Neural Network Based Predictive Control

Predictive control is the only ¿advanced¿ control strategy that is widely used in industry, because it is the only generic control scheme that can take hardware constraints explicitly into account. Often considered slow and only suitable for the control of processes with slow dynamics, recent research activities are aiming at a fast and efficient implementation of predictive control laws. This project will explore the possibility of efficient predictive control laws for highly nonlinear processes, using neural network representations and the idea of approximate predictive control. The efficiency of such controllers will be compared with linear parameter-varying control strategies.