| 4 | 0 | 23 |
| 下载次数 | 被引频次 | 阅读次数 |
为探究山区公路小客车借道超车冲突风险特性，选取典型山区公路的直道和弯道路段为研究对象，首先利用无人机采集小客车借道超车视频数据，采用二维碰撞时间(Two-Dimensional Time to Collision，TTC2D)指标评估超车全过程冲突风险；其次，基于11类机器学习算法构建小客车超车冲突风险识别模型并进行对比分析；最后，采用沙普利加性解释(SHapley Additive exPlanations，SHAP)算法揭示各特征对小客车借道超车冲突风险的影响机制。结果表明：直道路段超车冲突风险呈现“高速显性”特征，弯道路段呈现“低速隐性”特征；直道路段冲突风险识别模型整体性能优于弯道路段，其中类别提升(Categorical Boosting，CatBoost)模型在直道路段的准确率达96.8%，轻量梯度提升机(Light Gradient Boosting Machine，LightGBM)模型在弯道路段的准确率达93.0%；两车之间的速度差和车头间距是影响小客车超车冲突风险识别的关键因素，且被超车辆类型与超车冲突严重程度呈负相关。
Abstract：To investigate the conflict risk characteristics associated with passenger vehicles changing lanes to overtake on mountainous two-lane highways， both straight and curved sections of a typical mountain road were selected as study subjects. Initially， high-precision trajectory data for passenger vehicles were collected using Unmanned Aerial Vehicles (UAVs)， resulting in the extraction of 392 valid overtaking events. To comprehensively assess the overtaking conflict risk on mountainous roads， an improved Two-Dimensional Time to Collision (TTC2D) metric was employed， establishing a threshold of 2.01 seconds (the 15th percentile) to identify severe conflicts. Subsequently， twelve feature variables related to vehicle kinematics and vehicle type were selected， and passenger vehicle overtaking conflict risk identification models were constructed based on eleven machine learning algorithms， all rigorously evaluated through cross-validation. Finally， the SHapley Additive exPlanations (SHAP) algorithm was utilized to interpret the optimal models， revealing the intrinsic mechanisms and interaction effects of key features on overtaking risks. The results indicate: (1) Overtaking behaviors display distinct spatial variations; straight sections are characterized by higher speeds and larger gaps， presenting a “high-speed explicit” risk， while curved sections involve shorter gaps and greater acceleration fluctuations， exhibiting a “low-speed implicit” risk. (2) The conflict risk identification model for straight road sections outperforms that for curved road sections. Notably， the Categorical Boosting (CatBoost) model achieves the highest accuracy of 96.8% and an F-measure of 0.957 on straight sections， whereas the Light Gradient Boosting Machine (LightGBM) model attains an accuracy of 93.0% and an F-measure of 0.890 on curved sections. (3) SHAP analysis reveals that the headway distance and speed difference between the two vehicles are the most critical determinants of conflict risk. Risk severity exhibits a negative correlation with headway distance and a positive correlation with speed difference. Moreover， the type of overtaken vehicle inversely impacts risk severity， and a larger speed difference during the overtaking separation phase significantly accelerates the dissipation of risk.
[1] Mwesige G， Farah H， Koutsopoulos H N. Risk appraisal of passing zones on two-lane rural highways and policy applications[J]. Accident Analysis & Prevention， 2016， 90: 1–12.
[2] Ahmed M M， Abdel-Aty M A. The viability of using automatic vehicle identification data for real-time crash prediction[J]. IEEE Transactions on Intelligent Transportation Systems， 2011， 13(2): 459–468.
[3] 朱顺应， 蒋若曦， 王红， 等. 机动车交通冲突技术研究综述[J]. 中国公路学报， 2020， 33(2): 15–33.
[4] Hayward J C. Near-miss determination through use of a scale of danger[J]. Highway Research Record， 1972， 384: 24–34.
[5] Branzi V， Meocci M， Domenichini L， et al. A combined simulation approach to evaluate overtaking behaviour on two‐lane two‐way rural roads[J]. Journal of Advanced Transportation， 2021， 2021(1): 9973138.
[6] 刘照霞， 付锐， 牛世峰. 基于极值理论与智能网联信息的超车风险评估[J]. 吉林大学学报（工学版）， 2025， 55(3): 925–937.
[7] Mahmud S M S， Ferreira L， Hoque M S， et al. Overtaking risk modeling in two-lane two-way highway with heterogeneous traffic environment of a low-income country using naturalistic driving dataset[J]. Journal of Safety Research， 2022， 80: 380–390.
