Imitation of human arm movements by a humanoid robot using monocular vision

Abstract

A sociable robot must have the capability to imitate agents around it. In a human society, people generally teach new skills to other people by demonstration. Hence, our artificial partners should be able to learn from us by watching what we do. In this thesis, we programmed a humanoid robot to imitate human arm movements. The main problems we dealt with is the perception of the human arm movement and finding a corresponding motor sequence that will make the robot’s movement as much the same as human’s. We placed colored markers on the arm joints of the human demonstrator for tracking the arm. The physical difference between the human and the robot’s arms made it difficult to directly extract the necessary joint angles for the robot. Instead, we employed a shared representation of movement which is neither the joint values of the human nor the robot. We made our comparisons on that common model. We have conducted real world experiments where the robot watched a human demonstrator drawing horizontal and vertical lines and circle on the board. The robot successfully drew the lines and the circle.