Imitation of Human Upper-Body Motions by Humanoid Robots

Abstract

This paper presents a method for offline imitation of human upper-body motions. This approach is based on inverse kinematics with task classification which includes an end-effector tracking, joint limits avoidance and joint trajectories tracking, and the results are validated on the humanoid robot NAO. The whole process includes taking data from human using both markerless and marker-based motion capture systems, scaling down the human data to the robot size, calculating the joint angles, and iterating equations of inverse kinematics. After, the final values of joint angles are calculated, they are operated on the real robot using a publisher in Robot Operating System (ROS). On contrary to the studies in literature, we also focus on imitating human motions by a marker-based motion capture system, so we have additional joint orientation information and more accurate results.