LATIDIA · Robótica

SADP: Política de difusión consciente de subobjetivos para la manipulación a largo plazo aprendida de las demostraciones generadas por el modelo de base

arXiv:2605.16871v3 Tipo de anuncio: reemplazar Resumen: La manipulación de robots de horizonte largo requiere políticas para coordinar múltiples subobjetivos intermedios y determinar cuándo avanzar entre ellos. Sin embargo, la mayoría de las imitaciones aprenden

La noticia

arXiv:2605.16871v3 Announce Type: replace Abstract: Long-horizon robot manipulation requires policies to coordinate multiple intermediate subgoals and determine when to advance between them. However, most imitation learning methods are trained solely on task-level demonstrations, without explicitly modeling the active subgoal or its execution progress. This limitation is further exacerbated by the scarcity of subgoal-level supervision in standard robot learning datasets, which makes explicit subgoal-conditioned control and online transition modeling difficult to learn. To address this issue, this paper proposes Subgoal-Aware Diffusion Policy (SADP), a framework that leverages foundation models to autonomously generate subgoal-annotated demonstrations and trains diffusion policies on these datasets. SADP structures policy execution around explicit natural-language subgoals by conditioning