EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

BEVTraj: Map-free multimodal trajectory prediction with sensor-derived BEV context and sparse goal reasoning

Minsang Kong, Myeongjun Kim, Sang Gu Kang, Hejiu Lu, Yupeng Zhong, Sang Hun Lee

Journal of Computational Design and Engineering · 02.11.2026 · 13 (11) · 1-30

DOI: 10.1093/jcde/qwag088 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Computational Design and Engineering, online erschienen am 21.10.2026. Verfasst von Minsang Kong und 5 weiteren Autor:innen. Die Arbeit nennt 81 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Research Foundation of Korea, MSIT.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Trajectory prediction is a key task in autonomous driving, where a model must forecast plausible future motions of surrounding agents from dynamic histories and scene context. Many recent methods use high-definition (HD) maps to provide explicit road topology and semantic constraints, but such maps are costly to maintain, geographically limited, and difficult to update in changing environments. This paper presents Bird’s-Eye View Trajectory Prediction (BEVTraj), a map-free trajectory prediction framework that directly uses BEV features constructed from raw multi-sensor data. Since sensor-derived BEV features are dense and only implicitly structured, BEVTraj employs deformable attention to selectively aggregate agent-relevant spatial context from sparse locations rather than processing the entire BEV grid uniformly. In addition, a Sparse Goal Candidate Proposal module combines overlapping mode-wise anchor clusters with attention-based goal prediction to generate compact, scene-adaptive goal hypotheses. These goals are decoded and iteratively refined into multimodal trajectories, while trajectory-aware mode scoring and temporal consistency modeling improve confidence estimation and suppress abrupt, physically implausible refinement updates. Experiments on nuScenes and Argoverse 2 Sensor show that BEVTraj achieves competitive results under a map-free setting without using pre-built HD maps, while ablation studies and qualitative analyses illustrate the roles of BEV context aggregation and sparse goal reasoning. The source code is available at https://github.com/Kongminsang/bevtraj.

Förderung

  • National Research Foundation of Korea
  • MSIT

Zitieren

APA

Kong, M., Kim, M., Kang, S. G., Lu, H., Zhong, Y., & Lee, S. H. (2026). BEVTraj: Map-free multimodal trajectory prediction with sensor-derived BEV context and sparse goal reasoning. Journal of Computational Design and Engineering, 13(11), 1-30. https://doi.org/10.1093/jcde/qwag088

Vancouver

Kong M, Kim M, Kang SG, Lu H, Zhong Y, Lee SH. BEVTraj: Map-free multimodal trajectory prediction with sensor-derived BEV context and sparse goal reasoning. Journal of Computational Design and Engineering. 2026;13(11):1-30. doi:10.1093/jcde/qwag088

BibTeX
@article{Kong2026,
  author = {Kong, Minsang and Kim, Myeongjun and Kang, Sang Gu and Lu, Hejiu and Zhong, Yupeng and Lee, Sang Hun},
  title = {BEVTraj: Map-free multimodal trajectory prediction with sensor-derived BEV context and sparse goal reasoning},
  journal = {Journal of Computational Design and Engineering},
  year = {2026},
  volume = {13},
  number = {11},
  pages = {1--30},
  doi = {10.1093/jcde/qwag088}
}
Kostenlos starten →