高速公路路侧雷达布设间距对交通事故风险评估精度的影响

杨东锋; 戴杰; 张玥妍; 韩磊; 余荣杰

doi:10.3963/j.jssn.1674-4861.2023.02.003

高速公路路侧雷达布设间距对交通事故风险评估精度的影响

doi: 10.3963/j.jssn.1674-4861.2023.02.003

杨东锋^1,,
戴杰²,
张玥妍²,
韩磊³,
余荣杰^3, ,

1.
浙江杭绍甬高速公路有限公司杭州 310000
2.
浙江高信技术股份有限公司杭州 310000
3.
同济大学道路与交通工程教育部重点实验室上海 201804

基金项目:

国家自然科学基金项目 52172349

浙江省交通运输厅科技计划项目 2021047

详细信息

作者简介:
杨东锋(1977—)，硕士，高级工程师. 研究方向：交通机电控制. E-mail：105691451@qq.com

通讯作者:
余荣杰(1989—)，博士，教授. 研究方向：交通安全. E-mail：yurongjie@tongji.edu.cn

中图分类号: U491.3
计量
- 文章访问数: 701
- HTML全文浏览量: 284
- PDF下载量: 48
- 被引次数: 0
出版历程
- 收稿日期: 2022-09-28
- 网络出版日期: 2023-06-19

Effects of Spacing of Highway Roadside Millimeter-wave Radar Detectors on the Accuracy of a Crash Risk Evaluation Model

1.
Zhejiang Hangshaoyong Expressway Company Limited, Hangzhou 310000, China
2.
Zhejiang Highway Information Engineering Technology Company Limited, Hangzhou 310000, China
3.
Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, Shanghai 201804, China

摘要

摘要: 高速公路通过布设毫米波雷达等新型检测设备，实现交通状态的精准感知，并为主动交通管控提供支撑。然而检测设备布设成本高，其布设间距需综合考虑成本约束和交通状态感知成效。为探究路侧毫米波雷达布设间距对交通事故风险评估精度的影响，基于浙江省沪杭甬高速公路的实证数据开展研究。构建事故风险评估深度森林模型(deep forest，DF)，应用滑动时空窗提取交通运行特征，并通过多层级联随机森林的集成建立交通运行特征与事故风险的关联关系；考虑路侧毫米波雷达感知范围，构建不同雷达布设间距下的交通运行数据集，开展布设间距对事故风险评估模型精度的敏感性分析。研究结果表明：DF模型曲线面积值(area under curve，AUC)为0.849，事故样本分类准确率为80.9%，高于传统的卷积神经网络模型(AUC值为0.741，准确率为75.2%)、随机森林模型(AUC值为0.715，准确率为70.8%)；雷达布设间距与事故风险评估精度呈反比关系，且密集布设下模型精度提升的边际效应递减，当布设间距由1 500 m缩减至750 m时，事故风险评估模型AUC值呈显著上升趋势，由0.794提升至0.853，布设间距由750 m缩减至250 m时，AUC值无明显变化。综上，雷达布设间距为750 m可平衡布设成本和事故风险评估精度，成果可为高速公路车道级交通状态感知系统的规划设计提供决策依据。
- 交通安全 /
- 交通事故 /
- 高速公路雷达布设 /
- 深度森林 /
- 事故风险评估模型 /
- 卷积神经网络
Abstract: Freeways equipped with new sensing equipment such as millimeter-wave radar detectors can accurately monitor traffic operation and well support active traffic management measures. However, due to the high deployment expenditure, there is a need to consider the cost constraints and the effectiveness of traffic state detection. To investigate the impacts of millimeter-wave radar deployment spacing on crash risk evaluation performance, this study is conducted based on the empirical data of the Hangshaoyong highway in Zhejiang Province. A crash risk evaluation model based on deep forest (DF) is developed. Specifically, sliding spatio-temporal windows are employed to extract the features of traffic operation while the correlation relationships between the features and crash risk are established through the integrations of multi-layer cascaded random forests. Considering the sensing range of the millimeter-wave radar detectors, multiple traffic operation datasets are developed by assuming different deployment spacings. Sensitivity analyses of radar deployment spacing on the evaluation accuracy of crash risk are conducted. Analyses results show that: The area under curve (AUC) of DF model is 0.849 with 80.9% recall on crash samples, which is higher than traditional convolutional neural network model (AUC is 0.741, recall is 75.2%) and random forest model (AUC is 0.715, recall is 70.8%). An inverse relationship between radar deployment spacing and evaluation accuracy of crash risk is captured, and the marginal effects of the improvement to the model accuracy decreases under dense deployment conditions. If the radar deployment spacing is reduced from 1 500 m to 750 m, the AUC of crash risk evaluation model shows a substantial increase (from 0.794 to 0.853), but there is no obvious change in AUC values when the radar deployment spacing is reduced from 750 m to 250 m. In conclusion, the radar deployment spacing of 750 m can balance the deployment cost and the evaluation performance of crash risk, which could be used to support the decisions related to the installment of traffic sensing equipment.
- traffic safety /
- traffic accident /
- highway radar deployment /
- deep forest /
- traffic crash risk evaluation /
- convolutional neural network

