基于混合模型的自动驾驶车辆事故严重程度影响因素分析

阎莹; 王玉莹; 周旋; 袁华智; 王文璇

doi:10.19818/j.cnki.1671-1637.2025.01.013

基于混合模型的自动驾驶车辆事故严重程度影响因素分析

doi: 10.19818/j.cnki.1671-1637.2025.01.013

1.
长安大学运输工程学院，陕西西安 710064
2.
兰州理工大学土木工程学院，甘肃兰州 730050

基金项目:

国家自然科学基金项目 52362050

陕西省博士后科研项目 2023BSHEDZZ216

秦创原引用高层次创新创业人才项目 QCYRCXM2023-110

中国博士后科学基金项目 2024M752739

陕西省重点研发计划 2024GX-ZDCYL-02-14

详细信息

作者简介:
阎莹(1981-)，女，陕西西安人，长安大学教授，工学博士，从事道路交通安全、驾驶行为评估和事故预测建模研究

通讯作者:
王文璇(1992-)，女，山西临汾人，长安大学讲师，工学博士

中图分类号: U491.31
计量
- 文章访问数: 29
- HTML全文浏览量: 58
- PDF下载量: 16
- 被引次数: 0
出版历程
- 收稿日期: 2024-06-27
- 刊出日期: 2025-02-25

Analysis of autonomous vehicle accident severity factors based on hybrid model

1.
School of Transportation Engineering, Chang'an University, Xi'an 710064, Shaanxi, China
2.
School of Civil Engineering, Lanzhou University of Technology, Lanzhou 730050, Gansu, China

Funds:

National Natural Science Foundation of China 52362050

Postdoctoral Scientific Research Project of Shaanxi Province 2023BSHEDZZ216

Qinchuangyuan High-level Innovation and Entrepreneurship Talent Project QCYRCXM2023-110

China Postdoctoral Science Foundation 2024M752739

Key Research and Development Program of Shaanxi Province 2024GX-ZDCYL-02-14

More Information

Corresponding author: WANG Wen-xuan(1992-), female, assistant professor, PhD, wangwenxuan123@chd.edu.cn

Article Text (Baidu Translation)

