Learning Motor Control by Dancing YMCA

Abstract

To be able to generate desired movements a robot needs to learn which motor commands move the limb from one position to another. We argue that learning by imitation might be an efficient way to acquire such a function, and investigate favorable properties of the movement used during training in order to maximize the control system’s generalization capabilities. Our control system was trained to imitate one particular movement and then tested to see if it can imitate other movements without further training.