In this paper, a robust model prediction controller with state-feedback structure is constructed for a class of uncertain discrete nonlinear systems with multiple states time-delay, inputs time-delay and nonlinear perturbation. Firstly, an improved quadratic Lyapunov-Krasovskii functional is constructed by making full use of the upper and lower bounds of the time-delay information, and then the difficulty of solving min-max optimization problem is transformed into convex optimization problem with linear matrix inequality (LMI) constraints by utilizing LMI technique. Simultaneously, the sufficient conditions for the existence of robust predictive controller and the expression of the controller are given. Theoretically, it is proved that the designed controller can guarantee the closed-loop system is asymptotically stable. Finally, simulation results demonstrate the feasibility of the proposed scheme.