Social Interaction-Aware Motion Prediction of Surrounding Vehicles Based on Edge-Pruned Graph Convolutional Network for Autonomous Vehicles
IEEE Sensors Journal · 2026-04 · 26 (8) · 11667-11678
DOI: 10.1109/jsen.2026.3670964 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, 2026-04. Verfasst von Youngmin Yoon und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 68 Literaturangaben. Gefördert von: Chung-Ang University Research Grants in 2025, Faculty Research Fund of Sejong University in 2025.
Abstract (Originalsprache)
This article presents a method for predicting the motion of surrounding vehicles based on a deep neural network for autonomous vehicles. The proposed algorithm derives spatio-temporal information for the future motion of target vehicles, considering intervehicular interactions. Long short-term memories (LSTMs) are used to model the temporal dependencies of motion information. To effectively model intervehicular interactions, we construct the traffic scene as a graph with edge weights learned via a self-attention mechanism. We introduce an edge-pruned graph convolutional network (GCN). This utilizes a differentiable pruning mechanism to filter out low-importance edges and simplify the graph structure. This approach enhances the network’s focus on critical social interactions. For evaluation, we collected an in-house driving dataset using a test vehicle and utilized the public NGSIM dataset. Experimental results demonstrate that the proposed edge-pruned network achieves higher prediction accuracy than nonpruned baselines. Furthermore, the model shows improved performance in proactively detecting safety-critical maneuvers. Computational efficiency analysis confirms the suitability of the model for real-time implementation in autonomous driving applications.
Förderung
- Chung-Ang University Research Grants in 2025
- Faculty Research Fund of Sejong University in 2025
Zitieren
APA
Yoon, Y., & Jo, A. (2026). Social Interaction-Aware Motion Prediction of Surrounding Vehicles Based on Edge-Pruned Graph Convolutional Network for Autonomous Vehicles. IEEE Sensors Journal, 26(8), 11667-11678. https://doi.org/10.1109/jsen.2026.3670964
Vancouver
Yoon Y, Jo A. Social Interaction-Aware Motion Prediction of Surrounding Vehicles Based on Edge-Pruned Graph Convolutional Network for Autonomous Vehicles. IEEE Sensors Journal. 2026;26(8):11667-11678. doi:10.1109/jsen.2026.3670964
BibTeX
@article{Yoon2026,
author = {Yoon, Youngmin and Jo, Ara},
title = {Social Interaction-Aware Motion Prediction of Surrounding Vehicles Based on Edge-Pruned Graph Convolutional Network for Autonomous Vehicles},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {8},
pages = {11667--11678},
doi = {10.1109/jsen.2026.3670964}
}
EUVIMED