Joint UAV Movement Control and Load Balancing Based On Indirect Control in Air-Ground-Integrated Networks

Air–ground-integrated networks play a vital role in seamless coverage. The irregular topologies and dynamic loads require a comprehensive cooperation between terrestrial base stations (BSs) and unmanned aerial vehicles (UAVs) in air networks. In this letter, a joint UAV movement control and load balancing (LB) method is proposed to improve the load completion ratio (LCR). Specifically, a reinforcement learning (RL) method soft actor-critic (SAC) is utilized to control the UAV movements, while the LB is performed by tuning the cell individual offsets (CIOs) of BSs and UAVs. Moreover, to make the solution compatible to heterogeneous BSs and UAVs, an indirect step-and-slipping-window-based CIO tuning method is proposed to convert the actions of SAC into CIOs. Simulation results show that the proposed method outperforms the state-of-art baselines up to 29% in terms of the LCR.