Action Intent as a Bridge: Transferring Manipulation Knowledge from Human-Centric Videos to Robots
Action Intent as a Bridge: Transferring Manipulation Knowledge from Human-Centric Videos to Robots
Vision-Language-Action (VLA) models face limitations due to reliance on expensive real-robot data. This research introduces Action-Intent Transfer (AIT), a two-stage method that extracts clean action representations from human-centric videos, disentangles implicit noise, and aligns human and robot domains within a shared intent space. Extensive experiments in simulation and real-world environments show that AIT effectively transfers human manipulation knowledge and consistently outperforms prior baselines.
The paper proposes Action-Intent Transfer (AIT), a two-stage method that disentangles action intents from human-centric videos to bridge the gap between human actions and robot execution, demonstrating improved performance over prior baselines in both simulation and real-world settings.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
Source：arXiv Robotics — research abstracts · arxiv.org