Accident; analysis and prevention最新文献

Risk analysis of pedestrian crosswalks in airport drop-off zones based on integrated VISSIM–SSAM model 基于VISSIM-SSAM综合模型的机场落客区人行横道风险分析

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-21 DOI: 10.1016/j.aap.2025.108279

Meng-xin Qin , Tie-qiao Tang , Shang-wu Wen , Xiao-ting Yuan

{"title":"Risk analysis of pedestrian crosswalks in airport drop-off zones based on integrated VISSIM–SSAM model","authors":"Meng-xin Qin , Tie-qiao Tang , Shang-wu Wen , Xiao-ting Yuan","doi":"10.1016/j.aap.2025.108279","DOIUrl":"10.1016/j.aap.2025.108279","url":null,"abstract":"<div><div>Airport drop-off zones present a high-risk traffic environment due to limited space and frequent pedestrian–vehicle interactions. To address these challenges, this study develops an integrated simulation framework combining VISSIM and the SSAM to reconstruct microscopic pedestrian and vehicle behaviors, calibrate conflict detection thresholds, and evaluate safety performance in complex terminal settings. The framework enables the quantification of both the frequency and severity of pedestrian–vehicle conflicts under varying spatial and behavioral conditions. A set of scenario-specific TTC and PET thresholds is derived to enhance the accuracy of conflict identification. Based on these insights, two safety strategies are proposed: pedestrian signal control, which facilitates temporal separation, and spatial no-stopping zones, which provide physical buffers to reduce interaction intensity without compromising operational flow. Simulation experiments confirm the effectiveness of both interventions across key safety metrics. The proposed framework provides a transferable methodological basis for assessing pedestrian safety and testing mitigation strategies in high-density curbside environments such as airports and multimodal urban terminals.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"224 ","pages":"Article 108279"},"PeriodicalIF":6.2,"publicationDate":"2025-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145335021","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

Driver attention in urban intersections when crossing paths with cyclists 与骑自行车的人穿过道路时，驾驶员在城市十字路口的注意事项。

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-15 DOI: 10.1016/j.aap.2025.108276

Christer Ahlström , Katja Kircher , Fredrik Johansson , Anders Andersson , Johan Olstam

{"title":"Driver attention in urban intersections when crossing paths with cyclists","authors":"Christer Ahlström , Katja Kircher , Fredrik Johansson , Anders Andersson , Johan Olstam","doi":"10.1016/j.aap.2025.108276","DOIUrl":"10.1016/j.aap.2025.108276","url":null,"abstract":"<div><div>Conflicts arise when motor vehicles cross cycle paths during turns, often due to drivers neglecting over-the-shoulder glances. This paper examines the visual attention of car drivers at urban unsignalized intersections, considering turning direction, cross traffic, and cyclist approaching from behind.</div><div>A study with 44 participants was conducted using a fixed-base driving simulator equipped with an eXtended Reality (XR)-based visual system, which provided 360° immersion and enabled eye tracking for head movements and over-the-shoulder glances. Drivers were recruited based on urban cycling experience (inexperienced/experienced) and driving style (cautious/assertive).</div><div>Results showed that in 47.8 % of intersection approaches, participants failed to adequately check for cyclists from behind. Driver characteristics did not consistently reveal traits associated with this neglect. Although cautious drivers with urban cycling experience made fewer mistakes, all groups were generally poor at checking for cyclists from behind. Post-drive questionnaire results on rule knowledge showed that, across participants, only half of the requirements to check for cyclists approaching from behind were indicated correctly.</div><div>Drivers are more consistent in checking for cross-traffic, suggesting the neglect to check for cyclists is systemic rather than individual. The lack of awareness among drivers about their obligation to check for cyclists can be attributed to the less obvious nature of this traffic stream, as it comes from behind and lacks signs or warnings. Checking over the shoulder is physically more effortful, and cyclists may often pre-empt collisions, reinforcing drivers’ mental models to turn without checking. To improve safety, changes counteracting systemic biases against cyclists are needed.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108276"},"PeriodicalIF":6.2,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145306810","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

