Pub Date : 2025-12-31DOI: 10.1016/j.aap.2025.108387
Ling Deng , Chengcheng Xu , Pan Liu , Yuxuan Wang , Yanli Jiao , Kequan Chen
The primary objective of this paper is to improve the prediction accuracy of freeway secondary crashes by jointly modeling spatiotemporal dynamics and addressing the challenge of class imbalance. Using secondary-crash data from Interstate 5 (I-5) in California, we develop a Spatial-Temporal Gated Transformer Network (STGT-Net). STGT-Net employs a tri-branch encoder to model upstream, downstream, and differential traffic flows, and dual Transformer modules with a gated fusion mechanism to capture temporal and inter-feature dependencies. To address data rarity, we employ an LSTM-based Wasserstein GAN with gradient penalty (LSTM-WGAN-GP) to generate realistic, sequence-aware crash samples. Comparative results show that the secondary crash data generated by our method aligns more closely with real-world conditions than those produced by existing approaches. STGT-Net achieves substantial relative improvements of 13.08% in F1-score and 13.95% in Matthews Correlation Coefficient (MCC) compared with the best baseline. Finally, SHapley Additive exPlanations (SHAP)-based analysis reveals that upstream traffic variables contribute most strongly to the model’s predictions, with downstream and upstream–downstream differential indicators also providing informative cues.
本文的主要目标是通过对高速公路二次碰撞的时空动态建模和解决类别不平衡的挑战，提高高速公路二次碰撞的预测精度。利用加州5号州际公路（I-5）的二次事故数据，我们开发了一个时空门控变压器网络（STGT-Net）。STGT-Net采用三分支编码器对上游、下游和差分流量进行建模，并采用带门控融合机制的双Transformer模块来捕获时间和特征间的依赖关系。为了解决数据稀缺性问题，我们采用了基于lstm的带梯度惩罚的Wasserstein GAN （LSTM-WGAN-GP）来生成真实的、序列感知的崩溃样本。对比结果表明，我们的方法产生的二次碰撞数据比现有方法产生的数据更接近真实情况。与最佳基线相比，STGT-Net在f1评分和马修斯相关系数（MCC）方面取得了13.08%和13.95%的实质性相对改善。最后，基于SHapley加性解释（SHAP）的分析表明，上游交通变量对模型的预测贡献最大，下游和上下游差异指标也提供了信息线索。
{"title":"Spatial-temporal gated transformer network for freeway secondary crash prediction considering the impact of class imbalance","authors":"Ling Deng , Chengcheng Xu , Pan Liu , Yuxuan Wang , Yanli Jiao , Kequan Chen","doi":"10.1016/j.aap.2025.108387","DOIUrl":"10.1016/j.aap.2025.108387","url":null,"abstract":"<div><div>The primary objective of this paper is to improve the prediction accuracy of freeway secondary crashes by jointly modeling spatiotemporal dynamics and addressing the challenge of class imbalance. Using secondary-crash data from Interstate 5 (I-5) in California, we develop a Spatial-Temporal Gated Transformer Network (STGT-Net). STGT-Net employs a tri-branch encoder to model upstream, downstream, and differential traffic flows, and dual Transformer modules with a gated fusion mechanism to capture temporal and inter-feature dependencies. To address data rarity, we employ an LSTM-based Wasserstein GAN with gradient penalty (LSTM-WGAN-GP) to generate realistic, sequence-aware crash samples. Comparative results show that the secondary crash data generated by our method aligns more closely with real-world conditions than those produced by existing approaches. STGT-Net achieves substantial relative improvements of 13.08% in F1-score and 13.95% in Matthews Correlation Coefficient (MCC) compared with the best baseline. Finally, SHapley Additive exPlanations (SHAP)-based analysis reveals that upstream traffic variables contribute most strongly to the model’s predictions, with downstream and upstream–downstream differential indicators also providing informative cues.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"227 ","pages":"Article 108387"},"PeriodicalIF":6.2,"publicationDate":"2025-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145882003","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}
Pub Date : 2025-12-31DOI: 10.1016/j.aap.2025.108386
Zongni Gu , Ilir Bejleri , Binbin Peng
The escalating frequency of extreme heat events poses a potential threat to roadway safety, yet the spatial patterns of crash risk from more multi-dimensional perspectives remain underexplored. Using the Hazards-Exposure-Vulnerability-Adaptation paradigm, this study examines how traffic crash rates on extreme-heat days vary across roadway segments and road design characteristics in the City of Miami, Florida. This study analyzed traffic exposure and crash rates across three yearly extreme heat thresholds (90th, 95th, and 97th percentiles) from 2011 to 2015. A CatBoost model, interpreted via SHAP analysis, is used to identify key roadway and contextual features associated with higher or lower crash rates during extreme-heat days. The key findings are as follows: 1) Crash risk on extreme-heat days shows a threshold-dependent pattern across the examined percentiles. As the heat threshold intensifies from the 90th to the 97th percentile, the average network-wide crash rate increases, while the number of road segments with above-average crash rates follows a V-shaped pattern—first declining and then rising sharply at the highest threshold. This suggests that inherent adaptive characteristics of many roadways may be sufficient to moderate crash risk under moderately extreme heat but become increasingly inadequate once heat reaches very abnormally high threshold (e.g., the 97th percentile). 2) Models based on higher extreme-heat thresholds provide clearer insight into vulnerability patterns. Compared to the 90th and 95th percentile models, the 97th percentile model more clearly isolates roadway and contextual features most strongly associated with elevated crash rates on extreme-heat days, whereas lower thresholds appear more affected by noise from other coincident factors. 3) Roadway investment emerges as the primary adaptive factor associated with reduced risk. Physical attributes such as construction cost and geometric design are the most influential correlates of crash vulnerability on extreme-heat days, with higher-quality roadway investment linked to substantially lower crash rates. In contrast, the observed associations for safety control measures and emergency service accessibility are comparatively limited. These findings characterize which roadway environments are more vulnerable when extreme-heat conditions occur.
