In sliding mode control, state trajectory is divided to reaching mode and sliding mode. A method of robust model predictive control is employed to improve performance in reaching mode. Uncertain discrete time systems can be transformed to polytopic uncertain system. Then, the employed robust model predictive control can handle the robustness in uncertain polytopic system. The strong point of the proposed method is optimization of the cost function. As a result, the convergence rate to sliding surface is improved in reaching mode. Moreover, input constraint can be considered. A numerical example illustrates the effectiveness of the proposed method by comparison with existing methods.