Graph-Motif matrix enhanced conditional diffusion model for vehicle trajectory generation

Yuan Zhoulü; He Zhaocheng; Xie Chuanzhi

doi:10.11714/acta.snus.ZR20260066

您当前的位置：

首页 >

文章列表页 >

Graph-Motif matrix enhanced conditional diffusion model for vehicle trajectory generation

Research articles | 更新时间：2026-10-20

- Graph-Motif matrix enhanced conditional diffusion model for vehicle trajectory generation
- Acta Scientiarum Naturalium Universitatis Sunyatseni Vol. 65, Issue 5, Pages: 99-108(2026)
- 作者机构：
  
  1.中山大学智能工程学院 / 广东省智能交通系统重点实验室，广东深圳 518107
  2.鹏城实验室，广东深圳 518055
- 作者简介：
- 基金信息：
- DOI：10.11714/acta.snus.ZR20260066
  CLC： U12
- Received：20 March 2026，
  
  Revised：2026-05-01，
  
  Accepted：04 May 2026，
  
  Online First：26 June 2026，
  
  Published：25 September 2026
- 稿件说明：
移动端阅览
袁洲旅,何兆成,谢传智.图模体矩阵增强下的扩散模型轨迹生成方法[J].中山大学学报(自然科学版)(中英文),2026,65(05):99-108.

Yuan Zhoulü,He Zhaocheng,Xie Chuanzhi.Graph-Motif matrix enhanced conditional diffusion model for vehicle trajectory generation[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2026,65(05):99-108.
袁洲旅,何兆成,谢传智.图模体矩阵增强下的扩散模型轨迹生成方法[J].中山大学学报(自然科学版)(中英文),2026,65(05):99-108. DOI： 10.11714/acta.snus.ZR20260066.

Yuan Zhoulü,He Zhaocheng,Xie Chuanzhi.Graph-Motif matrix enhanced conditional diffusion model for vehicle trajectory generation[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2026,65(05):99-108. DOI： 10.11714/acta.snus.ZR20260066.

摘要

在实际智能交通系统中，一旦物联网出现故障就会引发数据中断，这凸显了轨迹数据生成算法的重要性。现有的两种轨迹生成方法均存在不足。基于物理模型的算法不能适应动态变化的交通场景，而数据驱动方法容易忽略车辆之间的交互影响。针对交叉口这类复杂场景，本文提出一种基于条件扩散模型的轨迹生成方法。该方法引入模体矩阵量化周围车辆对目标车辆的影响程度，使用图注意力机制准确识别关键邻居车辆信息，并将条件信息引入扩散模型的反向去噪过程中生成车辆轨迹。最后，在nuScenes数据集和自建轨迹数据集上开展了验证实验。结果显示：本文方法在多个精度评价指标上均优于现有主流模型，且更加符合运动学规律。

Abstract

In practical intelligent transportation systems，failures of IoT may lead to data interruption， underscoring the importance of trajectory data generation algorithms. Existing trajectory generation methods can be categorized into two types， both of which exhibit limitations. Physics-based algorithms are unable to adapt to dynamically changing traffic scenarios， whereas data-driven methods tend to overlook interactions among vehicles. To address complex scenarios such as intersections， this paper proposes a trajectory generation method based on a conditional diffusion model. The proposed method introduces a motif matrix to quantify the influence of surrounding vehicles on the target vehicle，and employs a graph attention mechanism to accurately identify key neighboring vehicle information， and the conditional information is incorporated into the reverse denoising process of the diffusion model to generate vehicle trajectories. Experiments conducted on the nuScenes dataset and a self-collected trajectory dataset demonstrate that the proposed method outperforms existing mainstream models across multiple accuracy evaluation metrics and is more consistent with kinematic principles.

关键词

Keywords

references

Altché F ， de la Fortelle A ， 2017 . An LSTM network for highway trajectory prediction ［C］// 2017 IEEE 20th International Conference on Intelligent Transportation Systems . Yokohama， Japan ： 353 - 359 .

Caesar H ， Bankiti V ， Lang A H ， et al ， 2019 . nuScenes： A multimodal dataset for autonomous driving ［C］// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle， Washington， USA ： 11618 - 11628 .

Choi D ， Min K ， et al ， 2022 . Hierarchical latent structure for multi-modal vehicle trajectory forecasting ［M］// Computer Vision-ECCV 2022 . Cham ： Springer Nature Switzerland： 129 - 145 .

Dendorfer P ， Elflein S ， Leal-Taixé L ， 2021 . MG-GAN： A multi-generator model preventing out-of-distribution samples in pedestrian trajectory prediction ［C］// 2021 IEEE/CVF International Conference on Computer Vision . Montreal， QC， Canada ： 13138 - 13147 .

Deo N ， Wolff E ， Beijbom O ， 2022 . Multimodal Trajectory Prediction Conditioned on Lane-Graph Traversals ［C］// Proceedings of the 5th Conference on Robot Learning . PMLR ： 203 - 212 .

Gellida-Coutiño C ， Rios-Cabrera R ， Maldonado-Ramirez A ， et al ， 2025 . Real-time multi-camera tracking for vehicles in congested， low-velocity environments： A case study on Drive-Thru Scenarios ［J］. Electronics ， 14 （ 13 ）： 2671 .

