RALib: A LearnLib extension for inferring EFSMs

Active learning of register automata infers extended finite state machines (EFSMs) with registers for storing values from a possibly infinite domain, and transition guards that compare data parameters to registers. In this paper, we present RALib, an extension to the LearnLib framework for automata learning. RALib provides an extensible implementation of active learning of register automata, together with modules for output, typed parameters, mixing different tests on data values, and directly inferring models of Java classes. RALib also provides heuristics for finding counterexamples as well as a range of performance optimizations. Compared to other tools for learning EFSMs, we show that RALib is superior with respect to expressivity, features, and performance.