UAV assisted Satellite-Air-Ground Communication: Performance Analysis

This work focuses on the trajectory optimization of unmanned aerial vehicles (UAVs) working as intermediate nodes in downlink satellite-air-ground communications. Here, the UAV is positioned in between the LEO satellite and terrestrial mobile users as aerial base stations. In this work, we optimize the UAVs trajectory, focus on the throughput maximization, and interference reduction in the satellite power transmission to the UAVs and ground users. To tackle these challenges, multi-agent deep reinforcement learning (DRL) is employed as a robust solution. The research introduces a sophisticated simulation environment encompassing multiple UAVs acting as DRL agents within an elaborate system model. This model incorporates mobility and transmission components, allowing crucial parameters calculation such as transmit rates, rewards, and states essential for reinforcement learning. Moreover, the simulation environment accounts for the dynamic nature of users locations, enabling agents to adapt and optimize trajectory planning under diverse conditions.