Abstract: Against the backdrop of accelerated intelligent development and large-scale operation in modern agriculture, efficiently accomplishing order allocation and route scheduling in multi-depot agricultural machinery cooperative service systems has become a critical technical challenge for improving resource utilization efficiency, reducing operational costs, and ensuring service quality. In practical scheduling scenarios, order allocation decisions and route planning processes are inherently coupled, and treating them as independent problems often leads to suboptimal solutions that fail to meet the requirements of large-scale, high-complexity agricultural service environments. To address these limitations, this paper proposes a multi-objective hierarchical optimization method for agricultural machinery scheduling that integrates multi-depot order allocation and internal route planning within a unified decision-making framework. From a system-level perspective, a bi-level optimization structure is constructed, in which the upper level focuses on allocating orders among multiple agricultural machinery depots, while the lower level optimizes detailed route planning and machinery scheduling within each individual depot, thereby explicitly capturing the interaction between allocation decisions and routing performance. At the upper level, a multi-objective order allocation model is formulated by simultaneously minimizing total scheduling cost and maximizing farmer satisfaction under realistic operational constraints, including machinery capacity limitations, service time windows, and task continuity requirements. An improved NSGA-II multi-objective evolutionary algorithm is developed to solve the upper-level model, incorporating problem-oriented genetic operators and solution evaluation mechanisms to enhance convergence performance, solution diversity, and overall stability. At the lower level, an enhanced route planning algorithm is designed to further optimize the internal scheduling schemes generated by the upper-level allocation results. This algorithm adopts a dual-layer chromosome encoding strategy and integrates heuristic initialization, adaptive evolutionary operators, and local search mechanisms, which effectively improve route feasibility and convergence efficiency under increasing task complexity. To comprehensively evaluate the effectiveness and practicality of the proposed hierarchical optimization approach, extensive simulation experiments are conducted based on real scheduling demands from an agricultural machinery cooperative in Henan Province, covering multiple order scales and service complexities. Comparative experiments are designed using traditional fixed allocation strategies and classical algorithms as benchmarks. The experimental results demonstrate that, under different order scales, the proposed collaborative optimization method consistently outperforms fixed allocation strategies in terms of comprehensive scheduling cost, achieving average cost reductions ranging from approximately 2.88% to 5.25% while maintaining a high level of service satisfaction. In medium-scale order scenarios, farmer satisfaction is increased by up to 18.86%, indicating that the proposed method effectively balances economic efficiency and service quality. Furthermore, multi-algorithm comparative experiments based on the hypervolume metric confirm that the improved NSGA-II algorithm exhibits superior performance in both convergence and diversity compared with other competing multi-objective evolutionary algorithms, demonstrating stronger global search capability and solution set robustness. In addition, the effectiveness of the improved internal route planning algorithm is quantitatively analyzed, and the results show that the enhanced strategy achieves consistent cost optimization across all tested order scales, with the maximum cost reduction reaching 8.32% in large-scale task scenarios. As the problem scale and complexity increase, the proposed route planning algorithm demonstrates improved adaptability and scalability, outperforming traditional genetic algorithms in terms of cost efficiency and solution stability. Moreover, the average satisfaction level is improved in most order-scale scenarios, with the maximum improvement exceeding 40%, reflecting the effectiveness of the introduced local search mechanisms and adaptive evolutionary strategies in reducing service delays and enhancing scheduling timeliness. Overall, the experimental results demonstrate that the proposed multi-objective hierarchical optimization framework effectively captures the coupling between order allocation and route planning, enabling coordinated optimization across decision levels and showing strong applicability to intelligent scheduling of multi-depot, multi-type agricultural machinery, while providing useful methodological support for intelligent agricultural machinery scheduling.