计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (1): 97-103.doi: 10.11896/jsjkx.250300132

• 数据库&大数据&数据科学 • 上一篇    下一篇

基于图注意力交互的行人轨迹预测方法

刘宏鉴, 邹丹平, 李萍   

  1. 上海交通大学电子信息与电气工程学院 上海 200240
  • 收稿日期:2025-03-24 修回日期:2025-05-22 发布日期:2026-01-08
  • 通讯作者: 邹丹平(dpzou@sjtu.edu.cn)
  • 作者简介:(liuhongjian@sjtu.edu.cn)
  • 基金资助:
    国家重点研发计划(2022YFB3903802)

Pedestrian Trajectory Prediction Method Based on Graph Attention Interaction

LIU Hongjian, ZOU Danping, LI Ping   

  1. School of Electronics, Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
  • Received:2025-03-24 Revised:2025-05-22 Online:2026-01-08
  • About author:LIU Hongjian,born in 1999,postgra-duate.His main research interest is perception,prediction and planning control of autonomous systems.
    ZOU Danping,born in 1982,Ph.D,professor,Ph.D supervisor.His main research interests include synchronous positioning and map construction,3D visual perception and autonomous systems.
  • Supported by:
    National Key R&D Program of China(2022YFB3903802).

PDF (PC)

13

摘要： 行人轨迹预测在自动驾驶领域和智慧交通领域均取得了显著的研究进展。由于行人的行为受到自身和环境因素的双重影响,其轨迹具有不确定性和复杂性,因此准确利用轨迹数据的交互特征生成多模态轨迹仍存在较大挑战。目前,该领域中的主要挑战是准确建模行人之间的时空交互。面对复杂的行人时空交互,提出了一种基于图注意力的时空图神经网络,其量化表示行人之间的空间交互并重点关注关键交互,从而将行人轨迹信息表示为有向时空图,利用图注意力机制提取空间位置特征和交互特征,同时结合自注意力机制在时间维度提取时间特征并融合时空特征信息,最后生成结合历史轨迹和交互信息的多模态未来轨迹。在ETH-UCY数据集上的实验表明,与最佳基线模型相比,所提出的方法在平均位移误差(ADE)和最终位移误差(FDE)方面分别降低3.4%和2.1%,并具有较短的推理时间,确保实现实时推理响应。可视化的结果表明,所提出的方法能够生成具有可接受性的未来行人轨迹,展现了良好的工程应用前景。

关键词: 轨迹预测, 时空图, 图神经网络, 图注意力, 时空交互

Abstract: Pedestrian trajectory prediction has made significant research progress in the fields of autonomous driving and intelligent transportation.Due to the dual influence of individual and environmental factors,pedestrian trajectories exhibit uncertainty and complexity.Accurately generating multimodal trajectories by leveraging the interactive features of trajectory data remains a challenge.One of the primary challenges in this field is the accurate modeling of spatial temporal interactions among pedestrians.To address the complexity of pedestrian spatial temporal interactions,this paper proposes a spatial temporal graph neural network based on graph attention.The proposed method quantitatively represents the spatial interactions between pedestrians,focusing on key interactions,and represents pedestrian trajectory information as a directed spatial temporal graph.The spatial position features and interaction features are extracted using a graph attention mechanism,while the temporal features are obtained using a self-attention mechanism.By integrating spatial temporal feature information,the model generates multimodal future trajectories based on historical trajectory data and interaction information.Experiments conducted on the publicly available ETH-UCY dataset demonstrate that the proposed method outperforms the baseline models,achieving improvements of 3.4% and 2.1% in ADE and FDE,respectively.Additionally,the proposed model has a shorter inference time,ensuring real-time inference responses.Visualization results further indicate that the predicted pedestrian trajectories are plausible and socially acceptable,demonstrating promising prospects for engineering applications.

