Amputees often feel disconnected from their bionic hands. AI could bridge the gap : Shots
Samoana Matagi was one of four participants in a study testing the capabilities of a new bionic hand. Here, Matagi wears the bionic hand on one arm and a common prosthetic called a body-powered hook on the other.
Dave Titensor/Utah Neurorobotics Lab
hide caption
toggle caption
Dave Titensor/Utah Neurorobotics Lab
Researchers have built a prosthetic hand that, thanks to artificial intelligence, can act much more like a natural hand.
The key is for the hand to recognize when the user wants to do something and then share control of the movements needed to accomplish the task.
The approach, which combined AI with special sensors, allowed four people missing one hand to simulate a drink in a cup, according to Marshall Troutresearcher at the University of Utah and lead author of the study.
When sensors and AI helped, participants could “very reliably” grab a cup and pretend to take a sip, Trout says. But without this shared control of the bionic hand, he said, they “crushed it or dropped it every time.”
success, describe in the magazine Natural communicationsis notable because “the ability to exert grip force is one of the things we really struggle with in prosthetics right now,” says John Downeyassistant professor at the University of Chicago, who was not involved in the research.
Such problems cause many amputees to become frustrated with their bionic hands and stop using them, he says.
A helping hand
The latest bionic hands are equipped with motors that allow them to rotate, move individual fingers and manipulate objects. They can also detect electrical signals from the muscles used to control these actions.
But as bionic hands have become more capable, they have also become more difficult for users to control, Trout says.
“The person has to sit still and really concentrate on what they’re doing,” he says, “which is really not the behavior of an intact hand.”
A natural hand, for example, requires very little cognitive effort to perform routine tasks like grasping an object or tying a shoelace. That’s because once a person sets the task in motion, most of the work is done by specialized circuits in the brain and spine that take over.
These circuits allow many tasks to be accomplished efficiently and automatically. Our conscious mind only intervenes if, for example, a shoelace breaks or an object is moved unexpectedly.
So Trout and a team of scientists decided to make a smart prosthetic that would act more like a person’s hand.
“I just know where my coffee cup is, and my hand will naturally squeeze it and make contact with it,” he says. “That’s what we wanted to recreate with this system.”
Using AI and a suite of sensors, researchers equipped a bionic hand so that it shares control with the brain when carrying out instructions.
Dave Titensor/Utah Neurorobotics Lab
hide caption
toggle caption
Dave Titensor/Utah Neurorobotics Lab
The team turned to AI to take over some of these subconscious functions. This meant detecting not only the signal coming from a muscle, but also the intention behind it.
For example, the AI control system has learned to detect the slightest contraction of a muscle flexing the hand.
“That’s when the machine’s controller kicks in and says, ‘Oh, I’m trying to grab something, I’m not sitting still,'” Trout says.
To make the approach work, scientists modified a bionic hand by adding proximity and pressure sensors. This allows the AI system to measure the distance to an object and assess its shape.
Meanwhile, pressure sensors on the fingertips tell the user how firmly their prosthetic hand is holding the object.
Share control
The idea of sharing control of a bionic hand responds to a reaction many people have when using a prosthesis with superhuman abilities, explains Jacob Georgesprofessor at the University of Utah and director of the Utah NeuroRobotics Lab.
“You can create a robotic hand that can perform tasks better than a human user,” he says. “But when you give that to someone, they don’t like it.”
That’s because the device feels foreign and out of their control, he says.
John Downey says that one of the reasons we feel connected to our own hands is that they are jointly controlled by our thoughts and by brainstem and spinal cord reflexes.
This means that the thinking part of our brain doesn’t have to worry about the details of each movement.
“All of our motor control involves reflexes that are subconscious,” Downey says, “so it will be important to come up with robotic imitations of these reflex loops.”
George says the smart bionic hand solves this problem.
“The machine does something and the human does something, and we combine those two things,” he says.
This is a crucial step toward creating prosthetic limbs that feel like an extension of the person’s body.
“Ultimately, when you create an embodied robotic hand, it becomes part of that user’s experience, it becomes a part of them and not just a tool,” George explains.
Even the most advanced bionic hands still need help from a human brain, Downey says.
For example, a person might use the same natural hand to gently thread a needle and then firmly lift a child.
“The dynamic range is way beyond what robots typically handle,” says Downey.
This is likely to change, as bionic limbs become more versatile and capable. What won’t change, scientists say, is humans’ desire to retain a sense of control over their artificial appendages.
