@article{ref1,
title="Vehicle trajectory forecasting network based on static scene context modulation for autonomous driving",
journal="Transactions of the Korean Society of Automotive Engineers",
year="2023",
author="Choi, Dooseop and Min, KyoungWook",
volume="31",
number="8",
pages="597-606",
abstract="In this paper, we are proposing a vehicle trajectory forecasting network based on static scene context modulation. First, in modeling the distribution over future trajectories efficiently via variational auto-encoder frameworks, we suggest using a transformer-based trajectory encoder that models the interaction between neighboring vehicles. The proposed encoder is trained to remove interaction between irrelevant vehicles, and model key interaction more efficiently. Moreover, to increase the diversity of generated trajectories, we propose using latent variables during the trajectory generation process in modulating static scene context. Then, we can use large-scale, real-world datasets like nuScenes in evaluating performance. Experimental results showed that the proposed model generates plausible and diverse future trajectories with the techniques proposed in this paper. Furthermore, it outperformed the baseline models in terms of prediction accuracy.  	 Keywords: Autonomous driving, Deep learning, Trajectory forecasting, Planning, Context modulation 키워드: 자율주행, 딥러닝, 궤적 예측, 판단, 컨텍스트 변조<p /> <p>Language: ko</p>",
language="ko",
issn="1225-6382",
doi="10.7467/KSAE.2023.31.8.597",
url="http://dx.doi.org/10.7467/KSAE.2023.31.8.597"
}