ETH Zurich Team Teaches Robotic Hand to Walk and Work Independently

Engineers at ETH Zurich’s Soft Robotics Lab have adapted an off-the-shelf robotic hand to crawl across varied surfaces, recover its balance and interact with objects without being attached to a robotic arm.

The team, led by Amir Kazemi, used a human-like hand with unequal fingers and an opposed thumb. Those features help robotic hands manipulate everyday objects but complicate walking because the fingertips have different reaches. Earlier walking-hand designs avoided that problem by using symmetrical fingers.

Researchers mounted a battery, motion sensors and a miniature computer on the hand, then trained it through reinforcement learning in a computer simulation. The work is described in a paper posted on the arXiv preprint server.

In physical tests, the hand crossed 14 indoor and outdoor surfaces, including carpet, gravel, grass and metal grating. It also righted itself in 21 of 25 trials that began with the device lying on either side.

The hand could use its fingers while supporting itself. Balancing on four fingers, it pressed the correct arrow key in 29 of 32 attempts. In another experiment, it approached a small block and pushed it across a table, placing it within two centimeters of a target in 15 tests.

The researchers said a mobile robotic hand could operate in confined spaces without requiring the arm that normally carries it to follow.