A computationally efficient constrained predictive control law

A predictive control algorithm is presented that uses a weighted sum of Linear quadaratic (LQ) optimal predictions and 'Mean Level' predictions. Each time step only a single weighting parameter is solved for using a Linear Program problem hence giving a low computational load. The controller achieves LQ-optimality when constraints are inactive, and remains feasible (under mild conditions) if constraints are active. © 1998 AACC