Tutorial: Automata learning in practice

Abstract

The paper reviews active automata learning with a particular focus on sources of redundancy. In particular, it gives an intuitive account of TTT, an algorithm based on three tree structures which concisely capture all the required information. This guarantees minimal memory consumption and it drastically reduces the length of membership queries, in particular in application scenarios like monitoring-based learning, where long counter examples arise. The essential steps and the impact of TTT are illustrated via experimentation with LearnLib, a free, open source Java library for active automata learning.