EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Embodied Intelligence Modeling: A Vision–Language–Action Model Based on the Diffusion Transformer Framework

Bin Ren, Yunxiang Wang

Journal of Robotics · 2026-01 · 2026 (1)

DOI: 10.1155/joro/1046193 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Robotics, online erschienen am 02.11.2026. Verfasst von Bin Ren und 1 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Humanoid robots in industrial settings face challenges such as high intensity, high repetitiveness, and high precision requirements, making it difficult for them to meet practical needs for stability and generalization. This paper proposes a 3B‐parameter vision–language–action (VLA) model for industrial production lines. This model deeply integrates the multi‐view perception network (MVP‐Net) multi‐view perception module into the VLA framework, effectively mitigating perception challenges such as occlusion, lighting interference, and viewpoint shifts in industrial scenarios through cross‐view feature fusion. It uses a large language model (LLM) to achieve cross‐modal semantic reasoning; its action‐generation module is based on the diffusion transformer framework, and the VLA under this framework was systematically evaluated across four real‐world production‐line tasks (grasping, assembly, extraction, and placement). Experimental results show that a success rate exceeding 90% was achieved on all tasks, significantly outperforming baseline models (diffusion policy [DP], Octo, and Open VLA).

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Ren, B., & Wang, Y. (2026). Embodied Intelligence Modeling: A Vision–Language–Action Model Based on the Diffusion Transformer Framework. Journal of Robotics, 2026(1). https://doi.org/10.1155/joro/1046193

Vancouver

Ren B, Wang Y. Embodied Intelligence Modeling: A Vision–Language–Action Model Based on the Diffusion Transformer Framework. Journal of Robotics. 2026;2026(1). doi:10.1155/joro/1046193

BibTeX
@article{Ren2026,
  author = {Ren, Bin and Wang, Yunxiang},
  title = {Embodied Intelligence Modeling: A Vision–Language–Action Model Based on the Diffusion Transformer Framework},
  journal = {Journal of Robotics},
  year = {2026},
  volume = {2026},
  number = {1},
  doi = {10.1155/joro/1046193}
}
Kostenlos starten →