Robust MPC for nonlinear multivariable systems

In this work, a robust predictive controller of uncertain nonlinear multivariable systems is developed. The control design is based on Multi-Input Multi-Output (MIMO) Nonlinear Auto Regressive Moving Average (NARMA) model. To cope with uncertain dynamic behavior of the system, the structured uncertainty is adopted. In fact, the main limitation of the robust predictive controllers is the computational burden leading to a lack of on line implementation. In this work, an efficient method is proposed. This method is based on transformation variables which reduce the initial non-convex problem to a convex programming. The efficiency of the proposed method is tested and compared with LMI and genetic algorithms optimizers on benchmark functions. The robustness of the proposed control law is experimented on three tanks system.