Dynamic ABS Placement: Using Human Mobility Predictions for Adaptive UAV Trajectories

Unmanned Aerial Vehicles (UAVs) as dynamic aerial-based Base Stations (ABS) show significant promise in enhancing coverage in next-generation wireless networks, especially for addressing temporary events, disasters, or hotspots. However, optimizing ABS deployment in response to ground users’ dynamic and spatial variations poses challenges due to complex air-to-ground channels and UAV interference. This paper investigates UAVs’ transformative role in 5G networks, highlighting their flexibility and cost-effectiveness in adapting to evolving coverage needs and addressing challenges in optimal 3D placement. We propose a solution using reinforcement learning algorithms to dynamically adjust ABS positions based on predicted user spatial distribution. Empirical assessments using real-life mobility data demonstrate the superiority of the transformer model in prediction of movement patterns, surpassing state-of-the-art counterparts by over 10%, and providing transparent explanations for predictions. This research underscores the potential of UAV-based ABS deployment to optimize user serviceability in dynamic 5G networks.