LEARNING FOR ADAPTIVE AND REACTIVE ROBOT CONTROL: A DYNAMICAL SYSTEMS APPROACH