In this paper, we address the simultaneous scheduling of machines and mixed multi-load automated guided vehicles(AGVs) in flexible manufacturing systems. To minimize the maximum completion time, an improved mixed discrete differential evolution algorithm with variable neighborhood search is proposed. The variation and crossover of the differential evolution algorithm are performed by discrete operations, and the optimal individual of each iteration is updated by variable neighborhood search. A three-layer coding structure based on tasks, machines and AGVs is proposed and rationalized. Variable neighborhood search based on critical path is proposed, and the order of tasks on the critical path, machines and AGVs of corresponding processes are adjusted in a targeted manner to improve the local search efficiency. A colony update strategy based on the external archiving is designed to replace the poorer individuals in the population from the external archiving to improve the global search efficiency. Finally, through experiments, the effectiveness of the algorithm improvement strategy and the effectiveness of the algorithm are proved, the effects of the number of AGVs, the capacity of AGV loading and the available time of AGV batteries on the algorithm solution results and the improvement are analyzed.