Learning Bimanual Coordination Patterns for Rhythmic Movements

Coordinated bimanual movements form the basis for many everyday motor skills. In human bimanual coordination there are several basic principles or default coordination modes, such as the preference for in-phase or anti-phase movements. The objective of our work is to make robots learn bimanual coordination in a way that they can produce variations of the learned movements without further training. In this paper we study an artificial system that learns bimanual coordination patterns with various phase differences, frequency ratios and amplitudes. The results allow us to speculate that when the relationship between the two arms is easy to represent, the system is able to preserve this relationship when the speed of the movement changes.