As an undergraduate mechanical engineering student at Tsinghua University in China, Xiaogang Hu, PhD, was drawn to the assistive technologies that could help restore movement for people who no longer have full use of their limbs.
“I asked, ‘What is the major problem in assistive robots used for rehabilitation? What are we missing?’” said Hu, now the Dorothy Foehr Huck and J. Lloyd Huck Chair in Neurorehabilitation and an associate professor of mechanical engineering at Penn State. “The response was, ‘We don’t really understand well enough how biological systems work, and this knowledge gap blocks effective communications between biological systems and engineered assistive systems.’”
This answer inspired Hu to set out to solve the field’s overarching problem, while keeping the people who needed the solutions at the center of his work. Hu’s research focuses on creating wearable assistive robotic systems for people with limited use of their limbs, especially their hands, with the goal of the user intuitively controlling the systems and devices on which they rely.
“We focus on developing assistive techniques in a way such that when the person controls those assistive devices, they would do it as if they are controlling their own biological limbs,” Hu said. “Interacting with assistive devices needs to be very natural. That’s one of the most critical factors that will determine whether the person will eventually use those systems or not. If the devices are difficult to use, people will just put them away.”
With that at the forefront of his research, Hu said he concentrates on how to predict what a user is trying to do through artificial intelligence (AI). The technology involves a wearable exoskeleton—for example, a glove with motors and cables—and electrodes that can be placed on the skin’s surface or implanted into the skin. The electrodes can pick up very weak electrical muscle signals, and their technique can then use different signal processing techniques such as AI and other types of computational assessments to figure out what the user intends to do. These signals can then be used to drive the exoskeletons.
“Our recent work is focused on the dexterity of the hand,” Hu said, explaining that allowing a person with disabilities to move individual fingers is very difficult and that most of the work in the field focuses on helping a hand open or close and do a two- or three-finger pinch. “But being able to move fingers individually really determines whether a person can live independently or not. If they don’t have good hand dexterity, they won’t be able to perform daily tasks like combing their hair, preparing meals, or interacting with their smartphones. We want to improve hand dexterity to help people live a more or less independent lifestyle.”
After graduate school, Hu worked as a researcher at Northwestern University’s Shirley Ryan AbilityLab and then at the University of North Carolina Chapel Hill. He said that the time spent in rehabilitation hospitals—where he interacted daily with patients who have amputations, strokes, amyotrophic lateral sclerosis or other neural conditions—reinforced the impact of this research.
“We would ask the patients a lot of questions such as what really bothers them when they go home, what can we do to make their lives better,” he said. “We might see a person who on day one couldn’t move their limb at all, and then we would run a clinical trial with intensive therapy, and a few months later, they are able to pick up a cup by themselves. That’s huge.”
Hu’s next research steps include two main thrusts. The first is to have the AI algorithm that predicts the person’s intended movement to adapt seamlessly over time to the user’s changing behavior or habit. “We want to have the algorithm slowly adapt with the person so that the algorithm prediction accuracy remains high,” Hu said.
The second thrust is to convey touch sensation such as the temperature, rigidity, or other properties of an object held in a hand back to the brain. “When people lack some of that touch sensation, they need to rely on visual information instead,” Hu said. “That the sense of touch is really critical when we use our hands and interact with objects. But assistive robots are not sending that information back to the brain via the nerves.”
Hu said that the nerves can be thought of as information pathways going back to the brain. “Even if the hand is not working properly, the pathways are still there,” he said. “We can stimulate that nerve or the pathways to generate artificial signals that will trick the brain into feeling that sensation, so that the skin will feel pulsing or pressure sensations. It’s really just mimicking how the biological systems work.”
Editor’s note: This story was adapted from materials provided by Penn State.