摘要

摘要: 为深入探究自动驾驶车辆事故中各因素之间的交互作用对事故严重程度的影响，提出考虑因素交互作用的可解释机器学习与树形增强朴素(TAN)贝叶斯网络混合模型，基于汇总后的加利福尼亚州2014~2023年自动驾驶车辆事故报告公开数据，确定了13个影响因素为自变量，并将因变量(事故严重程度)划分为无伤害、轻微和中重度事故3类；分别采用极端梯度提升、随机森林、K最近邻、支持向量机4种机器学习算法，建立了自动驾驶车辆事故严重程度分类预测模型，选择性能最优的分类预测模型作为基础模型，结合基础模型与沙普利加和解释(SHAP)算法，筛选关键因素并建立TAN贝叶斯网络，以提取9组交互因素对，并结合SHAP算法分析了单因素和交互效果显著的3组交互因素对。研究结果表明：碰撞对象的碰撞部位、车辆驾驶模式和主车碰撞前的运动状态分别对无伤害、轻微和中重度事故严重程度有重要影响；2020年后自动驾驶车辆尾部被碰撞时，无伤害事故中SHAP交互值均为正值，中重度事故中SHAP交互值均为负值；在自动驾驶车辆碰撞部位为正面与碰撞对象的碰撞部位为尾部两因素交互情况下，中重度事故中SHAP交互值仅有16.7%为正，在自动驾驶车辆碰撞部位为正面且碰撞对象碰撞部位也为正面两因素交互情况下，轻微事故中89.1%的SHAP交互值为正。可见，自动驾驶车辆能够及时改变运行轨迹，避免车辆正面碰撞，从而降低车辆碰撞的严重程度。
- 交通信息工程 /
- 事故严重程度 /
- 沙普利加和解释 /
- 树形增强朴素贝叶斯网络 /
- 自动驾驶车辆 /
- 因素交互作用
Abstract: To deeply investigate the impacts of interactions between various factors in autonomous vehicle accidents on the accident severity, a hybrid model of interpretable machine learning and tree-augmented naive (TAN) Bayes network was proposed considering the factor interactions. According to the summarized public dataset of autonomous vehicle accident reports in California from 2014 to 2023, 13 factors were determined to be the independent variables. The dependent variable (accident severity) was classified into three classes: no injury, minor, and moderate-to-major accidents. Four machine learning algorithms, namely, extreme gradient boosting, random forests, K-nearest neighbors, and support vector machines, were applied to establish a classification prediction model for autonomous vehicle accident severity. The classification prediction model with the optimal performance was selected as the base model. Then the base model was combined with the Shapley additive explanations (SHAP) algorithm to select the key factors and the TAN Bayes network was established to extract nine interaction factor pairs. Finally, the SHAP algorithm was employed to analyze the single factors as well as three groups of interaction factor pairs with significant interaction effects. Research results show that for no injury, minor and moderate-to-major accidents, the accident area of accident partner, vehicle driving mode, and subject vehicle pre-accident motion state have a significant impact on accident severity, respectively. After 2020, when the rear part of the autonomous vehicle is collided, the SHAP interaction values for no injury accidents are all positive, while those for moderate-to-major accidents are all negative. When the autonomous vehicle's accident area as the front and the object vehicle accident area as the rear, only 16.7% of the SHAP interaction values are positive in moderate-to-major accidents. If both the accident areas of the autonomous vehicle and object vehicle are the front, 89.1% of the SHAP interaction values are positive in minor accidents. Therefore, autonomous vehicles can timely change the running track to avoid vehicle frontal accidents and then reduce the severity of vehicle accidents.
- traffic information engineering /
- accident severity /
- Shapley additive explanation /
- tree-augmented naive Bayes network /
- autonomous vehicle /
- factor interactive effect

HTML全文

图 1 可解释机器学习与TAN贝叶斯网络混合模型框架

Figure 1. Hybrid model framework of interpretable machine learning and TAN Bayes network

下载: 全尺寸图片幻灯片

图 2 自动驾驶车辆碰撞示意

Figure 2. Schematic of autonomous vehicle accident

下载: 全尺寸图片幻灯片

图 3 分类预测模型混淆矩阵

Figure 3. Confusion matrices for classification prediction models

下载: 全尺寸图片幻灯片

图 4 因素重要度排序

Figure 4. Importance ranking of factors

下载: 全尺寸图片幻灯片

图 5 TAN贝叶斯网络结构

Figure 5. TAN Bayes network structure

下载: 全尺寸图片幻灯片

图 6 单因素SHAP贡献

Figure 6. SHAP contributions of single factors

下载: 全尺寸图片幻灯片

图 7 主车碰撞部位和车辆年份对无伤害事故的单因素效应

Figure 7. Single factor effects of accident area of subject vehicle and vehicle year on no injury accident

下载: 全尺寸图片幻灯片

图 8 主车碰撞部位和车辆年份对无伤害事故的交互效应

Figure 8. Interaction effects of accident area of subject vehicle and vehicle year on no injury accident

下载: 全尺寸图片幻灯片

图 9 主车碰撞部位和车辆年份对中重度事故的交互效应

Figure 9. Interaction effects of accident area of subject vehicle and vehicle year on moderate-to-major accident

下载: 全尺寸图片幻灯片

图 10 主车和碰撞对象碰撞部位对轻微事故的交互效应

Figure 10. Interaction effects of accident areas of subject vehicle and accident partner on minor accident

下载: 全尺寸图片幻灯片

图 11 主车和碰撞对象碰撞部位对中重度事故的交互效应

Figure 11. Interaction effects of accident areas of subject vehicle and accident partner on moderate-to-major accident

