Action Representation for Spatial Generalization in Robot World Models
Action Representation for Spatial Generalization in Robot World Models
This research investigates how action representations affect spatial generalization in robot world models.
The study evaluates action representations in world, object, and surface coordinate frames using a robotic peeling benchmark, finding that surface-relative coordinates improve consequence prediction but learned predictors remain vulnerable to coordinate frame uncertainty.
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.
prediction remains insufficiently understood. This study evaluates matched action representations defined in world, object, and surface coordinate frames using a controlled robotic peeling benchmark. Global pose and contact relocation experiments reveal a cost associated with limited training data when equivalent interactions receive different spatial descriptions. Removing information about peel thickness creates a
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org