A Specification Language for Consistent Model Generation based on Partial Models.

Automated graph generation has become a key component in many testing and benchmarking scenarios. For example, modeling tool qualification can be effectively supported by the direct synthesis of well-formed graph models as test inputs, systematic testing of cyber-physical systems requires different test environment models, and different optimization and design-space exploration approaches require the best models with respect to an objective function. In this paper, we propose a novel specification language for partial models used in consistent graph model generation. The language includes constructs to uniformly capture initial, intermediate and final results of the generation by combining partial models, graph predicates and model metrics with mutual dependencies between them. The formal semantics of the language is defined by using 4-valued Belnap-Dunn logic that explicitly marks inconsistent model elements as part of the partial model. The use of our language is illustrated in the context of a complex case study defined by NASA researchers.