Imitation Learning from Video by Leveraging Proprioception
This work addresses a practical problem for robotics and AI by enabling more flexible imitation learning from video, though it is incremental as it builds on prior observation-based methods.
The paper tackles imitation learning from visual demonstrations without action labels by incorporating proprioceptive states into policy learning, achieving significant performance improvements over existing methods on MuJoCo domains.
Classically, imitation learning algorithms have been developed for idealized situations, e.g., the demonstrations are often required to be collected in the exact same environment and usually include the demonstrator's actions. Recently, however, the research community has begun to address some of these shortcomings by offering algorithmic solutions that enable imitation learning from observation (IfO), e.g., learning to perform a task from visual demonstrations that may be in a different environment and do not include actions. Motivated by the fact that agents often also have access to their own internal states (i.e., proprioception), we propose and study an IfO algorithm that leverages this information in the policy learning process. The proposed architecture learns policies over proprioceptive state representations and compares the resulting trajectories visually to the demonstration data. We experimentally test the proposed technique on several MuJoCo domains and show that it outperforms other imitation from observation algorithms by a large margin.