Section: New Results

Composing Simulink and AADL

Participants : An Phung-Khac, Jean-Pierre Talpin, Benoit Combemale, Jean-Marc Jezequel.

The goal of this work is to improve an import function of the Polychrony environment proposed by the team. Particularly, Polychrony comprises a co-modeling tool supporting the import a high-level Simulink (functional) and AADL (architectural) specifications [21] . This import function is currently implemented by two different transformations, namely Simulink-to-Signal, and AADL-to-Signal. To integrate the Signal programs resulting from these transformations, some Signal interfaces are manually implemented. The composition of Simulink and AADL models thus depends on system designers who implement the interfaces, making difficult its maintenance and validation. To deal with this issue, the model composition approach proposed by the Triskell team, namely ModMap [37] , could be used to build a new Simulink and AADL model composition framework.

In ModMap, model composition is considered as a pair of a mapping and an interpretation. A mapping aligns concepts of two meta-models, while the interpretation describes the composition goal. As a model mapping framework, ModMap provides an extensible modeling language supporting the definition of generic mappings and the definition of interpretations. Together with this language, the ModMap kernel is also implemented as an extensible set of mapping processing functions. Model composition frameworks are then built by extending the language and the kernel according to specific composition purposes.

As mentioned above, we intend to apply the ModMap approach to the development of the Simulink and AADL model composition framework. To this end, we need to extend the ModMap mapping language to obtain an other one that allows system designers to align elements between Simulink and AADL models regarding the purpose of co-simulation in Signal. Then, a transformation, namely ModMap-to-Signal, needs to be implemented by extending the ModMap kernel. This transformation uses mappings provided by system designers as inputs to generate Signal interfaces. The three transformations (i.e., Simulink-to-Signal, AADL-to-Signal, and ModMap-to-Signal) form the new model composition framework. Comprared to the previous one, this framework will more automated. On the other hand, existing transformations will also be reused.