机效应模型，以解释数据中的过度分散和面板结构。结果显示，在交通限制实施前的30分钟内，撞车频率持续增加，尤其是在早晨。令人惊讶的是，与减少车辆数量会降低事故发生率的预期相反，在限制期间，事故发生率仍然很高。这些发现表明存在混淆的行为、空间或执法相关因素。此外，每月和每周的碰撞模式与季节变化、假日和交通量变化相关，强调了环境感知交通政策的重要性。统计证据表明，目前的车辆限制政策可能会无意中将风险集中在特定的时间窗口。这些见解强调了政策制定者需要通过同时考虑碰撞和交通数据、实施有针对性的公众意识运动、针对速度执法以及将执法动态纳入车辆限制计划的设计来完善城市交通战略。

{"title":"The effect of license plate number-based vehicle restrictions on crash frequency","authors":"Maria C. Valencia-Cardenas , John Jairo Posada-Henao , Jhan Kevin Gil-Marin , Carlos A. Gonzalez-Calderon , Alexander Paz","doi":"10.1016/j.aap.2025.108326","DOIUrl":"10.1016/j.aap.2025.108326","url":null,"abstract":"<div><div>This study investigates the effects of license plate-based vehicular circulation restrictions, known as “peak-and-plate” policies, on urban traffic crashes in Medellín, Colombia. Although such policies are widely implemented to reduce congestion and emissions, limited research has examined their unintended safety impacts on crash frequency before and during the restriction periods. Considering high-resolution crash data from 2008 to 2022, this study evaluates the temporal effects of these restrictions on crash frequency across different hours, months, and weekdays. Statistical preprocessing and hierarchical clustering techniques were applied to discover crash temporal patterns, and negative binomial, pooled, fixed effects, and random effects models were tested to account for overdispersion and panel structure in the data. Results reveal that crash frequencies increased consistently in the 30 min preceding the enforcement of circulation restrictions, particularly during the morning hours. Surprisingly, crash rates remained elevated during the restriction period, contrary to expectations that reduced vehicle volume would lower crash occurrences. These findings suggest the presence of confounding behavioral, spatial, or enforcement-related factors. Additionally, monthly and weekly crash patterns correlate with seasonal variations, holidays, and changes in traffic volume, emphasizing the importance of context-aware traffic policies. Statistical evidence suggests that current vehicle restriction policies may inadvertently concentrate risk in specific time windows. These insights highlight the need for policymakers to refine urban mobility strategies by considering crash and traffic data simultaneously, implementing targeted public awareness campaigns<em>,</em> targeting speed enforcement, and incorporating enforcement dynamics into the design of vehicle restriction programs.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108326"},"PeriodicalIF":6.2,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145562107","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Short-Term driving speed prediction under consecutive Variable speed Limits: An interpretable deep learning approach using Wide-Area trajectory data 连续可变速度限制下的短期驾驶速度预测：使用广域轨迹数据的可解释深度学习方法。

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-17 DOI: 10.1016/j.aap.2025.108316

