https://flecart.github.io/notes/inverse-reinforcement-learning/