Researchers at the University of California San Diego have built EgoThumb, a wearable robotic thumb that holds objects in place so that people can do some two-handed tasks with one hand. They describe it in a preprint on arXiv.
A small camera captures images of the wearer's hand and nearby objects, and an AI model uses them, with the thumb's joint positions, to decide when to grasp and release. The thumb has three joints, each moved by a pair of tendons, and its motors sit on the forearm to keep weight off the hand.
In lab tests, three wearers tried four tasks, including unscrewing a bottle cap and stirring tea, with 10 trials per task. Average success was 90% for the person who provided the training demonstrations and 92.5% and 87.5% for two new users, but it fell when objects were placed outside the areas used in training.
Co-first author Jiawei Fu said the thumb can learn to coordinate with a human hand “without requiring the user to explicitly command each grasp”. The team has released the hardware designs and software as open source.
Background
University of California San Diego
University
The University of California San Diego is a US university whose researchers developed the EgoThumb robotic thumb.
EgoThumb
Wearable robotic thumb
EgoThumb is a wearable robotic thumb that uses a camera and an AI model to decide when to grasp and release objects.



