Interactive humanoid robot arm imitation system using human upper limb motion tracking

Abstract

Robot arm imitation based on human upper limb motion tracking possesses huge potential application value, especially with the addition of hazardous environments. It provides an intuitive way of teaching a complicated humanoid robot arm to perform human-like behaviors. In this paper, we build an interactive imitation system composed of Kinect and a self-built humanoid robot arm, which could mimic human upper limb motions in real time. The Kinect device captures the positions and joint orientations of the participant's upper limb. A method of determining the credibility of estimated joints is put forward, and along with Savitzky-Golay smoothing filter, we decrease jitter caused by measurement errors of Kinect. Experiments indicate that the robot arm system has a certain flexibility to accomplish a variety of human upper limb motions.