面向智能驾驶测试的可变跟驰特性交通车建模方法

doi:10.19562/j.chinasae.qcgc.2024.11.002

摘要/Abstract

摘要：

提出一种面向智能驾驶测试的可变跟驰特性交通车建模方法。首先，通过对自然驾驶数据聚类分析，建立高真实交互个性化的跟驰模型，并利用模型输出耦合赋予多元权值，构建可用于智能驾驶测试的可变跟驰特性交通车模型；然后，通过建立交通车轨迹评价方法验证模型输出轨迹的合理性、多样性及真实性；最后，搭建联合仿真平台进行了所构建交通车模型对自动紧急制动（autonomous emergency braking， AEB）算法的应用测试。结果表明，本文构建的交通车模型可以输出不同跟驰特性下合理、多样且真实的轨迹，当轨迹数量达到60 000条时与真实自然驾驶速度轨迹匹配的平均均方根误差为0.427 m/s，且在不同交通车轨迹特性下待测系统行为响应不尽相同，通过权值系数的变化可以揭示待测系统响应的演化规律，并可实现待测系统性能的针对性测试。

关键词: 智能驾驶测试, 交通车建模, 可变跟驰特性, Transformer网络, 多元权值分配

Abstract:

A variable following characteristic traffic vehicle modeling method for intelligent driving system testing is proposed in this paper. Firstly， by clustering and analyzing natural driving data， a highly realistic interactive personalized car following model is established， and the model output coupling is used to assign multiple weights to construct a traffic vehicle model with variable car following characteristics that can be used for intelligent driving system testing. Then， by establishing the traffic vehicle trajectory evaluation method， the rationality， diversity and authenticity of the model's output trajectory are verified. Finally， a joint simulation platform is built to test the application of the constructed traffic vehicle model to the Automatic Emergency Braking （AEB） algorithm. The results show that the traffic vehicle model constructed in this paper can output reasonable， diverse， and realistic trajectories under different car following characteristics. When the number of trajectories reaches 60 000， the average root mean square error matched with the real natural driving speed trajectory is 0.427 m/s. Moreover， the behavioral response of the tested system varies under different traffic vehicle trajectory characteristics. By changing the weight coefficients， the evolution law of the tested system response can be revealed， and targeted testing of the tested system performance can be achieved.

Key words: intelligent driving system testing, traffic vehicle modeling, variable car following characteristic, Transformer network, multiple weight allocation

赵健,李文旭,朱冰,张培兴,汤瑞,李嘉胜. 面向智能驾驶测试的可变跟驰特性交通车建模方法[J]. 汽车工程, 2024, 46(11): 1952-1961.

Jian Zhao,Wenxu Li,Bing Zhu,Peixing Zhang,Rui Tang,Jiasheng Li. A Modeling Method for Traffic Vehicle with Variable Car Following Characteristic for Intelligent Driving System Testing[J]. Automotive Engineering, 2024, 46(11): 1952-1961.

图/表 17

图1

表1

表2

表3

表4

表5

表6

表7

图2

表8

图3

图4

图5

图6

图7

图8

图9

参考文献 9

1	HU X， ZHENG Z， CHEN D， et al. Processing， assessing， and enhancing the Waymo autonomous vehicle open dataset for driving behavior research［J］. Transportation Research Part C： Emerging Technologies， 2022， 134： 103490.
2	朱冰，张培兴，赵健，等. 基于场景的自动驾驶汽车虚拟测试研究进展［J］. 中国公路学报， 2019， 32（6）： 1-19.
	ZHU B， ZHANG P X， ZHAO J， et al. Review of scenario-based virtual validation methods for automated vehicles［J］. China Journal of Highway and Transport， 2019， 32（6）： 1-19.
3	MA L， QU S， REN J， et al. Mixed traffic flow of human-driven vehicles and connected autonomous vehicles： string stability and fundamental diagram［J］. Math. Biosci. Eng， 2023， 20： 2280-2295.
4	YAO Z， WU Y， WANG Y， et al. Analysis of the impact of maximum platoon size of CAVs on mixed traffic flow： an analytical and simulation method［J］. Transportation Research Part C： Emerging Technologies， 2023， 147： 103989.
5	XIE Y Z， ZHANG Y， DAI K P， et al. A real-time critical-scenario-generation framework for defect detection of autonomous driving system［J］. IET Intelligent Transport Systems， 2023： 1-15.
6	FENG S， SUN H， YAN X， et al. Dense reinforcement learning for safety validation of autonomous vehicles［J］. Nature， 2023， 615（7953）： 620-627.
7	ZHU B， ZHANG P X， ZHAO J， et al. Hazardous scenario enhanced generation for automated vehicle testing based on optimization searching method［J］. IEEE Transactions on Intelligent Transportation Systems， 2021， 23（7）： 7321-7331.
8	ZHU B， SUN Y， ZHAO J， et al. A critical scenario search method for intelligent vehicle testing based on the social cognitive optimization algorithm［J］. IEEE Transactions on Intelligent Transportation Systems， 2023， 24（8）： 7974-7986.
9	KRAJEWSKI R， BOCK J， KLOEKER L， et al. The highd Dataset： a drone dataset of naturalistic vehicle trajectories on German highways for validation of highly automated driving systems［C］. 2018 21st International Conference on Intelligent Transportation Systems （ITSC）， 2018.

序号	特征变量	单位
1	X₁-纵向速度绝对值的均值	m·s^-1
2	X₂-纵向速度绝对值的标准差	m·s^-1
3	X₃-纵向加速度绝对值的均值	m·s^-2
4	X₄-纵向加速度绝对值的标准差	m·s^-2
5	X₅-纵向冲击度的均值	m·s^-3
6	X₆-纵向冲击度的标准差	m·s^-3

主成分	特征值	贡献率	累计贡献率
1	3.113	51.9%	51.9%
2	1.067	17.8%	69.7%
3	0.970	16.2%	85.9%

变量	主成分
变量	1	2	3
X₁	0.011	-0.871	0.435
X₂	0.484	0.122	-0.017
X₃	0.518	-0.178	0.053
X₄	0.523	0.083	0.071
X₅	0.093	-0.425	-0.895
X₆	0.465	0.083	0.047

特征变量	保守型	一般型	激进型
X₁-纵向速度均值	23.182	31.007	36.739
X₂-纵向速度标准差	0.839	0.872	0.888
X₃-纵向加速度均值	0.189	0.252	0.293
X₄-纵向加速度标准差	0.118	0.137	0.136
X₅-纵向冲击度均值	0.003	0.004	0.006
X₆-纵向冲击度标准差	0.143	0.151	0.153

参数	数值
输入序列维度	4
输出序列维度	1
d_model	128
注意力机制头数	8
学习率	0.000 5
网络层数	3
隐藏层神经元个数	64
Dropout	0.1
迭代次数	30
Batch_Size	64