Learning Self-Correctable Policies and Value Functions from Demonstrations with Negative Sampling
This addresses sample efficiency and error correction in imitation learning for robotics control, representing an incremental improvement over prior methods.
The paper tackles the covariate shift problem in imitation learning by introducing conservatively-extrapolated value functions that enable self-correcting policies, and shows that their VINS algorithm corrects behavioral cloning mistakes on robotics tasks and significantly improves sample efficiency when initializing reinforcement learning.
Imitation learning, followed by reinforcement learning algorithms, is a promising paradigm to solve complex control tasks sample-efficiently. However, learning from demonstrations often suffers from the covariate shift problem, which results in cascading errors of the learned policy. We introduce a notion of conservatively-extrapolated value functions, which provably lead to policies with self-correction. We design an algorithm Value Iteration with Negative Sampling (VINS) that practically learns such value functions with conservative extrapolation. We show that VINS can correct mistakes of the behavioral cloning policy on simulated robotics benchmark tasks. We also propose the algorithm of using VINS to initialize a reinforcement learning algorithm, which is shown to outperform significantly prior works in sample efficiency.