A group of scientists from the Georgia Institute of Technology and Emory University have found that conventional ankle exoskeletons sometimes have no effect at all when it comes to maintaining balance while standing. According to the researchers, robotic motion aids respond far too late. Instead, they must intervene before the human reacts in order to restore balance after a disturbance.
In the study, "Exoskeletons need to react faster than physiological responses to improve standing balance," published in Science Robotics, the research team investigated the extent to which ankle exoskeletons have any effect at all in helping, say, elderly and physically impaired people achieve better balance so they don't fall.
Examination of a commercially available pair of ankle exoskeleton boots from Dephy, which was used as a proxy for the current technology, caused disillusionment. This is because the exoskeleton's support, which is activated with a physiological delay, did not help a person regain balance in standing after a disturbance.
The team tested this with various test subjects. They had to stand on a moving platform with their exoskeleton boots strapped on, which was abruptly moved to disturb the participants' balance while standing. The scientists tried this out under three different conditions: completely without exoskeleton support, with delayed exoskeleton support to the natural body response, and support that kicked in faster than the body's physical response.
The video shows how the effects of disturbances in balance are compensated by exoskeleton boots.
Generally, such exoskeletons are controlled by physiological signals from the wearer - such as muscle or brain activity. For example, muscle activity around the ankle is measured. When the muscle is tensed, the exoskeleton is then activated. The control system "steals" the human reaction, so to speak, and simply superimposes it.
Faster response needed
As the science team found, this is not enough when the balance is disturbed while standing. The exoskeleton is activated too late by the body's own reactions. The previous approaches of this type of control of exoskeletons are not suitable for maintaining balance while standing. According to the scientists, it takes about 150 milliseconds for the body to react to a loss of balance. In this time span, the exoskeleton must react in order to still be able to counteract.Display
The scientists realized this with electronic accelerometers, such as those used in smartphones. They used them to determine the disturbance and were thus able to trigger a faster response from the exoskeleton boots. But problems arose here as well, because the artificial fast support then overlapped the initial movement of the ankle and disrupted it as well. With additional active control via the muscle, the destabilization was even higher. However, the faster support alone brought an improvement: 9 percent of the larger disturbances could be compensated without the probands having to make a lunge.
Based on the results, the team is certain that a combination of global sensor and ankle sensor does not work because of the delay between the two. The researchers thus support theses from the 1970s that global physiological signals are better suited to make predictions about balance behavior than local signals from the ankle or leg. Nevertheless, this alone is not enough, as the results show. The team believes that machine learning approaches could help to detect and respond to disturbances at an early stage.