Learning Time-Invariant Reward Functions through Model-Based Inverse Reinforcement Learning