下载: 全尺寸图片幻灯片

图 12 碰撞对象碰撞前运动状态和碰撞部位对无伤害事故的交互效应

Figure 12. Interaction effects of pre-accident motion state and accident area of accident partner on no injury accident

下载: 全尺寸图片幻灯片

图 13 碰撞对象碰撞前运动状态和碰撞部位对轻微事故的交互效应

Figure 13. Interaction effects of pre-accident motion state and accident area of accident partner on minor accident

下载: 全尺寸图片幻灯片

图 14 碰撞对象碰撞前运动状态和碰撞部位对中重度事故的交互效应

Figure 14. Interaction effects of pre-accident motion state and accident area of accident partner on moderate-to-major accident

下载: 全尺寸图片幻灯片

表 1 变量描述

Table 1. Descriptions of variables

变量		描述	赋值	各类事故样本的数量比例/%
变量		描述	赋值	无伤害	轻微	中重度
事故时间特征	季节A	春季	0	19.51	22.61	17.95
		夏季	1	17.07	27.54	30.77
		秋季	2	29.27	28.70	20.51
		冬季	3	34.15	21.15	30.77
	是否为工作日B	非工作日	0	21.95	24.35	20.51
	是否为工作日B	工作日	1	78.05	75.65	79.49
	是否为高峰C	非高峰	0	63.41	72.46	67.95
	是否为高峰C	高峰	1	36.59	27.54	32.05
车辆特征	车辆年份D	2020年以前	0	53.66	48.12	50.00
	车辆年份D	2020年及以后	1	46.34	51.88	50.00
	驾驶模式E	人工	0	56.10	46.38	52.56
	驾驶模式E	自动	1	43.90	53.62	47.44
环境特征	天气F	晴朗	0	90.24	86.09	89.74
	天气F	非晴朗(雨/雾/多云等)	1	9.76	13.91	10.26
	照明G	白天	0	85.37	72.75	61.54
	照明G	夜间	1	14.63	27.25	38.46
基础设施特征	路面H	干燥	0	80.49	90.14	97.44
	路面H	湿滑	1	19.51	9.86	2.56
	道路状况I	无异常	0	87.80	90.72	93.59
	道路状况I	异常	1	12.20	9.28	6.41
碰撞细节特征	主车碰撞前的运动状态J	变道/转向	0	9.76	13.04	20.51
		直行前进	1	21.95	26.96	37.18
		减速/停车	2	46.34	45.80	33.34
		其他	3	21.95	14.20	8.97
	碰撞对象碰撞前的运动状态O	变道/转向	0	17.07	18.84	20.51
		直行前进	1	43.90	43.77	52.57
		减速/停车	2	9.76	5.22	0.00
		其他	3	29.27	32.17	26.92
	主车碰撞部位L	尾部	0	34.15	20.58	19.23
		侧面	1	9.76	15.65	16.67
		正面	2	4.87	5.22	7.69
		其他	3	51.22	58.55	56.41
	碰撞对象碰撞部位M	尾部	0	34.15	35.36	29.49
		侧面	1	19.51	25.22	32.05
		正面	2	31.71	12.75	20.51
		其他	3	14.63	26.67	17.95

下载: 导出CSV

表 2 模型预测结果

Table 2. Model prediction results

数据类型	模型	准确度	精确度	召回率	F₁评分
原始	XGB	0.882	0.777	0.882	0.826
	RF	0.860	0.823	0.860	0.839
	KNN	0.849	0.803	0.849	0.823
	SVM	0.796	0.811	0.796	0.803
SMOTE	XGB	0.952	0.957	0.952	0.951
	RF	0.921	0.928	0.921	0.922
	KNN	0.819	0.874	0.819	0.815
	SVM	0.930	0.930	0.930	0.930