Junhua Wang , Yiwei Ren , Ting Fu , Qiangqiang Shangguan

Existing research on driver behavior under Variable Speed Limits (VSLs) primarily relies on simulations and loop detector-based cross-sectional traffic data, with limited studies using real-world microscopic vehicle trajectory data. This study proposes an interpretable deep learning framework for short-term driving speed prediction under consecutive VSLs control. Using wide-area trajectory data from a 2.2 km segment of the Shanxi Wuyu Freeway with two successive VSL signs, driver behavior was quantitatively analyzed. A Convolutional Neural Network − Bidirectional Long Short-Term Memory (CNN-BiLSTM) model enhanced with a Multi-View Spatio-Temporal Attention Mechanism (MSTAM) was developed to predict short-term speeds and assess the influence of spatiotemporal features on driver responses. Results show that heavy vehicles consistently decelerate under all VSL strategies, while light vehicles display minimal adjustment at 100 km/h and 80 km/h limits but respond more significantly at 60 km/h, with greater inter-driver variability. Drivers in the left lane respond more promptly and decisively than those in the right lane, with shorter response distances and greater speed reductions under all VSL conditions. Additionally, the second VSL sign generally exhibited superior regulatory effectiveness compared to the first VSL sign. Compared to the baseline CNN-BiLSTM, the proposed MSTAM model reduces MAE by 17.2 % and RMSE by 23.5 %. The MSTAM model further captures cognitively consistent spatiotemporal attention patterns, focusing on regulatory zones near VSL signs, sustaining elevated attention in the left lane, and selectively recalling past behaviors to simulate adaptive driver responses. These findings offer a scientific foundation for enhanced VSL deployment and lane-specific speed control strategies.

现有的变速限制（VSLs）下的驾驶员行为研究主要依赖于模拟和基于环路检测器的横断面交通数据，很少有研究使用真实世界的微观车辆轨迹数据。本研究提出一个可解释的深度学习框架，用于连续VSLs控制下的短期驾驶速度预测。利用山西武玉高速公路2.2 km路段连续两个VSL标志的广域轨迹数据，对驾驶员行为进行了定量分析。建立了基于多视角时空注意机制（MSTAM）的卷积神经网络-双向长短期记忆（CNN-BiLSTM）模型，用于预测短期速度并评估时空特征对驾驶员反应的影响。结果表明，重型车辆在所有VSL策略下都能保持减速，而轻型车辆在100 km/h和80 km/h限速下的调整最小，但在60 km/h限速下的响应更显著，驾驶员之间的差异更大。在所有VSL条件下，左车道驾驶员的反应都比右车道驾驶员更迅速、更果断，反应距离更短，减速幅度更大。此外，与第一个VSL标志相比，第二个VSL标志通常表现出更好的调节效果。与基线CNN-BiLSTM相比，MSTAM模型的MAE降低了17.2%，RMSE降低了23.5%。MSTAM模型进一步捕捉认知上一致的时空注意模式，重点关注VSL标志附近的调节区域，保持左侧车道的高注意力，并选择性地回忆过去的行为来模拟自适应驾驶员反应。这些发现为增强VSL部署和特定车道速度控制策略提供了科学基础。

{"title":"Short-Term driving speed prediction under consecutive Variable speed Limits: An interpretable deep learning approach using Wide-Area trajectory data","authors":"Junhua Wang , Yiwei Ren , Ting Fu , Qiangqiang Shangguan","doi":"10.1016/j.aap.2025.108316","DOIUrl":"10.1016/j.aap.2025.108316","url":null,"abstract":"<div><div>Existing research on driver behavior under Variable Speed Limits (VSLs) primarily relies on simulations and loop detector-based cross-sectional traffic data, with limited studies using real-world microscopic vehicle trajectory data. This study proposes an interpretable deep learning framework for short-term driving speed prediction under consecutive VSLs control. Using wide-area trajectory data from a 2.2 km segment of the Shanxi Wuyu Freeway with two successive VSL signs, driver behavior was quantitatively analyzed. A Convolutional Neural Network − Bidirectional Long Short-Term Memory (CNN-BiLSTM) model enhanced with a Multi-View Spatio-Temporal Attention Mechanism (MSTAM) was developed to predict short-term speeds and assess the influence of spatiotemporal features on driver responses. Results show that heavy vehicles consistently decelerate under all VSL strategies, while light vehicles display minimal adjustment at 100 km/h and 80 km/h limits but respond more significantly at 60 km/h, with greater inter-driver variability. Drivers in the left lane respond more promptly and decisively than those in the right lane, with shorter response distances and greater speed reductions under all VSL conditions. Additionally, the second VSL sign generally exhibited superior regulatory effectiveness compared to the first VSL sign. Compared to the baseline CNN-BiLSTM, the proposed MSTAM model reduces MAE by 17.2 % and RMSE by 23.5 %. The MSTAM model further captures cognitively consistent spatiotemporal attention patterns, focusing on regulatory zones near VSL signs, sustaining elevated attention in the left lane, and selectively recalling past behaviors to simulate adaptive driver responses. These findings offer a scientific foundation for enhanced VSL deployment and lane-specific speed control strategies.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108316"},"PeriodicalIF":6.2,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145547749","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Modelling cognitive load using drift-diffusion models in pedestrian street-crossing: a method supported by neural evidence 用漂移-扩散模型模拟行人过马路的认知负荷：一种神经学证据支持的方法。

