基于障碍车辆轨迹预测的驾驶碰撞风险模型

杨厚新; 陆丽萍; 秦恒; 杨奥; 褚端峰

doi:10.3963/j.jssn.1674-4861.2025.01.004

基于障碍车辆轨迹预测的驾驶碰撞风险模型

doi: 10.3963/j.jssn.1674-4861.2025.01.004

杨厚新^1,,
陆丽萍^2, ,,
秦恒²,
杨奥²,
褚端峰³

1.
湖北省交通运输厅通信信息中心武汉 430030
2.
武汉理工大学计算机与人工智能学院武汉 430079
3.
武汉理工大学智能交通系统研究中心武汉 430063

基金项目:

国家自然科学基金面上项目 52172392

湖北省交通运输厅科技项目 2023-121-3-2

详细信息

作者简介:
杨厚新（1968—），本科，高级工程师. 研究方向：交通运输行业信息化研究. E-mail：331011781@qq.com

通讯作者:
陆丽萍（1977—），博士，副教授. 研究方向：无人驾驶技术. E-mail：luliping@whut.edu.cn

中图分类号: U471.15
计量
- 文章访问数: 27
- HTML全文浏览量: 14
- PDF下载量: 8
- 被引次数: 0
出版历程
- 收稿日期: 2024-06-18
- 网络出版日期: 2025-06-27

Driving Collision Risk Modeling with Trajectory Prediction of Obstacle Vehicles

1.
Communication and Information Center of Hubei Provincial Department of Transportation, Wuhan 430030, China
2.
School of Computer Science and Artificial Intelligence, Wuhan University of Technology, Wuhan 430079, China
3.
Intelligent Transportation Systems Research Center, Wuhan University of Technology, Wuhan 430063, China

摘要

摘要: 针对智能驾驶系统在驾驶风险预警中存在的动态交互特征捕捉不足、多模态轨迹预测精度有限以及碰撞风险量化物理指标过度单一等问题，研究了基于多模态轨迹预测与概率量化耦合的预见性碰撞风险评估模型。在轨迹预测部分，研究了分层图注意力网络，通过图注意力机制融合高精地图、车道线以及车辆历史轨迹特征，能够有效捕捉车辆行驶环境中的动态变化；针对传统模型中先预测再细化的两阶段解码结构，引入滑动窗口优化解码器，能够准确预测临近车辆的未来轨迹。在碰撞风险评估部分，研究了1种基于概率量化的碰撞风险评估方法，通过结合预测的未来轨迹与碰撞风险，估算自车与周边车辆发生碰撞的概率，实现对车辆危险行为的提前预警。实验结果表明：在Argoverse数据集上最小终点位移误差、最小平均位移误差和漏检率分别为0.785、1.157和0.126，与HiVT与LaneGCN相比，在终点预测方面误差分别减少了1%和15.1%。在城市交通能力仿真软件（simulation of urban mobility，SUMO）上验证预测风险与实际风险的偏差约为5%，从数据波动性上看，危险程度波动幅度为0.3，与碰撞时间（time to collision，TTC）方法和动态安全指数（dynamic safety index，DSI）方法相比，波动幅度分别减少33.3%和18.75%，在持续驾驶场景中展现出更优秀的风险评估水准；证明了基于障碍车辆轨迹预测的驾驶碰撞风险模型在预测未来潜在驾驶风险的准确性。
- 交通安全 /
- 预见性碰撞风险评估 /
- 车辆轨迹预测 /
- 图神经网络 /
- 风险概率量化
Abstract: In order to address critical challenges in intelligent driving systems, including insufficient dynamic interaction modeling, limited accuracy in multimodal trajectory prediction, and over-reliance on single physical metrics for collision risk quantification. A proactive collision risk assessment framework is proposed by integrating probabilistic quantification with multimodal trajectory prediction. For trajectory prediction, a hierarchical graph attention network is developed to capture dynamic environmental features through adaptive fusion of high-definition maps, lane geometries, and vehicle motion history. A sliding window-optimized decoder is introduced within the conventional two-stage prediction architecture to refine trajectory outputs. For risk assessment, a probabilistic collision quantification method is designed to calculate collision likelihood between ego and surrounding vehicles based on predicted trajectories. Results on the Argoverse dataset demonstrate state-of-the-art performance with minimum final displacement error (=0.785), average displacement error (=1.157), and miss rate (=0.126), achieving 1% and 15.1% error reduction in endpoint prediction compared to HiVT and LaneGCN respectively. simulation of urban mobility, SUMO simulations reveal 5% deviation between predicted and actual risks, with risk fluctuation amplitude reduced by 33.3% and 18.75% against time to collision (TTC) and dynamic safety index (DSI) methods. The proposed model shows enhanced stability in continuous driving scenarios (risk fluctuation=0.3) and demonstrates superior accuracy in forecasting potential collision risks through systematic integration of trajectory prediction and probabilistic analysis. These findings validate the framework's effectiveness in proactive safety warning for intelligent vehicles.
- traffic safety /
- predictive crash risk assessment /
- vehicle trajectory prediction /
- graph neural networks /
- Risk probability quantification

