Regular Model Checking Using Inference of Regular Languages

Regular model checking is a method for verifying infinite-state systemsbased on coding their configurations as words over a finite alphabet,sets of configurations as finite automata, and transitions as finitetransducers. We introduce a new general approach to regular modelchecking based on inference of regular languages. The method buildsupon the observation that for infinite-state systems whose behaviourcan be modelled using length-preserving transducers,  there is afinite computation for obtaining all reachable  configurations upto a certain length n.
These configurations are a (positive) sample of the reachableconfigurations of the given system, whereas~all other words up tolength n are a negative sample. Then, methods of inference of regularlanguages can be used to generalise the sample to the full reachabilityset (or an overapproximation of it). We have implemented our method ina prototype tool which shows that our approach is competitive on anumber of concrete examples. Furthermore, in contrast to all otherexisting regular model checking methods, termination is guaranteed ingeneral for all systems with regular sets of reachable configurations.The method can be applied in a similar way to dealing with reachabilityrelations instead of reachability sets too.