Specifying Uncertainty in Use Case Models.
Man Zhang,Tao Yue,Shaukat Ali,Bran Selic,Oscar Okariz,Roland Norgre,Karmele Intxausti
DOI: https://doi.org/10.1016/j.jss.2018.06.075
IF: 3.5
2018-01-01
Journal of Systems and Software
Abstract:Context Latent uncertainty in the context of software-intensive systems (e.g., Cyber-Physical Systems (CPSs)) demands explicit attention right from the start of development. Use case modeling-a commonly used method for specifying requirements in practice, should also be extended for explicitly specifying uncertainty. Objective: Since uncertainty is a common phenomenon in requirements engineering, it is best to address it explicitly by identifying, qualifying, and, where possible, quantifying uncertainty at the beginning stage. The ultimate aim, though not within the scope of this paper, was to use these use cases as the starting point to create test-ready models to support automated testing of CPSs under uncertainty. Method: We extend the Restricted Use Case Modeling (RUCM) methodology and its supporting tool to specify uncertainty as part of system requirements. Such uncertainties include those caused by insufficient domain expertise of stakeholders, disagreements among them, and known uncertainties about assumptions about the environment of the system. The extended RUCM, called U-RUCM, inherits the features of RUCM, such as automated analyses and generation of models, to mention but a few. Consequently, U-RUCM provides all the key benefits offered by RUCM (i.e., reducing ambiguities in requirements), but also, it allows specification of uncertainties with the possibilities of reasoning and refining existing ones and even uncovering unknown ones. Results: We evaluated U-RUCM with two industrial CPS case studies. After refining RUCM models (specifying initial requirements), by applying the U-RUCM methodology, we successfully identified and specified additional 306% and 512% (previously unknown) uncertainty requirements, as compared to the initial requirements specified in RUCM. This showed that, with U-RUCM, we were able to get a significantly better and more precise characterization of uncertainties in requirement engineering. Conclusion: Evaluation results show that U-RUCM is an effective methodology (with tool support) for dealing with uncertainty in requirements engineering. We present our experience, lessons learned, and future challenges, based on the two industrial case studies.