Predictive Adaptive Control of Nonlinear Multivariable Systems Using Fuzzy CMAC

CMAC computational model that is based on the cerebellum structure is known as a Neural Network with high computation and learning speed. Fuzzy CMAC, by introducing fuzzy reasoning to CMAC, converts it from a black box to a white box whose performance can be interpreted using fuzzy rules. Fuzzy CMAC compared to CMAC has higher approximation and modeling capability and allows for model gradient computation necessary for applications such as predictive control. This paper presents two nonlinear predictive control algorithms based on fuzzy CMAC. The first algorithm uses a numerical method for optimization computations and the second one uses model gradient. Furthermore, their performance is shown by applying them to control a nonlinear multivariable system.