DexPoint: Generalizable Point Cloud Reinforcement Learning for Sim-to-Real Dexterous Manipulation.
Yuzhe QinBinghao HuangZhao-Heng YinHao SuXiaolong WangPublished in: CoRR (2022)
Keyphrases
- point cloud
- reinforcement learning
- structure from motion
- surface reconstruction
- manipulation tasks
- point sets
- stereo camera
- point cloud data
- real world objects
- triangular mesh
- laser scanner
- haptic feedback
- computer vision
- learning algorithm
- neural network
- single image
- real robot
- optical flow
- object manipulation
- iterative closest point