Snap, snap. There\u2019s a new member of the Addams-inspired robotics family, and it\u2019s got a very unusual way of getting around. Shaped like a disembodied hand and walking on its own fingers, this quirky robot brings Thing to life in a whole new way.<\/p>\n
A team of researchers from Soft Robotics Lab, ETH Zurich, built an autonomous robotic hand that uses its fingers as legs. Inspired by Thing from The Addams Family, the robot learned how to navigate across different surfaces, press keys, push small objects, and recover from falls, all with the same set of fingers it walks on.<\/p>\n
You can watch the little guy in action in the video above. And yes, we agree it\u2019s pretty adorable.<\/p>\n
Busy hands <\/p>\n
The robot is based on a standard anthropomorphic right hand with five fingers and 20 actuator joints (four on each finger). The researchers installed a compact on-board module with a battery, sensors, and computing hardware onto the hand.<\/p>\n
Weighing in at only 1.8 pounds (818 grams), the detached robotic hand can operate autonomously without the rest of its body. \u201cA robotic hand with its own mobility could operate in confined workspaces without requiring the arm that normally carries it to follow,\u201d the researchers explain in a paper<\/a> available on the preprint server, arXiv. \u201cFor example, a larger robot could place the hand on a support surface near a restricted opening. The hand could then move toward a control or object, interact with it, and return for retrieval.\u201d<\/p>\n The researchers trained the hand using reinforcement learning in a specialized simulator with different parallel environments. Teaching it how to walk and carry out different tasks using its fingers presented a challenge: as one finger lifts to step or press a key, the others need to keep the entire hand supported.<\/p>\n Unlike the legs of a standard robot, the fingers are not identical, and the palm rests at a tilt. To help overcome this challenge, the researchers assigned each fingertip its own target position built around the hand\u2019s natural stance. Based on each task, the robot would decide on where and when each finger steps.<\/p>\n A hand with a plan <\/p>\n During a series of tests, the detached hand navigated through 14 different types of indoor and outdoor surfaces\u2014including asphalt, metal, tile, grass, gravel, and damaged stone. It also learned to carry out different tasks, including pressing keyboard commands and pushing a small cube toward a target.<\/p>\n