Application of Dynamic Fuzzy Q-learning with Forgetting Facility to Car Racing Game

Abstract

In fuzzy Q-learning, it is very difficult to design a state space for a given problem. We have, therefore, proposed a dynamic fuzzy Q-learning with generating, tuning and removing fuzzy sets and a pairs of state and actions using forgetting facility. We have also applied a conventional fuzzy Q-learning to a car agent of IEEE CEC 2007 Car Racing Competition, which is very difficult to define fuzzy sets of Q-table. In this paper, we apply the dynamic fuzzy Q-learning to Car Racing Game in order to acquire a appropriate state space. We have a good result and compare it to the conventional fuzzy Q-learning.