|Full text PDF:||http://elib.suub.uni-bremen.de/edocs/00104250-1.pdf|
Model-driven development (MDD) has been identified as a promising approach for developing software. By using abstract models of a system and by generating parts of the system out of these models, one tries to improve the efficiency of the overall development process and the quality of the resulting software. In the context of MDD the Unified Modeling Language (UML) and its related textual Object Constraint Language (OCL) have gained a high recognition. To be able to generate systems of high quality and to allow for interoperability between modeling tools, a well-defined semantics for these languages is required. This thesis summarizes published work in this context that employs an endogenous metamodeling approach to define the semantics of newer elements of the UML. While the covered elements are exhaustively used to define relations between elements of the metamodel of the UML, the UML specification leaves out a precise definition of their semantics. Our proposed approach uses models, not only to define the abstract syntax, but also to define the semantics of UML. By using UML and OCL for this, existing modeling tools can be used to validate the definition. The second part of this thesis covers work on the usage of UML and OCL models for runtime verification. It is shown how models can still be used at the end of a software development process, i. e., after an implementation has manually been added to generated parts, even though they are not used as central parts of the development process. This work also influenced the integration of protocol state machines into a modeling tool, which lead to publications about the runtime semantics of state machines and the capabilities to declaratively specify behavior using state machines.