【作者】 郭研；

【导师】 李南；

【作者基本信息】 南京航空航天大学 ， 管理科学与工程， 2013， 博士

【摘要】 在软件企业实践中，大多数的软件项目不是孤立的，而是与其它项目之间存在资源竞争和信息交流等各种联系，而这些软件项目群之间的资源竞争，加剧了软件项目管理的复杂性。因此研究资源受限软件项目群调度问题的理论与方法，有利于软件项目管理领域的发展建设。本文围绕着资源受限软件项目群调度中的多技能员工配置问题、多模式多资源均衡问题和基于模糊关键链项目群调度问题展开研究，主要内容有：(1)针对软件项目的特殊性，详尽分析了软件企业项目群管理的内容和过程，并建立了软件项目群调度框架模型，该模型以项目群管理系统为核心，包括了需求管理、过程管理、文档管理等模块。最后给出了软件项目群调度过程的具体步骤。(2)多技能员工配置问题是当前软件企业实际工作中广泛存在的一种项目调度问题。在该问题中，每个员工掌握的技能不同，每项任务所需的技能也不尽相同。无法用传统的资源受限项目调度算法去处理项目群多技能员工配置问题。本文针对软件项目群多技能员工配置问题的特点，建立了以项目群总工期和总费用为目标的调度模型，然后将云模型嵌入到基于Pareto的向量评价微粒群算法中，提出了一种新的基于云多目标微粒群算法。最后通过一个软件研发案例验证了该方法的有效性和可行性。(3)对于许多软件研发项目，每项任务可能有多种不同的执行模式，每种执行模式代表不同的资源需求与工期组合。对此类软件项目群进行资源均衡优化时，不仅要调整各项任务的实际开工时间，还要选取合适的执行模式，这就是软件项目群多模式多资源均衡问题。由于该问题中，任务工期是离散分布的，所以软件项目群多模式多资源均衡问题属于不确定环境下的离散软件项目群调度问题的范畴。本文提出了软件项目群多模式多资源均衡优化方法，该方法通过建立多目标优化模型，能同时对项目群工期、资源总量和资源方差进行优化。在算法设计上，我们将种群竞争模型嵌入到基于Pareto的向量评价微粒群算法中，提出了一种新的基于动态种群的多目标微粒群算法。该算法结合任务执行模式和开工时间设计了微粒编码，并可根据优化目标动态调整各子微粒群的规模。最后，通过“任务模式数相等”和“任务模式数不等”两个软件研发案例，测试了新算法的性能，结果表明该算法对于求解软件项目群多模式多资源均衡问题是有效的。(4)现有的关键链多项目调度方法一般基于概率论基础，要求给出任务工期的概率分布，这对于缺乏历史统计数据的软件项目是非常困难的，因而现有关键链多项目调度方法并不适用于不确定环境下的软件项目群调度问题。本文首先借鉴了模糊关键链管理领域的相关研究成果，分析了现有的任务工期估算方法的缺陷和不足，提出了基于德尔菲的任务工期模糊估算法。然后研究了模糊关键链项目群调度中能力约束缓冲的设置方法，该方法充分考虑了能力约束任务的复杂性、资源紧张度和安全时间。最后通过软件案例验证了该方法能以较高的可能性提供足够的错开时间，同时对项目群最终完工率影响很小。本论文的研究能够为软件企业优化配置各种项目资源提供理论依据，为相关调度软件的开发提供研究基础，进而提高软件项目群的综合效益。更多还原

【Abstract】 In the practice of software enterprises, most of software projects are not insulated but relatedbetween each other in resource competition and the exchange of information, and then the resourcecompetition of these software programmes has intensified the complexity of software projectmanagement. Therefore, it is of advantage to the development and construction of the field ofsoftware project management that to research theories and methods for resource-constrained softwareprogramme scheduling problem.This paper has worked on a series of questions for resource-constrained software programmescheduling problem on multi-skills employee dispatching, multi-mode multiple resources levling andprogramme scheduling based on fuzzy critical chain. The achievements and main works of the paperare as follows:(1) In accordance with the characteristics of software project, the content and process ofprogramme management in software enterprise are analysized detailed in this paper, and we proposethe frame model for software programme scheduling. This frame model is based on programmemanagement system, and is consisted of requirement management, process management anddocument management. Finally, the process of software programme scheduling is designed.(2) Multi-skilled employee dispatching is a problem that is encountered frequently in the practiceof software enterprise. In this problem, every employee has distinguished skills and each activity hasdifferent skill requirements. Traditional methods for resource-constrained project scheduling are notsuited for programme multi-skilled employee dispatching problem. For the characteristics ofmulti-skilled employee dispatching problem in software programme scheduling, a scheduling modelis established with the optimization object of minimum multi-project duration and minimum total cost.By applying cloud model into Vector Evaluated Particle Swarm Optimization Based on Pareto, anovel Cloud Multi-Objective Particle Swarm Optimization is utilized for solving this problem. Finally,the effectiveness and feasibility of this method are verified by a software development case.(3) For many projects in software development enterprise, each activity is executed in one ofseveral modes, and each mode represents a combination of its resource requirements and its duration.It is crucial to not only adjust the start time, but also select the execution mode of each activity whenminimizing the variation of resource utilization. This issue has affected multi-mode multiple resourceslevling problem in software programme scheduling. Because activity duration in this problem is discrete distribution, intrinsically, multi-mode multiple resource leveling problem in softwareprogramme scheduling belongs to the discrete software programme scheduling problem in theuncertain environment. The method for multi-mode multiple resources leveling problem in softwareprogramme scheduling is formulated in this paper, in this method; we establish a multi-objectivemodel to minimize programme duation, resource requirements and resource variance. By applyingpopulation competition model into vector evaluated particle swarm optimization based on Pareto, anovel multi-objective particle swarm optimization is utilized for solving this problem. In thisalgorithm, the swarm code is comprised of the execute mode and the start time of activities, and thepopulation of each sub-partcile swarm is adjusted automatically according to the optimaizationobjectives. Finally, two software R&D cases test the performance of new algorithm, and results showthat this method is suited for solving multi-mode multiple resource leveling problem in softwareprogramme scheduling.(4) The current method for critical chain multi-project scheduling based on probability theory,which requires giving the probability distribution of activity duration. It is difficult for softwareproject which lacks of history statistical data, and then traditional methods for critical chainmulti-project scheduling are not suited for software programme scheduling problem in the uncertainenvironment. This paper is based on relevant research on fuzzy critical chain, and analyses the defectsand deficiencies of the estimation method for activity duration. A novel fuzzy estimation technique foractivity duration based on Delphi is proposed in this paper. Then a new method to determine thecapacity constraint buffer size is discussed in fuzzy critical chain programme scheduling, thecomplexity, resource tightness and safety time of capacity constraint activity are considered in themethod. Finally, the result shows that the method is able to provide the sufficient staggered time andalmost the same on-time completion rate of the whole projects.This research will provide theory basis for the optimum assignment of project resource insoftware corporations and foundation for the development of relative scheduling software, which willimprove the comprehensive benefits of the software programme.更多还原