Minimum variance control of a class of nonlinear stochastic systems

The extension of the linear minimum variance controller to the nonlinear minimum variance controller is presented. The algorithm relies on the description of the underlying process by a nonlinear autoregressive moving average model and on the optimal (last squares) predictor to obtain an explicit control formula. It contains feedforward and feedback compensators which are all simple to be realized digitally. The theoretical basis of the procedures is presented along with simulation tests.