Robust model predictive control using neural networks

The paper deals with robust model predictive control designed using recurrent neural network. A dynamic neural network is trained to act as the one-step ahead predictor, which is then used successively to obtain k-step ahead prediction of the plant output. Based on the neural predictor, the control law is derived solving a constrained optimization problem. The robustness of the considered predictive scheme is derived using the concept of an error model. Based on the developed robust model, a optimization problem is redefined. Two solutions are portrayed. The first one is to change the cost function in order to consider the robust model of the plant, while the second one is to impose constraints on the process output using derived uncertainty bands.