Versatile Offline Imitation from Observations and Examples via Regularized State-Occupancy Matching
This addresses the problem of learning from expert observations in offline settings for robotics and AI applications, offering a versatile solution but is incremental as it builds on existing imitation learning frameworks.
The paper tackled offline imitation learning without expert actions by proposing SMODICE, a regression-based algorithm via state-occupancy matching, and demonstrated its effectiveness across three settings, significantly outperforming prior state-of-the-art methods in gridworld and high-dimensional benchmarks.
We propose State Matching Offline DIstribution Correction Estimation (SMODICE), a novel and versatile regression-based offline imitation learning (IL) algorithm derived via state-occupancy matching. We show that the SMODICE objective admits a simple optimization procedure through an application of Fenchel duality and an analytic solution in tabular MDPs. Without requiring access to expert actions, SMODICE can be effectively applied to three offline IL settings: (i) imitation from observations (IfO), (ii) IfO with dynamics or morphologically mismatched expert, and (iii) example-based reinforcement learning, which we show can be formulated as a state-occupancy matching problem. We extensively evaluate SMODICE on both gridworld environments as well as on high-dimensional offline benchmarks. Our results demonstrate that SMODICE is effective for all three problem settings and significantly outperforms prior state-of-art.