Hierarchical Imitation Learning for Stochastic Environments