A simple and fast geometric kinematic solution for imitation of human arms by a NAO humanoid robot

This paper addresses the imitation procedure of the human arms by a NAO humanoid robot. The exact position of the user arms are obtained based on the position of him/her shoulders, elbows and wrists, which are obtained by using the Microsoft Kinect as a 3-dimensional optical motion capture sensor. Communication between the kinect and computer server is done by using Robot Operating System (ROS) framework. Due to the differences between human and humanoid robot dimensions, captured motion should be mapped on the robot's dimensions. Mapped data is used for kinematic operation. In this paper, the inverse kinematic problem is solved by proposing an approach based on a novel simplified geometric method. In this approach, the distal form of the Denavit-Hartenberg convention of both arms is studied and a simple and fast geometric inverse kinematic solution is proposed. Finally, the proposed method is verified experimentally in Webots simulator and on a real NAO-H25 humanoid robot. Experiments are resulting in a coherent imitation between human arm and NAO with respect to time response.