Dynamic Backhaul-Aware UAV Deployment in 6G Integrated Access and Backhaul Networks

Unmanned Aerial Vehicles (UAVs) provide an effective solution for extending wireless network coverage, particularly in scenarios where terrestrial infrastructure is limited or temporarily unavailable. Within 6G Integrated Access and Backhaul (IAB) architectures, UAVs can operate as aerial base stations, simultaneously providing user access and maintaining backhaul connectivity to Ground Base Stations (GBSs), enabling rapid and flexible network deployment in dynamic environments. In this work, we investigate the problem of UAVs deployment in 6G IAB architectures, with the objective of maximizing user coverage, while maintaining reliable backhaul connectivity. We propose a deep reinforcement learning (DRL)-based trajectory optimization framework, with both discrete and continuous action control. Performance is evaluated under varying user densities and UAV coverage ranges. Results demonstrate that Double Deep Q-Network (DDQN) consistently outperforms other discrete-action approaches in terms of cumulative reward and adaptability, while the TD3-enhanced Deep Deterministic Policy Gradient (TD3-DDPG) achieves the best performance among continuousaction methods, surpassing Advantage Actor-Critic (A2C), and Distributionally Robust A2C (DR-A2C) in coverage efficiency and backhaul robustness.