TY - JOUR

T1 - GATraj

T2 - A graph- and attention-based multi-agent trajectory prediction model

AU - Cheng, Hao

AU - Liu, Mengmeng

AU - Chen, Lin

AU - Broszio, Hellward

AU - Sester, Monika

AU - Yang, Michael Ying

N1 - Funding Information: This work is partially supported by the MSCA European Postdoctoral Fellowships under the 101062870 – VeVuSafety project and partially performed in the framework of project KaBa (Kamerabasierte Bewegungsanalyse aller Verkehrsteilnehmer für automatisiertes Fahren) supported by the European Regional Development Fund at VISCODA company.

PY - 2023/11

Y1 - 2023/11

N2 - Trajectory prediction has been a long-standing problem in intelligent systems like autonomous driving and robot navigation. Models trained on large-scale benchmarks have made significant progress in improving prediction accuracy. However, the importance on efficiency for real-time applications has been less emphasized. This paper proposes an attention-based graph model, named GATraj, which achieves a good balance of prediction accuracy and inference speed. We use attention mechanisms to model the spatial–temporal dynamics of agents, such as pedestrians or vehicles, and a graph convolutional network to model their interactions. Additionally, a Laplacian mixture decoder is implemented to mitigate mode collapse and generate diverse multimodal predictions for each agent. GATraj achieves state-of-the-art prediction performance at a much higher speed when tested on the ETH/UCY datasets for pedestrian trajectories, and good performance at about 100 Hz inference speed when tested on the nuScenes dataset for autonomous driving. We conduct extensive experiments to analyze the probability estimation of the Laplacian mixture decoder and compare it with a Gaussian mixture decoder for predicting different multimodalities. Furthermore, comprehensive ablation studies demonstrate the effectiveness of each proposed module in GATraj.

AB - Trajectory prediction has been a long-standing problem in intelligent systems like autonomous driving and robot navigation. Models trained on large-scale benchmarks have made significant progress in improving prediction accuracy. However, the importance on efficiency for real-time applications has been less emphasized. This paper proposes an attention-based graph model, named GATraj, which achieves a good balance of prediction accuracy and inference speed. We use attention mechanisms to model the spatial–temporal dynamics of agents, such as pedestrians or vehicles, and a graph convolutional network to model their interactions. Additionally, a Laplacian mixture decoder is implemented to mitigate mode collapse and generate diverse multimodal predictions for each agent. GATraj achieves state-of-the-art prediction performance at a much higher speed when tested on the ETH/UCY datasets for pedestrian trajectories, and good performance at about 100 Hz inference speed when tested on the nuScenes dataset for autonomous driving. We conduct extensive experiments to analyze the probability estimation of the Laplacian mixture decoder and compare it with a Gaussian mixture decoder for predicting different multimodalities. Furthermore, comprehensive ablation studies demonstrate the effectiveness of each proposed module in GATraj.

KW - Autonomous driving

KW - Graph model

KW - Mixture density network

KW - Pedestrian

KW - Trajectory prediction

UR - http://www.scopus.com/inward/record.url?scp=85174064136&partnerID=8YFLogxK

U2 - 10.48550/arXiv.2209.07857

DO - 10.48550/arXiv.2209.07857

M3 - Article

VL - 205

SP - 163

EP - 175

JO - ISPRS Journal of Photogrammetry and Remote Sensing

JF - ISPRS Journal of Photogrammetry and Remote Sensing

SN - 0924-2716

ER -