A new supernumerary robotic thumb called EgoThumb improves the ability of a single hand to hold and manipulate small items. It is attached to the back of the user’s hand and adds a finger‑like thumb that moves under computer control.

EgoThumb’s compact mechanical design uses a dual‑tendon finger driven from the forearm, so the added thumb stays light. A visual camera that looks straight ahead from the palm captures the hand and the target object. The camera image and the thumb’s joint positions are fed to a vision‑based imitation‑learning model that predicts a short‑term plan for the thumb’s joints.

In laboratory tests EgoThumb achieved a 90 % success rate across four tasks—picking up a cube, unscrewing a cap, stirring tea, and transferring a ball—when used by the study’s operator. New participants who had never seen the device reached success rates of 87 % or higher on the same tasks.

The device keeps well when objects are placed in a larger search area, but performance drops for tasks that require long‑step planning, such as the ball transfer. The small camera mounted on the palm can blur during rapid hand movements and is protruding from the skin, which may affect comfort.

Overall, EgoThumb demonstrates that a lightweight, vision‑driven robotic thumb can extend a human hand’s reach and precision, but further work is needed to handle faster movements and more distant targets.