An H_∞ robust model predictive control strategy is proposed for time-varying uncertain nonlinear systems subject to input constraints. The provided uniform bound of linearized system matrices, the non-convex programming problem is converted into a convex one with LMI constraints, which decreases the computational complexity of controllers. The receding horizon principle and the H_∞ control method are used to minimize performance indexes on-line, and the close-loop system satisfies the performance and constraints. The method solving nonlinear model predictive control(NMPC) and the sufficient condition guaranteeing robust stability of NMPC are derived in form of LMIs. A simulation example verifies the effectiveness of the proposed strategy.