Fuzzy CMAC Model Predictive Control

Conventional model predictive controller (MFC) described by a linear model may not be able to cope with highly nonlinear process. This has led to the development of a number of nonlinear MPC (NMPC). Most of these methods are excellent in theory but it is difficult to apply them to the industrial cases because of tremendous computation. The fuzzy cerebellar model articulation controller (CMAC) neural network is adopted as the model of the predictive control to describe the nonlinear process. The quickness of prediction is guaranteed by the principle of the fuzzy CMAC neural network characterized simple architecture. A novel online learning algorithm is proposed. The errors between the historical actual output and predictive output are used to adjust the weigh of the hypercubers in feedback correcting. Finally, the predictive control based on fuzzy CMAC model is demonstrated by the nonlinear tape rectification.