面向自动驾驶汽车的信号交叉口行人多模态轨迹预测方法

doi:10.3969/j.issn.1674-8484.2024.05.007

摘要/Abstract

摘要：

为了提高自动驾驶汽车在人车混行交叉口场景下的行车安全性，提出了一种面向自动驾驶汽车的信号交叉口行人多模态轨迹预测方法。考虑社会生成对抗网络模型（SGAN）的社会属性，将行人历史轨迹作为输入，通过生成器与判别器交替训练，采用交叉熵损失函数进行模型优化，提出基于SGAN的行人轨迹预测模型；建立行人自驱力、行人间交互力、斑马线边界力和信号灯作用力的4种约束力模型，提出基于社会力模型（SFM）的行人轨迹预测模型，采用粒子群算法对SFM的不可测量参数进行标定；基于AdaBoost算法对SGAN和SFM的预测结果进行融合，通过多个弱学习器迭代训练并动态优化各模型权重，以提高模型预测准确性；实验基于西安市某交叉口行人数据进行对比验证。结果表明：相比于单一SFM模型和单一SGAN模型，该文方法的平均位移误差（ADE）和最终位移误差（FDE）分别提高了约21.7%和10.5%，尤其在绕行超越、结伴等复杂行为场景中，该方法能够实现更精准的行人轨迹预测。

关键词: 自动驾驶, 城市交叉口, 行人, 轨迹预测, 融合模型

Abstract:

A multi-modal trajectory prediction method of pedestrians at signalized intersections for autonomous vehicles was proposed to improve the driving safety of autonomous vehicles in the mixed traffic with pedestrian and vehicles. Firstly, considering the social attributes of the Social Generative Adversarial Network (SGAN) model, the pedestrian history trajectory was taken as the model input, the generator and discriminator were trained alternately, and the cross-entropy loss function was used to optimize the model, and then a pedestrian trajectory prediction model based on SGAN was proposed. Secondly, four binding force models based on pedestrian self-drive, pedestrian interaction, zebra crossing boundary force and traffic light force were established, and then a pedestrian trajectory prediction model based on Social Force Model (SFM) was proposed. The particle swarm optimization algorithm was used to calibrate the non-measurable parameters of SFM. Finally, based on the AdaBoost algorithm, the prediction results of SGAN and SFM were fused, and the weights of each model were iteratively trained and optimized dynamically by multiple weak learners to improve the prediction accuracy of the model. Based on the pedestrian data of an intersection in Xi'an city, the experimental analysis and verification were carried out. The results show that the average displacement error (ADE) and final displacement error (FDE) of the proposed method are increased by about 21.7% and 10.5%, respectively, compared with the single SFM model and the single SGAN model. The proposed model can realize more accurate pedestrian trajectory prediction.

Key words: autonomous driving, urban intersections, pedestrian, trajectory prediction, integration model

中图分类号:

U461.99

瞿广跃, 杨澜, 袁梦, 房山, 刘松岩. 面向自动驾驶汽车的信号交叉口行人多模态轨迹预测方法[J]. 汽车安全与节能学报, 2024, 15(5): 689-701.

QU Guangyue, YANG Lan, YUAN Meng, FANG Shan, LIU Songyan. A multimodal trajectory prediction method of pedestrians at signalized intersections for autonomous vehicles[J]. Journal of Automotive Safety and Energy, 2024, 15(5): 689-701.

