Publication
arXiv
Stage
Preprint
What we read
Summary of the abstract
Authors
Yuyao Liu, Jiayuan Mao, David Hsu, Leslie Pack Kaelbling, Tomás Lozano-Pérez
Universities and research institutions
Not yet supplied in verified metadata; the Brief does not guess.

What the paper reports

The RAPID system automatically generates, verifies, and refines robot programs from a single visual human demonstration, using an object-centric relational program representation and trajectory-optimization primitives. It was evaluated in simulation on eight tasks and on a real Franka arm, showing generalization across pose, shape, material, and environment.

Why it matters

This abstract-only study demonstrates a new method for converting demonstrations into reusable robot programs, but its preprint status and simulated-plus-limited real-world tests mean results should be interpreted cautiously and not generalized to all robots or tasks.

RAPID introduces an automated loop that creates and tests robot programs from just one human demonstration. It uses object-level representations to capture the strategy rather than exact motions, and expresses actions as trajectory-optimization primitives that act on objects.

In experiments, RAPID generalizes across different object poses, shapes, materials, and environments, and it was deployed on a real arm for eight nonprehensile tasks, illustrating potential but not universal applicability.

What this does not tell us

This article summarizes abstract-only research and notes preprint status; conclusions are not validated by full peer-reviewed publication or broad real-world deployment.

Original sources · 1
  1. RAPID: Robot Agentic Programming from Demonstrations ↗arXiv · 2026-09-24

Check the original paper for its authors, methods, version and access terms.