AI helps prosthetic hand grasp more naturally ^{CEFR B2}

Everyday tasks such as reaching for a mug rely on an automatic, subconscious sense of how fingers move and touch objects. People who use prosthetic hands often lose that skill and must consciously control each finger, which makes simple actions slow and mentally tiring. A University of Utah team used AI to restore much of that ease.

The researchers fitted a TASKA Prosthetics artificial hand with custom fingertips that measured pressure and included optical proximity sensors intended to reproduce a fine sense of touch. The sensors were sensitive enough to detect an effectively weightless cotton ball being dropped onto a finger. A neural network was trained on the proximity data so each finger moved to the correct distance to form a stable grasp; because each digit has its own sensor, fingers act in parallel to secure different shapes.

To prevent conflict between user and machine, the team developed a bioinspired shared-control approach that balances human and AI input. As Marshall Trout explains, the machine improved the user's precision while making tasks easier, effectively augmenting natural control so users do not have to think about simple actions. Jacob A. George adds that offloading part of grasping to the prosthesis makes control more intuitive and dexterous.

The study, published in Nature Communications, tested the system with four participants whose amputations fell between the elbow and wrist. Participants showed greater grip security, greater precision and less mental effort, and they completed fine motor activities such as picking up small objects and drinking from a plastic cup. The team is exploring implanted neural interfaces and plans to blend those interfaces with the enhanced sensors and intelligent control. Additional coauthors are from the University of Utah and the University of Colorado, Boulder; funding came from the National Institutes of Health and the National Science Foundation.