A fast algorithm for a stabilizing receding horizon controller

A fast algorithm for a stabilizing receding horizon controller is proposed to reduce the computational load from O(N/sup 3/) to O(N), N being the control horizon. The algorithm can be used for solving a prototypical model-based predictive control problem. The algorithm is based on a constrained least-squares formulation which permits treatment of the cost-functional, the plant output predictions, and the terminal constraints in a unified manner. It employs fast Givens and Gauss transformations as basic numerical tools resulting in a simple and numerically reliable implementation. >