Abstract: For the control of a multiple-input multiple-output (MIMO) system with actuator constraint, we propose a new fast anti-windup predictive compensation control algorithm based on the sliding mode control method. First, we construct the equivalent unconstrained augmented model of the system with input constraint. Then, based on the assumption that the control effort's change rate is bounded, we design the strategy of the searching compensation control effort in the bounded region. It reduces the disturbance in predictive optimization, compresses the optimization space, and ensures rolling optimization with guaranteed cost. The proposed method has distinct physical meaning, and it applies to both linear and nonlinear time invariable systems. Simulation results show that the proposed method can improve the control performance of a system with input constraints, and possesses such characteristics as less memory occupation and calculation.