A unified experimental framework for estimating collision rates and occupant injury severity across different levels of driving automation 一个统一的实验框架，用于估计不同驾驶自动化水平下的碰撞率和乘员伤害严重程度。

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-14 DOI: 10.1016/j.aap.2025.108273

Jiajie Shen , Detong Qin , Zijian He , Hanshuo Wang , Xiangdong Ji , Yajun Zhang , Qing Zhou , Bingbing Nie

{"title":"A unified experimental framework for estimating collision rates and occupant injury severity across different levels of driving automation","authors":"Jiajie Shen , Detong Qin , Zijian He , Hanshuo Wang , Xiangdong Ji , Yajun Zhang , Qing Zhou , Bingbing Nie","doi":"10.1016/j.aap.2025.108273","DOIUrl":"10.1016/j.aap.2025.108273","url":null,"abstract":"<div><div>Safety performance of autonomous vehicles is crucial for their adoption and market acceptance. However, the lack and imbalance of real-world accident data have prevented a rigorous verification of the safety performance of autonomous vehicles across manufacturers and automation levels in safety–critical scenarios. This study aimed to bridge this gap by establishing a unified experimental framework that enables fair comparisons of vehicle safety across automation levels under comparable road scenarios, similar urgency levels and consistent evaluation metrics. Vehicles at different automation levels were evaluated in simulated highway scenarios, with hazard-triggering algorithms generating safety–critical events and occupant injury severity estimated under specific collision conditions. The results show that in our designed safety–critical scenarios, vehicles operating at automation levels 2, 3, and 4 have collision rates of 24.3%, 21.4%, and 14.1%, respectively, with corresponding probabilities of severe occupant injuries of 11.1%, 21.6%, and 9.2%. Among the findings, Level 3 autonomous vehicles can reduce collision rates but may result in more severe occupant injuries compared to Level 2 vehicles, thus leading to a comparable unified safety benefit. Level 4 autonomous vehicles show improved safety benefits over Level 2, primarily due to lower collision rates, while the severity of occupant injuries remains similar once a collision occurs. This study offers a unified experimental framework to robustly evaluate safety performance of autonomous vehicles in safety–critical scenarios, and support large-scale deployment of autonomous vehicles in the future.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108273"},"PeriodicalIF":6.2,"publicationDate":"2025-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145297847","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

An online adaptive learning approach for predicting multi-type traffic participants’ microscopic behavior 多类型交通参与者微观行为预测的在线自适应学习方法。

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-14 DOI: 10.1016/j.aap.2025.108250

Meng Li , Tao Chen , Hanggai Chen , Yicheng Zhang , Yan Zhang

{"title":"An online adaptive learning approach for predicting multi-type traffic participants’ microscopic behavior","authors":"Meng Li , Tao Chen , Hanggai Chen , Yicheng Zhang , Yan Zhang","doi":"10.1016/j.aap.2025.108250","DOIUrl":"10.1016/j.aap.2025.108250","url":null,"abstract":"<div><div>Predicting multi-type traffic participant behavior in dynamic transportation hubs remains challenging. Existing deep-learning frameworks lack robust online learning and autonomous error-correction, limiting their reliability under evolving conditions. Most state-of-the-art models exhibit poor cross-scenario generalization, requiring intensive retraining for new settings. Prediction errors also propagate over time due to absent real-time correction.</div><div>To address these limitations, this paper introduces an adaptive framework integrating online learning with probabilistic error correction. Key innovations include: (1) an Extended Kalman Filter-based module for real-time trajectory correction; (2) a hierarchical graph encoder enabling transfer learning with minimal retraining; and (3) unified node–edge-plane modeling for multimodal context fusion. Validated using real-world hub data and redesigned benchmark experiments, our method significantly outperforms existing approaches in unseen scenarios, positioning it as a promising solution for real-time behavioral prediction in modern traffic systems.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108250"},"PeriodicalIF":6.2,"publicationDate":"2025-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145297868","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

