异构时间与异构行为融合的需求建模和需求确认方法研究

Methods and technologies in Requirements Engineering (RE) and Model Driven Engineering (MDE) have become essential factors in successful software development. With the emergence of Cyber Physical Systems (CPS), software development demands higher ability to deal with the complexity, the close interactions between software and physical environment, the integration of heterogenous time and behaviors, and diverse needs from various stakeholders. MDE uses models as abstractions of complex problems and reduces time-to-market by enabling automatic model transformations. Good quality of early models is essential for a successful application. The guarantee of that quality is assured through requirements validation and verification. Works in RE have made great contributions in this area. However, RE and MDE have evolved separately and in distinct communities. To leverage the effectiveness of requirements validation and verification efficiently, we aim to bridge the gap between them by requirements modeling, validation and verification through light-weighted formal specifications. Based on a series of concrete problems identified during previous practice, we propose to construct a coordination specification for requirements model which focus on the coordination with physical environment, the integration of heterogenous modules, and the integration between time and behaviors. We propose an intelligent and effective method of performing requirements validation and verification on the specification based on the combination of machine learning techniques and recommender systems, including the following steps: generate questions which represent design intensions and validation requests; construct customized self-adapted questionnaires for different stakeholders; collect stakeholders’ answers and analyze their effects, and generate the agreed requirements specification after the reconfirmation with stakeholders; perform model checking on the specification for its correctness and schedulability; integrate the conformed requirements specification into models, execute the model with its operational semantics and constrain later modifications.

信息物理融合系统的涌现对已成功应用的需求工程和模型驱动工程提出了更高的要求：软件更复杂、与环境交互更紧密、时间与行为的融合更高、涉及到更多样的利益相关者。模型驱动可以处理复杂性、提高开发效率。其中，初始模型满足利益相关者的期望是应用成功的前提。然而，目前尚无成熟的方法支持需求工程与模型驱动工程的相互应用。急需研究以上要求下的需求建模、需求确认与初始模型的衔接。本课题拟构建支持异构时间和异构行为融合的需求规约，研究该规约与需求模型和初始模型的关系，并提供相应需求确认方法来完成以上衔接。具体包括：定义融合异构时间和异构行为的需求规约语言及建模方法；根据规约的语法和语义、基于机器学习和推荐系统技术生成反映建模者意图的需求确认问题，构建面向利益相关者的个性化的可动态配置的问卷，进行需求确认；基于模型检查技术量化分析需求确认结果生成改进后的需求规约，将该规约融入模型提供行为语义。

针对信息物理融合系统的特点，软件更复杂、与环境的交互更紧密、异构时间与异构行为的融合要求更高、涉及到更多样的利益相关者，在已有的需求工程和模型系统工程工作基础上：定义了融合异构时间和异构行为的需求规约语言及建模方法；组织利益相关者关心的行为需求，表达为时间事件因果关系，基于规约的语法和语义，利用模型检查技术生成需求确认问题；基于深度学习和推荐系统技术生成反映建模者意图的需求确认问题，构建面向利益相关者的个性化的可动态配置的问卷，进行需求确认；基于模型检查技术进行模型验证，完成规约的安全性、可调度性、最差执行时间等定性和定量的分析。以上的模型的构建、规约的生成、模型验证以及需求确认都支持迭代开发，上一阶段的结果可用于下一阶段模型的精化。通过需求的验证、确认和反馈，完成了需求初始模型与后续基于模型的设计和开发的衔接。