Researchers at North Carolina State University have created a robotic exoskeleton that helps users walk, run, and climb stairs. The device leverages artificial intelligence to be ready to use as soon as a new user puts it on.
While it has so far only been tested with able-bodied individuals, it could have future applications for people with disabilities as well as the military and industrial work.
Existing exoskeleton technologies need to be trained on a new person for upwards of an hour before they can support a new wearer’s movement. But, this new device is different.
Dr. Hao Su, an Associate Professor of Mechanical and Aerospace Engineering at NC State, leads the research group that published the study. He explained that their device still needs to be trained, but by using AI and a computer-simulated human, the training is more efficient.
“So, instead of testing on a physical human with a physical robot, we can test in the digital world,” said Su.
To eliminate the need for training, Su’s research group created the simulation using videos of people walking. The software that controls the exoskeleton can then use AI to learn how to support a human’s movements within the simulation.
“So, we don't need to do this with a physical person,” Su said. “We can just do this in the simulation environment.”
Not only does training the exoskeleton in the virtual world work well enough, but it is also more efficient. Greater efficiency could eventually make the technology cheaper to implement and thus more accessible.
The device is not quite ready for the market. It can only assist with walking, running, and climbing stairs, and isn’t able to transition between those activities, or assist with things like standing up from a sitting position.
Another limitation is that the exoskeleton is only one-size-fits-all for able-bodied people. However, Su said that adapting the technology for people with mobility impairments is well within the realm of possibility. He and his group think they could adapt the technology by creating a simulation of a person with a specific impairment and training the exoskeleton software using that simulation.
“We want to just make this even more versatile by building on this learning in simulation framework,” Su said. “So, I think, in about maybe one year, the robot will be more versatile and smarter. It can probably assist all the common activities a human does.”