HTML全文

图 1 高速公路研究路段

Figure 1. The section of highway for study

下载: 全尺寸图片幻灯片

图 2 毫米波雷达布设示意图

Figure 2. Millimeter-wave radar deployment schematic

下载: 全尺寸图片幻灯片

图 3 事故前交通流提取规则

Figure 3. Pre-crash traffic flow extraction rule

下载: 全尺寸图片幻灯片

图 4 多粒度扫描流程

Figure 4. The procedure of multi-grained scanning

下载: 全尺寸图片幻灯片

图 5 级联森林流程

Figure 5. The procedure of cascade forests

下载: 全尺寸图片幻灯片

图 6 CNN模型架构

Figure 6. The CNN model architecture

下载: 全尺寸图片幻灯片

图 7 对事故风险影响的交通流特征前5名

Figure 7. Top 5 traffic flow characteristics that impact crash risk

下载: 全尺寸图片幻灯片

图 8 不同雷达布设间距的实验场景

Figure 8. Experimental scenarios in different radar intervals setting

下载: 全尺寸图片幻灯片

图 9 不同雷达布设间距的模型AUC

Figure 9. Model AUC values in different radar intervals setting

下载: 全尺寸图片幻灯片

表 1 毫米波雷达采集的交通流数据示例

Table 1. Examples of traffic flow data collected by millimeter-wave radars

检测时间	检测器编号	车道编号	方向	平均速度/(km/h)	车辆数	大车数
检测时间	2020-01-01 T00:00:12	102202101	1	上行	100.4	6	0
2020-01-01 T00:00:12	102202101	2	上行	91.6	3	1
2020-01-01 T00:00:12	102202101	3	上行	67.4	6	4
2020-01-01 T00:01:12	102202101	1	上行	106.8	5	0
2020-01-01 T00:01:12	102202101	2	上行	87.1	5	0
2020-01-01 T00:01:12	102202101	3	上行	75.8	3	3
2020-01-01 T00:02:12	102202101	1	上行	104.2	3	0
2020-01-01 T00:02:12	102202101	2	上行	94.1	1	0
2020-01-01 T00:02:12	102202101	3	上行	74.7	5	4

