Dynamic fuzzy Q-learning with facility of tuning and removing fuzzy rules

In Q-learning, it is very difficult to design a state space for given problems. We propose a dynamic fuzzy Q-learning with facility of tuning and removing fuzzy rules to resolve it. We dynamically construct fuzzy state space of the continuous attributes, that is, we have no initial rules and gradually generate new fuzzy rules with the states of fuzzy sets and tune the center values and widths of fuzzy sets with TD (Temporal Difference) error with removing unnecessary fuzzy sets and rules. We apply the method to the pursuit problem in the continuous environment.