Model Checking At Scale: Automated Air Traffic Control Design Space Exploration

Many possible solutions, differing in the assumptions and implementations of the components in use, are usually in competition during early design stages. Deciding which solution to adopt requires considering several trade-offs. Model checking represents a possible way of comparing such designs, however, when the number of designs is large, building and validating so many models may be intractable.During our collaboration with NASA, we faced the challenge of considering a design space with more than 20,000 designs for the NextGen air traffic control system. To deal with this problem, we introduce a compositional, modular, parameterized approach combining model checking with contract-based design to automatically generate large numbers of models from a possible set of components and their implementations. Our approach is fully automated, enabling the generation and validation of all target designs. The 1,620 designs that were most relevant to NASA were analyzed exhaustively. To deal with the massive amount of data generated, we apply novel data-analysis techniques that enable a rich comparison of the designs, including safety aspects. Our results were validated by NASA system designers, and helped to identify novel as well as known problematic configurations.