{"title":"Roadway traffic crash during extreme heat days: insights from hazards-exposure-vulnerability-adaptation","authors":"Zongni Gu , Ilir Bejleri , Binbin Peng","doi":"10.1016/j.aap.2025.108386","DOIUrl":"10.1016/j.aap.2025.108386","url":null,"abstract":"<div><div>The escalating frequency of extreme heat events poses a potential threat to roadway safety, yet the spatial patterns of crash risk from more multi-dimensional perspectives remain underexplored. Using the Hazards-Exposure-Vulnerability-Adaptation paradigm, this study examines how traffic crash rates on extreme-heat days vary across roadway segments and road design characteristics in the City of Miami, Florida. This study analyzed traffic exposure and crash rates across three yearly extreme heat thresholds (90th, 95th, and 97th percentiles) from 2011 to 2015. A CatBoost model, interpreted via SHAP analysis, is used to identify key roadway and contextual features associated with higher or lower crash rates during extreme-heat days. The key findings are as follows: 1) Crash risk on extreme-heat days shows a threshold-dependent pattern across the examined percentiles. As the heat threshold intensifies from the 90th to the 97th percentile, the average network-wide crash rate increases, while the number of road segments with above-average crash rates follows a V-shaped pattern—first declining and then rising sharply at the highest threshold. This suggests that inherent adaptive characteristics of many roadways may be sufficient to moderate crash risk under moderately extreme heat but become increasingly inadequate once heat reaches very abnormally high threshold (e.g., the 97th percentile). 2) Models based on higher extreme-heat thresholds provide clearer insight into vulnerability patterns. Compared to the 90th and 95th percentile models, the 97th percentile model more clearly isolates roadway and contextual features most strongly associated with elevated crash rates on extreme-heat days, whereas lower thresholds appear more affected by noise from other coincident factors. 3) Roadway investment emerges as the primary adaptive factor associated with reduced risk. Physical attributes such as construction cost and geometric design are the most influential correlates of crash vulnerability on extreme-heat days, with higher-quality roadway investment linked to substantially lower crash rates. In contrast, the observed associations for safety control measures and emergency service accessibility are comparatively limited. These findings characterize which roadway environments are more vulnerable when extreme-heat conditions occur.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"227 ","pages":"Article 108386"},"PeriodicalIF":6.2,"publicationDate":"2025-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145882096","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}
Pub Date : 2025-12-29DOI: 10.1016/j.aap.2025.108383
Dong Liu , Hui Li , Yongkang Chen , Jun Yi , Ye Tian , Renxi Wang
Highly automated mining trucks significantly enhance operational safety and efficiency. However, until full automation is realized, remote takeover remains a necessary safeguard. Consequently, designing appropriate takeover request (TOR) is critical for improving driving safety and user experience. Nevertheless, existing research has seldom considered the design of audiovisual TORs under different lighting conditions or their adaptation for non-passenger vehicles. To address this gap, this study conducted a remote takeover experiment on mining trucks, involving 48 human participants, to investigate the effects of nine distinct audiovisual TORs on takeover performance, attention, driving stress, and subjective perceptions under two lighting conditions (daytime and nighttime). Statistical analysis was performed using Generalized Estimating Equation (GEE). The results indicate that both lighting conditions and the type of audiovisual TOR significantly influenced the dependent variables. Furthermore, TORs exhibited differential effectiveness depending on the lighting condition. During nighttime, drivers demonstrated greater reliance on the auditory channel, adopting a strategy of visual concentration and more conservative driving. Although heightened tension responses were observed, drivers exhibited better stress regulation capabilities. Speech-based cues effectively enhanced driving safety, particularly during daytime. TORs relying solely on beep sounds were prone to inducing tension and reducing acceptability. Excessive visual elements could distract attention, impairing takeover performance and situational awareness, although they partly alleviated physiological stress. The optimal TOR configuration was condition-specific: a combination of text and speech was most effective during daytime, whereas a combination of icon, beep, and speech yielded the best results at night. This study provides valuable insights for optimizing TOR design in remote takeover interfaces for mining trucks.
