SysML, SoaML, AADL, MARTE and others are flexible general purpose modeling approaches for systems. They favor freedom of choice. While they often provide different modeling views, these views are not connected such that overall system consistency can be ensured throughout all potential development phases. This hinders separation of roles that is required for successful system composition and therefore is in contrast with the overall needs for modeling in RobMoSys.
The focus of RobMoSys is on composability and consistency of the different views such that the different roles contribute in a consistent and composable way to the system under specification and development. This requires more elaborate structures to connect the different views in a consistent way. This can be achieved via superordinated meta-model structures and via model-to-model transformations.
Of course, the structures of RobMoSys will be inspired by, for example, the above approaches wherever appropriate. The RobMoSys structures might enable linking the different modeling views of the mentioned modeling approaches.
For example, AADL requires more abstract, yet consistent, modeling views on top, while other approaches such as SysML might be subprofiled, thus providing more detailed, yet again consistent, robotic-specific views underneath. Many of the (meta-model) structures and abstractions in RobMoSys focus on transformations (and exchange of knowledge) between well known and widely accepted modeling views.
Within the context of UML the term “semantic variation point” has been coined to express the purposeful semantic ambiguity for certain UML elements. Because UML is a general purpose modeling language, this semantic ambiguity makes sense and can be narrowed within the derived domain-specific models using e.g. the UML profile mechanism. Moreover, even the domain-specific models can still expose some semantic variability that is closed within concrete realizations (e.g. through code generation or reference implementations). In this sense, RobMoSys as well offers different levels of abstraction for modeling where the higher levels (such as e.g. the block-port-connector) are more general purpose (leaving open some semantic variability) and lower (i.e. domain-specific) abstraction levels (such as e.g. the RobMoSys composition structures) that narrow this semantic variability.