图/表 17

参考文献 27

[1]	李克强, 熊辉, 刘金鑫. 面向弱势道路使用者的多目标运动轨迹预测方法[J]. 中国公路学报, 2022, 35(1): 298-315.
	LI Keqiang, XIONG Hui, LIU Jinxin. Multi-objective motion trajectory prediction method for vulnerable road users[J]. Chin J Highw Transport, 2022, 35(1): 298-315. (in Chinese)
[2]	Sharma N, Dhiman C, Indu S. Pedestrian intention prediction for autonomous vehicles: A comprehensive survey[J]. Neurocomput, 2022, 508: 120-152.
[3]	Fox E, Sudderth E B, Jordan M I, et al. Bayesian nonparametric inference of switching dynamic linear models[J]. IEEE Trans Sign Proce, 2011, 59(4): 1569-1585.
[4]	Kooij J F P, Schneider N, Flohr F, et al. Context-based pedestrian path prediction[C]// Proc Comput Visi-ECCV 2014: 13th Euro Conf, Part VI 13. Zurich, Switzerland, 2014: 618-633.
[5]	Rashid M M, Seyedi M R, Jung S. Simulation of pedestrian interaction with autonomous vehicles via social force model[J]. Simul Model Pract Theo, 2024, 132: 102901.
[6]	Lorenzo J, Parra I, Wirth F, et al. Rnn-based pedestrian crossing prediction using activity and pose-related features[C]// 2020 IEEE Intel Vehi Symp (IV). Las Vegas, NV, USA, IEEE, 2020: 1801-1806.
[7]	XUE Hao, Huynh D Q, Reynolds M. PoPPL: Pedestrian trajectory prediction by LSTM with automatic route class clustering[J]. IEEE Trans Neural Network Learn Syst, 2020, 32(1): 77-90.
[8]	Gupta A, Johnson J, LI Feifei, et al. Social gan: Socially acceptable trajectories with generative adversarial networks[C]// Proc IEEE Conf Comput Visi Patt Recog. Salt Lake City, UT, USA, 2018: 2255-2264.
[9]	Eiffert S, LI Kunming, SHAN Mao, et al. Probabilistic crowd GAN: Multimodal pedestrian trajectory prediction using a graph vehicle-pedestrian attention network[J]. IEEE Robot Auto Letts, 2020, 5(4): 5026-5033.
[10]	QUAN Ruijie, ZHU Linchao, WU Yu, et al. Holistic LSTM for pedestrian trajectory prediction[J]. IEEE Trans Imag Proce, 2021, 30: 3229-3239.
[11]	XU Yi, WANG Lichen, WANG Yizhou, et al. Adaptive trajectory prediction via transferable GNN[C]// Proc IEEE / CVF Conf Comput Visi Patt Recog. New Orleans, LA, USA, 2022: 6520-6531.
[12]	Dendorfer P, Elflein S, Leal-Taixé L. Mg-gan: A multi-generator model preventing out-of-distribution samples in pedestrian trajectory prediction[C]// Proc IEEE/CVF Int’l Conf Comput Visi. Montreal, Canada, 2021: 13158-13167.
[13]	KE Kexin, YANG Jian, LIU Yingjie, et al. Social Lode: Human trajectory prediction with latent odes[C]// ICASSP 2024-2024 IEEE Int’l Conf Acoust, Speech Sign Proc (ICASSP), Seoul, South Korea, IEEE, 2024: 5360-5364.
[14]	CHEN Hao, LIU Yinhua, HU Chuan, et al. Vulnerable road user trajectory prediction for autonomous driving using a data-driven integrated approach[J]. IEEE Trans Intel Transport Syst, 2023, 24(7): 7306-7317.
[15]	Sheykhfard A, Haghighi F, Kavianpour S, et al. Risk assessment of pedestrian red-light violation behavior using surrogate safety measures: Influence of human, road, vehicle, and environmental factors[J]. IATSS Res, 2023, 47(4): 514-525.
[16]	周于涛, 吴华意, 成洪权, 等. 结合自注意力机制和结伴行为特征的行人轨迹预测模型[J]. 武汉大学学报(信息科学版), 2020, 45(12): 1989-1996.
