Proved Metamodels as Backbone for Software Adaptation

In this paper we demonstrate the error-prone status of the UML 2.3 metamodel relating to state machines. We consequently provide a corrected version based on formal proofs written and processed with the help of the Coq system prover. The purpose of the proposed research is to support dynamical adaptation by means of models at runtime. Software components are internally endowed with complex state machines (models) realizing their behavior. Adaptation amounts to dynamically changing the state machines' structure (for instance, adding a new state). This occurs via SimUML, a state machine execution engine that is constructed on the top of a metamodel resulting from correctness proofs.