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-16 DOI: 10.1016/j.aap.2025.108310

Siwei Ma , Xuedong Yan , Lu Ma , Jac Billington , Natasha Merat , Gustav Markkula

When pedestrians are cognitively loaded, this influences their street-crossing behaviour, leading to negative impact on overall road safety. However, the mechanisms underpinning this impact remain debated, and this study seeks to further investigate them through modelling and electroencephalography. We conducted a computer-based pedestrian crossing experiment, and employed drift-diffusion models to quantitatively analyse how cognitive load impacts pedestrian decision-making. To further test the models’ validity, we analysed centro-parietal positive potential (CPP), a neural signal associated with evidence accumulation, to investigate whether this neural evidence aligned with the evidence accumulation predicted by the models. In our experiment, participants encountered a simulated scenario with a car approaching under four different time-to-arrival (TTA) conditions. In half the trials, participants performed cognitive tasks while deciding when to cross the street. Results showed that cognitive load weakened the effect of TTA on the probability of crossing before the car, increased response times, raised the probability of collision, and attenuated CPP amplitude. The best-performing model, which captured all of these effects, accumulated evidence based on utility estimates, but with a lower responsiveness to these utilities during cognitive load. This model also showed the strongest correlation between its evidence traces and the CPP amplitude, both with and without cognitive load. These findings support the hypothesis that cognitive load reduces responsiveness to perceptual evidence (at least in non-automatised tasks), making it a strong candidate for explaining both our results and existing research on the effects of cognitive load in other tasks.

当行人认知负荷过重时，这会影响他们过马路的行为，从而对整体道路安全产生负面影响。然而，支持这种影响的机制仍然存在争议，本研究试图通过建模和脑电图进一步研究它们。我们进行了基于计算机的行人过街实验，并采用漂移-扩散模型定量分析了认知负荷对行人决策的影响。为了进一步验证模型的有效性，我们分析了与证据积累相关的神经信号中枢-顶叶正电位（CPP），以研究这种神经证据是否与模型预测的证据积累相一致。在我们的实验中，参与者遇到了一个模拟场景，一辆汽车在四种不同的到达时间（TTA）条件下接近。在一半的试验中，参与者在决定何时过马路时执行认知任务。结果表明，认知负荷减弱了TTA对车前穿越概率的影响，增加了反应时间，提高了碰撞概率，减弱了CPP振幅。表现最好的模型捕获了所有这些影响，根据效用估计积累了证据，但在认知负荷期间对这些效用的反应较低。该模型还显示，无论有无认知负荷，其证据痕迹与CPP振幅之间都存在最强的相关性。这些发现支持了认知负荷降低对感知证据的反应的假设（至少在非自动化任务中），使其成为解释我们的结果和现有的关于认知负荷对其他任务影响的研究的有力候选。

