Sequence Model Imitation Learning with Unobserved Contexts

We consider imitation learning problems where the expert has access to a per-episode context that is hidden from the learner, both in the demonstrations and at test-time. While the learner might not be able to accurately reproduce expert behavior early on in an episode, by considering the entire history of states and actions, they might be able to eventually identify the context and act as the expert would. We prove that on-policy imitation learning algorithms (with or without access to a queryable expert) are better equipped to handle these sorts of asymptotically realizable problems than off-policy methods and are able to avoid the latching behavior (naive repetition of past actions) that plagues the latter. We conduct experiments in a toy bandit domain that show that there exist sharp phase transitions of whether off-policy approaches are able to match expert performance asymptotically, in contrast to the uniformly good performance of on-policy approaches. We demonstrate that on several continuous control tasks, on-policy approaches are able to use history to identify the context while off-policy approaches actually perform worse when given access to history.