{"title":"Enhancing safety in remote takeover of highly automated mining trucks: What are the different effects of audiovisual takeover requests under daytime and nighttime conditions?","authors":"Dong Liu , Hui Li , Yongkang Chen , Jun Yi , Ye Tian , Renxi Wang","doi":"10.1016/j.aap.2025.108383","DOIUrl":"10.1016/j.aap.2025.108383","url":null,"abstract":"<div><div>Highly automated mining trucks significantly enhance operational safety and efficiency. However, until full automation is realized, remote takeover remains a necessary safeguard. Consequently, designing appropriate takeover request (TOR) is critical for improving driving safety and user experience. Nevertheless, existing research has seldom considered the design of audiovisual TORs under different lighting conditions or their adaptation for non-passenger vehicles. To address this gap, this study conducted a remote takeover experiment on mining trucks, involving 48 human participants, to investigate the effects of nine distinct audiovisual TORs on takeover performance, attention, driving stress, and subjective perceptions under two lighting conditions (daytime and nighttime). Statistical analysis was performed using Generalized Estimating Equation (GEE). The results indicate that both lighting conditions and the type of audiovisual TOR significantly influenced the dependent variables. Furthermore, TORs exhibited differential effectiveness depending on the lighting condition. During nighttime, drivers demonstrated greater reliance on the auditory channel, adopting a strategy of visual concentration and more conservative driving. Although heightened tension responses were observed, drivers exhibited better stress regulation capabilities. Speech-based cues effectively enhanced driving safety, particularly during daytime. TORs relying solely on beep sounds were prone to inducing tension and reducing acceptability. Excessive visual elements could distract attention, impairing takeover performance and situational awareness, although they partly alleviated physiological stress. The optimal TOR configuration was condition-specific: a combination of text and speech was most effective during daytime, whereas a combination of icon, beep, and speech yielded the best results at night. This study provides valuable insights for optimizing TOR design in remote takeover interfaces for mining trucks.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"227 ","pages":"Article 108383"},"PeriodicalIF":6.2,"publicationDate":"2025-12-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145861727","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}
The global adoption of e-bikes improves urban mobility but can impact traffic safety. Braking behavior is a vital task with a high association with e-bike cyclists’ safety, especially under mixed traffic conditions. Although traditional studies on cyclists’ traffic safety analyses have been conducted based on crash data in transportation research, most studies focus on crash severity rather than the specific behavior risks associated with different traffic conditions. This paper uses field experiment data to study the factors that affect the observed braking behavior of e-bike cyclists under mixed and non-mixed traffic conditions. To simultaneously account for this flexibility to model varying effects across traffic conditions and capture unobserved heterogeneity, a partially constrained random parameters logit model with heterogeneity in the means is estimated. The analysis controls cyclists’ characteristics, behavioral factors and roadway characteristics. Our main findings indicate that safety is affected by mixed traffic conditions, which has implications for infrastructure design and cyclist characteristics. Different head-turning and handlebar turning behavior reflect cyclists’ risk perception and risk avoidance responses under various road conditions. Overall, accounting for unobserved heterogeneity and cross-condition correlation provides a more nuanced understanding of e-bike cyclists’ braking behavior.
