Inference of safe approximate models for system composition

Abstract

The scope of this thesis is in the field of software engineering and formal methods. A core issue in component-based software engineering is how to rigorously build a system from off-the-shelf software components. This is done by automatic component adaptation and composition, which need behavioural models of all components. We have developed an approximate model inference technique for black box asynchronous software components. In contrast to existingtechniquesthatassumeonlycontrollablebehaviour, ourmethodaddressesthe challenges raised by uncontrollable events, which cannot be enabled or disabled at will. The models inferred are safe with respect to black box controllability, i.e. they can be reliably used to control the real components.