Gu T ， Chen G ， Li J ， et al ， 2022 . Stochastic trajectory prediction via motion indeterminacy diffusion ［C］// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition . New Orleans， LA， USA ： 17092 - 17101 .

Gupta A ， Johnson J ， Li F F ， et al ， 2018 . Social GAN： Socially acceptable trajectories with generative adversarial networks ［C］// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Salt Lake City， UT， USA ： 2255 - 2264 .

Ho J ， Jain A ， Abbeel P ， 2020 . Denoising diffusion probabilistic models ［J］. Adv Neural Inf Process Syst ， 33 ： 6840 - 6851 .

Klein L A ， 2024 . Roadside sensors for traffic management ［J］. IEEE Intell Transp Syst Mag ， 16 （ 4 ）： 21 - 44 .

Leon F ， Gavrilescu M ， 2021 . A review of tracking and trajectory prediction methods for autonomous driving ［J］. Mathematics ， 9 （ 6 ）： 660 .

Li L ， Wang X ， Yang D ， et al ， 2024 . Real-time heterogeneous road-agents trajectory prediction using hierarchical convolutional networks and multi-task learning ［J］. IEEE Trans Intell Veh ， 9 （ 2 ）： 4055 - 4069 .

Li Z ， Liang H ， Wang H ， et al. ， 2023 . A multi-modal vehicle trajectory prediction framework via conditional diffusion model： A coarse-to-fine approach ［J］. Knowl Based Syst ， 280 ： 110990 .

Luo C ， Sun L ， Dabiri D ， et al ， 2020 . Probabilistic multi-modal trajectory prediction with lane attention for autonomous vehicles ［C］// 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems . 2370 - 2376 .

Mao W ， Xu C ， Zhu Q ， et al ， 2023 . Leapfrog diffusion model for stochastic trajectory prediction ［C］// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Vancouver， BC， Canada ： 5517 - 5526 .

Newell G F ， 1961 . Nonlinear effects in the dynamics of car following ［J］. Oper Res ， 9 （ 2 ）： 209 - 229 .

Nichol A Q ， Dhariwal P ， 2021 . Improved denoising diffusion probabilistic models ［C］// Proceedings of the 38th International Conference on Machine Learning . PMLR ： 8162 - 8171 .

Phan-Minh T ， Grigore E C ， Boulton F A ， et al ， 2020 . CoverNet： Multimodal behavior prediction using trajectory sets ［C］// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Virtual Online ： 14062 - 14071 .

Sliwa B ， Piatkowski N ， Wietfeld C ， 2020 . The channel as a traffic sensor： Vehicle detection and classification based on radio fingerprinting ［J］. IEEE Internet Things J ， 7 （ 8 ）： 7392 - 7406 .

Song X ， Pi R ， Lv C ， et al ， 2021 . Augmented multiple vehicles’ trajectories extraction under occlusions with roadside LiDAR data ［J］. IEEE Sens J ， 21 （ 19 ）： 21921 - 21930 .

Tashiro Y ， Song J ， Song Y ， et al ， 2021 . CSDI： Conditional score-based diffusion models for probabilistic time series imputation ［J］. Adv Neural Inf Process Syst ， 34 ： 24804 - 24816 .

Wen D ， Xu H ， He Z ， et al ， 2024 . Density-Adaptive model based on motif matrix for multi-agent trajectory prediction ［C］// 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Seattle， WA， USA ： 14822 - 14832 .

Xu C F ， Li T ， Tang C ， et al ， 2022 . PreTraM： Self-supervised pre-training via connecting trajectory and map ［EB/OL］.［ 2026-03-14 ］. https：//www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990034.pdf https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136990034.pdf .

Xu P ， Hayet J B ， Karamouzas I ， 2023 . Context-aware timewise VAEs for real-time vehicle trajectory prediction ［J］. IEEE Robot Autom Lett ， 8 （ 9 ）： 5440 - 5447 .

Yuan Y ， Weng X ， Ou Y ， et al ， 2021 . AgentFormer： Agent-aware transformers for socio-temporal multi-agent forecasting ［C］// 2021 IEEE/CVF International Conference on Computer Vision . Montreal， QC， Canada ： 9793 - 9803 .

Zhao J ， Knoop V L ， Sun J ， et al ， 2023 . Unprotected left-turn behavior model capturing path variations at intersections ［J］. IEEE Trans Intell Transp Syst ， 24 （ 9 ）： 9016 - 9030 .

Zhong Z ， Luo Y ， Liang W ， 2022 . STGM： Vehicle trajectory prediction based on generative model for spatial-temporal features ［J］. IEEE Trans Intell Transp Syst ， 23 （ 10 ）： 18785 - 18793 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Long-time dynamics for a Kirchhoff beam equation with degenerate nonlocal strong damping

Optimization of H-κ stacking using a flood-fill-based topological approach for sedimentary layer characterization

Spectra and norms of algebraic combinations of projection pairs

Solvability of a superlinear Neumann problem at resonance in the first eigenvalue

Measurement and allocation methods for carbon emissions from vehicles

Related Author

No data

Related Institution

No data

Postal code：510275
Tel：020-84112585，84113223 Email：xuebaozr@mail.sysu.edu.cn
Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19 京公网安备11010802024621
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.

⁰