A relationship-based approach to model integration

A key problem in model-based development is integrating a collection of models into a single, larger, specification as a way to construct a functional system, to develop a unified understanding, or to enable automated reasoning about properties of the resulting system. In this article, we suggest that the choice of a particular model integration operator depends on the inter-model relationships that hold between individual models. Based on this observation, we distinguish three key integration operators studied in the literature—merge, composition and weaving—and describe each operator along with the notion of relationship that underlies it. We then focus on the merge activity and provide a detailed look at the factors that one must consider in defining a merge operator, particularly the way in which the relationships should be captured during merge. We illustrate these factors using two merge operators that we have developed in our earlier work for combining models that originate from distributed teams.