Intelligent Aerial Relay Deployment for Enhancing Connectivity in Emergency Communications

Guaranteeing stable connectivity for emergency communications in rural and remote areas (e.g., to transfer high-quality video for remote surgery) has been challenging for any cellular network generation. With many initiatives proposed to expand the network coverage in 6 G, emergency communications are supposed to be enhanced significantly. For example, with line-of-sight propagation and high mobility advantages, aerial-assisted vehicular networks are expected to be key technologies in 6 G to enable broadband connectivity from the sky to emergency vehicles in rescue missions. However, deploying a group of low-altitude short = ABS, long=Aerial Base Stations (ABSs) to provide connectivity for emergency vehicles remains an open issue due to the challenge of maintaining the trade-off between connectivity quality guarantee and efficient flights. This work presents a hybrid Deep Reinforcement Learning-based scheme with Accumulative Training, namely EVRELAY, to address the problem. Based on a pre-trained signal map, the system can provide the best trajectories for deploying high-capability ABSs. Each DRL agent can dynamically adjust the ABS's movements to serve the maximum number of EVs based on the average signal power, remaining capacity (available data rate and energy), and collision avoidance with surrounding obstacles. The simulation results show that our method can maintain the highest overall connectivity coverage and data rate, 7% and 25% better than the existing methods while maintaining 10% lower energy consumption. The system is particularly efficient in large-scale deployment scenarios with many emergency vehicles departing simultaneously.