RPG, 오프라인 데이터로 로봇 조작 능력 95% 성공률 달성
Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents
- 매체
- arXiv
- 게시
- 중요도
- 4/5
- 화제 지수
- 6
한국어 요약
RPG 프레임워크는 시뮬레이션에서 연습하고 진짜 로봇에 적용하기 전에 오프라인 데이터셋에서 능력을 습득한다. 15라운드 후 22개 조작 태스크에서 성공률을 28.6%에서 95.0%로 끌어올렸으며, 실제 하드웨어에서 30번 시험 모두 성공했다.
같은 소식, 다른 매체 · 4곳
원문 발췌 · English
Building reliable robot capabilities across diverse tasks requires substantial human effort to develop and maintain skills, design rewards, and integrate perception with control. We present Reconstruct, Practice, Go Real (RPG), a framework for autonomous improvement of robot execution systems without updating model weights. RPG identifies manipulation capabilities in an offline dataset and constructs related practice tasks in simulation. During practice, RPG uses execution feedback, privileged simulator state, and available dataset videos to diagnose failures. It develops new reusable symbolic skills, refines existing skills, and revises the system prompt based on these diagnoses. Cross-task evaluation tests individual candidate changes and merged revisions before they are retained for reuse. At test time, a multimodal LLM uses the resulting system prompt and skill library to coordinate perception and robot control. On held-out initializations of 22 manipulation tasks, RPG improves task success from 28.6% after the first practice round to 95.0% after 15 rounds, outperforming all evaluated baselines, including ASPIRE (75.5%) and CaP-Agent0 powered by GPT-6 Astra Pro (60.0%). After a common calibration and hardware-adaptation procedure, the frozen system succeeds in all 30 physical trials, with ten trials on each of three tasks. Project Website: https://rpg-robot.github.io/