Federated Digital Twins for Space Systems

A federated digital twin architecture provides a path for managing complexity and computational costs, while also addressing privacy and security constraints. In contrast to a monolithic architecture, which represents a complex system via a single tightly integrated digital twin, a federated architecture comprises multiple subsystem-level digital twins that are coupled through well-defined communication interfaces. These digital twins exchange information with each other while maintaining their own models and data, potentially with built-in privacy and security constraints. This paper proposes a method based on generalized Sobol’ sensitivity analysis to characterize the coupling strength between subsystem-level digital twins, which is a first step towards designing a federated digital twin architecture. Leveraging uncertainty quantification, the proposed approach identifies approximate communication strategies that improve computational efficiency while controlling the error in the federated digital twins. We illustrate the potential of this approach for representing a satellite through a federated digital twin architecture suitable for delivering improved operational health monitoring and control capabilities, which in turn lead to improved spacecraft performance, robustness, and reliability.