	ZHOU Yutao, WU Huayi, CHENG Hongquan, et al. Pedestrian trajectory prediction model combining self-attention mechanism and companion behavior characteristics[J]. Geomat Info Sci Wuhan Univ, 2020, 45(12): 1989-1996. (in Chinese)
[17]	YANG Junheng, ZANG Xiaodong, CHEN Weiying, et al. Improved social force model based on pedestrian collision avoidance behavior in counterflow[J]. Phys A: Statist Mech Its Appl, 2024, 642: 129762.
[18]	LIU Yenchen, Jafari A, Shim J K, et al. Dynamic modeling and simulation of electric scooter interactions with a pedestrian crowd using a social force model[J]. IEEE Trans Intel Transport Syst, 2022, 23(9): 16448-16461.
[19]	ZENG Weiliang, Nakamura H, CHEN Peng. A modified social force model for pedestrian behavior simulation at signalized crosswalks[J]. Procedia-Social Beha Sci, 2014, 138: 521-530.
[20]	HU Xiangmin, CHEN Tao, SONG Yushan. Anticipation dynamics of pedestrians based on the elliptical social force mode l[J]. Chaos: An Interdisc J Nonlin Sci, 2023, 33: 073102.
[21]	ZENG Weiliang, CHEN Peng, Nakamura H, et al. Application of social force model to pedestrian behavior analysis at signalized crosswalk[J]. Transport Res Part C: Emerg Tech, 2014, 40: 143-159.
[22]	曾浩, 蔡万翰, 任志刚, 等. 基于动态搜索策略离散粒子群算法的稀疏阵列约束优化[J]. 电子学报, 2023, 51(4): 783-791.
	ZENG Hao, CAI Wanhan, REN Zhigang, et al. Sparse array constraint optimization based on discrete particle swarm optimization based on dynamic search strategy[J]. Acta Electron Sini, 2023, 51(4): 783-791. (in Chinese)
[23]	Belghit A, Lazri M, Ouallouche F, et al. Optimization of one versus All-SVM using AdaBoost algorithm for rainfall classification and estimation from multispectral MSG data[J]. Advan Space Res, 2023, 71(1): 946-963.
[24]	刘延钊, 黄志球, 沈国华, 等. 基于决策树和BN的自动驾驶车辆行为决策方法[J]. 系统工程与电子技术, 2022, 44(10): 3143-3154.
	LIU Yanzhao, HUANG Zhiqiu, SHEN Guohua, et al. Behavioral decision-making method of autonomous vehicles based on decision tree and BN[J]. Syst Engi Electron, 2022, 44(10): 3143-3154. (in Chinese)
[25]	Klusowski J M, Tian P M. Large scale prediction with decision trees[J]. J Am Statist Asso, 2024, 119(545): 525-537.
[26]	XU Yanchao, SHAO Wenbo, LI Jun, et al. SIND: A drone dataset at signalized intersection in China[C]// 2022 IEEE 25th Int’l Conf Intel Transport Syst (ITSC), Macau, China, IEEE, 2022: 2471-2478.
[27]	桑海峰, 王金玉, 陈旺兴, 等. 基于第一视角的非自回归行人轨迹预测模型[J]. 电子学报, 2023, 51(5): 1266-1272.
	SANG Haifeng, WANG Jinyu, CHEN Xinwang, et al. Non-autoregressive pedestrian trajectory prediction model based on first-person perspective[J]. Acta Electron Sinica, 2023, 51(5): 1266-1272. (in Chinese)

参数	标定值
行人作用强度系数，A_p	0.83
行人距离影响系数，B_p	1.89
斑马线外约束力强度，A_b	0.45
斑马线距离影响系数，B_b	0.94
斑马线内约束力强度，A_br	0.22
斑马线距离影响系数，B_br	0.77
信号灯作用强度系数，A_s	0.15
信号灯距离影响系数，B_s	0.21

参数	标定值
行人作用强度系数，A_p	0.83
行人距离影响系数，B_p	1.89
斑马线外约束力强度，A_b	0.45
斑马线距离影响系数，B_b	0.94
斑马线内约束力强度，A_br	0.22
斑马线距离影响系数，B_br	0.77
信号灯作用强度系数，A_s	0.15
信号灯距离影响系数，B_s	0.21

Model	ADE / m	FDE / m
SFM	0.42	0.46
SGAN	0.36	0.41
融合模型	0.32	0.37

Model	ADE / m	FDE / m
SFM	0.42	0.46
SGAN	0.36	0.41
融合模型	0.32	0.37