Mitigating Covariate Shift in Imitation Learning via Offline Data Without Great Coverage

Open in new window