Two-loop robust model predictive control with improved tube for industrial applications

Abstract

In this paper, a two-loop Model Predictive Controller (MPC) with an improved tube is proposed for industrial application with bounded uncertainties subject to input and state constraints. This scheme attempts to remove some existing obstacles against exploiting MPC in industrial applications, such as (i) risk and cost of a new controller replacement, (ii) difficulties of attaining a precise open-loop model of an industrial system and (iii) high computational burden of MPC methods. To this end, tube conservatism and calculation burden are reduced using the transient response of the error dynamics. Thus the feasible region of the MPC is enlarged and its computation time is reduced. To reduce modelling difficulties, the investigated approach does not require the open-loop model dynamics of the system and utilises the closed-loop model instead. On the other hand, it allows the existing inner-loop controller to remain unchanged without any manipulations, which results in eliminating a new controller replacement risk and cost. Additionally, for the proposed control method, robust stability and recursive feasibility are guaranteed without terminal gradients. Finally, an illustrative example is carried out in the simulation results to show the effectiveness of the proposed approach.