A method learns domain-invariant 3D point cloud representations from simulation and limited real snapshots to train grasping policies entirely in simulation, achieving 10% better performance than 2.5D baselines without real grasping data.
Title resolution pending
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
-
Data-Efficient Learning for Sim-to-Real Robotic Grasping using Deep Point Cloud Prediction Networks
A method learns domain-invariant 3D point cloud representations from simulation and limited real snapshots to train grasping policies entirely in simulation, achieving 10% better performance than 2.5D baselines without real grasping data.