A value-guided MPC policy trained on 2 million synthetic trajectories improves closed-loop 6-DoF grasping in clutter and adapts to object perturbations.
After the set abstraction layers, the output passes through three fully connected layers with sizes 512,256, and128
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Grasp-MPC: Closed-Loop Visual Grasping via Value-Guided Model Predictive Control
A value-guided MPC policy trained on 2 million synthetic trajectories improves closed-loop 6-DoF grasping in clutter and adapts to object perturbations.