Customizable ‘Smart’ Exoskeleton Learns from Your Steps
Previously, the median energy reductions carried out by other research teams were 14.5 percent, the use of manually adjusted ankle exoskeletons on both legs and 22.8 percent, using an exosuit that acted in Both hips and both ankles using preprogrammed parameters.
However, the human-in-the-loop CMU algorithm performed well and was not based on previous programming.
“This algorithm was so good that it has discovered a strategy to help reduce energy costs with a single device,” Jackson said. “It was very good.” [10 best inventions that changed the world]
The challenge of exo-skeletons is that even though they are meant to help a person, they can prevent the movement, Jackson said. To begin with, each device comes with its own weight from a few ounces to a few kilos, and the user must carry that weight. Exo-skeletons are also designed to apply force to body parts, but if the force time is off, the person may need to use more energy to move, Jackson said. And it is counterproductive.
During the optimization of the recent study phase, each participant wore an ankle exoskeleton and a mask designed to measure levels of oxygen and carbon dioxide (CO2). These measures refer to the amount of energy consumed by the person. As each person entered a treadmill at a steady pace, the exoskeleton has been applied a set of different modes of assistance ankles and toes.
These models were a combination of applied force and strength. For example, forces could be applied early in a position (when the heel touches the ground first) in the middle position (when the foot is flat) or the last position (when the foot is wrapped in the toe). During these changing positions, more or more force could be applied.
The algorithm was tested responses of the participants to 32 different models, which change every 2 minutes. Then it is measured if the model was easier or more difficult for the person to walk.
At the end of the session, which lasted for more than an hour, the algorithm produces a unique support model optimized for each individual.
“As for the overall shape of the models, there is great variability, which speaks to the importance of tailoring these strategies to each person, rather than applying the same to everyone,” Jackson said.
He added that the device may have worked well, not only because it was “learning”, but also because, by changing the method of attendance, the person who used it also taught.
“We believe this forces people to explore different ways of coordinating their efforts to better interact with the device,” Jackson said. This helps guide the person on how best to use the device and make the most of it. “It’s a two-lane street,” he said.
Other team members plan to test how the algorithm could be expanded to create an exoskeleton at six joints, designed to be used over the entire lower half of the body.