Phone scans can personalize robot pickup policies through simulation

Splat2Act converts captured indoor spaces and object images into simulated demonstrations that adapt a vision-language-action policy without real-world teleoperation.

Big Tech
Xiaomeng Ye · Krishna Jaganathan · Gyubeum Lim · Gunjan Chhablani · Muhammad Zubair Irshad · Zsolt Kira

Georgia Institute of Technology · Waymo LLC · Toyota Research Institute

Research Digest··2 min read
Ye et al.

Splat2Act reconstructs building-scale environments using 3D Gaussian Splatting for visual appearance and separate collision geometry for physical simulation.

Why this paper

From Toyota Research Institute and 2 others

In one line

A phone scan and object images can replace real demonstrations for personalizing a VLA policy, yielding 63.9% real-world pickup success.

What we could check

  • ·No code link found
  • ·No weights link found
  • ·No dataset link found
  • ·No compute details found
  • ·No stated limitations found
  • ✓Reports numbers on named benchmarks (4 benchmarks)

Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.

§

Research Digest

Written by software from the reporting listed above, scored by an automated standards desk, and published without a person reading it first. If something here is wrong, tell the editor and it will be put right.

How we workSubscribe