{"title":"Determinants influencing risks in e-bike cyclists under mix traffic condition: a partially constrained random parameters approach using experimental study data","authors":"Yuntong Zhou , Xin Gu , Mohamed Abdel-Aty , Yanyan Chen","doi":"10.1016/j.aap.2025.108364","DOIUrl":"10.1016/j.aap.2025.108364","url":null,"abstract":"<div><div>The global adoption of e-bikes improves urban mobility but can impact traffic safety. Braking behavior is a vital task with a high association with e-bike cyclists’ safety, especially under mixed traffic conditions. Although traditional studies on cyclists’ traffic safety analyses have been conducted based on crash data in transportation research, most studies focus on crash severity rather than the specific behavior risks associated with different traffic conditions. This paper uses field experiment data to study the factors that affect the observed braking behavior of e-bike cyclists under mixed and non-mixed traffic conditions. To simultaneously account for this flexibility to model varying effects across traffic conditions and capture unobserved heterogeneity, a partially constrained random parameters logit model with heterogeneity in the means is estimated. The analysis controls cyclists’ characteristics, behavioral factors and roadway characteristics. Our main findings indicate that safety is affected by mixed traffic conditions, which has implications for infrastructure design and cyclist characteristics. Different head-turning and handlebar turning behavior reflect cyclists’ risk perception and risk avoidance responses under various road conditions. Overall, accounting for unobserved heterogeneity and cross-condition correlation provides a more nuanced understanding of e-bike cyclists’ braking behavior.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"227 ","pages":"Article 108364"},"PeriodicalIF":6.2,"publicationDate":"2025-12-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145852944","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}
Pub Date : 2025-12-26DOI: 10.1016/j.aap.2025.108360
Ying Luo , Xiaomeng Li , Ashish Bhaskar , Dong Ngoduy , Mohammed Elhenawy , Sebastien Glaser
Surrogate Safety Measures (SSMs) have demonstrated substantial potential in identifying collision risks and supporting traffic safety evaluation. However, conventional SSMs are primarily designed for individual vehicle–level risk assessment, which limits their ability to capture inter-vehicle risk coupling and the mechanisms of potential risk propagation within a vehicle platoon. To address this limitation, this study proposes a novel Cascading Risk Measure (CRM) to characterize potential cascading risks and enhance platoon-level safety assessment. First, we introduce the concept of critical deceleration, representing the maximum deceleration disturbance from a leading vehicle that a follower can tolerate, to characterize each vehicle’s risk state. Building on this, we develop a recursive mathematical framework that captures inter-vehicle coupling and the propagation of cascading risks by establishing recursive relationships between the critical decelerations of successive vehicles. This formulation reveals both the direction and the fundamental manner of risk propagation, thereby offering clear physical interpretability. In addition, the recursive framework ensures that CRM naturally degenerates into its single-vehicle form (denoted as DCRM) when no cascading risk is present, thereby generalizing the single-vehicle view, ensuring full compatibility with conventional SSMs, and providing a coherent and interpretable bridge between vehicle-level and platoon-level risk evaluation. The framework also guarantees complete coverage of cascading risks under arbitrary disturbance-propagation scenarios, ensuring that cascading effects are not omitted by design. Validation results show that, compared with existing SSMs, CRM enables earlier identification of peak platoon risk and generates smoother, more consistent risk profiles by better aligning with the evolution of platoon-level risk. These advantages are quantitatively supported by performance gains of 36.18% in MAE, 17.56% in sMAPE, and 22.22% in RMSSD, as well as an 18.84% increase in Kendall’s relative to the best-performing baseline. CRM also effectively captures the impact of platoon size and vehicle spatial distribution—factors to which conventional SSMs are largely insensitive. These findings highlight CRM’s potential for supporting proactive platoon-level risk warning and management.
{"title":"Enhancing vehicle platoon safety assessment: A novel cascading risk measure","authors":"Ying Luo , Xiaomeng Li , Ashish Bhaskar , Dong Ngoduy , Mohammed Elhenawy , Sebastien Glaser","doi":"10.1016/j.aap.2025.108360","DOIUrl":"10.1016/j.aap.2025.108360","url":null,"abstract":"<div><div>Surrogate Safety Measures (SSMs) have demonstrated substantial potential in identifying collision risks and supporting traffic safety evaluation. However, conventional SSMs are primarily designed for individual vehicle–level risk assessment, which limits their ability to capture inter-vehicle risk coupling and the mechanisms of potential risk propagation within a vehicle platoon. To address this limitation, this study proposes a novel Cascading Risk Measure (CRM) to characterize potential cascading risks and enhance platoon-level safety assessment. First, we introduce the concept of critical deceleration, representing the maximum deceleration disturbance from a leading vehicle that a follower can tolerate, to characterize each vehicle’s risk state. Building on this, we develop a recursive mathematical framework that captures inter-vehicle coupling and the propagation of cascading risks by establishing recursive relationships between the critical decelerations of successive vehicles. This formulation reveals both the direction and the fundamental manner of risk propagation, thereby offering clear physical interpretability. In addition, the recursive framework ensures that CRM naturally degenerates into its single-vehicle form (denoted as DCRM) when no cascading risk is present, thereby generalizing the single-vehicle view, ensuring full compatibility with conventional SSMs, and providing a coherent and interpretable bridge between vehicle-level and platoon-level risk evaluation. The framework also guarantees complete coverage of cascading risks under arbitrary disturbance-propagation scenarios, ensuring that cascading effects are not omitted by design. Validation results show that, compared with existing SSMs, CRM enables earlier identification of peak platoon risk and generates smoother, more consistent risk profiles by better aligning with the evolution of platoon-level risk. These advantages are quantitatively supported by performance gains of 36.18% in MAE, 17.56% in sMAPE, and 22.22% in RMSSD, as well as an 18.84% increase in Kendall’s <span><math><msub><mrow><mi>τ</mi></mrow><mrow><mi>b</mi></mrow></msub></math></span> relative to the best-performing baseline. CRM also effectively captures the impact of platoon size and vehicle spatial distribution—factors to which conventional SSMs are largely insensitive. These findings highlight CRM’s potential for supporting proactive platoon-level risk warning and management.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"227 ","pages":"Article 108360"},"PeriodicalIF":6.2,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145839755","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}