{"title":"Modelling cognitive load using drift-diffusion models in pedestrian street-crossing: a method supported by neural evidence","authors":"Siwei Ma , Xuedong Yan , Lu Ma , Jac Billington , Natasha Merat , Gustav Markkula","doi":"10.1016/j.aap.2025.108310","DOIUrl":"10.1016/j.aap.2025.108310","url":null,"abstract":"<div><div>When pedestrians are cognitively loaded, this influences their street-crossing behaviour, leading to negative impact on overall road safety. However, the mechanisms underpinning this impact remain debated, and this study seeks to further investigate them through modelling and electroencephalography. We conducted a computer-based pedestrian crossing experiment, and employed drift-diffusion models to quantitatively analyse how cognitive load impacts pedestrian decision-making. To further test the models’ validity, we analysed centro-parietal positive potential (CPP), a neural signal associated with evidence accumulation, to investigate whether this neural evidence aligned with the evidence accumulation predicted by the models. In our experiment, participants encountered a simulated scenario with a car approaching under four different time-to-arrival (TTA) conditions. In half the trials, participants performed cognitive tasks while deciding when to cross the street. Results showed that cognitive load weakened the effect of TTA on the probability of crossing before the car, increased response times, raised the probability of collision, and attenuated CPP amplitude. The best-performing model, which captured all of these effects, accumulated evidence based on utility estimates, but with a lower responsiveness to these utilities during cognitive load. This model also showed the strongest correlation between its evidence traces and the CPP amplitude, both with and without cognitive load. These findings support the hypothesis that cognitive load reduces responsiveness to perceptual evidence (at least in non-automatised tasks), making it a strong candidate for explaining both our results and existing research on the effects of cognitive load in other tasks.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108310"},"PeriodicalIF":6.2,"publicationDate":"2025-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145538191","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-factor coupled friction coefficient map spatiotemporal modeling and driving risk fusion for rainy roads 雨天道路多因素耦合摩擦系数图时空建模与驾驶风险融合

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-15 DOI: 10.1016/j.aap.2025.108313

Lintao Yang , Xing Cui , Huizhao Tu , Hao Li , Yiik Diew Wong

The reduction in tire-road friction coefficient (TRFC) during rainy weather is a major cause of traffic accidents. TRFC values result from the spatiotemporal coupling of road, vehicle, and environmental factors, yet existing estimation methods struggle to fully integrate these factors, hindering the accuracy of TRFC-related driving risk assessments. To address this, a framework for spatiotemporal coupling of these factors at the lane level to generate friction coefficient maps is proposed and two fusion methods for multi-driving risks are established for rainy roads. A tire-fluid-road simulation model is developed to output a TRFC dataset for training a surrogate prediction model. Lanes are divided into grids to align TRFC-related factors, which are then input into the surrogate model to estimate TRFC and create friction coefficient maps. Three TRFC-related driving risks (hydroplaning, rear-end collisions, and sideslip) are analyzed and normalized to construct multi-risk maps, evaluated using max-risk and weighted-sum fusion strategies. The proposed method was validated using rainy-day car-following trajectory data on an urban expressway in Shanghai. Results show that heavy rainfall and high speeds reduce TRFC levels and increase its variability between wheel and non-wheel paths, augmenting the hydroplaning and sideslip risks, while reduced TRFC increases safe following distance and rear-end collision risk as vehicle convergence. The risk fusion results evidence that max-risk fusion excels in scenarios with a dominant risk, while weighted-sum fusion suits scenarios with multiple high-risk types. This study offers a lane-level driving risk assessment for rainy roads, providing insights for developing safety measures in wet road conditions.

