基于XGBoost算法的危险场景驾驶行为模式分析及安全评估

魏田正; 魏雯; 李海梅; 刘浩学; 朱彤; 刘斐

doi:10.3963/j.jssn.1674-4861.2022.05.006

基于XGBoost算法的危险场景驾驶行为模式分析及安全评估

doi: 10.3963/j.jssn.1674-4861.2022.05.006

魏田正^1,,
魏雯²,
李海梅¹,
刘浩学^{1, 3},
朱彤^2, ,,
刘斐¹

1.
长安大学汽车学院西安 710064
2.
长安大学运输工程学院西安 710064
3.
长安大学汽车运输安全保障技术交通行业重点实验室西安 710064

基金项目:

国家重点研发计划项目 2019YFE0108000

长安大学研究生科研创新实践项目 300103722024

详细信息

作者简介:
魏田正（1996—），博士研究生. 研究方向：交通安全与驾驶员行为.E-mail：2019022010@chd.edu.cn

通讯作者:
朱彤（1977—），博士，副教授. 研究方向：交通安全与驾驶员行为. E-mail：zhutong@chd.edu.cn

中图分类号: X951
计量
- 文章访问数: 816
- HTML全文浏览量: 316
- PDF下载量: 88
- 被引次数: 0
出版历程
- 收稿日期: 2022-07-25
- 网络出版日期: 2022-12-05

An Analysis of Driving Behavior Model and Safety Assessment Under Risky Scenarios Based on an XGBoost Algorithm

WEI Tianzheng^1
,,
WEI Wen²,
LI Haimei¹,
LIU Haoxue^{1, 3},
ZHU Tong^{2
, ,},
LIU Fei¹

1.
School of Automobile, Chang'an University, Xi'an 710064, China
2.
College of Transportation Engineering, Chang'an University, Xi'an 710064, China
3.
Key Laboratory of Automotive Transportation Safety Enhancement Technology of Ministry of Transport, Chang'an University, Xi'an 710064, China

摘要

摘要: 危险感知能力对驾驶人的驾驶行为模式具有重要影响。为准确评估驾驶人的危险感知能力、提升危险感知水平判别的准确度，提出了基于模拟驾驶技术的危险感知能力影响分析方法和基于极端梯度提升树（XGBoost）算法的危险感知水平判别模型。通过设计3种常见交通冲突场景，采集模拟驾驶中驾驶人的多维度驾驶行为特征数据，并分析危险感知能力与驾驶行为的相关关系。通过模拟实验发现：驾驶人对行人的危险感知能力较弱，易发生碰撞事故；驾驶人在危险场景中的车速（p＝0.01）、制动反应位置（p < 0.01）以及反应时间（p < 0.01）与危险感知水平之间存在显著负相关关系。在相关性分析的基础上，利用XGBoost算法识别能反映驾驶人危险感知能力的重要特征变量，并构建以制动反应位置、反应时间、车速、刹车深度，以及加速度为指标的驾驶人危险感知水平判别模型；通过与LightGBM、支持向量机（SVM），以及逻辑回归（LR）等算法分类预测性能的对比分析，评价危险感知模型的判别精度，结果表明：基于XGBoost算法的危险感知水平判别模型的判别准确率为84.8%、F1值为83.4%、AUC值为0.959，优于LightGBM（准确率为78.8%、F1值为76.7%、AUC值为0.924）、SVM（准确率为57.6%、F1值为42.2%、AUC值为0.859），以及LR算法（准确率为69.7%、F1值为65.5%、AUC值为0.836）。所提方法可为判别驾驶人危险感知能力及其对驾驶行为模式的影响提供可靠手段。
- 交通安全 /
- 驾驶行为 /
- 危险感知 /
- 机器学习 /
- XGBoost
Abstract: Hazard perception is a critical factor of a driving behavior model. A simulator-based method and an extreme gradient boosting tree(XGBoost)algorithm are proposed, in order to study the impacts of hazard perception on driving behaviors and improve the accuracy of hazard perception. Three typical scenarios of traffic conflicts are simulated, and a large amount of driving behavior data are collected. The correlation between hazard perception and driving behavior models is discussed under the three scenarios. The correlation analysis reveals that when hazard perception(e.g., dangerous behaviors of pedestrians)is weak, and the vehicle speed(p=0.01), braking reaction position(p < 0.01), and reaction time(p < 0.01)are significantly negatively correlated with the drivers'hazard perception. Based on the correlation analysis, the XGBoost algorithm is used to identify important features determining the capability of hazard perception of drivers. Then, a discriminant model of hazard perception is proposed with following the indicators, such as braking reaction position, reaction time, vehicle speed, braking depth, and acceleration. Compared the proposed method with Light Gradient Boosting Machine(LightGBM), Support Vector Machine(SVM), and Logistic Regression(LR)algorithms, it is found that the accuracy of the XGBoost-based method is 84.8%, its F1-score is 83.4%, and the area under the receiver operating characteristic Curve(AUC)is 0.959, which is better than the LightGBM(accuracy is 78.8%, F1-score is 76.7%, and AUC is 0.924), SVM(accuracy is 57.6%, F1-score is 42.2%, and AUC is 0.859)and LR algorithm(accuracy is 69.7%, F1-score is 65.5%, and AUC is 0.836). In conclusion, the proposed method can provide a more reliable way for understanding the capability of hazard perception of drivers and its impacts on driving behavior models.
- traffic safety /
- driving behavior /
- hazard perception /
- machine learning /
- XGBoost

