自适应软件设计模式组合理论研究

The development of self-adaptive software is facing the challenges of high development complex, low reusability of existing mature solution and high relaiablity on expert knowledge. With growing demands for more systmatic software development, this project proposes a new development model based on design pattern composition. In this model, design patterns are treated as basic modeling units for the software design. The selected design patterns are composed togheter with the Model Driven Architecture(MDA). This model can effectively increase the granularity of design patterns and increase the accuracy of design pattern compostion, and reduces the uncertainty existed in the desgin pattern compositon process. In order to support this model, this project is going to study the pattern description language to formized design pattern into models, pattern composition language to describe user inputs and the engine for autonmatic pattern composition. Compared to the existing pattern-based development process, our model can effectively reduce the complexity introduced from the high dependence of expert knowledge in manual pattern composition. Based on the UML Profile, a new design pattern description language will be designed to provide role-based pattern element description. This language also support the description of pattern interface for compostion. With the introduction of pattern composition language, this project allows users to precicesly control the pattern composition porcess. Those user inputs then is converted to a fomral user-input evolution model to guild model transformeration process. Pattern composition strategies and rules will be studied to provide effective and correct model composition. A model transformation engine will be designed and implemented with QVT transformation language and will be integrate with existing UML design tools. A prototype is scheduled to be developed with the gained theoretical results. The characteristics of this prototype will be analyzed and evaluated from both qualitative and quantitative perspectives by applying it in several self-adaptive software system development process.

针对当前自适应软件的开发复杂度高，已有的设计模式复用困难，严重依赖专家知识等问题，提出一种新的基于设计模式组合的自适应软件开发模式，将设计模式作为基本的建模单元进行软件设计，并利用模型驱动架构将选择的设计模式进行准确组合, 以提升模式复用的粒度和设计模式的组合精度，降低设计模式组合使用中的随意性和不确定性。本项目将设计基于UML扩展的设计模式描述语言，以支持基于角色设计模式的模型元素描述并提供演化和组合支持；研究模式组合语言，提供用户对于设计模式组合的描述能力，并将用户的需求映射到规范的演化模型；研究设计模式组合的策略并制定规范、严格的组合规则，以检测和解决设计模式组合过程中出现的约束违反问题；开发基于QVT模型转换语言的模型转换引擎，实现模型的自动组合并与选定的UML设计工具无缝兼容。系统开发原型的有效性将通过几个具体开发实例进行验证。

本课题主要是针对已有的设计模式复用困难，严重依赖专家知识等问题，提出一种基于设计模式组合的自适应软件开发模式。首次将运行时软件架构的研究与机器人软件的开发相结合，提出了一种支持在线重配置的机器人架构模型开发方式。在自适应软件设计中，为了满足开发过程中用户的增量性需求，针对适应逻辑的组合难题，提出了把部分行为模型的形式化方法引入到适应行为的描述中，研究了适应行为模型的形式化描述方法以及一致性模型融合方法、模型融合可视化。最后，课题组扩展了本研究中的构建适应行为的本体模型来实现适应行为的在线融合的思路到软件自适应测试领域，并取得了一系列的研究成果。