Intrinsically-motivated reinforcement learning for control with continuous actions