The exoskeleton robot uses machine learning to help people stand up.

Scientists have developed a lightweight exoskeleton robot that, by learning, can guess exactly how the user wants to move, according to EurekAlert!. The results of the experiment are published in the journal IEEE Robotics and Automation Letters.

The exoskeleton robot is made from carbon fiber for the lower body. The part that attaches to the thighs weighs only 604 grams, and the one that attaches to the shins is 206 grams. The developers used artificial intelligence � a technique known as PU learning (unmarked learning) � to get the exoskeleton robot to correctly read the user's intentions � based on measurements of their muscle activity. PU allows the use of ambiguous data: the method mixes positive action options - the machine is sure of them - and ambiguous ones: they can be both positive and negative, which allows artificial intelligence to learn and more accurately understand the human intention.

During the experiment, the participants performed various movements that could start in the same way: standing up, crossing their legs, leaning forward, and changing positions in the chair. The exoskeleton used machine learning to guess when they were actually trying to get up and then provided assistance in moving.

The experiment was successful. As the authors of the work note, the new exoskeleton robot turned out to be more sensitive than traditional devices that use only well-defined data and cannot always accurately guess the movement that a person wants to make.