An integrated active–passive safety strategy for automobiles based on driver state recognition and injury risk prediction 基于驾驶员状态识别和伤害风险预测的汽车主被动集成安全策略

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-11 DOI: 10.1016/j.aap.2025.108271

Jing Huang, Xinyu Huang

{"title":"An integrated active–passive safety strategy for automobiles based on driver state recognition and injury risk prediction","authors":"Jing Huang, Xinyu Huang","doi":"10.1016/j.aap.2025.108271","DOIUrl":"10.1016/j.aap.2025.108271","url":null,"abstract":"<div><div>This study proposes an integrated active–passive safety strategy based on driver state recognition and injury risk prediction, aiming to enhance vehicle safety by dynamically coordinating the operation of the autonomous emergency braking (AEB) system and occupant restraint systems. First, injury prediction and driver state recognition models were developed using machine learning and deep learning techniques, respectively, based on real-world traffic accident data and physiological signals. These predictive outcomes were then incorporated into a fuzzy control algorithm to optimize the AEB system, enabling it to dynamically adjust activation timing according to varying driver states and potential injury risks. Experimental results demonstrate that the optimized AEB system effectively adapts braking initiation based on driver responsiveness and injury severity, significantly improving collision avoidance performance. Furthermore, by integrating passive safety mechanisms, the control parameters of seatbelts and airbags were optimized, resulting in a 30.60% reduction in the head injury criterion (HIC) and a 22.44% decrease in the weighted injury criterion (WIC). This study provides novel insights and methodological approaches for the integrated optimization of intelligent vehicle safety systems, offering both theoretical and practical value.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108271"},"PeriodicalIF":6.2,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145263636","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

A physics-informed attention model for integrated driving risk assessment 综合驾驶风险评估的物理信息注意模型

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-09 DOI: 10.1016/j.aap.2025.108266

Tianle Lu , Gaoyuan Kuang , Dongyang Xu , Shaobing Xu , Yiran Luo , Qingfan Wang , Shi Shang , Qing Zhou , Bingbing Nie

{"title":"A physics-informed attention model for integrated driving risk assessment","authors":"Tianle Lu , Gaoyuan Kuang , Dongyang Xu , Shaobing Xu , Yiran Luo , Qingfan Wang , Shi Shang , Qing Zhou , Bingbing Nie","doi":"10.1016/j.aap.2025.108266","DOIUrl":"10.1016/j.aap.2025.108266","url":null,"abstract":"<div><div>Accurate and stable quantification of driving risk is critical for enhancing the safety performance of autonomous vehicles (AVs). Such quantification not only effectively prevents traffic accidents but also, in scenarios where collisions are unavoidable, enables timely and targeted occupant protection by accurately assessing potential severity. This study proposes a physics-informed integrated risk assessment model (PIRAM), which fuses collision probability and severity predictions into a unified integrated driving risk (IDR) metric. First, an integrated driving risk prediction dataset (IDRPD) was constructed using driving simulator experiments. Multiple hazardous driving scenarios were simulated within CARLA’s virtual environment to collect data on drivers’ safety–critical decision-making behavior. Then, a neural network model combining data-driven methods with physics-based constraints was developed. Specifically, the model employs an attention mechanism to capture spatiotemporal dependencies in vehicle trajectory and map information, and integrates a dynamic bicycle model as a physical constraint to guide predictions in accordance with fundamental physical laws, thereby significantly improving both prediction stability and accuracy. Experimental results and case studies conducted on the IDRPD demonstrate that PIRAM outperforms several baseline models, increasing prediction accuracy for collision probability and severity by 7.9 % and 3.2 %, respectively, and enhancing prediction stability by 10.3 % and 5.9 %, respectively. Furthermore, PIRAM enables earlier risk warnings by an average of 0.5 s. These advancements offer a reliable quantitative basis for occupant protection strategies in AVs and underscore PIRAM’s substantial potential to improve safety in autonomous driving applications.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108266"},"PeriodicalIF":6.2,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145263638","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