Key words: Trajectory prediction, Spatial temporal graph, Graph neural networks, Graph attention, Spatial temporal interaction

中图分类号: 

  • TP391
[1]GAO J,SUN C,ZHAO H,et al.Vectornet:Encoding hd maps and agent dynamics from vectorized representation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:11525-11533.
[2]YU B,YIN H,ZHU Z.Spatio-temporal graph convolutional networks:A deep learning framework for traffic forecasting[J].arXiv:1709.04875,2017.
[3]易虹宇,杨智宇,杜力.基于变分自动编码器的车辆轨迹预测研究[J].重庆工商大学学报(自然科学版),2024,41(2):60-65.
[4]DAAMEN W,HOOGENDOORN S P.Experimental research of pedestrian walking behavior[J].Transportation Research Record,2003,1828(1):20-30.
[5]ALAHI A,GOEL K,RAMANATHAN V,et al.Social lstm:Human trajectory prediction in crowded spaces[C]//Procee-dings of the IEEE Conference on Computer Vision and Pattern Recognition.2016:961-971.
[6]GUPTA A,JOHNSON J,FEI-FEI L,et al.Social gan:Socially acceptable trajectories with generative adversarial networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018:2255-2264.
[7]MOHAMED A,QIAN K,ELHOSEINY M,et al.Social-stgcnn:A social spatio-temporal graph convolutional neural network for human trajectory prediction[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:14424-14432.
[8]HOU L,LIU J,YU X,et al.Review of graph neural network[J].Computer Science,2024,51(6):282-298.
[9]CAO J,CHEN Y,LI H,et al.A survey of pedestrian trajectory prediction based on graph neural network[J].Computer Engineering & Science,2023,45(6):1040-1053.
[10]SHI L,WANG L,LONG C,et al.SGCN:Sparse graph convolution network for pedestrian trajectory prediction[C]//Procee-dings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2021:8994-9003.
[11]ZHANG X,ANGELOUDIS P,DEMIRIS Y.Dual-branch spatio-temporal graph neural networks for pedestrian trajectory prediction[J].Pattern Recognition,2023,142:109633.
[12]VELICKOVIC P,CUCURULL G,CASANOVA A,et al.Graph attention networks[J].arXiv:1710.10903,2017.
[13]ZHAO H,WANG Y,DUAN J,et al.Multivariate time-seriesanomaly detection via graph attention network[C]//2020 IEEE International Conference on Data Mining(ICDM).IEEE,2020:841-850.
[14]VASWANI A,SHAZEER N,PARMAR N,et al.Attention is all you need[J].arXiv:1706.03762,2017.
[15]DOSOVITSKIY A.An image is worth 16×16 words:Transfor-mers for image recognition at scale[J].arXiv:2010.11929,2020.
[16]BRODY S,ALON U,YAHAV E.How attentive are graph attention networks?[J].arXiv:2105.14491,2021.
[17]BUSBRIDGE D,SHERBURN D,CAVALLO P,et al.Relational graph attention networks[J].arXiv:1904.05811,2019.
[18]SHAO K,ZHANG Y,WEN Y,et al.DTI-HETA:prediction of drug-target interactions based on GCN and GAT on heterogeneous graph[J].Briefings in Bioinformatics,2022,23(3):bbac109.
[19]WANG X,HE X,CAO Y,et al.Kgat:Knowledge graph attention networkfor recommendation[C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Disco-very & Data Mining.2019:950-958.
[20]LERNER A,CHRYSANTHOU Y,LISCHINSKI D.Crowds by example[C]//Computer graphics forum.Oxford,UK:Blackwell Publishing Ltd,2007:655-664.
[21]PELLEGRINI S,ESS A,SCHINDLER K,et al.You’ll never walk alone:Modeling social behavior for multi-target tracking[C]//2009 IEEE 12th International Conference on Computer Vision.IEEE,2009:261-268.
[22]KOSARAJU V,SADEGHIAN A,MARTÍN-MARTÍN R,et al.Social-bigat:Multimodal trajectory forecasting using bicycle-gan and graph attention networks[C]//Proceedings of the 33rd International Conference on Neural Information Processing Systems,2019:137-146.
[23]MANGALAM K,GIRASE H,AGARWAL S,et al.It is not the journey but the destination:Endpoint conditioned trajectory prediction[C]//Computer Vision-ECCV 2020:16th European Conference,Glasgow,UK,August 23-28,2020,Proceedings,Part II 16.Springer International Publishing,2020:759-776.
[24]CHEN W,SANG H,WANG J,et al.DSTCNN:Deformablespatial-temporal convolutional neural network for pedestrian trajectory prediction[J].Information Sciences,2024,666:120455.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发