Efficient Disaster-Resilient Network Slicing for URLLC in UAV-Enabled MEC Using Federated Learning

Upon natural or industrial disasters, the reliability and continuity of communications becomes an absolute priority to ensure the availability of critical services. In this context, network slicing emerges as a promising solution for isolating quality of service needs. In addition, URLLC (Ultra-Reliable Low Latency Communication) services are essential to enable effective coordination for real-time emergency management and remote equipment control. The integration of UAVs (Unmanned Aerial Vehicles) as flying MECs (Multi-access Edge Computing) to provide the essential resources for network slicing in disaster scenarios has a substantial effect on the scope of emergency operations management. UAVs involvement facilitates extending connectivity to remote areas, where mobile infrastructure is unavailable. However, due to inefficient node selection for resource allocation, the number of satisfied slicing requests may be reduced, which could put lives at risk in disaster-affected areas. The use of UAVs to ensure network slicing and energy transmission jointly poses a significant challenge given that the UAVs' energy is limited. In response to these challenges, we propose a new multi-objective optimization model enabling URLLC services. We also propose a proactive model that predicts the required workload in the network in terms of computing and storage capacity using federated learning. Our optimization model demonstrates a significant improvement compared to existing approaches, with a remarkable reduction in latency of 50%.