EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Diffusion Models for End-to-End Autonomous Driving: A Survey of Perception, Prediction, Planning, and Control

Junsu Choi, Juhyung Park, Sungjin Lee

IEEE Access · 2026 · 14 · 95498-95518

DOI: 10.1109/access.2026.3704955 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Junsu Choi und 2 weiteren Autor:innen. Die Arbeit nennt 135 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Soonchunhyang University.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents a systematic survey of End-to-End (E2E) autonomous driving studies that leverage Diffusion Models. The conventional perception–prediction–planning–control pipeline facilitates verification, but suffers from inter-module information loss and error accumulation. Existing E2E approaches simplify the driving stack, yet they have been criticized for multimodal averaging, difficulty in incorporating constraints, domain shift, and open-loop/closed-loop discrepancies. Diffusion-based E2E methods sample multiple candidates through conditional generation and incorporate safety, rule compliance, and ride comfort constraints at inference time via guidance. Based on the point of application, this paper categorizes related studies into (T1) perception, (T2) prediction, (T3) planning, (T4) control, and (T5) scenario diffusion, and compares representative datasets and evaluation protocols. It further discusses key challenges, including real-time inference, safety verification, long-horizon rollout stability, and bias in synthetic scenarios.

Förderung

  • Soonchunhyang University

Zitieren

APA

Choi, J., Park, J., & Lee, S. (2026). Diffusion Models for End-to-End Autonomous Driving: A Survey of Perception, Prediction, Planning, and Control. IEEE Access, 14, 95498-95518. https://doi.org/10.1109/access.2026.3704955

Vancouver

Choi J, Park J, Lee S. Diffusion Models for End-to-End Autonomous Driving: A Survey of Perception, Prediction, Planning, and Control. IEEE Access. 2026;14:95498-95518. doi:10.1109/access.2026.3704955

BibTeX
@article{Choi2026,
  author = {Choi, Junsu and Park, Juhyung and Lee, Sungjin},
  title = {Diffusion Models for End-to-End Autonomous Driving: A Survey of Perception, Prediction, Planning, and Control},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {95498--95518},
  doi = {10.1109/access.2026.3704955}
}
Kostenlos starten →