下载: 导出CSV

表 3 交互因素对的相关性

Table 3. Correlations of interaction factor pairs

交互因素对	相关性	交互因素对	相关性	交互因素对	相关性
(C, G)	0.157	(D, G)	0.188	(B, D)	0.166
(L, D)	0.216	(L, M)	0.438	(O, M)	0.314
(J, M)	0.067	(J, E)	0.055	(A, E)	0.059

下载: 导出CSV

参考文献(35)

[1]	王润民, 朱宇, 赵祥模, 等. 自动驾驶测试场景研究进展[J]. 交通运输工程学报, 2021, 21(2): 21-37. doi: 10.19818/j.cnki.1671-1637.2021.02.003 WANG Run-min, ZHU Yu, ZHAO Xiang-mo, et al. Research progress on test scenario of autonomous driving[J]. Journal of Traffic and Transportation Engineering, 2021, 21(2): 21-37. doi: 10.19818/j.cnki.1671-1637.2021.02.003
[2]	MAHASHHASH B O M, YUSOFF M I N, YAZID M R M, et al. Factors influencing injury severity in road traffic collisions: a comprehensive analysis from Libya[J]. International Journal on Transport Development and Integration, 2023, 7(4): 303-310.
[3]	吴静婷, 潘义勇, 施颖. 农村公路交叉口侧面碰撞事故伤害严重程度分析[J]. 大连交通大学学报, 2022, 43(6): 13-18, 44. WU Jing-ting, PAN Yi-yong, SHI Ying. Severity analysis of side crash accidents at rural highway intersections based on mixed logit[J]. Journal of Dalian Jiaotong University, 2022, 43(6): 13-18, 44.
[4]	胡郁葱, 韦湖, 曾强. 基于空间广义有序Probit模型的高速公路事故严重程度分析[J]. 华南理工大学学报(自然科学版), 2023, 51(1): 114-122. HU Yu-cong, WEI Hu, ZENG Qiang. Analysis of freeway crash severity based on spatial generalized ordered Probit model[J]. Journal of South China University of Technology (Natural Science Edition), 2023, 51(1): 114-122.
[5]	OLAYODE I O, DU Bo, SEVERINO A, et al. Systematic literature review on the applications, impacts, and public perceptions of autonomous vehicles in road transportation system[J]. Journal of Traffic and Transportation Engineering (English Edition), 2023, 10(6): 1037-1060.
[6]	AHMAD N, AHMED A, WALI B, et al. Exploring factors associated with crash severity on motorways in Pakistan[J]. Proceedings of the Institution of Civil Engineers—Transport, 2022, 175(4): 189-198.
[7]	潘义勇, 李烁. 考虑时空不稳定性的行人事故严重程度分析[J/OL]. 吉林大学学报(工学版), 2024. https://doi.org/10.13229/j.cnki.jdxbgxb.20231051. PAN Yi-yong, LI Shuo. Analysis of pedestrian accident injury severities considering spatiotemporal[J/OL]. Journal of Jilin University (Engineering and Technology Edition), 2024. https://doi.org/10.13229/j.cnki.jdxbgxb.20231051.
[8]	DING Sheng-xuan, ABDEL-ATY M, BARBOUR N, et al. Exploratory analysis of injury severity under different levels of driving automation (SAE Levels 2 and 4) using multi-source data[J/OL]. ArXiv, 2023, https://arxiv.org/abs/2303.17788.
[9]	李贵阳, 张福明, 王永岗. 基于SVM模型的山区高速公路多车事故影响因素分析[J]. 武汉理工大学学报(交通科学与工程版), 2020, 44(6): 1046-1051. doi: 10.3963/j.issn.2095-3844.2020.06.020 LI Gui-yang, ZHANG Fu-ming, WANG Yong-gang. Influencing factors analysis of multiple vehicle accidents in mountainous expressway based on SVM model[J]. Journal of Wuhan University of Technology (Transportation Science and Engineering), 2020, 44(6): 1046-1051. doi: 10.3963/j.issn.2095-3844.2020.06.020
