Robust adaptive decoupling design for generalized predictive control with neural network

A novel robust decoupling method with multivariable generalized predictive control (MGPC) for a class of nonlinear systems is presented in an adaptive version. System cross-coupling action and the nonlinear actors are identified on-line by a neural network, which is then compensated in the control algorithm using the feedforward technique to realize robust decoupling. The identified result is also taken as a modifying signal of parameter estimate so that the equivalent model matches the real system well. Simulation results show the effectiveness of the proposed algorithm.