Learning Teleoperation of an Assistive Humanoid Platform by Intact and Upper-Limb Disabled Users

With the advent of highly dexterous robotic arms, assistive platforms for home healthcare are gaining increasing attention from the research community. Control of the many degrees of freedom of such platforms, however, must be ensured uniformly, both for non-disabled and disabled users, in order to give them as much autonomy as possible. Nine users, including two upper-limb disabled, were asked to complete highly complex bimanual tasks by teleoperating a humanoid robot with biosignals. The users were equipped with a light and wearable interface consisting of a body tracking device for guiding the torso and arms and two electromyography armbands for controlling the hands by means of interactive machine learning. All users were able to complete the required tasks, and learning curves are visible in completion time metric.