Event

ETH Zurich Uses WUJI Humanoid Hand to Achieve Self-Supporting Walking and Manipulation

1 reports · 1 sources

In the paper "Fingers as Legs," the ETH Zurich team used reinforcement learning to enable the WUJI humanoid dexterous hand to crawl, turn, and recover from falls using its own fingers, without the need for a separate locomotion mechanism. Training was completed in simulation, and the simulation environment was calibrated using measurement data from the physical hardware. The physical system has its own power supply and computation, allowing it to operate as a standalone platform. In manipulation demonstrations, the hand continuously executed keyboard keypress commands without vision; in another demonstration, it used top-mounted visual feedback to push an object toward a target position. The paper's authors are Amirhossein Kazemipour, Hehui Zheng, and Robert Katzschmann.