HTML全文

图 1 驾驶模拟器

Figure 1. Driving simulator

下载: 全尺寸图片幻灯片

图 2 危险场景示意图

Figure 2. Schematic diagram of risk scenarios

下载: 全尺寸图片幻灯片

图 3 3种类型交通冲突场景的TTC值箱图

Figure 3. TTC box plot of three types of traffic conflict scenarios

下载: 全尺寸图片幻灯片

图 4 制动反应位置与车速交互作用分析

Figure 4. Analysis of interaction between braking reaction position and vehicle speed

下载: 全尺寸图片幻灯片

图 5 刹车深度与反应时间交互作用分析

Figure 5. Analysis of interaction between brake depth and reaction time

下载: 全尺寸图片幻灯片

图 6 特征重要性

Figure 6. Importance of features

下载: 全尺寸图片幻灯片

图 7 特征变量个数对模型性能的影响分析

Figure 7. Analysis of the influence of the number of characteristic variables on model performance

下载: 全尺寸图片幻灯片

图 8 危险感知水平预测混淆矩阵图

Figure 8. Confusion matrix plot of hazard perception prediction level

下载: 全尺寸图片幻灯片

图 9 ROC曲线图

Figure 9. ROC curve

下载: 全尺寸图片幻灯片

表 1 风险场景描述

Table 1. Risk scenario description

场景	风险类型	交通冲突类型	示意图	场景描述
1	显性危险	轿车与轿车	图 2（a）	被试车辆右侧有1辆同向行驶的轿车在打转向灯3 s后换道至中间车道
2	隐性危险	轿车与轿车	图 2（b）	被试车辆直行方向为绿灯，在其右前方有1辆轿车正准备从施工区隔离围挡后驶出
3	显性危险	轿车与行人	图 2（c）	被试车辆直行方向为绿灯，在右侧交叉口处人行横道上有1名行人正准备冲入人行横道
4	隐性危险	轿车与行人	图 2（d）	被试车辆驶近前方人行横道时，有1名被树木遮挡的行人突然冲入人行横道
5	显性危险	轿车与非机动车	图 2（e）	被试车辆右前方有1名骑行者的行驶方向被施工区阻碍，骑行者换道至中间车道
6	隐性危险	轿车与非机动车	图 2（f）	被试车辆直行方向为绿灯，在其右前方公交车后有1名骑行者突然从公交车后冲出

下载: 导出CSV

表 2 危险场景TTC值的总体情况

Table 2. Overall situation of TTC values in risky scenarios

危险场景类型	均值	标准差	中位数	百分位数/%
危险场景类型	均值	标准差	中位数	25	75
显性危险	4.14	1.80	3.52	2.60	5.68
隐性危险	3.60	2.47	2.90	1.64	5.47

下载: 导出CSV

表 3 Spearman相关性分析结果

Table 3. Spearman correlation analysis results

特征	相关系数	显著性（双尾）p
车速/（km/h）	-0.204	0.010
加速度/（m/s²）	-0.055	0.491
制动反应位置/m	-0.743	0.000
刹车深度	-0.107	0.178
转向盘旋转率/（1/s）	0.015	0.847
制动车速/（km/h）	-0.078	0.328
反应时间/s	-0.606	< 0.001

