面向不确定性需求的智能化软件架构演化优化和自适应技术研究

Software engineering requirements will change at any time and anywhere, and there are a lot of different requirement change types, however it is very difficult to predict when requirements change will happen, and also very difficult to capture the causes why the requirement changes. With the occurrence of new computing paradigms such as mobile computing, cloud computing etc. and they are widely adapted in different application domains, software engineering needs to deal with the more and more complex objects, develop different kinds of software products, and satisfy the requirement of different running environments, which leads directly to the dynamic uncertainty of software engineering requirement becomes an obvious and serious problem, and further leads to the uncontrolled risk in existing project. In this project proposal, we try to find an intelligent architecture by deep and wide study to deal with uncertain requirement efficiently and effectively, and we promote our first task from identifying and the collecting uncertain requirements, and then use modularization techniques to isolate uncertain requirements, and then use feature models to describe uncertain requirements, and then use traceability matrix to record and manage all information and data produced during the procedures and steps for abstracting scenario from the requirement use cases or use case maps, extracting features from scenarios, mapping features to responsibilities, and mapping responsibilities to architectural elements etc., and then use machine learning to obtain all kinds of architectural knowledges for building knowledge base, and then use knowledge base to make decisions for online planning, online refactoring, online evolving and optimizing etc., and then use features and MAPE-K reference model to realize the self-adaptation of intelligent architecture. The expected research results will include patents, papers, books, software tools and other documents, which forms a set of useful methods and techniques which can be used to guide how to reduce or avoid the project risk brought by uncertain requirements in the future develop projects.

软件需求会随时随地发生变更，而且变更的类型多种多样，变更的根源难以预测，即需求具有动态不确定性。而随着移动计算、云计算等新型计算范型的出现，软件工程面临的处理对象更加复杂、软件的形态更加多样、运行的环境更加多变，需求动态不确定性问题显得尤其突出，由此带来的风险更加难以控制。本课题拟从不确定性需求的认知与获取出发，研究如何利用智能化架构来处理不确定性需求的方法：即研究如何利用模块化技术对不确定性需求进行模块化处理、如何利用特征模型对不确定性需求建模、如何利用轨迹矩阵记录和跟踪管理不确定性需求到软件架构的映射过程、如何利用机器学习来获取多源异构软件架构知识、如何利用知识库来实现软件架构在线演化和优化、如何利用特征和MAPE-K来实现架构自适应等。本课题的预期研究成果将形成比较完整的智能化架构演化和优化技术理论和方法体系，为在未来软件系统开发过程中如何降低或避免不确定需求导致的风险提供技术支持。

软件需求会随时随地发生变更，而且变更的类型多种多样，变更的根源难以预测，即需求具有动态不确定性。而随着移动计算、云计算等新型计算范型的出现，软件工程面临的处理对象更加复杂、软件的形态更加多样、运行的环境更加多变，需求动态不确定性问题显得尤其突出，由此带来的风险更加难以控制。本课题拟从不确定性需求的认知与获取出发，研究如何利用智能化架构来处理不确定性需求的方法：即研究如何利用模块化技术对不确定性需求进行模块化处理、如何利用特征模型对不确定性需求建模、如何利用轨迹矩阵记录和跟踪管理不确定性需求到软件架构的映射过程、如何利用机器学习来获取多源异构软件架构知识、如何利用知识库来实现软件架构在线演化和优化、如何利用特征和MAPE-K来实现架构自适应等。通过四年的研究，本项目在软件架构演化优化，基于上下文感知的API推荐、面向模式的自适应重构以及相关技术的应用方面取得了一系列创新的成果，发表CCF推荐的高水平论文19篇（另有4篇论文在等待评审结果），申报中国发明专利28项（其中3项已经获得授权），出版专著1部，登记软件著作权6项，完成智能化软件架构专题报告1份，指导博士论文4篇、硕士论文31篇。研究成果形成比较完整的智能化架构演化和优化技术理论和方法体系，为在未来软件系统开发过程中如何降低或避免不确定需求导致的风险提供技术支持。