Design of Multi-Model Controller Based on Neural Network Sets

A multi-model controller based on the neural network sets is proposed for the time-variant nonlinear systems with uncertainty and time-delay in this paper. The local-models and the local-controllers are built to neural network sets. The output of the controller is the weighted sum of the local-controllers' outputs, and by this way the smooth-switch among the local-models is realized. The robustness of the controller is also enhanced by using of the neural network sets. The controller is applied to the three-tank level system to demonstrate its efficacy.