Efficient constrained model predictive control

Abstract

Algebraic model predictive control is a more efficient predictive control algorithm in which the optimal control problem can be reduced down to a single prediction point, thus removing any additional and redundant calculations. The drawback of the reduction is lost system information and hence degraded performance while the system is constrained. This paper introduces a method of improving the constraint handling performance of the system by introducing additional points at which the constraints are checked. The computational efficiency is retained by recognizing that the key constraint is the peak location of the response, with a method shown for calculating its location. The proposed method is demonstrated on a numerical simulation of a second order system and linear aircraft model with performance and computational efficiency compared.