下载: 导出CSV

表 2 交通流特征变量

Table 2. Traffic Flow Characteristic Variables

变量名称	单位	变量描述
AS(U/C/D)T	km/h	第T个时间片段下上游/事故/下游路段的速度平均值
SS(U/C/D)T	km/h	第T个时间片段下上游/事故/下游路段的速度标准差
DS(U/C/D)T	km/h	第T个时间片段下上游/事故/下游路段的相邻车道间速度差绝对值的均值
SCDT	km/h	第T个时间片段下事故路段及其下游的速度差绝对值
SCUT	km/h	第T个时间片段下事故路段及其上游的速度差绝对值
AF(U/C/D)T	veh/min	第T个时间片段下上游/事故/下游路段的流量平均值
SF(U/C/D)T	veh/min	第T个时间片段下上游/事故/下游路段的流量标准差
BF(U/C/D)T	%	第T个时间片段下上游/事故/下游路段的大车率
FCDT	veh/min	第T个时间片段下事故路段及其下游的流量差绝对值
FCUT	veh/min	第T个时间片段下事故路段及其上游的流量差绝对值

下载: 导出CSV

表 3 模型主要参数设置

Table 3. Model parameter settings of each model

模型	最优模型参数设置
Logit	Sigmoid函数分类阈值=0.26
SVM	核函数=高斯核函数，核函数系数gamma=1/110
RF	子树数量=200，最大深度=5
CNN	单次训练样本量=64，优化器=Adam优化器，学习率=0.001，学习率衰减率=0，训练轮次=100
DF	每层包含随机森林数=4，每个随机森林中子树数量=100，级联森林层数=4（模型训练自动生成）

下载: 导出CSV

表 4 模型评价指标对比

Table 4. Comparison of model evaluation indicators

模型	ACC	recall	AUC
Logit	0.669	0.667	0.692
SVM	0.672	0.681	0.703
RF	0.721	0.708	0.715
CNN	0.793	0.752	0.741
DF	0.812	0.809	0.849

下载: 导出CSV

表 5 DF模型变量权重表

Table 5. The variable weights of DF model 单位: %

变量	时间片段
变量	6	5	4	3	2
ASD	0.84	0.69	1.37	2.62	3.44
ASC	0.87	0.75	1.08	2.71	4.06
ASU	0.88	0.80	0.91	1.43	3.05
SSD	0.70	0.60	0.75	0.85	1.05
SSC	0.89	0.78	0.81	1.35	1.49
SSU	0.87	0.77	0.73	0.69	1.10
DSD	0.70	0.67	0.82	0.80	0.87
DSC	0.92	0.96	0.90	0.95	1.19
DSU	0.69	1.19	0.90	1.08	1.09
AFD	0.67	0.70	0.74	0.63	0.64
AFC	0.88	0.72	0.85	0.72	0.67
AFU	0.64	0.70	0.64	0.61	0.57
SFD	0.67	0.58	0.65	0.58	0.71
SFC	0.70	0.65	0.70	0.80	0.67
SFU	0.53	0.65	0.57	0.58	0.57
BFD	0.68	0.65	0.73	0.65	0.76
BFC	0.76	0.74	0.83	0.98	1.05
BFU	0.76	0.78	1.00	0.99	0.96
SCD	0.61	0.59	0.73	1.00	1.76
SCU	0.66	0.68	0.76	0.84	1.33
FCD	0.57	0.61	0.58	0.54	0.59
FCU	0.69	0.66	0.72	0.58	0.56
均值	0.74	0.72	0.81	1.00	1.28

下载: 导出CSV

参考文献(27)