雨天时胎路摩擦系数的降低是造成交通事故的主要原因。TRFC值是道路、车辆和环境因素时空耦合的结果，但现有的估算方法难以充分整合这些因素，影响了TRFC相关驾驶风险评估的准确性。为解决这一问题，提出了在车道层面上对这些因素进行时空耦合以生成摩擦系数图的框架，并建立了两种雨天道路多重驾驶风险的融合方法。开发了轮胎-流体-道路仿真模型，输出TRFC数据集用于训练代理预测模型。车道被划分为网格，以对齐与TRFC相关的因素，然后将其输入代理模型以估计TRFC并创建摩擦系数图。对三种与trfc相关的驾驶风险（打滑、追尾和侧滑）进行分析和归一化，构建多风险图，并使用最大风险和加权和融合策略进行评估。利用上海市某城市高速公路雨天车辆跟踪轨迹数据对该方法进行了验证。结果表明：强降雨和高速降低了车轮与非车轮路径之间的TRFC水平，增加了其变异性，增加了打滑和侧滑风险，而减少的TRFC增加了车辆收敛时的安全跟随距离和追尾碰撞风险。风险融合结果表明，最大风险融合适用于风险占主导地位的场景，而加权和融合适用于高风险类型较多的场景。该研究为雨天道路提供了车道级驾驶风险评估，为在潮湿道路条件下制定安全措施提供了见解。

{"title":"Multi-factor coupled friction coefficient map spatiotemporal modeling and driving risk fusion for rainy roads","authors":"Lintao Yang , Xing Cui , Huizhao Tu , Hao Li , Yiik Diew Wong","doi":"10.1016/j.aap.2025.108313","DOIUrl":"10.1016/j.aap.2025.108313","url":null,"abstract":"<div><div>The reduction in tire-road friction coefficient (TRFC) during rainy weather is a major cause of traffic accidents. TRFC values result from the spatiotemporal coupling of road, vehicle, and environmental factors, yet existing estimation methods struggle to fully integrate these factors, hindering the accuracy of TRFC-related driving risk assessments. To address this, a framework for spatiotemporal coupling of these factors at the lane level to generate friction coefficient maps is proposed and two fusion methods for multi-driving risks are established for rainy roads. A tire-fluid-road simulation model is developed to output a TRFC dataset for training a surrogate prediction model. Lanes are divided into grids to align TRFC-related factors, which are then input into the surrogate model to estimate TRFC and create friction coefficient maps. Three TRFC-related driving risks (hydroplaning, rear-end collisions, and sideslip) are analyzed and normalized to construct multi-risk maps, evaluated using max-risk and weighted-sum fusion strategies. The proposed method was validated using rainy-day car-following trajectory data on an urban expressway in Shanghai. Results show that heavy rainfall and high speeds reduce TRFC levels and increase its variability between wheel and non-wheel paths, augmenting the hydroplaning and sideslip risks, while reduced TRFC increases safe following distance and rear-end collision risk as vehicle convergence. The risk fusion results evidence that max-risk fusion excels in scenarios with a dominant risk, while weighted-sum fusion suits scenarios with multiple high-risk types. This study offers a lane-level driving risk assessment for rainy roads, providing insights for developing safety measures in wet road conditions.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108313"},"PeriodicalIF":6.2,"publicationDate":"2025-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145518958","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SET-DGCN: An end-to-end electroencephalography-based fatigue detection method for young drivers SET-DGCN：基于端到端脑电图的年轻驾驶员疲劳检测方法

IF 6.2 1区工程技术 Q1 ERGONOMICS

Accident; analysis and prevention

Pub Date : 2025-11-14 DOI: 10.1016/j.aap.2025.108311

Yang Cao , Tiantian Chen , Ke Han , Hyungchul Chung , Zhaoguo Huang , Hongliang Ding

Driver fatigue poses a critical threat to global road safety, particularly among young drivers. Nevertheless, policy-level interventions remain fragmented due to the lack of reliable and deployable detection technologies. Bridging this gap requires accurate, interpretable, and real-time fatigue monitoring systems capable of informing practical decision-making in transportation safety management. To address this challenge, we propose an end-to-end EEG-based fatigue detection model, Scale-Enhanced Transformer and Dynamic Graph Convolutional Network (SET-DGCN). The model captures multi-scale temporal dependencies and spatial brain-region interactions by integrating convolutional embeddings, attention mechanisms, and learnable graph structures. Extensive evaluations on both a driving simulation dataset and the publicly available SEED-VIG dataset confirm that SET-DGCN outperforms mainstream convolutional neural network (CNN)-based, graph convolutional network (GCN)-based, and Transformer-based models in terms of accuracy and F1-score, while maintaining strong cross-subject generalization. To enhance both interpretability and application relevance, a component-level attribution method (COAR) is employed to evaluate the functional contribution of model modules, while SHapley Additive exPlanations (SHAP) analysis is used to uncover brain region–specific patterns across fatigue stages. Based on these neural insights, a set of multi-level policy and design recommendations is proposed, ranging from infrastructure enhancements to adaptive in-vehicle systems and individualized interventions, to provide a comprehensive framework for mitigating fatigue among young drivers in real-world transportation contexts.