Zone-specific real-time traffic conflict risk modeling for freeway tunnels: a CrossTabNet approach 高速公路隧道特定区域的实时交通冲突风险建模：一种crossstabnet方法

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-09 DOI: 10.1016/j.aap.2025.108274

Jieling Jin , Jipu Li , Shan Tian , Qing Ye

{"title":"Zone-specific real-time traffic conflict risk modeling for freeway tunnels: a CrossTabNet approach","authors":"Jieling Jin , Jipu Li , Shan Tian , Qing Ye","doi":"10.1016/j.aap.2025.108274","DOIUrl":"10.1016/j.aap.2025.108274","url":null,"abstract":"<div><div>This study proposes a zone-specific, real-time traffic conflict risk modeling framework specifically designed for freeway tunnels. The framework integrates traffic conflict analysis, refined tunnel segmentation, and interpretable deep learning to address limitations in traditional collision data. Vehicle trajectory data are utilized to derive surrogate safety measures based on traffic conflicts. A refined five-zone tunnel classification—pre-entrance, entrance, interior, exit, and post-exit—is adopted by extending existing zoning frameworks. This facilitates more precise spatial attribution of risk patterns in real-time conflict analysis. To model complex, interdependent risk factors, a CrossTabNet architecture is developed. This innovative structure combines a feature interaction layer with a TabNet encoder, enabling the model to capture high-order nonlinear relationships between traffic variables while maintaining interpretability through sparse attention mechanisms. The proposed model demonstrates superior predictive performance compared to established machine learning and deep learning methods. Notably, zone-specific models significantly outperform a global model trained on all data, emphasizing the necessity of localized modeling for effective tunnel safety assessment. Global sensitivity analysis reveals that the standard deviation of upstream traffic flow consistently contributes positively to conflict risk across all zones, highlighting the critical role of flow variability. Other significant features vary by tunnel segment, reflecting distinct local dynamics. These findings provide valuable insights for implementing adaptive, zone-targeted traffic safety interventions in freeway tunnel environments.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108274"},"PeriodicalIF":6.2,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145263637","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

Human-machine cooperation strategies in game-theoretic scenarios within mixed traffic: A simulator study on driving styles 混合交通博弈论情景下的人机合作策略：驾驶风格模拟器研究。

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-08 DOI: 10.1016/j.aap.2025.108254

Yutong Zhang, Shiqi Wu, Danneil Mubbala, Na Du

{"title":"Human-machine cooperation strategies in game-theoretic scenarios within mixed traffic: A simulator study on driving styles","authors":"Yutong Zhang, Shiqi Wu, Danneil Mubbala, Na Du","doi":"10.1016/j.aap.2025.108254","DOIUrl":"10.1016/j.aap.2025.108254","url":null,"abstract":"<div><div>As human-driven vehicles (HVs) and automated vehicles (AVs) increasingly share roadways, understanding their interactions is essential for traffic safety and efficiency. This driving simulator study using game-theoretic scenarios investigates how AV and human driving styles influence decision-making in mixed traffic. Our findings show that AV driving styles had a significantly stronger impact in parallel scenarios. AV driving styles had a stronger impact in parallel interactions: aggressive AVs led to passive yet riskier human maneuvers, with shorter time-to-collision and higher lateral deceleration. Regarding different drivers, conservative drivers showed greater maximum counter-steering rate and lateral deceleration to adjust their intentions and avoid risks. Scenario types significantly influenced drivers’ strategies. Drivers showed a higher tendency to defect in head-on scenarios by asserting their right of way. Trajectory clustering reflected differences in proactive versus reactive adjustments in specific scenarios. These findings highlight the need for adaptive AV strategies to foster safe and cooperative mixed-traffic interactions.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108254"},"PeriodicalIF":6.2,"publicationDate":"2025-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145257001","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

Generation of naturalistic and critical boundary scenarios: A bi-level adaptive deep reinforcement learning method 自然和关键边界场景的生成：一种双层自适应深度强化学习方法。