[1]	交通运输部, 2021年交通运输行业发展统计公报[R]. 北京: 交通运输部, 2022. Ministry of Transport. Statistical bulletin on the development of the transportation industry(2021)[R]. Beijing: Ministry of Transport, 2022. (in Chinese)
[2]	焦蕴平. 向交通强国奋进[J]. 中国公路, 2019(19): 16-19. https://www.cnki.com.cn/Article/CJFDTOTAL-GLZG201919010.htm JIAO Y P. Forge towards a country with strong transportation network[J]. China Highway, 2019(19): 16-19. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-GLZG201919010.htm
[3]	王少飞, 谯志, 付建胜, 等. 智慧高速公路的内涵及其架构[J]. 公路, 2017, 62(12): 170-175. https://www.cnki.com.cn/Article/CJFDTOTAL-GLGL201712045.htm WANG S F, QIAO Z, FU J S, et al. Connotation and architecture of smart expressway[J]. Highway, 2017, 62(12): 170-175. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-GLGL201712045.htm
[4]	王虹. 新基建模式下智慧高速的"破"与"建"[J]. 中国交通信息化, 2021(7): 22-26. https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202107003.htm WANG H. The"break"and"build"of smart highway in the new infrastructure model[J]. China ITS Journal, 2021(7): 22-26. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202107003.htm
[5]	公安部交通管理局. 中国道路交通事故统计年报(2021)[R]. 北京: 公安部交通管理局, 2022. Ministry of Public Security, Transportation Bureau. The road traffic accidents statistics report in China(2021)[R]. Beijing: Ministry of Public Security, Transportation bureau, 2022. (in Chinese)
[6]	王锐, 高磊. 智慧高速主动交通管控策略探究[J]. 中国交通信息化, 2022(增刊): 109-111. https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC2022S1028.htm WANG R, GAO L. Exploration of active traffic control strategy on smart highway[J]. China ITS Journal, 2022(SUP1): 109-111. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC2022S1028.htm
[7]	崔录库, 刘金伟. 主动交通管理在智慧高速中的应用探讨[J]. 中国交通信息化, 2021(12): 131-133. https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202112021.htm CUI L K, LIU J W. Exploring the application of active traffic management in smart highway[J]. China ITS Journal, 2021 (12): 131-133. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202112021.htm
[8]	赵祥模, 高赢, 徐志刚, 等. IntelliWay-变耦合模块化智慧高速公路系统一体化架构及测评体系[J]. 中国公路学报, 2023, 36(1): 176-201. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL202301015.htm ZHAO X M, GAO Y, XU Z G, et al. IntelliWay: An integrated architecture and testing methodology for intelligent highway using varied coupling modularization[J]. China Journal of Highway and Transport, 2023, 36(1): 176-201. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL202301015.htm
[9]	ABDEL-ATY M A, CAI Q, ELURU N, et al. Integrated freeway/arterial active traffic management[R]. Orlando: Department of Transportation, 2019.
[10]	刘星良, 单珏, 刘唐志, 等. 基于交通流稳定性系数的高速公路交通事故实时风险预测[J]. 交通信息与安全, 2022, 40 (4): 71-81. doi: 10.3963/j.jssn.1674-4861.2022.04.008 LIU X L, C Y, LIU T Z, et al. Real-time forecast models for traffic accidents on expressways using stability coefficients of traffic flow[J]. Journal of Transport Information and Safety, 2022, 40(4): 71-81. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2022.04.008
[11]	王西. 新基建背景下智慧高速多元融合感知技术应用浅谈[J]. 中国交通信息化, 2020(6): 125-126. https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202006017.htm WANG X. Discussion on the application of smart highway multi-integration sensing technology under the background of new infrastructure construction[J]. China ITS Journal, 2020(6): 125-126. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202006017.htm
[12]	ROSHANDEL S, ZHENG Z, WASHINGTON S. Impact of real-time traffic characteristics on freeway crash occurrence: systematic review and meta-analysis[J]. Accident Analysis & Prevention, 2015, 79: 198-211.
