Abstraction and Counterexample-Guided Construction of ω-Automata for Model Checking of Step-Discrete Linear Hybrid Models

Abstract

For the verification of reactive hybrid systems existing approaches do not scale well w.r.t. large discrete state spaces, since their excellence mostly applies to data computations. However, especially control dominated models of industrial relevance in which computations on continuous data are comprised only of subsidiary parts of the behavior, these large discrete state spaces are not uncommon. By exploiting typical characteristics of such models, the herein presented approach addresses step-discrete linear hybrid models with large discrete state spaces by introducing an iterative abstraction refinement approach based on learning reasons of spurious counterexamples in an ω-automaton. Due to the resulting exclusion of comprehensive classes of spurious counterexamples, the algorithm exhibits relatively few iterations to prove or disprove safety properties. The implemented algorithm was successfully applied to parts of industrial models and shows promising results.