🛰️ Daily AI Frontier
‹ back to 2026-09-25

RAPID: Robot Agentic Programming from Demonstrations

Research LLM Agents

Ranking

Overall 82
Content 95
Popularity N/A

No observed public metrics; popularity remains neutral/archived.

Merged summary

TL;DR - RAPID is an agentic system that turns a single visual human demonstration into reusable robot programs, automatically deriving task specifications, action primitives, and a verification environment. It matters because the generated programs generalize across object and scene variations, including contact-rich manipulation on a real robot arm.

  • Uses an iterative agentic loop to generate, execute, verify, and refine robot code.
  • Represents demonstrations as object-centric relational programs rather than reproducing specific motions.
  • Implements action primitives through trajectory optimization and adapts them at runtime using scene-specific geometric constraints.
  • Evaluated in simulation and on a Franka arm across eight nonprehensile tasks, with additional simulation tests on LIBERO-Pro prehensile tasks.

Sources (1)

RAPID: Robot Agentic Programming from Demonstrations

arXiv cs.RO Yuyao Liu, Jiayuan Mao, David Hsu, Leslie Pack Kaelbling, Tomás Lozano-Pérez 2026-09-24 arXiv:2609.30249
Public signals N/A
Providers: Hugging Face · N/A OpenAlex · N/A Publisher · N/A Semantic Scholar · N/A X · N/A Fetched 2026-09-26 14:14:06.907823 UTC

TL;DR - RAPID is an agentic system that turns a single visual human demonstration into reusable robot programs, automatically deriving task specifications, action primitives, and a verification environment. It matters because the generated programs generalize across object and scene variations, including contact-rich manipulation on a real robot arm.

  • Uses an iterative agentic loop to generate, execute, verify, and refine robot code.
  • Represents demonstrations as object-centric relational programs rather than reproducing specific motions.
  • Implements action primitives through trajectory optimization and adapts them at runtime using scene-specific geometric constraints.
  • Evaluated in simulation and on a Franka arm across eight nonprehensile tasks, with additional simulation tests on LIBERO-Pro prehensile tasks.
item →