Inverse Reinforcement Learning algorithms and features for robot navigation in crowds: An experimental comparison | Synapse