This paper investigates the joint optimization of movement and user association for multiple UAV base stations (UAV-BSs) to provide emergency services to indoor users. In this paper, we propose a multi-agent deep Q-network (MADQN) reinforcement learning algorithm to address this problem. Moreover, we apply a self-attention mechanism to achieve the objective more effectively. Simulation results demonstrate that the proposed algorithm outperforms conventional algorithms, such as random action and MADQN without a self-attention mechanism.
This work was partly supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. RS-2024-00359330, Design of Low Earth Orbit Satellite Communication System) and Institute of Information & communications Technology Planning & Evaluation (IITP) grant funded by the Korea government (MSIT) (No. RS-2022-II220704, Development of 3D-NET Core Technology for High-Mobility Vehicular Service).
