Abstract
In fuzzy Q-learning, it is very difficult to design a state space for a given problem. We, therefore, proposed a dynamic fuzzy Q-learning with facility of turning and removing fuzzy rules to resolve it. We implemented a dynamic fuzzy Q-learning library in Java, which is composed of 3 components, initialization, selection of action and learning. In the initialization, a user defines learning parameters, fuzzy sets and selection methods, and give an initial Q-table. In the selection of action, the library selects actions for environment data using the Q-table and a selection method. In the learning, the library update Q-values, and tunes and removes fuzzy rules. A user can define your own fuzzy sets and selection methods. The library will find many applications to solve problems using Q-learning.
