Learning Optimal Polices through Interactive Imitation Learning

Abstract

Imitation learning solves reinforcement learning problems with reference to some teacher information. While the typical method of behavioral cloning could not be applied to long-term tasks due to covariate shifts, interactive imitation learning solves this problem by obtaining online feedback from a teacher model. On the other hand, in the existing methods of interactive imitation learning, students could not learn the optimal policies when the teacher differed from the optimal for the student. In this study, we propose a novel method to solve this problem while providing an organized review of interactive imitation learning.