Pub Date : 2025-12-24DOI: 10.1016/j.aap.2025.108365
Jie Pan , Yongjun Shen , Chengyu He , Jing Shi
Automated vehicles (AVs) face a critical need to adopt socially compatible behaviors and cooperate with human-driven vehicles (HVs) in heterogeneous traffic environments. However, existing lane-changing decision frameworks for AVs rarely account for the dynamic trust levels of HVs, thereby impeding the accurate prediction of HV behaviors due to inherent uncertainties. This study proposes a trust-aware, game-theoretic lane-changing decision framework for AVs. First, a trust-aware multi-vehicle coalition game for heterogeneous traffic environment is developed, which encompasses a fully cooperative game among AVs and a partially cooperative model for HVs based on each HV’s real-time trust level. Second, an online evaluation method to dynamically estimate HV’s trust during lane-change interactions is designed, guiding AVs to select appropriate cooperative maneuvers. Finally, social compatibility objectives are considered by minimizing disruption to surrounding vehicles and enhancing the predictability of AV behavior, ensuring the overall lane-changing strategy remains both human-friendly and context-adaptive. To test the proposed model, a human-in-the-loop experiment was conducted in a highway on-ramp scenario. Results showed that the AV adjusted its lane-changing strategy accordingly for human drivers of varying driving styles and trust levels. Moreover, the proposed model facilitated safe, efficient, and comfortable merging behavior for the AV. Ablation studies revealed that incorporating the trust mechanism yields higher average lane-change speeds than models without it, demonstrating improved efficiency while maintaining safety in heterogeneous traffic. This research contributes to enhance the interpretability of HV-AV interaction and promotes the design of more transparent and adaptive lane-changing strategies in automated driving systems.
{"title":"TGLD: A trust-aware game-theoretic lane-changing decision framework for automated vehicles in heterogeneous traffic","authors":"Jie Pan , Yongjun Shen , Chengyu He , Jing Shi","doi":"10.1016/j.aap.2025.108365","DOIUrl":"10.1016/j.aap.2025.108365","url":null,"abstract":"<div><div>Automated vehicles (AVs) face a critical need to adopt socially compatible behaviors and cooperate with human-driven vehicles (HVs) in heterogeneous traffic environments. However, existing lane-changing decision frameworks for AVs rarely account for the dynamic trust levels of HVs, thereby impeding the accurate prediction of HV behaviors due to inherent uncertainties. This study proposes a trust-aware, game-theoretic lane-changing decision framework for AVs. First, a trust-aware multi-vehicle coalition game for heterogeneous traffic environment is developed, which encompasses a fully cooperative game among AVs and a partially cooperative model for HVs based on each HV’s real-time trust level. Second, an online evaluation method to dynamically estimate HV’s trust during lane-change interactions is designed, guiding AVs to select appropriate cooperative maneuvers. Finally, social compatibility objectives are considered by minimizing disruption to surrounding vehicles and enhancing the predictability of AV behavior, ensuring the overall lane-changing strategy remains both human-friendly and context-adaptive. To test the proposed model, a human-in-the-loop experiment was conducted in a highway on-ramp scenario. Results showed that the AV adjusted its lane-changing strategy accordingly for human drivers of varying driving styles and trust levels. Moreover, the proposed model facilitated safe, efficient, and comfortable merging behavior for the AV. Ablation studies revealed that incorporating the trust mechanism yields higher average lane-change speeds than models without it, demonstrating improved efficiency while maintaining safety in heterogeneous traffic. This research contributes to enhance the interpretability of HV-AV interaction and promotes the design of more transparent and adaptive lane-changing strategies in automated driving systems.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"227 ","pages":"Article 108365"},"PeriodicalIF":6.2,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145808393","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}
Pub Date : 2025-12-24DOI: 10.1016/j.aap.2025.108366
Haowei Xu, Yang Li, Hongmao Qin, Yougang Bian, Shu Jiang, Wenwei Que, Jianzuo Man
The rapid evolution of Connected Autonomous Vehicles (CAVs) introduces a growing array of tasks and scenarios, significantly complicating the safe operation of Autonomous Driving Systems (ADS). These systems must navigate rare and unpredictable challenges, such as extreme weather, road irregularities, and unconventional behaviors from other traffic participants. Traditional safety specifications fall short of defining the safe operating conditions required for such complexities. Hence, the paper introduces the Operational Design Condition (ODC) framework that encompasses the status of drivers and passengers, vehicle status, ODD, and other conditions. This survey reviews current approaches to defining and describing the Operational Design Domain (ODD) and provides an in-depth look at the ODC framework for ADS. Moreover, this survey formulates a closed-loop ODC development process by summarizing existing research and leveraging established hazard analysis techniques to set ODC boundaries. In addition, it also discusses ODC monitoring and extension methods, emphasizing the necessity for ongoing research and development. It calls for advancements in standardization, the creation of a more dynamic and adaptable ODC design, the granular ODC mapping and management, the deeper integration of Vehicle-to-Everything (V2X) technologies, and the evolution of ODC testing and validation methods. These efforts are crucial to ensure the safe and effective operation of Autonomous Vehicles (AVs) across various scenarios.
