Towards Engineering Digital Twins by Active Behaviour Mining

In the context of Confirm, the Irish Research Centre on Smart Manufacturing, field demonstrators are used to show new techniques to industrial partners, various kinds of students, and the general public alike. Considering the robotics demonstrator for the Digital Thread concept used in Confirm, which is a small cyberphysical system based on the UR3 cobot and a web controller for it, we apply Active Automata Learning in order to obtain a Digital Twin for it. Behavior mining done in this fashion is nowadays uncommon, but it has various advantages over, e.g., models obtained with popular AI techniques in that the AAL models are accurate deterministic behavioural explanations for the system behaviour at the chosen level of abstraction, and they may be further amenable to formal verification, e.g., by model checking, in order to establish properties of interest. This extension has the effect of showcasing the Digital Twin concept, the AAL technique, the use of model checking, and the importance of working with formal models that are amenable to these technologies. We then reflect on the nature of the models and their uses and meaning, from the point of view of the comments and questions we receive in the demonstrations. We also consider the use of a feature-based approach to modelling the systems and their interactions, which is a further aspect for which the demonstrator could be used, with a special attention to the aspects of this work, like AAL and the feature based and feature interaction research, that connect directly with the collaboration with and the research of Bengt Jonsson.