驾驶员疲劳对全球道路安全构成严重威胁，尤其是对年轻驾驶员而言。然而，由于缺乏可靠和可部署的检测技术，政策层面的干预措施仍然是零散的。弥合这一差距需要准确、可解释和实时的疲劳监测系统，能够为运输安全管理的实际决策提供信息。为了解决这一挑战，我们提出了一种端到端的基于脑电图的疲劳检测模型，缩放增强变压器和动态图卷积网络（SET-DGCN）。该模型通过集成卷积嵌入、注意机制和可学习的图结构来捕获多尺度时间依赖性和空间脑区域相互作用。对驾驶模拟数据集和公开可用的SEED-VIG数据集的广泛评估证实，SET-DGCN在准确性和f1分数方面优于主流的基于卷积神经网络（CNN）、基于图卷积网络（GCN）和基于transformer的模型，同时保持了强大的跨学科泛化。为了提高可解释性和应用相关性，采用组件级归因方法（COAR）来评估模型模块的功能贡献，而使用SHapley加性解释（SHAP）分析来揭示跨疲劳阶段的大脑区域特定模式。基于这些神经系统的见解，作者提出了一套多层次的政策和设计建议，从基础设施增强到自适应车载系统和个性化干预，为减轻现实交通环境中年轻司机的疲劳提供了一个全面的框架。

{"title":"SET-DGCN: An end-to-end electroencephalography-based fatigue detection method for young drivers","authors":"Yang Cao , Tiantian Chen , Ke Han , Hyungchul Chung , Zhaoguo Huang , Hongliang Ding","doi":"10.1016/j.aap.2025.108311","DOIUrl":"10.1016/j.aap.2025.108311","url":null,"abstract":"<div><div>Driver fatigue poses a critical threat to global road safety, particularly among young drivers. Nevertheless, policy-level interventions remain fragmented due to the lack of reliable and deployable detection technologies. Bridging this gap requires accurate, interpretable, and real-time fatigue monitoring systems capable of informing practical decision-making in transportation safety management. To address this challenge, we propose an end-to-end EEG-based fatigue detection model, Scale-Enhanced Transformer and Dynamic Graph Convolutional Network (SET-DGCN). The model captures multi-scale temporal dependencies and spatial brain-region interactions by integrating convolutional embeddings, attention mechanisms, and learnable graph structures. Extensive evaluations on both a driving simulation dataset and the publicly available SEED-VIG dataset confirm that SET-DGCN outperforms mainstream convolutional neural network (CNN)-based, graph convolutional network (GCN)-based, and Transformer-based models in terms of accuracy and F1-score, while maintaining strong cross-subject generalization. To enhance both interpretability and application relevance, a component-level attribution method (COAR) is employed to evaluate the functional contribution of model modules, while SHapley Additive exPlanations (SHAP) analysis is used to uncover brain region–specific patterns across fatigue stages. Based on these neural insights, a set of multi-level policy and design recommendations is proposed, ranging from infrastructure enhancements to adaptive in-vehicle systems and individualized interventions, to provide a comprehensive framework for mitigating fatigue among young drivers in real-world transportation contexts.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"225 ","pages":"Article 108311"},"PeriodicalIF":6.2,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145500152","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0