Compositional Learning for Synchronous Parallel Automata

Abstract Automata learning is an approach for extracting a model in the shape of an automaton from a black-box system. This approach has recently gained much attention in both industry and academia. In this paper, we introduce a compositional automata learning algorithm for systems comprising synchronous parallel components. Our algorithm assumes no prior knowledge about the number of components, their individual alphabets, and the synchronizing alphabets. The learning process is automatic and figures out the alphabet symbols on-the-fly during learning the components. We prove that the proposed algorithm terminates and correctly learns the individual components. We use a number of case studies from the industrial automotive domain and synthetic benchmarks to evaluate the performance of the proposed algorithm. The experimental results show that the algorithm requires significantly fewer input symbols and resets to learn the system compositionally.