Inverse reinforcement learning