Despite the marked improvement of prosthetic technology in recent years, many prosthetic users still find that commercially available prosthetics can be unintuitive, making everyday activities a challenge.
To address this problem, a team of mechanical engineers from the University of Utah designed a new approach to prosthetic movement using artificial intelligence (AI) to imitate a prosthetic user's leg motion. This new approach promises to make walking, even in new environments, more comfortable and smoother. This study was published in the journal Science Robotics in July 2020.
The study highlighted that previous iterations of powered prosthetics relied on pre-programmed behaviors based on the movements of non-amputees. Although this approach is effective, the authors of the study say that it limits prosthetic use to pre-mapped obstacles and areas, which would be great if the user doesn't need to walk in new environments. The existing approach requires a prosthetist to retrain the prosthesis to function in new terrain, impractical.
So, the team turned to AI that could quickly adapt and learn how to move in real-time.
How an AI-controlled prosthetic leg works
The AI is programmed to collect information about the user's hip movements. It will then use the data to reinterpret the motion of the residual limb. One thousand times in a second, the system gauges the residual limb's movement and then plans the bionic joints' corresponding movements.
To test the AI-controlled prosthetic leg in different settings, the team selected three amputee participants. They experienced different scenarios, such as walking across a short stretch of the lab floor and stepping over obstacles or continuous walking on a treadmill while tossing obstacles of varying sizes.
A video released along with the study (see video below) showed that participants wearing the AI prosthesis could easily clear obstacles of increasing size than the participants who used non-powered prostheses.
The video also shows that having the ability to bend the prosthetic knee effectively can help prosthetic users avoid potentially damaging compensating actions, such as swinging the leg over an obstacle.
The study's authors acknowledge that they still have a long way to go before this research becomes commercially available. Although the prosthesis's neural interfaces and AI are essential, they now need to focus on optimizing lightweight designs, which is also crucial for the leg's effectivity.As AI systems evolve, a new generation of prosthetics will soon allow users to control their prosthesis at a granular level. The future of prosthetics is looking very bright.