UAVs and Federated Learning-Enabled Digital Twins of Vehicles in Dynamic Disaster Environments
摘要
A digital twin is a virtual model of a physical system designed to ensure the quality of the physical experience. One can preferably use privacy-preserving federated learning (FL) to model digital twins of vehicular networks. However, extensive communication via the terrestrial network reduces the performance of FL, particularly in high-mobility environments like the Internet of Vehicles (IoV). IoV is a network of connected vehicles that share data among vehicles and other infrastructures, such as Roadside Units (RSUs). To overcome these limitations, unmanned aerial vehicles (UAVs) provide on-demand communication resources, especially in disaster areas. For example, in flood, earthquake, fire, and landslide scenarios, UAVs can quickly restore connectivity and support real-time data processing, enhancing the responsiveness of emergency services. We provide a high-level architecture on how UAVs assist the FL and digital twinning process in a disaster scenario. Moreover, we propose a two-layer architecture for FL-based digital twins of vehicles that strengthens communications. The emerging use cases of UAV-based FL for the digital twin of vehicles are presented along with significant challenges and potential solutions.