Federated Learning for Enablement of Digital Twin

Creation, maintenance, and update of digital twins are costly and time-consuming mechanisms. The required effort can be optimized with the use of LiDAR technologies, which support the process of collecting data related to spatial information such as location, geometry, and position. Sharing such data in multi-stakeholder environments is hindered due to competition, confidentiality, and security requirements. Multi-stakeholder environments favor the use of decentralized creation and update mechanisms with reduced data exchange. Such mechanisms are facilitated by Federated Learning, where the learning process is performed at the data owner's location. Two case studies are presented in this paper, where LiDAR is used to extract information from industrial equipment as a part of the creation of a digital twin.