|Title||Uncertainty-wise Cyber-Physical Systems Testing|
|Project(s)||U-Test: Testing Cyber-Physical Systems under Uncertainty: Systematic, Extensible, and Configurable Model-based and Search-based Testing Methodologies, MBT4CPS: Model-Based Testing For Cyber-Physical Systems|
|Publication Type||PhD Thesis|
|Year of Publication||2018|
|Degree awarding institution||The University of Oslo|
|Number of Pages||292|
|Publisher||The University of Oslo|
|Keywords||Cyber-Physical System, Model-based Testing, Uncertainty|
A Cyber-Physical Systems (CPS), as an integration of computing, communication, and control for making intelligent and autonomous systems, has been widely applied in various safety-critical domains, e.g., avionics and automotive. However, uncertainty is inherent in CPSs due to various reasons such as unpredictable environment under which the CPSs are operated. And, uncertainties may cause irreparable accidents once they cannot be handled properly by CPSs. Therefore, it is crucial to identify uncertainties in CPSs and test CPSs under the uncertainties, to ensure that CPSs are capable of handling the uncertainties during their actual operations, i.e., making CPSs less uncertain.
Towards this direction, five contributions were made in the thesis corresponding to five papers respectively: (C1) a conceptual model, named as U-Model, for helping develop a systematic and comprehensive understanding of uncertainty in CPSs; (C2) an use case modeling methodology, named as U-RUCM, for identifying, qualifying, and, where possible, quantifying uncertainty in requirements engineering; (C3) a test modeling methodology, named as UncerTum, for supporting the construction of test ready models with the explicit representation of uncertainties in CPSs; (C4) an evolution framework, named as UncerTolve, for interactively evolving test ready models specified with UncerTum based on real operational data; and (C5) a testing framework, named as UncerTest, for testing CPSs in the presence of uncertainties in their operating environments in a cost-effective manner using model-based and search-based testing techniques.
Based on our evaluations of the five contributions with the industrial CPS case studies, we observed that U-Model, as the foundation for this research, is sufficiently complete for characterizing and classifying uncertainties in CPSs. Then, the U-Model based modeling methodologies U-RUCM and UncerTum offer solutions to enable the identification and specification of uncertainties at two critical phases of a system development lifecycle: requirements engineering and testing. Furthermore, UncerTolve can successfully evolve model elements of the test ready models specified with UncerTum and calculate objective uncertainty measurements based on real operational data. Last, UncerTest managed to cost-effectively test CPSs in the presence of uncertainties and proactively identify unknown uncertainties by introducing the sources of the uncertainties into the test environments during test case execution.