Learnlib tutorial: An open-source java library for active automata learning

Abstract

Active automata learning is a promising technique to generate formal behavioral models of systems by experimentation. The practical applicability of active learning, however, is often hampered by the impossibility of realizing so-called equivalence queries, which are vital for ensuring progress during learning and finally resulting in correct models. This paper discusses the proposed approach of using monitoring as a means of generating counterexamples, explains in detail why virtually all existing learning algorithms are not suited for this approach, and gives an intuitive account of TTT, an algorithm designed to cope with counterexamples of extreme length. The essential steps and the impact of TTT are illustrated via experimentation with LearnLib, a free, open source Java library for active automata learning.