Extended linearised predictive control: Practical control algorithms for non-linear systems

Abstract

In this paper two predictive controllers for non-linear systems are presented. Both methods — ELPC and MELPC — are based on a local linearisation of the process at each sampling instant, thus, a priori, should perform better than the controllers designed in terms of a single linearisation about an operating point. On the other hand, their linear formulation leads to computation times much shorter than those provided by non-linear predictive control methods. After their formulation, ELPC and MELPC are compared with a linear predictive controller on a three-degrees-of-freedom robot, showing their good performance and computation times features.