随着人工智能技术的发展，智能调度系统也应用到了更多领域。在工程领域中，土方工程作为建筑工程中至关重要的一环，也在进行智能化调度的升级和转变。为了提高机械配合效率和应对动态风险能力，多机协同的智能化调度方案有着至关重要的作用。因此，本研究的目标是基于土方施工工序、施工现场信息和机械参数信息，有效的解决土方施工中多机协作的静态以及动态调度问题。

本研究是根据土方施工机械调度的需求，对于静态的事前调度和受天气等影响产生新到达施工任务的动态调度分别生成多目标优化的调度方案，重点分析了土方施工机械调度的特点、影响因素以及存在的问题，总结了工程问题和实际需求，依此形成了土方施工多机柔性作业调度的约束条件和调度规则。在此基础上，分别建立了土方施工多机柔性作业静态调度和动态调度问题模型，开发了适用于静态和动态两种土方施工场景下的调度算法。在静态调度场景下，采用非支配排序遗传算法（NSGA-Ⅲ）求解多目标优化调度方案，提高求解的多样性和准确性，分析了不同算法参数对于方案的影响，并在不同规模的问题情景下验证算法的适用性。在动态调度场景下，采用深度强化学习 DDQN 算法生成动态的调度方案，分析不同参数以及调度规则对结果的影响，实验确定了最优参数组合和调度规则，同时对比 DQN 算法验证 DDQN 具有求解速度快、准确性高的优势。最后，综合施工现场的调度需求设计了一套能够解决多机柔性作业调度问题的土方智能施工调度系统，具有工程信息管理、施工方案生成、任务管理、质量监测等功能，起到指导施工的作用。
土方施工多机柔性作业调度是施工智能化发展中至关重要的一环。本研究提出了多机柔性作业调度问题模型，明确约束条件和调度规则，采用合适的算法分别对静态和动态调度问题进行求解，为土方施工机械调度提出了解决方案。

关键词：土方施工；多机调度；非支配排序遗传算法；深度强化学习；土方智能施工调度系统

With the development of artificial intelligence technology, intelligent scheduling system has also been applied to more fields. In the field of engineering, earthwork, as a crucial part of construction engineering, is also undergoing intelligent scheduling upgrading and transformation. In order to improve the efficiency of machinery cooperation and the ability to deal with dynamic risks, the intelligent scheduling scheme of multi-machine cooperation plays a vital role. Therefore, the goal of this study is to effectively solve the static and dynamic scheduling problem of multi-machine cooperation in earthwork construction based on earthwork construction process, construction site information and mechanical parameter information.
According to the needs of earthmoving construction machinery scheduling, this study generates a multi-objective optimization scheduling scheme for static pre-scheduling and dynamic scheduling of newly arrived construction tasks affected by weather, etc., focuses on analyzing the characteristics, influencing factors and existing problems of earthmoving construction machinery scheduling, summarizes the engineering problems and actual needs, and forms the constraints and scheduling rules for multi-machine flexible operation scheduling of earthmoving construction. On this basis, the static scheduling and dynamic scheduling problem models of multi-machine flexible operation of earthmoving construction are established, and the scheduling algorithms suitable for static and dynamic earthmoving construction scenarios are developed. In the static scheduling scenario, the nondominated ranking genetic algorithm (NSGA-III) is used to solve the multi-objective optimization scheduling scheme to improve the diversity and accuracy of the solution, the influence of different algorithm parameters on the scheme is analyzed, and the applicability of the algorithm is verified under different scale problem scenarios. In the dynamic scheduling scenario, the DDQN algorithm of deep reinforcement learning is used to generate a dynamic scheduling scheme, analyze the influence of different parameters and scheduling rules on the results, find the optimal parameter combination and scheduling rules, and compare the DQN algorithm to verify that DDQN has the advantages of fast solution speed and high accuracy. At the same time, a set of earthmoving intelligent construction scheduling system that can solve the problem of multi-machine flexible operation scheduling is designed based on the scheduling requirements of the construction site, which has the functions of engineering information management, construction plan generation, task management, quality monitoring and so on, and plays a role in guiding construction.
The flexible operation scheduling of multiple machines in earthwork construction is a crucial link in the development of intelligent construction. This study proposes a multimachine flexible job scheduling problem model, defines the constraints and scheduling rules, and uses appropriate algorithms to solve the static and dynamic scheduling problems respectively, providing a solution for earthwork construction machinery scheduling.