RoboticsPhys.org Technologyabout 3 hours ago
Robot Hand Learns Nimble Moves Safely
3 min read
Researchers at UC Berkeley and Davis train a 16-joint LEAP robotic hand in simulation to manipulate objects. A separate software safeguard reduces drops from 16 to four over 1,200 sequences, though it lowers target-reaching success from 95.2% to 91.5%. The team published these results on arXiv in 2026.
Level
Hype check
Tap to vote and see what everyone thinks.
