基于用户需求的特征模型配置分析与优化方法研究

发布时间：2018-08-02 10:47

【摘要】：软件产品线工程是一种在公共核心资源的基础上,按照规定方式开发软件产品家族的方法。该方法具有提高产品质量、降低开发成本以及缩短上市时间等优势。在软件产品线工程中,特征模型是描述产品家族中共性和可变性特征以及特征之间关系的需求模型,它代表了特定领域中所有可能的产品配置空间,是实现基于领域目标和利益相关者的需求进行软件产品定制开发的基础。基于特征模型的配置过程面临以下挑战:1)如何实现对大型软件产品线的特征模型配置进行有效分析验证,给出系统行为属性的满足程度,如系统响应等功能需求;2)如何对特征模型的配置过程进行优化,使其满足特定的目标函数,如最小化成本等。针对特征模型配置分析验证问题,本文以软件产品线模型检测理论为基础,提出了一种基于特征切片的软件产品线多值模型检测方法。该方法首先根据用户需求定义切片准则,利用特征模型的语义抽取出与切片准则相关的特征依赖集和特征排斥集;然后,将切片结果映射到表示软件产品线行为模型的特征迁移系统上,引入三值逻辑对其迁移关系进行抽象;最后,使用多值模型检测工具对抽象的特征迁移系统进行验证,给出系统行为属性的满足程度。实验结果表明该方法的有效性。针对特征模型配置过程优化问题,本文在现有优化方法以及原子集概念的基础上,提出一种遗传算法优化特征模型配置过程的方法。该方法首先使用原子集算法优化特征模型,并对模型的整体约束信息进行建模,作为遗传算法中划分有效配置和无效配置的评估函数。然后,通过将有效配置和无效配置进行交叉,自适应选择以及变异等遗传操作,选出最优的有效配置进行输出。实验结果表明,在特征模型的整体约束函数以及交叉和变异等遗传算子的相互配合下,种群能够加速向最优解收敛。最后,基于本文提出的特征切片方法和配置过程优化方法,设计与实现了自动化配置工具。该工具具有基于基数的特征建模、特征模型的图形化展示、特征切片以及基于特征的配置优化等功能,建立了面向用户需求的软件产品线特征模型配置验证与优化过程。
[Abstract]:Software product line engineering is a method of developing software product family based on common core resources. The method has the advantages of improving the product quality, reducing the development cost and shortening the time to market. In the software product line engineering, feature model is the requirement model to describe the commonness and variability of product family and the relationship between features. It represents all possible product configuration spaces in a particular domain. It is the foundation of customized software development based on domain goals and stakeholder requirements. The configuration process based on the feature model faces the following challenge: 1) how to effectively analyze and verify the feature model configuration of the large-scale software product line, and give the degree of satisfaction of the behavior attributes of the system. How to optimize the configuration process of the feature model to satisfy the specific objective function, such as minimizing the cost, etc. Aiming at the problem of feature model configuration analysis and verification, this paper presents a method of software product line multi-valued model detection based on feature slicing based on the theory of software product line model detection. Firstly, the feature dependency set and feature exclusion set related to the slice criterion are extracted by using the semantics of the feature model according to the slice criterion defined by the user, and then, The slice results are mapped to the feature migration system which represents the behavior model of the software product line, and the ternary logic is introduced to abstract the migration relationship. Finally, the abstract feature migration system is verified by the multi-valued model detection tool. The degree of satisfaction of the behavior attributes of the system is given. The experimental results show that the method is effective. Based on the existing optimization methods and the concept of atomic set, a genetic algorithm (GA) method is proposed to optimize the process of feature model configuration. Firstly, the atomic set algorithm is used to optimize the feature model, and the global constraint information of the model is modeled as an evaluation function in genetic algorithm to divide the effective and invalid configurations. Then, the optimal efficient configuration is selected for output by crossover, adaptive selection, mutation and other genetic operations. The experimental results show that the population converges to the optimal solution with the cooperation of the global constraint function of the characteristic model and the genetic operators such as crossover and mutation. Finally, an automatic configuration tool is designed and implemented based on the feature slicing method and the configuration process optimization method proposed in this paper. The tool has the functions of cardinal-based feature modeling, graphical display of feature model, feature slicing and feature-based configuration optimization. The process of software product line feature model configuration verification and optimization is established.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.5

【相似文献】