Automated abstraction in Model-Driven Engineering

OData support
Dr. Horváth Ákos
Department of Measurement and Information Systems

Modeling is considered as one of the most elementary disciplines in engineering. Its purpose is to overcome complexity, increase understandability or ease design. In software development, the model-driven engineering paradigm follows this concept by envisioning a process starting form high-level system models and through several well-defined abstraction and refinement steps to automatically derive source code, documentation or configuration artifacts.

In Model-driven Engineering (MDE), the concept of providing view-models (e.g., reliability model, hardware architecture etc.) automatically derived from detailed system models to highlight relevant, domain specific information is a widely used and accepted approach. This view creation approach can be also called model abstraction. Synchronization and maintenance between these system (source) models and their different view-models are usually done by batch transformations. Unfortunately, as these transformations lack of incremental execution support, they have to be rebuild the various view-models whenever a change happens in their corresponding system model.

The well-known derived features give us the ability to define special features automatically computed from different parts of the model, where their value is kept as real object and updated incrementally based on the changes in the model.

In my thesis, I specify a framework, that can incrementally synchronize abstract view models defined by derivation rules using a set of queries. These view models consist of derived objects as an extension of derived features automatically materialized from the source model. These objects are tightly following the lifecycle of the source model and resulting in full featured models that can be inputs for any other model transformation. Then I successfully integrated the developed framework to the Viewers graph visualizer layer of the EMF-IncQuery incremental graph pattern matching framework.

The framework is built upon the Eclipse Modeling Framework (EMF) considered as the de facto industry standard for modeling, and the declarative pattern definition capabilities of EMF-IncQuery is used to define model queries. Finally, the usability of the approachis evaluated on an ongoing avionics research project.


Please sign in to download the files of this thesis.