The Need for Semantic Extension of SysML to Model the Problem Space

Requirements in natural language, like shall statements, while commonly used, present inherent limitations in terms of accuracy and precision. Modeling requirements within a model-based systems engineering MBSE) framework shows promise to cope with these issues. Common approaches include either the definition of textual requirements as model objects or the flagging of system models as requirements. The first approach inherits the weaknesses of natural language. We show in this paper that the second approach necessarily leads to a poor set of requirements. We therefore argue that modeling languages, in particular SysML, need to be semantically extended to adequately model the problem space. We demonstrate with a specific example that simply flagging model elements as requirements is not effective to model the problem space. In fact, we show that such an approach produces a deficient definition of the problem space, since it inherently discards solutions that could otherwise be potentially acceptable to solve the problem that is being addressed. In addition, we leverage this example to discuss potential semantic extensions of SysML that could enable adequate modeling of the problem space that fulfills the formal conditions of good requirements.