A group of researchers from the Ecole Polytechnique Fédérale de Lausanne has created a new prosthesis for the hand, which allows the control of the fingers and automation thanks to machine learning software. This ensures a better grip and better handling of objects.
The study, published in Nature Machine Intelligence, describes how this robot hand is constructed by combining two concepts that belong to two different sectors: one comes from neuro-engineering, the other related to robotics.
As Aude Billard, EPFL Learning Algorithms and Systems Laboratory, manager of the study, explains: “The robot hand can react within 400 milliseconds. It can react and stabilize the object before the brain can actually perceive that the object is slipping.”
In order to move the hand and thus the objects, the amputee has to perform a series of movements as soon as he has applied the correct structure. In this way, the software-automatic learning algorithm is trained and used by the robot hand to move. After training, the algorithm can understand how to move the brain, something that allows the robot to perform many more movements and more natural ways.
“Because muscle signals can be noisy, we need a machine learning algorithm that extracts meaningful activities from muscles and interprets them in movements,” says Katie Zhuang, first author of the study, suggesting how much software’s approach to artificial intelligence was necessary to obtain such functions.