Humanoid robots imitation of human motion using offline and real-time adaptation techniques

Abstract

This paper describes offline and real-time motion conversion approaches for imitating the motion of a human being using a humanoid robot. In both cases, the human motion is measured with marker points captured by the 3D motion capture system Vicon. A similar physical morphology between the shape of the robot and the human is assumed. This paper also demonstrates that the input measurements of the human motion to remain as close as possible to the real motion. In the offline approach, we use the iCub humanoid robot, while, for the real-time approach, James robot is used with some modification of the existing imitation and limiting human motion techniques. During the realisation of this work, we had a particular attention to obtain a 'natural' appearance of the motion. The proposed strategies, based on the obtained results, can be capable of accomplishing offline and real-time responses.