Neural network-based fuzzy control surface implementation

This paper proposes a new design methodology of two-input and one-output fuzzy logic controller by training an Artificial Neural Network (ANN) that approximates a fuzzy control surface resulting from a basic fuzzy controller. The main purpose of this approach is to fully exploit the Artificial Neural Network (ANN) feature by translating the expertise of controlling the plant into two stages. In the first stage, our methodology mathematically presents the fuzzy rules and the procedure of obtaining a fuzzy control surface. In the second stage, we map the resultant fuzzy control surface with an ANN model that can be easily calculated. We have implemented the trained ANN with the LabVIEW2009 program to control the car parking system, whose simulation results established the validity of the proposed controller.