UAV-Assisted Relay Communication: A Multi-Agent Deep Reinforcement Learning Approach

Unmanned Aerial Vehicles (UAVs) are increasingly employed across diverse sectors for their exceptional agility, adaptability, and swift deployment. This research delves into the application of UAVs as relay nodes within expansive cellular networks, where ground sensors (GSs) transmit their packets to ground base stations (GBSs) via multiple UAV relays. To ensure UAVs' safe navigation and consistent connectivity with GBSs, we introduce a constraint on the control link's outage time between a UAV and its corresponding GBS during flight. The focus of this paper is on the design of UAVs' trajectory and relaying strategy to minimize the overall mission completion duration, taking into account energy usage limitations and UAV control link outage constraints. Leveraging the collaboration of multiple UAVs, we propose a QMIX-based UAV relaying strategy and assess its efficacy via comprehensive simulations. Comparing with two benchmark algorithms, our proposed algorithm demonstrates a reduction in mission completion time and an increase in overall reward.