HTML全文

图 1 分段轨迹预测网络

Figure 1. Segmented trajectory prediction network

下载: 全尺寸图片幻灯片

图 2 编码部分

Figure 2. Encoding section

下载: 全尺寸图片幻灯片

图 3 风险来源车辆的局部坐标系

Figure 3. Local coordinate system of risk source vehicles

下载: 全尺寸图片幻灯片

图 4 临界碰撞距离O_AO_B^'

Figure 4. Critical collision distance O_AO_B^'

下载: 全尺寸图片幻灯片

图 5 计算示例

Figure 5. Calculation example

下载: 全尺寸图片幻灯片

图 6 示例场景

Figure 6. Example scenario

下载: 全尺寸图片幻灯片

图 7 示例场景对应的碰撞风险指数CRI_t

Figure 7. The collision risk index corresponding to the example scenario

下载: 全尺寸图片幻灯片

图 8 跟车场景中2车速度与相对距离

Figure 8. The speed and relative distance of two cars in the following scene

下载: 全尺寸图片幻灯片

图 9 跟车场景中的碰撞概率与碰撞风险指数

Figure 9. Collision probability and collision risk index in following scenarios

下载: 全尺寸图片幻灯片

图 10 跟车场景下不同碰撞风险评估方法的对比图

Figure 10. Comparison diagram of different collision risk assessment methods in car following scenarios

下载: 全尺寸图片幻灯片

图 11 跟车场景下与DSI方法的对比图

Figure 11. Comparison with the DSI method in a car-following scenario.

下载: 全尺寸图片幻灯片

图 12 切入场景各车速度与距离

Figure 12. Cutting into the scene, the speed and distance of each vehicle

下载: 全尺寸图片幻灯片

图 13 切入场景下不同碰撞风险评估方法的对比图

Figure 13. Comparison diagram of different collision risk assessment methods in different scenarios

下载: 全尺寸图片幻灯片

图 14 切入场景下不同碰撞风险评估方法的对比图

Figure 14. Comparison diagram of different collision risk assessment methods in different scenarios

下载: 全尺寸图片幻灯片

表 1 不同模型在Argoverse测试集中的结果对比

Table 1. Comparison of results of different methods in the argverse test set 单位: m

模型名	minADE6	minFDE6	MR6
LaneGCN^[4]	0.870	1.362	0.168
DenseTNT^[3]	0.882	1.282	0.126
SSL-Lanes^[5]	0.840	1.249	0.132
TPCN^[20]	0.815	1.244	0.133
HiVT^[6]	0.774	1.169	0.127
本文模型	0.785	1.157	0.126

下载: 导出CSV

参考文献(22)

