An improved discrete differential evolution algorithm with variable neighborhood search was proposed for solving simultaneous scheduling of machines and AGVs in flexible manufacturing systems. With the optimization goal of making the maximum completion time minimum, considering the dual resource constraints of machines and AGVs, the corresponding mathematical model was established. The three-layer coding structure of operation,machine and AGV was employed to schedule machines and AGVs simultaneously. In order to improve the global search capability, the differential evolution algorithm generated new individuals by improved mutation and crossover operators, and introduced the acceptance criterion of solution in simulated annealing algorithm to select next generation. Furthermore, a variable neighborhood search was performed on the optimal individual in each iteration of the algorithm in order to enhance the local search capability. Finally, the effectiveness, stability and superiority of the improved differential evolution algorithm were proved by calculation and comparison of examples.