Graph-Based Aggregation of Drones–Digital Twins Entanglement Metrics

In recent years Digital Twins (DT) have emerged as a valuable technology for modelling, monitoring and controlling any type of Smart Objects in Real-Time and operating them efficiently in distributed systems, where the coordinations and the interactions between the various Phistcal Twins (PTs) are crucial for the completition of the assigned tasks. In this context, aggregating entanglement metrics from each DT-PT pair is essential to both understand effects on the whole distributed system, the causes of entanglement deterioration, and eventually select the best policies that should be applied to mitigate the global impacts. This thesis studies the problem of entanglement metrics aggregation in the Drone Swarm scenario, for which we implemented both Drone DTs and Swarm DT ad used them to aggregate entanglement data from each DT: in presence of synchronization deterioration that affects one or more DT, entanglement metrics aggregation is needed to both estimate how the overall impact on swarm operations and identify the factors that most influenced the entanglement degradation. In order to achieve this, we represent the drones swarm and its DTs as a graph, where drones, payloads, communication protocols, operational areas, and their corresponding DTs are modeled as nodes. Graph edges encode both the properties of each drone, by linking it to the nodes which represents its attributes and components, and the interactions among DTs during the mission. Relevant properties, including DT-PT entanglement metrics, are associated with both nodes and edges. The resulting graph is dynamic, as it is continuously updated in real-time to reflect changes in the drone swarm and it's DT. Moreover, for analysis purposes, it is possible to extract subgraphs representing the physical layer, the digital layer of the system and their interactions. This representations is then exploited to compute both graph-based and statistical metrics aimed at aggregating entanglement metrics from each PT-DT pair into a global-level assessment. Thanks to the graph structure, the proposed approach also enables the evaluation of entanglement degradation while accounting for the importance of each drone, which is defined by the interactions of its corresponding Digital Twin. In addition, the graph makes it possible to determine whether degradation phenomena are randomly distributed or concentrated in specific regions of the system, thus supporting the identification of possible root causes. We evaluated the proposed metrics through simulated scenarios in which entanglement degradation affects DTs with different levels of importance, as well as cases in which anomalies are associated with a specific operational area or with a server shared by a subset of drones.