LATIDIA · Robótica

Aprendizaje de la dinámica de contacto a través del tacto: Redes neuronales de gráficos condicionales de acción para la inserción de clavijas robóticas

arXiv: 2509.12151v3Tipo de anuncio: reemplazar Resumen: Presentamos un modelo basado en la física que predice el movimiento del efector final del robot y el par de fuerza de reacción en la manipulación rica en contacto. El modelo representa t

La noticia

arXiv:2509.12151v3 Announce Type: replace Abstract: We present a learnable physics-based model that predicts motion of the robot end effector and reaction force-torque in contact-rich manipulation. The model represents the end effector and the environment as interacting meshes in a graph structure, and conditions its prediction explicitly on the applied control input. It predicts object-level pose update directly, while the reaction torque emerges from a per-vertex force field. Training is self-supervised using only joint encoder and force-torque data while the robot is randomly touching the environment without task context. In simulation, our model transfers to peg insertion with unseen concave geometry, where an MPC agent using it reaches up to 98%