New AI system tunes robotic knee to help amputees walk within minutes

People who receive robotic limbs take time to adapt to movement styles and can’t really start using them right away. However, researchers have made an AI system that can tune a robotic knee to make amputees walk within minutes.

To help amputees walk right away, researchers from the North Carolina State University have crafted an AI system that can tune a robotic knee, letting them walk on level ground within 10 minutes.

In the experiments, the AI system made use of trial-and-error algorithm to recognize patterns in sensor data where the training data was recorded from amputees walking in brief sessions. For setting initial safety limits, the researchers imposed constraints on the AI system to avoid situations that could cause the wearer to fall. Also, the AI system learnt patterns that match stable, smooth walking motions, explained Venture Beat.

Brain implant enables paralyzed patients to communicate through texts

As the robotic limb is exercised, the AI model looks at various parameters defining the relationship between force and motion in the limb. This includes the stiffness of a robotic joint, or the range or vertical motion allowed in a foreleg.

“Our body does weird things when we have a foreign object on our body,” Jennie Si, co-author of the paper, told IEEE Spectrum. “In some sense, our computer reinforcement learning algorithm learns to cooperate with the human body.”

However, the system still needs more work to be done such as scientists would have to verify that the AI system works with stairs, inclines and rough terrain, as per Engadget.

“If you wanted to make this clinically relevant, there are many, many steps that we have to go through before this can happen. So far it’s really just to show it’s possible — by itself that’s very, very exciting,” co-author Helen Huang told IEEE Spectrum.