下载: 导出CSV

表 4 描述性统计分析

Table 4. Descriptive statistical analysis

指标	危险感知能力
指标	低	标准差	中	标准差	高	标准差
车速/（km/h）	47.68	12.10	44.66	8.09	41.80	8.30
纵向加速度/（m/s²）	-0.41	0.44	-0.48	0.45	-0.41	0.30
制动反应位置/m	107.52	36.86	80.71	19.98	36.42	20.54
刹车深度	0.11	0.08	0.11	0.05	0.08	0.05
转向盘旋转率/（1/s）	0.02	0.02	0.02	0.02	0.02	0.02
制动车速/（km/h）	59.88	11.60	59.01	9.75	57.46	8.94
反应时间/s	0.77	0.38	0.48	0.56	0.29	0.42

下载: 导出CSV

表 5 最优模型参数

Table 5. Optimal model parameters

参数	范围	参数最优值
learning_rate	[0，1，0.01]	0.1
max_depth	[1，15，1]	5
n_estimators	[0，1 000，100]	200
subsample	[0，1，0.1]	0.7
min_child_weight	[1，10，1]	3
reg_alpha	[0，0.05，0.001]	0.001
gamma	[0，1，0.1]	0.28
colsample_bytree	[0，1，0.1]	0.8

下载: 导出CSV

表 6 模型分类性能

Table 6. Performance of model classification

模型	accuracy	precision	recall	F1值	AUC
XGBoost	0.848	0.839	0.850	0.834	0.959
LightGBM	0.788	0.775	0.777	0.767	0.924
LR	0.697	0.660	0.656	0.655	0.836
SVM	0.576	0.452	0.487	0.422	0.859

下载: 导出CSV

参考文献(21)

