AI Co-Pilot for Bionic Hands Transforms How Amputees Control Prosthetics with Intuitive Grasping

AI Co-Pilot Addresses Critical Control Problem in Bionic Hands

Up to 50 percent of amputees abandon advanced bionic hands, citing poor controls and excessive cognitive burden as primary reasons for discontinuation1

1

. Researchers at the University of Utah have developed an AI co-pilot system that fundamentally changes how amputees interact with prosthetic limbs. Led by engineering professor Jacob A. George and postdoctoral researcher Marshall Trout from the Utah NeuroRobotics Lab, the team published their findings in Nature Communications2

2

.

The challenge stems from a lack of autonomy in current prosthetic designs. Natural hands rely on elaborate reflexes and feedback loops that operate within 60 to 80 milliseconds, before conscious awareness even registers1

1

. Controlling a natural hand involves managing 27 major joints and 20 muscles simultaneously, yet this happens effortlessly through subconscious brain circuits. Most commercially available bionic hands lack these automatic responses, forcing users to micromanage every movement through apps or surface electromyography signals from remaining muscles.

Proximity and Pressure Sensors Enable Natural Grasping Ability

The research team modified a commercial bionic hand manufactured by TASKA Prosthetics by replacing fingertips with silicone-wrapped proximity and pressure sensors1

1

. These optical proximity sensors can detect objects approaching and measure the precise grasp force needed to hold items without crushing or dropping them. The sensors proved sensitive enough to detect an effectively weightless cotton ball being dropped on the fingertips3

3

.

The team trained an artificial neural network by repeatedly moving the hand back and forth to touch objects, collecting extensive training data on various grip patterns1

1

. Because each finger has its own sensor and can "see" in front of it, each digit works in parallel to form a stable grasp across any object2

2

. This approach enables the AI to recognize different objects and switch between grip types while controlling each finger independently, achieving natural conforming movements.

Shared Control System Balances Human-Machine Control

The breakthrough lies in how control is distributed between user and machine. Earlier autonomous prostheses required users to manually switch autonomy on and off, creating an awkward experience1

1

. The University of Utah team implemented a bioinspired approach involving shared control that operates subtly in the background. "What we don't want is the user fighting the machine for control. In contrast, here the machine improved the precision of the user while also making the tasks easier," Trout explained2

2

.

The AI detects the tiniest muscle twitch that flexes the hand, interpreting this as intention to grasp. "That's when the machine controller kicks on, saying, 'Oh, I'm trying to grasp something, I'm not just sitting still,'" Trout noted4

4

. Users remain in charge at all times and can adjust grip strength or release objects at will. This balance creates what George describes as an embodied experience where the prosthetic becomes part of the user's sense of self rather than just a tool4

4

.

Testing Shows Dramatic Improvement in Dexterity and Reduces Cognitive Burden

The research team conducted studies with four participants whose amputations fell between the elbow and wrist. They performed everyday tasks requiring fine motor control, such as picking up a paper cup to drink from it or transferring an egg from a plate without breaking it1

1

. Without artificial intelligence assistance, participants succeeded in only one or two attempts out of 10. With the AI system activated, success rates jumped to 80 or 90 percent1

1

.

Participants demonstrated greater grip security, greater grip precision, and less mental effort when working with the intelligent system2

2

. Critically, they accomplished these improvements without extensive training or practice, using different gripping styles for various objects3

3

. Simple tasks like drinking from a plastic cup, which requires precise grasp force to avoid crushing or dropping, became manageable again. "By adding some artificial intelligence, we were able to offload this aspect of grasping to the prosthesis itself," George said2

2

.

Future Integration with Neural Interfaces and Tactile Feedback

The team is exploring implanted neural interfaces that would allow individuals to control prostheses with their thoughts while receiving tactile feedback3

3

. Next steps involve blending these technologies so enhanced sensors can improve tactile function while the intelligent prosthesis integrates seamlessly with thought-based control. John Downey, an assistant professor at the University of Chicago not involved in the research, notes that providing robotic imitations of subconscious reflex loops will be important as the field advances4

4

.

The dynamic range required for human tasks—from gently threading a needle to firmly lifting a child—remains a challenge for robotics4

4

. As advanced bionic hands become increasingly versatile and capable, the research suggests that maintaining human control while augmenting it with machine intelligence offers the most promising path forward. The work received funding from the National Institutes of Health and the National Science Foundation3

3

.