Efficient Sampling-Based Maximum Entropy Inverse Reinforcement Learning with Application to Autonomous Driving