[10]	陈治. 考虑天气和道路条件影响的道路交通事故严重程度分析及预测[D]. 北京: 北京交通大学, 2023. CHEN Zhi. Analysis and prediction of the severity of road traffic accidents considering the influence of weather and road conditions[D]. Beijing: Peking University, 2023.
[11]	IRANITALAB A, KHATTAK A. Comparison of four statistical and machine learning methods for crash severity prediction[J]. Accident Analysis & Prevention, 2017, 108: 27-36.
[12]	单永航, 张希, 胡川, 等. 基于集成学习的交通事故严重程度预测研究与应用[J]. 计算机工程, 2024, 50(2): 33-42. SHAN Yong-hang, ZHANG Xi, HU Chuan, et al. Traffic accident severity prediction research and application based on ensemble learning[J]. Computer Engineering, 2024, 50(2): 33-42.
[13]	严利鑫, 胡鑫辉, 刘清梅, 等. 道路交通事故严重程度预测及致因分析[J]. 华东交通大学学报, 2024, 41(5): 65-73. YAN Li-xin, HU Xin-hui, LIU Qing-mei, et al. Road traffic accident severity prediction and causation analysis[J]. Journal of East China Jiaotong University, 2024, 41(5): 65-73.
[14]	邱宏, 余德新, 王晓蓉, 等. Logistic回归模型中交互作用的分析及评价[J]. 中华流行病学杂志, 2010, 31(7): 812-814. QIU Hong, YU De-xin, WANG Xiao-rong, et al. Interaction between continuous variables in logistic regression model[J]. Chinese Journal of Epidemiology, 2010, 31(7): 812-814.
[15]	OLSZEWSKI P, SZAGAŁA P, WOLA AN'G SKI M, et al. Pedestrian fatality risk in accidents at unsignalized zebra crosswalks in Poland[J]. Accident Analysis & Prevention, 2015, 84: 83-91.
[16]	SUN Zhi-yuan, 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.
[17]	彭志鹏, 潘恒彦, 王永岗. 基于贝叶斯网络的网约车交通事故致因机理分析[J]. 东北大学学报(自然科学版), 2023, 44(1): 145-152. PENG Zhi-peng, PAN Heng-yan, WANG Yong-gang. Analyzing the causes of traffic accidents of online ride-hailing cars using the Bayesian network[J]. Journal of Northeastern University (Natural Science), 2023, 44(1): 145-152.
[18]	YU Rong-jie, ABDEL-ATY M, AHMED M. Bayesian random effect models incorporating real-time weather and traffic data to investigate mountainous freeway hazardous factors[J]. Accident Analysis & Prevention, 2013, 50: 371-376.
[19]	张道文, 王朝健, 蒋骏, 等. 考虑因素交互作用的车-车事故严重程度分析[J]. 汽车安全与节能学报, 2022, 13(4): 643-650. doi: 10.3969/j.issn.1674-8484.2022.04.005 ZHANG Dao-wen, WANG Chao-jian, JIANG Jun, et al. Analysis of the severity of vehicle to vehicle accidents considering the interaction of factors[J]. Journal of Automotive Safety and Energy, 2022, 13(4): 643-650. doi: 10.3969/j.issn.1674-8484.2022.04.005
[20]	王健宇, 陈献天, 焦朋朋, 等. 考虑建成环境的交通事故严重程度致因交互效应研究[J]. 交通运输系统工程与信息, 2024, 24(2): 272-280. WANG Jian-yu, CHEN Xian-tian, JIAO Peng-peng, et al. Interactive effect on traffic accident severity considering built environment[J]. Journal of Transportation Systems Engineering and Information Technology, 2024, 24(2): 272-280.
[21]	胡立伟, 薛刚, 李林育, 等. 高原地质及气象环境下公路交通风险致因耦合分析[J]. 中国公路学报, 2018, 31(1): 110-119. HU Li-wei, XUE Gang, LI Lin-yu, et al. Analysis of coupling of highway traffic risks in geological and meteorological environment of plateau regions[J]. China Journal of Highway and Transport, 2018, 31(1): 110-119.
