A comparison of the computational efficiency of generalised function MPC using active set methods

Abstract

This paper considers the computational efficiency of the recently developed generalised function parameterisations predictive control algorithms using active set methods. Alternative parameterisations have been shown to improve the volume of the feasible region when the number of degrees of freedom is limited. However, earlier work also demonstrates that some of the structure of optimisation problem is lost when using these parameterisations and thus requires more computation operations per step. This paper considers the dense problem structure arising from removing any redundant constraints and then considers the computational efficiency using the minimal constraint sets. Extensive simulation results suggest there can still be benefits from using more general parameterisations, although less significant then indicated by the number of degrees of freedom alone.