A requirements engineering methodology combining models and controlled natural language

The use of models in requirements engineering (RE) for software-intensive embedded systems is considered beneficial. The main advantages of requirements models as documentation format are that they facilitate requirements understanding and foster automatic analysis techniques. However, natural language (NL) is still the dominant documentation format for requirements specifications, particularly in the domain of embedded systems. This is due to the facts that NL-based requirements can be used within legally binding documents and are more appropriate for reviews than models. In order to bridge the gap between both of these documentation formats, this paper proposes a model-driven RE methodology that makes use of requirements models along with a controlled natural language. The methodology combines the advantages of model-based and NL-based documentation by means of a bidirectional multi-step model transformation between both documentation formats. We illustrate the approach by means of an automotive example, explain the particular steps of the model transformation, and present performance results.