{"title":"Ensuring safe operation of autonomous vehicle: A comprehensive survey of Operational Design Condition","authors":"Haowei Xu, Yang Li, Hongmao Qin, Yougang Bian, Shu Jiang, Wenwei Que, Jianzuo Man","doi":"10.1016/j.aap.2025.108366","DOIUrl":"10.1016/j.aap.2025.108366","url":null,"abstract":"<div><div>The rapid evolution of Connected Autonomous Vehicles (CAVs) introduces a growing array of tasks and scenarios, significantly complicating the safe operation of Autonomous Driving Systems (ADS). These systems must navigate rare and unpredictable challenges, such as extreme weather, road irregularities, and unconventional behaviors from other traffic participants. Traditional safety specifications fall short of defining the safe operating conditions required for such complexities. Hence, the paper introduces the Operational Design Condition (ODC) framework that encompasses the status of drivers and passengers, vehicle status, ODD, and other conditions. This survey reviews current approaches to defining and describing the Operational Design Domain (ODD) and provides an in-depth look at the ODC framework for ADS. Moreover, this survey formulates a closed-loop ODC development process by summarizing existing research and leveraging established hazard analysis techniques to set ODC boundaries. In addition, it also discusses ODC monitoring and extension methods, emphasizing the necessity for ongoing research and development. It calls for advancements in standardization, the creation of a more dynamic and adaptable ODC design, the granular ODC mapping and management, the deeper integration of Vehicle-to-Everything (V2X) technologies, and the evolution of ODC testing and validation methods. These efforts are crucial to ensure the safe and effective operation of Autonomous Vehicles (AVs) across various scenarios.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"227 ","pages":"Article 108366"},"PeriodicalIF":6.2,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145832028","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}
Pub Date : 2025-12-21DOI: 10.1016/j.aap.2025.108361
Seyed Ahmadreza Almasi , Jingzhen Yang
Traffic crashes often exhibit strong spatial dependence that is insufficiently captured by the Empirical Bayes (EB) method recommended in the Highway Safety Manual (HSM). This study proposes a Spatially Adaptive Empirical Bayes (SA-EB) framework that integrates advanced spatial models, including Geographically Weighted Poisson Regression (GWPR) and Multiscale Geographically Weighted Regression (MGWR), with Crash Modification Factors (CMFs) to enhance the prediction accuracy of expected crash frequencies across rural divided multilane highways (RDMHs), and generates CMF-adjusted, spatially weighted forecasts for both past and future conditions. The framework was calibrated and validated using crash and roadway data from 1071 km of highways in Hamadan Province, Iran, encompassing 2995 crashes recorded between 2017 and 2019. A dynamic overdispersion parameter (ranging from 0.3 to 0.6) was incorporated to capture spatial variability in crash dispersion. Results revealed substantial spatial heterogeneity in crash predictors: a 1 % increase in roadway slope corresponded to a 3.5-unit rise in crash frequency, while a 1 km/h increase in speed deviation led to approximately 4.5 additional crashes per segment. Implementation of SA-EB–guided geometric improvements reduced predicted crash frequencies by about 20 %, outperforming the conventional EB model. Overall, the SA-EB framework advances both the theoretical understanding and practical application of spatial safety modeling by providing transportation agencies with a data-driven and location-sensitive tool to identify high-risk segments and optimize Highway Safety Improvement Program (HSIP) investments across rural highway networks worldwide.
