Prostheses get better, but not as fast as you think. They are not as smart as our real limbs that do things (controlled by the brain) that automatically extend to catch us when we fall. This particular "stumbling reflex" was the subject of an interesting study by Vanderbilt, in which the subjects had to fall down very often.
The problem that the team wants to alleviate is simply that users of prosthetic limbs may fall off more than most advise, and when they fall, it can be very difficult to recover because of an artificial leg – especially at Amputations over the knee – not as responsive as a natural leg.
The idea, explained senior researcher and mechanical engineer Professor Michael Goldfarb, is to determine what exactly enters a stumbling block and how it can be artificially reproduced.
"A person who stumbles performs different acts depending on various factors, not all of which are known. The reaction is changing, as the strategy most likely to prevent a fall depends heavily on the "initial conditions" at the time of the trip, "he told TechCrunch in an e-mail. "We hope to create a model whose factors determine the nature of the stumbling response. So when a trip occurs, we can use the various sensors on one leg of a robotic prosthesis to artificially reconstruct the reflex and achieve an effective response, in line with the biological reflex loop.
The experimental setup looked like this. Subjects were placed on a treadmill and instructed to walk normally. special glasses prevented them from looking down, arrows on a display held them straight, and a simple mental task (counting backwards from seven) occupied their brains.
Waiting for the best opportunity to put a literal stumbling block on the treadmill so the person can trip over it.
When this happened, the person inevitably stumbled, though a harness prevented them from actually falling and injuring themselves. But when they stumbled, their movements were meticulously captured by a movement detection device.
After 196 stumbling blocks and 190 stumbling blocks, the researchers had collected a lot of data on how exactly people move to recover from stumbling. Where do her knees go relative to her ankles? How do you tilt your feet? How much power does the other foot take on?
How exactly this data would be integrated into a prosthesis depends to a large extent on the nature of the artificial limb and the conditions of the user. However, having this data and possibly providing it to a machine learning model helps to expose patterns that can be used to inform about prosthetic emergency movements.
They can also be used for robotics: "The model can be used directly for programming reflections in a biped," said Goldfarb. These human-like movements performed by robots could be even more human if based directly on the original. There is no hurry – they may be a bit too human.
The research describing the system and the dataset and releasing it for free to anyone who wants to use it appeared in the Journal of NeuroEngineering and Rehabilitation.