Robot cognitive control with a neurophysiologically inspired reinforcement learning model