Environment Design for Inverse Reinforcement Learning

arXiv:2210.14972v2 Announce Type: replace
Abstract: Learning a reward function from demonstrations suffers from low sample-efficiency. Even with abundant data, current inverse reinforcement learning methods that focus on learning from a single environment can fail to handle slight changes in the environment dynamics. We tackle these challenges through adaptive environment design. In our framework, the learner repeatedly interacts with the expert, with the former selecting environments to identify the reward function as quickly as possible from the expert's demonstrations in said …

abstract arxiv challenges cs.ai cs.lg current data design dynamics efficiency environment focus framework function low reinforcement reinforcement learning sample the environment through type