Neuroevolutionary reinforcement learning for generalized helicopter control