[1]	JIYANG G, CHEN S, HANG Z, et al. VectorNet: encoding HD maps and agent dynamics from vectorized representation[C]. The IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), Seattle, USA: IEEE, 2020.
[2]	ZHAO H, GAO J, LAN T, et al. Tnt: Target-driven trajectory prediction[C]. Conference on Robot Learning, London, UK: PMLR, 2021.
[3]	GU J, SUN C, ZHAO H. Densetnt: end-to-end trajectory prediction from dense goal sets[C]. The IEEE/CVF International Conference on Computer Vision, Montreal, Canada: IEEE, 2021.
[4]	LIANG M, YANG B, HU R, et al. Learning lane graph representations for motion forecasting[C]. 16^th European Conference, Glasgow, UK: ECCV, 2020.
[5]	BHATTACHARYYA P, HUANG C, CZARNECKI K. Ssl-lanes: self-supervised learning for motion forecasting in autonomous driving[C]. Conference on Robot Learning, Seoul, Korea: PMLR, 2023.
[6]	ZHOU Z, YE L, WANG J, et al. Hivt: hierarchical vector transformer for multi-agent motion prediction[C]. The IEEE/ CVF Conference on Computer Vision and Pattern Recognition, New Orleans, USA : IEEE, 2022.
[7]	FAHMY H M, ABD EL GHANY M A, BAUMANN G. Vehicle risk assessment and control for lane-keeping and collision avoidance at low-speed and high-speed scenarios[J]. IEEE Transactions on Vehicular Technology, 2018, 67 (6): 4806-4818. doi: 10.1109/TVT.2018.2807796
[8]	WU B, YAN Y, NI D, et al. A longitudinal car-following risk assessment model based on risk field theory for autonomous vehicles[J]. International Journal of Transportation Science and Technology, 2021, 10(1): 60-68. doi: 10.1016/j.ijtst.2020.05.005
[9]	CHIA W M D, KEOH S L, MICHALA A L, et al. Real-time recursive risk assessment framework for autonomous vehicle operations[C]. 93^rd Vehicular Technology Conference (VTC2021-Spring), Helsinki, Finland: IEEE, 2021.
[10]	王建强, 吴剑, 李洋. 基于人-车-路协同的行车风险场概念、原理及建模[J]. 中国公路学报, 2016(1): 105-114. WANG J Q, WU J, LI Y. Concept, principle and modeling of driving risk field based on driver-vehicle-road interaction[J]. China Journal of Highway and Transport, 2016(1): 105-114. (in Chinese)
[11]	LI L, GAN J, ZHOU K, et al. A novel lane-changing model of connected and automated vehicles: using the safety potential field theory[J]. Physica A: Statistical Mechanics and its Applications, 2020, 559: 125039. doi: 10.1016/j.physa.2020.125039
[12]	田野, 裴华鑫, 晏松, 等. 车路协同环境下行车风险场模型的扩展与应用[J]. 清华大学学报(自然科学版), 2022 (3): 447-457. TIAN Y, PEI H X, YAN S, et al. Extended driving risk field model for i-VICS and its application[J]. Journal of Tsinghua University(Science and Technology), 2022(3): 447-457. (in Chinese)
[13]	KATRAKAZAS C, QUDDUS M, CHEN W H. A new integrated collision risk assessment methodology for autonomous vehicles[J]. Accident Analysis & Prevention, 2019, 127: 61-79.
[14]	LEDENT P, PAIGWAR A, RENZAGLIA A, et al. Formal validation of probabilistic collision risk estimation for autonomous driving[C]. 9^th IEEE International Conference on Cybernetics and Intelligent SystemsRobotics, Automation and Mechatronics(RAM), Bangkok, Thailand: IEEE, 2019.
[15]	STRICKLAND M, FAINEKOS G, AMOR H B. Deep predictive models for collision risk assessment in autonomous driving[C]. 2018 IEEE International Conference on Robotics and Automation(ICRA), Brisbane, Australia: IEEE, 2018.
[16]	汪磊, 李蕊君, 王菲茵. 基于QAR数据与互信息法的进近风险评估模型[J]. 交通信息与安全, 2024, 42(4): 21-29, 41. doi: 10.3963/j.jssn.1674-4861.2024.04.003 WANG Q, LI R J, WANG F Y. Approach risk assessment model based on QAR data and mutual information method[J]. Journal of Transport Information and Safety, 2024, 42(4): 21-29, 41. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2024.04.003
[17]	赵睿, 李云, 胡宏宇, 等. 基于V2I通信的交叉口车辆碰撞预警方法[J]. 吉林大学学报(工学版), 2023, 53(4): 1019-1029. ZHAO R, LI Y, HU H Y, et al. Vehicle collision warning method at intersection based on V2I communication[J]. Journal of Jilin University(Engineering Science), 2023, 53(4): 1019-1029. (in Chinese)
[18]	KRAMER B, NEUROHR C, BÜKER M, et al. Identification and quantification of hazardous scenarios for automated driving[C]. 7^th International Symposium on Model-based Safety and assessment. Cham, Germany: Springer International Publishing, 2020.
[19]	CHANG M F, LAMBERT J, SANGKLOY P, et al. Argoverse: 3D tracking and forecasting with rich maps[C]. The IEEE/CVF conference on computer vision and pattern recognition, Long Beach, USA: IEEE, 2019.
[20]	YE M, CAO T, CHEN Q. Tpcn: temporal point cloud net-works for motion forecasting[C]. The IEEE/CVF Conference on Computer Vision and Pattern Recognition: IEEE, 2021.
[21]	褚端峰, 彭赛骞, 胡海洋, 等. 预见性驾驶风险场模型[J]. 机械工程学报, 2024, 60(10): 160-170. CHU D F, PENG S Q, HU H Y, et al. Predictive driving risk field model[J]. Journal of Mechanical Engineering, 2024, 60(10): 160-170. (in Chinese)
[22]	LOPEZ A, PABLO, BEHRISCH, et al. Microscopic traffic simulation using sumo[C]. 21^st International Conference on Intelligent Transportation Systems (ITSC), Maui, Hawaii, USA: IEEE, 2018.