Challenges of Model-Based Inverse Reinforcement Learning