[8] Hou Jia， List G F， Guo Xiuchen. New algorithms for computing the time-to-collision in freeway traffic simulation models[J]. Computational Intelligence and Neuroscience， 2014， 2014(1): 761047.
[9] Guo Hongyu， Xie Kun， Keyvan-Ekbatani M. Modeling driver’s evasive behavior during safety–critical lane changes: two-dimensional time-to-collision and deep reinforcement learning[J]. Accident Analysis & Prevention， 2023， 186: 107063.
[10] Pan Jie， Shen Yongjun. Assessing driving risk at the second phase of overtaking on two-lane highways for young novice drivers based on driving simulation[J]. International Journal of Environmental Research and Public Health， 2022， 19(5): 2691.
[11] 卢梦媛， 戢晓峰， 徐迎豪， 等. 基于生存分析的山区公路借道超车风险暴露时间模型[J/OL]. 吉林大学学报（工学版）: 1–10[2025–09–08]. https://link.cnki.net/doi/10.13229/j.cnki.jdxbgxb.20250170.
[12] Oh T， Kang H C， Li Zhibin. Exploring safe overtaking behavior on two‐lane two‐way road using multiagent driving simulators and traffic simulation[J]. Journal of Advanced Transportation， 2024， 2024(1): 8242764.
[13] Shi Xiupeng， Wong Y D， Li M Z F， et al. A feature learning approach based on XGBoost for driving assessment and risk prediction[J]. Accident Analysis & Prevention， 2019， 129: 170–179.
[14] Fadhil S A， Al-Bayatti A H. Developing a new driver assistance system for overtaking on two-lane roads using predictive models[J]. Periodica Polytechnica Transportation Engineering， 2022， 50(4): 400–413.
[15] Lin Lei， Wang Qian， Sadek A W. A novel variable selection method based on frequent pattern tree for real-time traffic accident risk prediction[J]. Transportation Research Part C: Emerging Technologies， 2015， 55: 444–459.
[16] Wu Dan， Lee J J， Li Ye， et al. A surrogate model-based approach for adaptive selection of the optimal traffic conflict prediction model[J]. Accident Analysis & Prevention， 2024， 207: 107738.
[17] Fisher A， Rudin C， Dominici F. All models are wrong， but many are useful: Learning a variable's importance by studying an entire class of prediction models simultaneously[J]. Journal of Machine Learning Research， 2019， 20(177): 1–81.
[18] Zhang Xiaolong， Zhao Xiaohua， Bian Yang， et al. Interactive effects analysis of road， traffic， and weather characteristics on shared e-bike speeding risk: a data-driven approach[J]. Accident Analysis & Prevention， 2024， 207: 107755.
[19] Ribeiro M T， Singh S， Guestrin C. “Why should I trust you?” Explaining the predictions of any classifier[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery， 2016: 1135–1144.
[20] Lundberg S M， Lee S I. A unified approach to interpreting model predictions[J]. Advances in Neural Information Processing Systems， 2017， 30: 4765–4774.
[21] 陈英， 朱加勋， 田毅， 等. 基于K–Means–XGBoost–SHAP集成化框架的交通事故严重程度致因分析[J]. 中国公路学报， 2025， 38(5): 209–223.
[22] 温惠英， 何梓琦， 李秋灵， 等. 高速公路货车换道冲突预测及其影响因素分析[J]. 吉林大学学报(工学版)， 2024， 54(10): 2827–2836.
[23] Ahmed M T， Wang Ling， Ma Wanjing， et al. Real-time sideswipe conflict risk prediction on the expressway[J]. Journal of Transportation Safety & Security， 2024， 16(12): 1550–1576.
[24] Sun Zhiyuan， Wang Duo， Gu Xin， et al. A hybrid approach of random forest and random parameters logit model of injury severity modeling of vulnerable road users involved crashes[J]. Accident Analysis & Prevention， 2023， 192: 107235.
[25] 戢晓峰， 薛唯， 卢梦媛， 等. 不平衡数据驱动的山区公路货车移动遮断险态跟驰行为识别模型[J]. 安全与环境学报， 2024， 24(8): 3015–3027.
基本信息:
DOI：10.13637/j.issn.1009-6094.2025.1663
中图分类号:U492.8
引用信息:
[1]戢晓峰,庞兴任,郭雅诗,等.基于机器学习的山区公路小客车超车冲突风险识别[J].安全与环境学报().DOI:10.13637/j.issn.1009-6094.2025.1663.
基金信息:
国家自然科学基金项目（72561013）; 云南省研究生导师团队建设项目（2024）; 云南省外国专家项目（202605AP120006）
2026-08-18
2026-08-18
2026-08-18