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-07 DOI: 10.1016/j.aap.2025.108269

Junjie Zhou , Lin Wang , Qiang Meng , Xiaofan Wang

{"title":"Generation of naturalistic and critical boundary scenarios: A bi-level adaptive deep reinforcement learning method","authors":"Junjie Zhou , Lin Wang , Qiang Meng , Xiaofan Wang","doi":"10.1016/j.aap.2025.108269","DOIUrl":"10.1016/j.aap.2025.108269","url":null,"abstract":"<div><div>The complexities of real-world driving environments, coupled with a limited availability of naturalistic and critical test scenarios, have long hindered unbiased and effective comprehensive performance evaluations. In this work, we propose a bi-level adaptive deep reinforcement learning (BADRL) framework designed to generate realistic and diverse critical boundary scenarios. The method involves training AI-driven background agents to impartially assess the overall performance of autonomous vehicles. By leveraging naturalistic driving data, these agents acquire realistic driving behaviors via a neural model that encapsulates naturalistic driving patterns. To enrich the authenticity and diversity of the test scenarios, a wide array of traffic participants, encompassing vehicles, pedestrians, and bicycles, are meticulously modeled and portrayed to engage in intricate interactive behaviors with the tested autonomous vehicles. To address the challenges of high-dimensional environments, we introduce a scenario complexity model that assesses relative complexity in real time. This model enables the upper-level neural network in BADRL to dynamically escalate scenario complexity, with the resulting scenarios subsequently processed by lower-level models to optimize the actions of primary traffic participants. The BADRL method enables online real-time generation of naturalistic and critical boundary scenarios. Extensive simulation experiments validate the effectiveness of the BADRL approach in diverse driving environments, with results indicating an improvement in the efficiency of critical boundary scenario generation by approximately 15.89 % compared to state-of-the-art methods.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108269"},"PeriodicalIF":6.2,"publicationDate":"2025-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145249241","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

HAD-Gen: Human-like and diverse driving behavior modeling for controllable scenario generation HAD-Gen：类似人类的多样驾驶行为建模，用于可控场景生成。

IF 6.2 1区工程技术

Accident; analysis and prevention Pub Date : 2025-10-06 DOI: 10.1016/j.aap.2025.108270

Cheng Wang , Lingxin Kong , Massimiliano Tamborski , Stefano V. Albrecht

{"title":"HAD-Gen: Human-like and diverse driving behavior modeling for controllable scenario generation","authors":"Cheng Wang , Lingxin Kong , Massimiliano Tamborski , Stefano V. Albrecht","doi":"10.1016/j.aap.2025.108270","DOIUrl":"10.1016/j.aap.2025.108270","url":null,"abstract":"<div><div>Simulation-based testing has emerged as an essential tool for verifying and validating autonomous vehicles (AVs). However, contemporary methodologies, such as deterministic and imitation learning-based driver models, struggle to capture the variability of human-like driving behavior. Given these challenges, we propose HAD-Gen, a general framework for realistic traffic scenario generation that simulates diverse human-like driving behaviors. The framework first clusters the vehicle trajectory data into different driving styles according to safety features. It then employs maximum entropy inverse reinforcement learning on each of the clusters to learn the reward function corresponding to each driving style. Using these reward functions, the method integrates offline reinforcement learning pre-training and multi-agent reinforcement learning algorithms to obtain general and robust driving policies. Multi-perspective simulation results in highway scenarios show that our proposed scenario generation framework can generate diverse, human-like driving behaviors with strong generalization capability. The proposed framework achieved a 90.96% goal-reaching rate, an off-road rate of 2.08%, and a collision rate of 6.91% in new unseen driving scenarios, outperforming prior approaches by over 20% in goal-reaching performance. The source code is released at <span><span>https://github.com/RoboSafe-Lab/HAD-Gen</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":6926,"journal":{"name":"Accident; analysis and prevention","volume":"223 ","pages":"Article 108270"},"PeriodicalIF":6.2,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145243429","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