在以人力资源为主要资源的项目中，员工请假会使项目执行过程中资源可用量下降，导致事先制订的基线调度计划遭到破坏，采用反应型调度方法可以对被破坏的基线调度计划进行有效修复。针对员工请假情形的多技能分布式资源受限多项目调度问题（MS-DRCMPSP），建立包括初始局部调度、全局协调决策和修复调度的三阶段决策模型，初始局部调度模型和全局协调决策模型用于制订初始的基线调度计划，当员工请假导致基线调度计划遭到破坏时，管理者根据对活动和员工的决策偏好修复被破坏的基线调度计划，设计包含等待和调整策略的动态策略制订修复调度计划；基于MPSPLIB问题库改编生成多技能多项目调度问题算例，开展详细的实验研究。研究结果表明：考虑修复前后员工工作时间偏差最小的修复目标时，可使用等待或动态策略进行修复；考虑平均项目延期及修复前后活动开始时间偏差最小的修复目标时，可使用调整或动态策略进行修复；综合考虑活动和员工两方面的修复目标时，可直接使用动态策略完成修复调度计划。求解大规模算例时，本文设计的基于冲突活动的softmax评分机制的修复算法优于序贯博弈算法及其余启发式算法，优化结果最大可提升14.48%。

In the distributed decision-making environment， the shared resources are considered multi-skilled staff， independently dispatched by the respective project decision-makers from their interests. The shared multi-skilled staff， as the only link between multiple projects， are uniformly coordinated and distributed by the coordination manager. An effective coordination mechanism based on staff characteristics is designed to allocate shared multi-skilled staff for multi-project activities and determine the activity’s start time. This problem is called Multi-Skilled Distributed Resource Constrained Multi-Project Scheduling Problem （MS-DRCMPSP）. In these staff-oriented projects， the staff leave will reduce the available resources during project execution， destroying the pre-established baseline scheduling plan. How to repair the damaged baseline scheduling plan with a reactive scheduling method is studied.To solve the distributed resource-constrained multi-project scheduling problem with the multi-staff leave， a three-stage decision model including initial local scheduling， global coordination decision-making， and repaired schedule is established. The first two models are used to formulate the initial baseline scheduling plan. The initial local scheduling stage takes minimizing the project makespan as the optimization goal， and the global coordination decision-making model takes minimizing the average project delay （APD） as the optimization goal. When the baseline scheduling plan is damaged due to staff leave， the manager designs the repaired objective according to the decision preference of activities and staff to repair the damaged baseline scheduling plan. A “dynamic” strategy， including “waiting” and “adjusting” strategies is designed to develop the repaired scheduling plan， and a softmax scoring mechanism based on conflict activities is designed to apply in the repaired scheduling plan. The softmax scoring mechanism includes a series of evaluation factors： activity duration， slack time， resource demand， and several immediate activities.An example of a multi-skilled scheduling problem is generated based on MPSPLIB and experimental research is carried out on this example. The research shows that when considering the repaired objectives of staff working time deviation before and after the repairing， the “waiting” or “dynamic” strategy can be applied； When considering the minimizing repaired objectives of the APD and the deviation of activities start time before and after repairing， the “adjusting” or “dynamic” strategy can be applied； When considering the minimizing repaired objectives of both activities and staff， the “dynamic” strategy can be directly used to complete the repaired scheduling plan； When solving large-scale examples， the three-stage algorithm of softmax scoring mechanism based on conflicting activities is superior to sequential game algorithm and other heuristic algorithms. The optimization results can be improved by 14.48% at most.The vacancy of the multi-skilled scheduling problem in the distributed decision-making environment is made up for， especially in the case of uncertain resource availability. In addition， it provides a suitable repaired strategy for managers' different preferences and the related reference for subsequent relevant researches.