Causal Imitation Learning under Temporally Correlated Noise
This addresses the problem of robust imitation learning for AI/robotics practitioners when expert data is noisy, though it is incremental as it builds on existing instrumental variable regression techniques.
The paper tackled imitation learning from expert data corrupted by temporally correlated noise, which causes spurious correlations and poor policy performance, and developed two algorithms (DoubIL and ResiduIL) that outperformed behavioral cloning on simulated control tasks.
We develop algorithms for imitation learning from policy data that was corrupted by temporally correlated noise in expert actions. When noise affects multiple timesteps of recorded data, it can manifest as spurious correlations between states and actions that a learner might latch on to, leading to poor policy performance. To break up these spurious correlations, we apply modern variants of the instrumental variable regression (IVR) technique of econometrics, enabling us to recover the underlying policy without requiring access to an interactive expert. In particular, we present two techniques, one of a generative-modeling flavor (DoubIL) that can utilize access to a simulator, and one of a game-theoretic flavor (ResiduIL) that can be run entirely offline. We find both of our algorithms compare favorably to behavioral cloning on simulated control tasks.