Skip to main content Skip to main navigation


Towards Machine Learning-based Digital Twins in Cyber-Physical Systems

Felix Theusch; Lukas Seemann; Achim Guldner; Stefan Naumann; Ralph Bergmann
In: Gianfranco Lombardo; Marco Picone; Diego Reforgiato Recupero; Giuseppe Vizzari (Hrsg.). Proceedings of The First Workshop on AI for Digital Twins and Cyber-Physical Applications (AI4DT&CP). IJCAI Workshop on AI for Digital Twins and Cyber-Physical Applications (AI4DT&CP-2023), located at IJCAI International Joint Conference on Artificial Intelligence 2023, August 19, Macao, Macao, CEUR, 2023.


The use of Artificial Intelligence, and especially Machine Learning methods, promise to play key roles in the development of Digital Twins due to their outstanding properties in processing large IoT data streams. However, so far, there is a lack of research on the systematisation of Machine Learning-based Digital Twins (MLDTs) as well as on their methodological development and implementation processes in productive environments. The scientific literature describes various applications of MLDTs - even if they are not called this way - and specialised methods and architectures, but a generic reference model is still missing. Therefore, this paper proposes a systematisation of the characteristics of MLDTs and their specific challenges. Furthermore, a first proposal of a process model for the systematic development of MLTDs according to the Machine Learning Operations (MLOps) paradigm is presented as a tentative instance of a future reference model for MLDTs. We incorporate established software development methods as well as insights gained from the examination of several industrial applications in the field of water resource management, one of which we present during the paper. We expect that the process model allows practitioners to consistently develop and maintain MLDTs and researchers to find potentials and research gaps.