Abstract We develop a research agenda for the field of automata learning. Automata learning algorithms infer state-machines from observations. The study of such algorithms began in the 1970s and until today has led to a wide range of different learning models, learnability results, and learning algorithms for many different classes of automata as well as to many different applications of automata learning, e.g., specification generation, learning-based testing, and black-box verification. As the field still stratifies and learning algorithms and new applications are conceived, it will be helpful to consolidate and integrate individual obtained results into a coherent set of principles of automata learning and techniques for devising learning algorithms. We aim to provide a step in this direction by conducting a survey of active automata learning methods, focusing on different application scenarios (application domains, environments, and desirable guarantees) and the overarching challenges that emerge from these. We identify concrete research questions through a (short) bibliographic study highlighting the state of the art and the technical implications that are derived from the overarching challenges.
Fortz et al. (2026) studied this question.