Towards a Framework for the Whole-Body Teleoperation of a Humanoid Robot in Healthcare Settings

The use of robotic systems for doctor-patient interaction during Covid-19 and in post-pandemic phases has been proven useful. On the other hand, in current implementations, teleoperating a robot in critical contexts such as the medical scenario may induce a high mental workload on the operator, mainly due to the need to adapt to the remote control of a complex robot, and the reduced environmental awareness. Furthermore, robotic platforms for telemedicine do not usually offer the possibility of establishing physical contact with the patient, which may indeed be useful to show how to assume a certain posture, or to guide a specific movement. The aim of this work is to overcome these limitations, by creating a framework in which the arms, the head, and the base of a humanoid robot can be easily teleoperated with a rapid learning curve and a low mental workload for the operator. The proposed approach is based on the real-time human pose estimation of the operator, which is calculated in real-time and transformed into correspondent skeleton joint angles, used as input to control the upper body joints of the Softbank Robotics robot Pepper. Experiments with users have been performed to check the effectiveness of the imitation system, by verifying the similarity between the human and robot pose and measuring its usability and perceived workload.