Using Multiple Models to Imitate Drumming

A. Tidemann and P. Öztürk (Norway)


Artificial Intelligence, Robot Design and Architecture, Motor control and learning, Drumming


The ability to learn motor skills by imitating other individ uals is an important part of human cognition. The research presented in this paper aims to implement this cognitive ability in robots. The goal is to facilitate programming of robots in a way that is natural to humans. To achieve this goal, a multi-modal architecture for learning motor skills by imitation is implemented. The system imitates human drumming behaviour. The goal is to imitate both the play ing style (i.e. the “groove”) as well as the arm movements. The virtual drummer can then be used in a musical setting, as it will both look and sound human. Echo State Networks are used to implement the architecture, and it self-organizes the control of the simulated robot. The ability of the system to self-organize is evaluated, showing promising results.

Important Links:

Go Back