A Robotic System For Imitating Human Percussionists

Robot musicians have the potential to revolutionise the way humans perceive and create music. Recent breakthroughs in this field have tended to focus more on the digital generation of music. Instead, we address how a musician’s physical embodiment can be translated to a robotic arm. Robots endowed with human-like musical capability open the possibility for wider applications such as human-robot bands, musical education and musical art. Prior work in this area tends to rely on pre-programmed actuation which is limited to simple motion and sound. In this paper, we propose a robotic system capable of imitating a human musician, with a focus on percussion instruments. Our system consists of a method for recording the human demonstration, a compact continuous representation of the demonstrated motion and a motion reproduction method which considers the dynamic constraints of the robot. We present results of our system and show that it is capable of closely reproducing the motion of the human percussionist.