This study aims to realize the joint deployment of multiple agricultural machinery stations, particularly for the real-time job orders. A mathematical model with a fuzzy time window was also established to minimize the total scheduling time and the number of dispatching agricultural machinery. Some factors were comprehensively considered, such as farmers' satisfaction, the cooperation of multiple agricultural machinery stations, the number of orders, the area of farmland, and the location coordinates. An improved genetic method (GA) with excellent parent genes was designed to fulfill the task of multi machine station responding to the demand of multi farmland. At the same time, the agricultural machinery was allocated in the shortest time to implement the operation requirements of each farmland, according to the shortest path. A case study was carried out to verify the model and the visual interface, including three stations of agricultural machinery and 35 operation orders of farmland in a certain area around Wuhan, Hubei Province of China. The results showed that an excellent searching and stable convergence were achieved in the scheduling system of agricultural machinery. Specifically, the reduction rate of the total scheduling distance was 9.89%, and the reduction rate of the number of agricultural machinery was 15.38%, when the fuzzy membership degree was 0.8. An optimal number of real-time orders accepted by a single farm station was not more than 20, according to the actual situation of the agricultural machinery quantity in each station. Furthermore, the improved GA presented a better performance than the hybrid genetic in general, indicating the less calculation time of the deployment, the more reasonable allocation of tasks, and the reduced scheduling distance. The multi-site and multi-machine cooperative instant repose scheduling was also considered the joint deployment agricultural machinery and fuzzy time window in the modeling. There was a higher accuracy of the scheduling operation on agricultural machinery, and the fully considered satisfaction of farmers, even though the complexity of model increased, compared with the scheduling operation at a single agricultural machinery station. In the scheduling algorithm, the crossover and mutation operators were improved to reduce the risk of the operation data falling into the local optimal solution with the less running time. Consequently, the scheme can be widely expected to completely deal with agricultural machinery scheduling under complex backgrounds, fully meeting the cooperative operation of multiple agricultural machinery stations for the real-time operation needs of farmers. This finding can provide a strong support to the cost-saving and high efficiency of operation on agricultural machinery in modern agriculture. [ABSTRACT FROM AUTHOR]