Requirements traceability is broadly recognized as a critical element of any rigorous software development process, especially for building safety-critical software (SCS) systems. Model-driven development (MDD) is increasingly used to develop SCS in many domains, such as automotive and aerospace. MDD provides new opportunities for establishing traceability links through modeling and model transformations. Architecture Analysis and Design Language (AADL) is a standardized architecture description language for embedded systems, which is widely used in avionics and aerospace industries to model safety-critical applications. However, there is a big challenge to automatically establish the traceability links between requirements and AADL models in the context of MDD, because requirements are mostly written as free natural language texts, which are often ambiguous and difficult to be processed automatically. To bridge the gap between natural language requirements (NLRs) and AADL models, we propose an approach to generate the traceability links between NLRs and AADL models. First, we propose a requirement modeling method based on the restricted natural language, which is named as RM-RNL. The RM-RNL can eliminate the ambiguity of NLRs and barely change engineers’ habits of requirement specification. Second, we present a method to automatically generate the initial AADL models from the RM-RNLs and to automatically establish traceability links between the elements of the RM-RNL and the generated AADL models. Third, we refine the initial AADL models through patterns to achieve the change of requirements and traceability links. Finally, we demonstrate the effectiveness of our approach with industrial case studies and evaluation experiments.

中文翻译：

---

一种在受限自然语言需求和 AADL 模型之间生成可追溯性的方法

需求可追溯性被广泛认为是任何严格软件开发过程的关键要素，尤其是对于构建安全关键软件 (SCS) 系统而言。模型驱动开发 (MDD) 越来越多地用于开发许多领域的 SCS，例如汽车和航空航天。MDD 为通过建模和模型转换建立可追溯性链接提供了新的机会。架构分析和设计语言 (AADL) 是嵌入式系统的标准化架构描述语言，广泛用于航空电子和航空航天行业，以对安全关键应用程序进行建模。然而，在 MDD 上下文中自动建立需求和 AADL 模型之间的可追溯性链接是一个很大的挑战，因为需求大多写成自由的自然语言文本，这通常是模棱两可的，难以自动处理。为了弥合自然语言需求 (NLR) 和 AADL 模型之间的差距，我们提出了一种在 NLR 和 AADL 模型之间生成可追溯性链接的方法。首先，我们提出了一种基于受限自然语言的需求建模方法，称为RM-RNL。RM-RNL 可以消除 NLR 的歧义，几乎不会改变工程师对需求规范的习惯。其次，我们提出了一种从 RM-RNL 自动生成初始 AADL 模型并在 RM-RNL 的元素和生成的 AADL 模型之间自动建立可追溯性链接的方法。第三，我们通过模式细化最初的 AADL 模型，以实现需求和可追溯性链接的变化。最后，