More Scalable LTL Model Checking via Discovering Design-Space Dependencies ( D^3 D 3 )

Abstract

Modern system design often requires comparing several models over a large design space. Different models arise out of a need to weigh different design choices, to check core capabilities of versions with varying features, or to analyze a future version against previous ones. Model checking can compare different models; however, applying model checking off-the-shelf may not scale due to the large size of the design space for today’s complex systems. We exploit relationships between different models of the same (or related) systems to optimize the model-checking search. Our algorithm, \(D^{3}\), preprocesses the design space and checks fewer model-checking instances, e.g., using nuXmv. It automatically prunes the search space by reducing both the number of models to check, and the number of LTL properties that need to be checked for each model in order to provide the complete model-checking verdict for every individual model-property pair. We formalize heuristics that improve the performance of \(D^{3}\). We demonstrate the scalability of \(D^{3}\) by extensive experimental evaluation, e.g., by checking 1,620 real-life models for NASA’s NextGen air traffic control system. Compared to checking each model-property pair individually, \(D^{3}\) is up to 9.4\(\times \) faster.