LATIDIA · Robótica

MARCOS: Recuperación de fallas y monitoreo de habilidades incorporadas para la Loco-Manipulación Humanoide

arXiv: 2609.22538v1Tipo de anuncio: nuevo Resumen: los planificadores del modelo de lenguaje grande (LLM) pueden descomponer las instrucciones en lenguaje natural y seleccionar habilidades de robot reutilizables, pero elegir la habilidad correcta no garantiza el éxitof

La noticia

arXiv:2609.22538v1 Announce Type: new Abstract: Large language model (LLM) planners can decompose natural-language instructions and select reusable robot skills, but choosing the correct skill does not guarantee successful physical execution. This gap is especially important in humanoid loco-manipulation, where errors during approach, grasping, transport, or placement can invalidate the remainder of a long-horizon plan. We present FRAMES, a failure-aware supervisory framework for the Unitree G1 humanoid that operates above the CEER whole-body controller. A Planner Agent selects subtasks through parameterized mid-level skills, while a vision-language-model-based Monitor Agent evaluates each skill using temporal multi-view observations and structured robot and contact evidence. Detected failures stop the active skill and provide grounded feedback