A U-Net-style CNN trained on synthetic multi-modal grasp data from the Genesis simulator enables a real quadruped robot to navigate to and precisely grasp objects in a loco-manipulation task.
Keras applications: Mobilenet,
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
-
Optimizing Grasping in Legged Robots: A Deep Learning Approach to Loco-Manipulation
A U-Net-style CNN trained on synthetic multi-modal grasp data from the Genesis simulator enables a real quadruped robot to navigate to and precisely grasp objects in a loco-manipulation task.