Discrete-time receding horizon optimal control is employed in model-based anti-windup augmentation. The optimal control formulation enables designs that minimize the mismatch between the unconstrained closed-loop response with a given controller and the constrained closed-loop response with anti-windup augmentation. Recently developed techniques for off-line computation of the constrained linear regulator's solution, which is piecewise affine, facilitate implementation. The resulting sampled-data, anti-windup closed-loop system's properties are established and its performance is demonstrated on a simulation example. (C) 2003 Elsevier Ltd. All rights reserved.