According to the varying scope of parameters, multiple models are set up for a kind of discrete time nonlinear system, and the corresponding controllers based on these models are given. At every sample time, the best model is selected by using an index switching function based on the estimation error of every model, and the controller based on this model is switched. “Localization”method is used to decrease the computing quantity of multiple model adaptive controller without the loss of control property. It can be proved that the closed-loop system is stable when different controllers are switched
each other. Due to the existence of multiple models, the control performance is improved greatly.