A computationally advantageous reparameterisation of a robust model predictive control scheme

Abstract

In this paper, a two step reparameterisation is proposed to reduce the computational cost of an online constraint tightening model predictive control (MPC) scheme for control of linear systems subject to time-varying state constraints and a bounded disturbance. The method is an exact reparameterisation based on well-established linear algebra techniques and does not cause degradation in the performance when compared to the original problem. Applying the reparameterisation, a smaller problem is to be solved at each step which yields computational benefits as demonstrated via a numerical example.