Zhang, Xinyu (Tongji University), Zhou, Zewei (Tongji University), Ji, Yangjie (Tongji University), Xing, Jiaming (Tongji University), Wang, Zhaoyi (Tongji University), Huang, Yanjun (Tongji University)

Co-HTTP: Cooperative Trajectory Prediction with Heterogeneous Graph Transformer for Autonomous Vehicles

Scheduled for presentation during the Invited Session "Data-driven and Learning-based Control Techniques for Intelligent Vehicles" (FrAT1), Friday, September 27, 2024, 12:10−12:30, Salon 1

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

This paper proposes a cooperative prediction framework named Co-HTTP which enables collaboration between vehicle-side prediction (Veh-Pred) and infrastructure-side prediction (Infra-Pred). First, the cleaned infrastructure's historical data is conveyed to the autonomous vehicles (AVs) for initial collaboration. Then, this paper predicts the target vehicles based on the driving intentions of AVs besides historical trajectories. Additionally, the intentions of agents around are all embedded into a heterogeneous graph neural network (GNN), updated by Transformer layers in a heterogeneous way, to further enhance vehicle-infrastructure cooperation. Finally, Co-HTTP is evaluated on V2X-Seq dataset, demonstrating that its prediction performance is better than the current state-of-the-art method on this dataset. As a learnable cooperative prediction paradigm, Co-HTTP will contribute to the improvement of autonomous vehicles.