| Iterative Trajectory Prediction Model Based on Interactive Agent |
| Hongpeng Tian, Xiaopei Zhang, Dan Cui |
| College of Computer Science and Technology, Xi’an University of Science and Technology, Xi’an, 710600, China |
|
|
|
Received: May 29, 2024; Revised: August 7, 2024 Accepted: August 21, 2024. Published online: October 7, 2024. |
|
|
|
| ABSTRACT |
|
Though many methods attempt to model how agents interact with each other and their environment, they often fail to capture the dynamic nuances of these interactions. To address this issue, this paper proposes an interactive iterative prediction model based on Transformer (IIPM-BT) that can distinguish between agents and accurately model the interactions between them. The encoder in IIPM-BT includes Map-Transformer and InterAgent-Transformer modules. The Map-Transformer module combines local maps to provide a real-time context for trajectory prediction. The InterAgent-Transformer module captures the interactive information between agents to understand the dynamic relationship between multiple agents. The decoder employs an iterative prediction strategy to refine future trajectories. This model uses the Waymo Open Motion Dataset to train and evaluate. Compared with the HDGT, our model performs significantly better in ADE, MR and MAP indicators, which are reduced by 7.02%, 21.43% and improved by 35.71% respectively. Experimental results show that the model has good performance. |
| Key Words:
Autonomous driving · Deep learning · Interactive agent · Attention mechanism · Iterative prediction |
|
Corresponding Author.
Hongpeng Tian , Email. 
Preview 
Full text via DOI 
Download Citation
Print
