Megadose AI progress, ranked and analyzed.

Fingers as Legs: Learning Self-Supported Locomotion and Manipulation with an Anthropomorphic Hand

· HF Daily Papers ·
The hand crawls, steers, recovers from falls, and still uses its fingers to act on objects.

The paper trains an anthropomorphic robotic hand to use its fingers for both self-supported movement and manipulation. Its reinforcement learning setup accounts for unequal fingers and uses a simulator calibrated from hardware measurements. On hardware, the untethered hand performs crawling, steering, fall recovery, keyboard commands without vision, and object pushing with overhead visual feedback. HF Daily Papers' note

score 5

Categories: Research