The paper proposes a model recovery anti-windup (MRAW) scheme for linear time-invariant and discrete-time systems under magnitude and rate saturation. The method is a modified, discrete-time counterpart of the algorithm presented in . As it is usual in the MRAW framework the AW compensator contains the exact copy of the plant in order that the ideal (unsaturated) behavior can be preserved in the states. The compensator is a controller that aims to push the plant towards this intended behavior. The design of this control action can be reduced to a construction of a stabilizing state feedback acting on the saturated plant. In  this feedback is a linear one, which is designed by convex optimization by enlarging the ellipsoidal approximation of the invariant domain. This paper presents a different, set-theoretic approach, which is based on the precise construction of the maximal control invariant set. The proposed control is a nonlinear one generated by point wise convex optimization.