Joint Optimization of Multi-UAV Deployment and User Association via Deep Reinforcement Learning for Long-Term Communication Coverage

The flexible deployment and strong adaptability of unmanned aerial vehicles (UAVs) have great significance in improving communication quality and expanding communication network coverage. However, due to the complex constraints and time-varying characteristics of the communication environment, the dynamic deployment of multi-UAV is challenging to ensure the reliable operation of the UAV-assisted communication systems. To address this challenge, we present a novel approach called long-term communication coverage for ground users through jointly optimizing the multi-UAV deployment and user association (LTCC-UDUA) using deep reinforcement learning (DRL). First, we model the multi-UAV deployment and user association as a decentralized partially observable Markov decision process. Subsequently, we formulate a reward function that accounts for both the communication fairness index for ground users (GUs) and the total system throughput. Finally, continuous optimization of the multi-UAV's movement trajectories is performed in a centralized training and distributed execution manner. Simulation results demonstrate that our proposed LTCC-UDUA scheme significantly outperforms two commonly used baseline methods in terms of GUs associated, fairness index, and total throughput.