SHF: Small: Scenario-Based Validation of Design Models

Colorado State University, Fort Collins CO

Investigators

Abstract

Developing complex software systems that play critical roles in organizations and societies is currently labour intensive and costly. There is a need for research on the means to significantly reduce the cost and effort of developing complex software systems. Model driven engineering (MDE) research is primarily concerned with reducing software development costs through the use of technologies that support rigorous analysis of software models and automated transformation of verified models to dependable implementations. Complexity is tackled through (1) the use of models that describe complex systems at multiple levels of abstraction and from a variety of perspectives, and (2) automated support for transforming and analyzing models. Software developers use multi-modeling notations such as the Unified Modeling Language (UML) to manage design complexity. A challenge in using multi-models is ensuring that structural and behavioral properties are consistently described across the different models. The highly iterative nature of software design makes manual consistency checking of multi-models tedious and error prone. Furthermore, iterative, incremental development of models requires support for analyzing incomplete models. While heavyweight formal analysis techniques are useful for analyzing detailed, complete design models of highly-critical systems, lightweight analysis techniques that allow developers to do ``just enough'' analysis with available information are needed to support iterative, incremental software development. The primary aim of this project is to develop a lightweight consistency analysis method that supports iterative, incremental development of software design multi-models. The research aims to produce an analysis method that provides modelers with meaningful feedback on the consistency of UML design multi-models as they evolve in an iterative and incremental process. The scenario-based UML design analysis method that will be developed extends the applicability of existing UML static analysis tools such as USE and OCLE to the analysis of behavior. The approach will allow developers to automatically check a design multi-model against a set of scenarios describing desirable and undesirable behaviors. What will make this analysis method different from other rigorous analysis methods is its tolerance for incompleteness. Specifically, the research aims to produce a method that will provide useful feedback on consistency when only partial descriptions of behavior are provided in multi-models. In addition to the above, the static analysis method will be integrated with a dynamic analysis tool called UMLAnT that allows developers to animate scenarios captured by UML class models. The support for both static and dynamic model analysis and the ability to analyze incomplete models can significantly reduce the cost and effort currently associated with producing dependable implementations from models in iterative, incremental design processes.