Abstract. Model checking is increasingly popular for hardware and, more recently, software verification. In this paper we describe two different approaches to extend the benefits of model checking to systems whose behavior is specified by graph transformation systems. One approach is to encode the graphs into the fixed state vectors and the transformation rules into guarded commands that modify these state vectors appropriately to enjoy all the benefits of the years of experience incorporated in existing model checking tools. The other approach is to simulate the graph production rules directly and build the state space directly from the resultant graphs and derivations. This avoids the preprocessing phase, and makes additional abstraction techniques available to handle symmetries and dynamic allocation. In this paper we compare these approaches on the basis of three case studies elaborated in both of them, and we evaluate the results. Our conclusion is that the first approach outperforms the second if the dynamic and/or symmetric nature of the problem under analysis is limited, while the second shows its superiority for inherently dynamic and symmetric problems.

Abstract. In the paper, we present a tool for model checking dynamic consistency properties in arbitrary well-formed instance models of any modeling language defined visually by metamodeling and graph transformation techniques. Our tool first translates such high-level specifications into a tool independent abstract representation of transition systems defined by a corresponding metamodel. From this intermediate representation the input language of the back-end model checker tool (i.e., SPIN in our case) is generated automatically. Keywords: visual modeling languages, metamodeling, graph transformation, model checking, formal verification. Towards a Formal Analysis of Modeling Languages in MDAAs the Model Driven Architecture (MDA) is becoming more and more widespread in the design process of IT systems, there is an increasing need for efficiently developing modeling languages and their model instances within a single modeling framework. For instance, UML itself (from version 2.0) is evolving into a family of modeling languages from a single and monolith language.The definition of such modeling languages is frequently based on a combination of visual metamodeling techniques and well-formedness constraints expressed in the Object Constraint Language (OCL) that allow an object-oriented specification of the static semantics of the language. Since the current MOF standard does not provide an appropriate means to precisely specify the dynamic operational semantics of a language, many approaches (e.g., [1,5]) facilitate the use of graph transformation for that purpose. Graph transformation rules provide a visual and pattern based manipulation of the target user model fitting well to best engineering practices.However, as the use of visual modeling techniques alone does not guarantee the correctness of a design, model checking tools (like SPIN [2]) are frequently used to mechanically analyze the functional correctness of system models based on UML statecharts (see e.g., [3]). Unfortunately, these approaches does not scale up well for ensuring consistency between multiple modeling languages as projecting modeling languages into a common semantic domain for verification purposes (as done, e.g., in [1]) can be difficult even for domain experts since input languages of model checker tools are very low-level. Moreover, a new tool has to be developed to extend the consistency framework for handling a new modeling language.

No abstract