[13]	杜豫川, 刘成龙, 吴荻非, 等. 新一代智慧高速公路系统架构设计[J]. 中国公路学报, 2022, 35(4): 203-214. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL202204017.htm DU Y C, LIU C L, WU D F, et al. Framework of the new generation of smart highway[J]. China Journal of Highway and Transport, 2022, 35(4): 203-214. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL202204017.htm
[14]	金峰, 顾永鑫, 胡飞. 沪杭甬高速毫米波雷达事件检测能力分析[J]. 中国交通信息化, 2022(8): 122-124, 139. https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202208012.htm JIN F, GU Y X, HU F. Analysis of millimeter wave radar event detection capability for the Huhangyong highway[J]. China ITS Journal, 2022(8): 122-124, 139. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JTXC202208012.htm
[15]	CHEN H, DOUGHERTY M, KIRBY H. An investigation of detector spacing and forecasting performance using neural networks[C]. 4th World Congress on Intelligent transport systems, Berlin, German: Mobility for everyone, 1997.
[16]	KWON J, PETTY K, VARAIYA P. Probe vehicle runs or loop detectors? effect of detector spacing and sample size on accuracy of freeway congestion monitoring[J]. Transportation research record, 2007, 2012(1): 57-63.
[17]	刘政威. 考虑交通事件检测的固定型交通检测器布设方法研究[D]. 南京: 东南大学, 2011. LIU Z W. Research on the fixed traffic detectors deployment method considering traffic event detection[D]. Nanjing: Southeast University, 2011. (in Chinese)
[18]	CAO Q, LI Z, MA Y, et al. Optimal layout of heterogeneous sensors for traffic accidents detection and prevention[J]. IEEE transactions on instrumentation and measurement, 2022, 71: 1-13.
[19]	刘昕, 刘志远, 聂品, 等. 微观交通仿真模型参数标定研究综述[J]. 铁道科学与工程学, 2022, 19(11): 3179-3189. https://www.cnki.com.cn/Article/CJFDTOTAL-CSTD202211007.htm LIU X, LIU Z Y, NIE P, et al. A survey of microscopic traffic simulation calibration methods[J]. Journal of Railway Science and Engineering, 2022, 19(11): 3179-3189. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-CSTD202211007.htm
[20]	HOSSAIN M, ABDEL-ATY M, QUDDUS M A, et al. Real-time crash prediction models: State-of-the-art, design pathways and ubiquitous requirements[J]. Accident Analysis & Prevention, 2019, 124: 66-84.
[21]	YU R, ABDEL-ATY M. Utilizing support vector machine in real-time crash risk evaluation[J]. Accident Analysis & Prevention, 2013, 51: 252-259.
[22]	孙剑, 孙杰. 城市快速路实时交通流运行安全主动风险评估[J]. 同济大学学报(自然科学版), 2014, 42(6): 873-879. https://www.cnki.com.cn/Article/CJFDTOTAL-TJDZ201406008.htm SUN J, SUN J. Proactive assessment of real-time traffic flow accident risk on urban expressway[J]. Journal of Tongji University(Natural Science), 2014, 42(6): 873-879. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-TJDZ201406008.htm
[23]	ZHOU Z H, FENG J. Deep forest: towards an alternative to deep neural networks[C]. IJCAI Conference, Melbourne, Australia: IJCAI, 2017.
[24]	夏恒, 汤健, 乔俊飞. 深度森林研究综述[J]. 北京工业大学学报, 2022, 48(2): 182-196. https://www.cnki.com.cn/Article/CJFDTOTAL-BJGD202202010.htm XIA H, TANG J, QIAO J F. Review of deep forest[J]. Journal of Beijing University of Technology, 2022, 48(2): 182-196. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-BJGD202202010.htm
[25]	YU R, WANG Y, ZOU Z, et al. Convolutional neural networks with refined loss functions for the real-time crash risk analysis[J]. Transportation Research Part C: Emerging Technologies, 2020(119): 102740.
[26]	LI P, ABDEL-ATY M, YUAN J. Real-time crash risk prediction on arterials based on LSTM-CNN[J]. Accident Analysis & Prevention, 2020(135): 105371.
[27]	高珍, 高屹, 余荣杰, 等. 连续数据环境下的道路交通事故风险预测模型[J]. 中国公路学报, 2018, 31(4): 280-287. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL201804033.htm