{"title":"A spatially adaptive empirical Bayes framework with dynamic dispersion parameters for enhanced crash frequency prediction across rural highway networks","authors":"Seyed Ahmadreza Almasi , Jingzhen Yang","doi":"10.1016/j.aap.2025.108361","DOIUrl":"10.1016/j.aap.2025.108361","url":null,"abstract":"<div><div>Traffic crashes often exhibit strong spatial dependence that is insufficiently captured by the Empirical Bayes (EB) method recommended in the Highway Safety Manual (HSM). This study proposes a Spatially Adaptive Empirical Bayes (SA-EB) framework that integrates advanced spatial models, including Geographically Weighted Poisson Regression (GWPR) and Multiscale Geographically Weighted Regression (MGWR), with Crash Modification Factors (CMFs) to enhance the prediction accuracy of expected crash frequencies across rural divided multilane highways (RDMHs), and generates CMF-adjusted, spatially weighted forecasts for both past and future conditions. The framework was calibrated and validated using crash and roadway data from 1071 km of highways in Hamadan Province, Iran, encompassing 2995 crashes recorded between 2017 and 2019. A dynamic overdispersion parameter (ranging from 0.3 to 0.6) was incorporated to capture spatial variability in crash dispersion. Results revealed substantial spatial heterogeneity in crash predictors: a 1 % increase in roadway slope corresponded to a 3.5-unit rise in crash frequency, while a 1 km/h increase in speed deviation led to approximately 4.5 additional crashes per segment. Implementation of SA-EB–guided geometric improvements reduced predicted crash frequencies by about 20 %, outperforming the conventional EB model. Overall, the SA-EB framework advances both the theoretical understanding and practical application of spatial safety modeling by providing transportation agencies with a data-driven and location-sensitive tool to identify high-risk segments and optimize Highway Safety Improvement Program (HSIP) investments across rural highway networks worldwide.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"226 ","pages":"Article 108361"},"PeriodicalIF":6.2,"publicationDate":"2025-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145809074","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}
Pub Date : 2025-12-20DOI: 10.1016/j.aap.2025.108357
Jianglin Lu , Chunjiao Dong , Xuedong Yan , Jingjun Li , Zhenzhen Lu
Electric vehicle (EV)-related risk and uncertainty pose critical challenges for urban traffic management. Fine-grained crash risk prediction at 1 km × 1 km and hour-of-day resolution remains difficult due to rapidly evolving, strongly spatiotemporally heterogeneous crash patterns. Crash risk research spans risk measurement, prediction modeling, and factor selection, with a move toward interpretable nonlinear hybrid methods, yet temporal dynamics and local heterogeneity remain insufficiently modeled. This study addresses these limitations by first constructing a Spatio-Temporal Adaptive Network Kernel Density Estimation (ST-ANKDE) method that combines network-constrained proximity, cyclic time weighting, severity weighting, and adaptive bandwidths, and then developing a Multiscale Geographically and Temporally Weighted Regression–Extreme Gradient Boosting (MGTWR-XGBoost) method to learn local heterogeneity and nonlinear effects. To capture the influence of preceding periods and adjacent grids, we introduce temporal and spatial weighted crash risk variables (T-AccRisk and S-AccRisk). These are analyzed alongside road-network density, built-environment variables, socioeconomic variables, and EV-specific infrastructure variables. An empirical case study on 14,818 EV crashes shows that ST-ANKDE effectively captures crash risk dynamics, with a mean value of 6.57, and reveals pronounced spatiotemporal heterogeneity. The results show that MGTWR-XGBoost, enhanced by S-AccRisk and T-AccRisk to capture spatiotemporal dependence, achieves MAE = 1.54 and RMSE = 2.06 and outperforms standalone machine learning and other hybrid methods; road-network density, built-environment features, population density, and EV infrastructure coefficients exhibit significant spatiotemporal heterogeneity. Moreover, SHapley Additive exPlanations (SHAP) further analyzes nonlinear effects. These findings enable grid-level early warning, priority targeting of high-risk periods/locations, and data-driven deployment of enforcement and infrastructure for EV safety management.
