Data-Driven Personalization of Body-Machine Interfaces to Control Diverse Robot Types

Abstract Body-Machine Interfaces (BoMIs) for robotic teleoperation can improve a user’s experience and performance. However, the implementation of such systems needs to be optimized on each robot independently, as a general approach has not been proposed to date. Here, we present a novel machine learning method to generate personalized BoMIs from an operator’s spontaneous body movements. The method captures individual motor synergies that can be used for the teleoperation of robots. The proposed algorithm applies to people with diverse behavioral patterns to control robots with diverse morphologies and degrees of freedom, such as a fixed-wing drone, a quadrotor, and a robotic manipulator.