RoboRSI: Stable, efficient, and reusable robot self-evolution in complex real-world environments
RoboRSI tries to make robot self-repair reusable by tying each fix to a specific skill branch before release.
The paper introduces Top-Down Skill Refinement, which breaks tasks into compound, atomic, and base skills with defined responsibilities and input-output contracts. A Manager, Planner, Engineer, and Reviewer handle planning, execution, diagnosis, and validation, with humans steering objectives and corrections. On a mobile manipulator, the system developed multi-object household cleanup over 104 rounds. In simulation, it led reported results on LIBERO, LIBERO-PRO, LIBERO-Plus, and RoboTwin by 2.7 to 11.0 percentage points over the strongest baseline. ArXiv · AI/CL/LG's note
The paper introduces Top-Down Skill Refinement, which breaks tasks into compound, atomic, and base skills with defined responsibilities and input-output contracts. A Manager, Planner, Engineer, and Reviewer handle planning, execution, diagnosis, and validation, with humans steering objectives and corrections. On a mobile manipulator, the system developed multi-object household cleanup over 104 rounds. In simulation, it led reported results on LIBERO, LIBERO-PRO, LIBERO-Plus, and RoboTwin by 2.7 to 11.0 percentage points over the strongest baseline. ArXiv · AI/CL/LG's note
score 5