robotics
Classical Planning Skipped the Easy VLA Steps
SkipVLA ran the learned policy for contact and a motion planner for free space, completing tasks up to 2.5× faster.
Summary
SkipVLA ran the learned policy for contact and a motion planner for free space, completing tasks up to 2.5× faster.
A generalist robot policy is useful when meaning and contact matter, but expensive for every centimeter of free-space travel. SkipVLA reuses a frozen vision-language backbone to predict target poses, sends collision-free motion between them to a classical planner, and queries the VLA only for grasping and placing. Across 13 LIBERO tasks and three physical pick-and-place tasks on a six-degree-of-freedom arm, the hybrid matched task success while cutting completion time by as much as 2.5 times and reducing energy use.
Why it matters
SkipVLA ran the learned policy for contact and a motion planner for free space, completing tasks up to 2.5× faster.
Limits and context
- SkipVLA reuses a frozen vision-language backbone to predict target poses, sends collision-free motion between them to a classical planner, and queries the VLA only for grasping and placing.
Key claims
SkipVLA ran the learned policy for contact and a motion planner for free space, completing tasks up to 2.5× faster.
Qualification: SkipVLA reuses a frozen vision-language backbone to predict target poses, sends collision-free motion between them to a classical planner, and queries the VLA only for grasping and placing.
Evidence: source-2026-09-18-010
Sources
- arXiv preprint 2609.20648arXiv · primary research
Corrections
No corrections have been recorded for this story.