[22]	PARSA B, MOVAHEDI A, TAGHIPOUR H, et al. Toward safer highways, application of XGBoost and SHAP for real-time accident detection and feature analysis[J]. Accident Analysis & Prevention, 2020, 136: 105405.
[23]	BREIMAN L. Random forests[J]. Machine Learning, 2001, 45: 5-32. doi: 10.1023/A:1010933404324
[24]	ZHANG Xiu-ling, WANG Shou, WU Zi-yun, et al. Unsupervised image clustering algorithm based on contrastive learning and K-nearest neighbors[J]. International Journal of Machine Learning and Cybernetics, 2022, 13(9): 2415-2423. doi: 10.1007/s13042-022-01533-7
[25]	CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20: 273-297.
[26]	LUNDBERG S M, LEE S I. A unified approach to interpreting model predictions[C]//NIPS. 31st Conference on Neural Information Processing Systems. Long Beach: NIPS, 2017: 4766-4775.
[27]	LUNDBERG S M, ERION G, CHEN H, et al. From local explanations to global understanding with explainable AI for trees[J]. Nature Machine Intelligence, 2020, 2(1): 56-67. doi: 10.1038/s42256-019-0138-9
[28]	成卫, 马铭炜, 张小龙. 基于贝叶斯网络的高速公路交通事故严重程度预测及致因分析[J]. 重庆交通大学学报(自然科学版), 2023, 42(7): 87-95. CHENG Wei, MA Ming-wei, ZHANG Xiao-long. Prediction and cause analysis of freeway traffic accident severity based on Bayesian network[J]. Journal of Chongqing Jiaotong University (Natural Science), 2023, 42(7): 87-95.
[29]	潘义勇, 徐翔宇. 数据不平衡的MobileViT网络交通事故严重程度预测模型[J/OL]. 吉林大学学报(工学版), 2023. https://doi.org/10.13229/j.cnki.jdxbgxb.20230614. PAN Yi-yong, XU Xiang-yu. Model for predicting severity of accidents based on MobileViT Network considering imbalanced data[J/OL]. Journal of Jilin University (Engineering and Technology Edition), 2023. https://doi.org/10.13229/j.cnki.jdxbgxb.20230614.
[30]	ZHENG Ou, ABDEL-ATY M, WANG Zi-jin, et al. AVOID: autonomous vehicle operation incident dataset across the globe[J/OL]. ArXiv, 2023. https://arxiv.org/abs/2303.12889.
[31]	PENG Yi-chuan, LI Chong-yi, WANG Ke, et al. Examining imbalanced classification algorithms in predicting real-time traffic crash risk[J]. Accident Analysis & Prevention, 2020, 144: 105610.
[32]	KUTELA B, DAS S, DADASHOVA B. Mining patterns of autonomous vehicle crashes involving vulnerable road users to understand the associated factors[J]. Accident Analysis & Prevention, 2022, 165: 106473.
[33]	GOSWAMY A, ABDEL-ATY M, ISLAM Z. Factors affecting injury severity at pedestrian crossing locations with rectangular RAPID flashing beacons (RRFB) using XGBoost and random parameters discrete outcome models[J]. Accident Analysis & Prevention, 2023, 181: 106937.
[34]	RICHARDS D C. Relationship between speed and risk of fatal injury: pedestrians and car occupants[R]. London: Department for Transport, 2010.
[35]	CHANNAMALLU S S, KERMANSHACHI S, PAMIDIMUKKALA A. Impact of autonomous vehicles on traffic crashes in comparison with conventional vehicles[C]//ASCE. International Conference on Transportation and Development 2023. Reston: ASCE, 2023: 39-50.