{"title":"Exploring spatiotemporal heterogeneity and nonlinear effects in electric vehicle crash risk prediction: A hybrid modeling approach","authors":"Jianglin Lu , Chunjiao Dong , Xuedong Yan , Jingjun Li , Zhenzhen Lu","doi":"10.1016/j.aap.2025.108357","DOIUrl":"10.1016/j.aap.2025.108357","url":null,"abstract":"<div><div>Electric vehicle (EV)-related risk and uncertainty pose critical challenges for urban traffic management. Fine-grained crash risk prediction at 1 km × 1 km and hour-of-day resolution remains difficult due to rapidly evolving, strongly spatiotemporally heterogeneous crash patterns. Crash risk research spans risk measurement, prediction modeling, and factor selection, with a move toward interpretable nonlinear hybrid methods, yet temporal dynamics and local heterogeneity remain insufficiently modeled. This study addresses these limitations by first constructing a Spatio-Temporal Adaptive Network Kernel Density Estimation (ST-ANKDE) method that combines network-constrained proximity, cyclic time weighting, severity weighting, and adaptive bandwidths, and then developing a Multiscale Geographically and Temporally Weighted Regression–Extreme Gradient Boosting (MGTWR-XGBoost) method to learn local heterogeneity and nonlinear effects. To capture the influence of preceding periods and adjacent grids, we introduce temporal and spatial weighted crash risk variables (T-AccRisk and S-AccRisk). These are analyzed alongside road-network density, built-environment variables, socioeconomic variables, and EV-specific infrastructure variables. An empirical case study on 14,818 EV crashes shows that ST-ANKDE effectively captures crash risk dynamics, with a mean value of 6.57, and reveals pronounced spatiotemporal heterogeneity. The results show that MGTWR-XGBoost, enhanced by S-AccRisk and T-AccRisk to capture spatiotemporal dependence, achieves MAE = 1.54 and RMSE = 2.06 and outperforms standalone machine learning and other hybrid methods; road-network density, built-environment features, population density, and EV infrastructure coefficients exhibit significant spatiotemporal heterogeneity. Moreover, SHapley Additive exPlanations (SHAP) further analyzes nonlinear effects. These findings enable grid-level early warning, priority targeting of high-risk periods/locations, and data-driven deployment of enforcement and infrastructure for EV safety management.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"226 ","pages":"Article 108357"},"PeriodicalIF":6.2,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145802869","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}
Pub Date : 2025-12-20DOI: 10.1016/j.aap.2025.108363
Shuchao Cao , Yuhang Wang , Guang Zeng , Xiaolian Li , Peng Wang
Smartphone use while walking has become increasingly prevalent, which significantly affects pedestrian safety and increases the risk of traffic accidents, especially in avoidance scenarios. Therefore, to reveal the avoidance mechanism and gait features under mobile phone use distractions, a series of walking experiments under varying distraction conditions including no phone use, text-browsing, message-sending, and video-watching were conducted. Multiple types of data such as physiological signals including the electrocardiogram (ECG) and electrodermal activity (EDA), and high-precision motion trajectories were synchronously obtained. To characterize the entire avoidance process, three critical positions including the start point, maximum point and end point are identified in this paper. The results show that distracted pedestrians initiate evasive maneuvers when they are close to obstacles, whereas non-distracted pedestrians prefer to avoid obstacles in advance. Moreover, the right-side avoidance strategy is adopted by most pedestrians under different distraction levels. Physiological signal indicates that distracted pedestrians experience elevated levels of psychological stress. Besides, the walking speed is significantly decreased in distracted circumstances. The gait analysis demonstrates that distracted pedestrians exhibit shorter step length, larger step width and longer step time during the avoidance phase. Furthermore, the real-time messaging task consumes large cognitive resources of pedestrians, which has the most pronounced impact on pedestrian safety. The study elucidates how the use of smartphones affects the avoidance mechanism of pedestrians, which can help develop intervention measures and safety strategies to reduce the risk of distracted walking in public places.
{"title":"Avoidance behavior and movement characteristics of pedestrians under mobile phone distraction","authors":"Shuchao Cao , Yuhang Wang , Guang Zeng , Xiaolian Li , Peng Wang","doi":"10.1016/j.aap.2025.108363","DOIUrl":"10.1016/j.aap.2025.108363","url":null,"abstract":"<div><div>Smartphone use while walking has become increasingly prevalent, which significantly affects pedestrian safety and increases the risk of traffic accidents, especially in avoidance scenarios. Therefore, to reveal the avoidance mechanism and gait features under mobile phone use distractions, a series of walking experiments under varying distraction conditions including no phone use, text-browsing, message-sending, and video-watching were conducted. Multiple types of data such as physiological signals including the electrocardiogram (ECG) and electrodermal activity (EDA), and high-precision motion trajectories were synchronously obtained. To characterize the entire avoidance process, three critical positions including the start point, maximum point and end point are identified in this paper. The results show that distracted pedestrians initiate evasive maneuvers when they are close to obstacles, whereas non-distracted pedestrians prefer to avoid obstacles in advance. Moreover, the right-side avoidance strategy is adopted by most pedestrians under different distraction levels. Physiological signal indicates that distracted pedestrians experience elevated levels of psychological stress. Besides, the walking speed is significantly decreased in distracted circumstances. The gait analysis demonstrates that distracted pedestrians exhibit shorter step length, larger step width and longer step time during the avoidance phase. Furthermore, the real-time messaging task consumes large cognitive resources of pedestrians, which has the most pronounced impact on pedestrian safety. The study elucidates how the use of smartphones affects the avoidance mechanism of pedestrians, which can help develop intervention measures and safety strategies to reduce the risk of distracted walking in public places.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"226 ","pages":"Article 108363"},"PeriodicalIF":6.2,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145788093","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号