Robot learns shifting actions that directly boost grasping probability in clutter, trained on 25k grasps and 2.5k shifts, achieving 274 picks per hour in bin picking and generalizing to novel objects.
Dex-Net 1.0: A cloud-based network of 3d objects for robust grasp planning using a Multi-Armed Bandit model with correlated rewards
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Robot Learning of Shifting Objects for Grasping in Cluttered Environments
Robot learns shifting actions that directly boost grasping probability in clutter, trained on 25k grasps and 2.5k shifts, achieving 274 picks per hour in bin picking and generalizing to novel objects.