[1]	张旭欣, 王雪松, 马勇, 等. 驾驶行为与驾驶风险国际研究进展[J]. 中国公路学报, 2020, 33(6): 1-17. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL202006002.htm ZHANG X X, WANG X S, MA Y, et al. International research progress on driving behavior and driving risks[J]. China Journal of Highway and Transport, 2020, 33(6): 1-17. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL202006002.htm
[2]	TARLOCHAN F, IBRAHIM M I M, GABEN B. Understanding traffic accidents among young drivers in Qatar[J]. International Journal of Environmental Research and Public Health, 2022, 19(1): 514-527. doi: 10.3390/ijerph19010514
[3]	吴艳霞, 刘剑, 黄帅, 等. 雨天高速公路纵坡对驾驶员心率及行车速度影响[J]. 交通信息与安全, 2021, 39(4): 35-42. doi: 10.3963/j.jssn.1674-4861.2021.04.005 WUYX, LIU J, HUANG S, et al. Influences of longitudinal slopes of highways on drivers'heart rate and driving speeds on rainy days[J]. Journal of Transport Information and Safety, 2021, 39(4): 35-42. (in Chinese). doi: 10.3963/j.jssn.1674-4861.2021.04.005
[4]	乔洁, 徐鑫, 刘传攀, 等. 驾驶人危险感知能力影响因素及干预方式综述[J]. 中国安全科学学报, 2022, 32(2): 34-41. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202202006.htm QIAO J, XU X, LIU C P, et al. Review on affecting factors and intervention methods of drivers'hazard perception ability[J]. China Safety Science Journal, 2022, 32(2): 34-41. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202202006.htm
[5]	郭双, 王君, 常若松. 驾驶员驾驶经验对驾驶愤怒的影响[J]. 辽宁师范大学学报(社会科学版), 2015, 38(4): 481-485. https://www.cnki.com.cn/Article/CJFDTOTAL-LNSS201504010.htm GUO S, WANG J, CHANG R S. Effects of drivers'driving experience on driving anger[J]. Journal of Liaoning Normal University(Social Science Edition), 2015, 38(4): 481-485. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-LNSS201504010.htm
[6]	吴玲. 基于生理与行为指标的高速公路特长隧道环境驾驶人风险特性研究[D]. 西安: 长安大学, 2018. WU L. Study on driver risk characteristics in extra-long highway tunnels based on physiological and behavioral indicators[D]. Xi'an: Chang'an University, 2018. (in Chinese).
[7]	刘鹏飞, 马昌喜, 杨东, 等. 危险货物车辆司机速度选择影响因素研究[J]. 中国安全科学学报, 2019, 29(2): 133-139. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK201902024.htm LIU P F, MA C X, YANG D, et al. Research on factors influencing driving speed selection of hazardous materials vehicle drivers[J]. China Safety Science Journal, 2019, 29(2): 133-139. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK201902024.htm
[8]	杨倩盈, 孙龙. 心理控制源和驾驶安全态度对年轻驾驶员驾驶行为的影响[J]. 人类工效学, 2020, 26(4): 35-38. https://www.cnki.com.cn/Article/CJFDTOTAL-XIAO202004008.htm YANG Q Y, SUN L. Effects of traffic locus of control and risk behavior attitude on young drivers'driving behavior[J]. Chinese Journal of Ergonomics, 2020, 26(4): 35-38. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-XIAO202004008.htm
[9]	巩建国. 基于事故数据的驾驶人行为深度分析与改善对策研究[J]. 汽车与安全, 2020(5): 68-71. https://www.cnki.com.cn/Article/CJFDTOTAL-QCAQ202005024.htm GONG J G. Driver's behavior in-depth analysis and improvement countermeasures based on accident data[J]. Auto & Safety, 2020(5): 68-71. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-QCAQ202005024.htm
[10]	江亮, 贺宜. 电动两轮车风险驾驶行为及事故影响因素分析[J]. 吉林大学学报(工学版), 2019, 49(4): 1107-1113. https://www.cnki.com.cn/Article/CJFDTOTAL-JLGY201904011.htm JIANG L, HE Y. Risky driving behavior and influencing factors analysis for electric two-wheeler[J]. Journal of Jilin University(Engineering and Technology Edition), 2019, 49(4): 1107-1113. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-JLGY201904011.htm
[11]	LIU Y, WANG X, GUO Y. The moderating effects of emotions on the relationship between self-reported individual traits and actual risky driving behaviors[J]. Psychology Research and Behavior Management, 2021, 14(4): 423-447.
[12]	陶达, 张瑞, 曲行达. 性别、年龄和人格特质对风险驾驶行为的影响[J]. 深圳大学学报(理工版), 2016, 33(6): 646-652. https://www.cnki.com.cn/Article/CJFDTOTAL-SZDL201606014.htm TAO D, ZHANG R, QU X D. The effects of gender, age and personality traits on risky driving behaviors[J]. Journal of Shenzhen University(Science and Engineering), 2016, 33(6): 646-652. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-SZDL201606014.htm
[13]	POLLATSEK A, NARAYANAAN V, PRADHAN A, et al. Using eye movements to evaluate a PC-based risk awareness and perception training program on a driving simulator[J]. Human Factors, 2006, 48(3): 447-464.
[14]	CASTRO C, VENTSISLAVOVA P, GARCIA-FERNANDEZ P, et al. Risky decision-making and hazard prediction are negatively related and could be assessed independently using driving footage[J]. Psychology Research and Behavior Management, 2021, 14(6): 857-876.
[15]	HILL A, HORSWILL M S, WHITING J, et al. Computer-based hazard perception test scores are associated with the frequency of heavy braking in everyday driving[J]. Accident Analysis & Prevention, 2019, 122(8): 207-214.
[16]	秦雅琴, 梁锺月, 贾现广, 等. 驾驶人风险感知类型预测模型研究[J]. 安全与环境学报, 2019, 19(4): 1266-1273. https://www.cnki.com.cn/Article/CJFDTOTAL-AQHJ201904023.htm QIN Y Q, LIANG Z Y, JIA X G, et al. A predictive and forecast model of the risk perception type of the vehicle drivers[J]. Journal of Safety and Environment, 2019, 19(4): 1266-1273. (in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-AQHJ201904023.htm
[17]	HORSWILL M S, HILL A, BUCKLEY L, et al. An online hazard perception training course reduces heavy braking, speeding, and over-revving rates during everyday driving[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2022, 87(3): 54-68.
[18]	中华人民共和国住房和城乡建设部. 城市道路工程设计规范: CJJ 37—2012[S]. 北京: 中国建筑工业出版社, 2012. Ministry of Housing and Urban-Rural Development of the People's Republic of China. Code for design of urban road engineering: CJJ 37—2012[S]. Beijing: China Architecture & Building Press, 2012. (in Chinese).
[19]	中华人民共和国住房和城乡建设部. 城市道路交通标志和标线设置规范: GB51038—2015[S]. 北京: 中国计划出版社. 2015. Ministry of Housing and Urban-Rural Development of the People's Republic of China. Code for layout of urban road traffic signs and markings: GB51038—2015[S]. Beijing: China Planning Press. 2015. (in Chinese).
[20]	把余韬. 基于行为和生理指标的驾驶风险分析与模式识别研究[D]. 北京: 清华大学, 2015. BA Y T. Risk factors identification during driving based on behavioral and physiological measures[D]. Beijing: Tsinghua University, 2015. (in Chinese).
[21]	王至恒. 基于车辆运行参数的驾驶危险状态辨识仿真研究[D]. 西安: 长安大学, 2018. WANG Z H. Study on simulation of identification of driving hazard based on vehicle operating parameters[D]. Xi'an: Chang'an University, 2018. (in Chinese).