Transfer Learning for Fuzzy Actor-Critic Learning via Fuzzy Rule Transfer

Reusing knowledge from previous experience to accelerate learning is the central idea in transfer learning. This notion also applies to reinforcement learning where the agents acquire knowledge via interacting with the environment. In this thesis, we explore transfer learning in the fuzzy reinforcement learning domain, particularly in the environment of differential games. We present a novel approach to transfer knowledge between similar tasks which use the Fuzzy Actor Critic Learning (FACL) algorithm. Specifically, we propose a fuzzy rule transfer (FRT) method to map fuzzy rules between source and target tasks.