robotics
One Robot World Model Handled Cabinets, Rope and Cloth
PointCast tracks identified 3D points instead of committing to one object's mesh or topology.
Summary
PointCast tracks identified 3D points instead of committing to one object's mesh or topology.
PointCast predicts future trajectories for sets of identified 3D points on objects and a robot end effector. Separate checkpoints using the same 19.8-million-parameter architecture were best on three of four simulated regimes and second on rigid objects; on a real teleoperation dataset, it had the lowest mean error in four of six categories. The work unifies a representation and training recipe, not one universal checkpoint for every object.
Why it matters
PointCast tracks identified 3D points instead of committing to one object's mesh or topology.
Limits and context
- The work unifies a representation and training recipe, not one universal checkpoint for every object.
Key claims
PointCast tracks identified 3D points instead of committing to one object's mesh or topology.
Qualification: The work unifies a representation and training recipe, not one universal checkpoint for every object.
Evidence: source-2026-09-24-008
Sources
- arXiv preprint 2609.28393arXiv · primary research
Corrections
No corrections have been recorded for this story.