The team of Bernhard Steffen at Dortmund and the team of Bengt Jonsson at Uppsala are collaborating to adapt and extend automata learning techniques (also called regular inference) for automatically deriving behavioral models of components from legacy code or observations of system behavior. Potential applications are to derive (timed or untimed) models of environments of component-based system for modelling and analysis. As a basis for experimentation, Dortmund has developed LearnLib [BRS06], a library for automata learning, which a flexible modular structure that can be configured to exploit specific properties of applications, in order to make automata learning scalable to realistic settings. This collaboration has been ongoing during previous year of ARTIST2. During Y3 of ARTIST2, the work has concentrated on extending regular inference techniques to simple classes of infinite-state models. The motivation is to handle parameters that, e.g., are identifiers of connections, objects, etc. During Y4, the intention is to apply these techniques to a larger industrial protocol module. In autumn 2006, Therese Berg from Uppsala visited Dortmund, and during spring 2007, Harald Raffelt from Dortmund visited Uppsala.

(c) Artist Consortium, All Rights Reserved - 2006, 2007, 2008, 2009

Réalisation Axome - Création de sites Internet