Energy-Efficient Multi-UAV Network using Multi-Agent Deep Reinforcement Learning

With the explosive growth in mobile data traffic, unmanned aerial vehicles (UAVs) has received much attention in recent years. While UAV offers a number of benefits, the degeneration in energy efficiency due to the hardware limitation has become a major challenge in achieving the primary goals viz., 100-fold increase in the throughput in 6G wireless communication. In this paper, we propose multi-agent actor-critic (MA-AC) based joint trajectory design and resource allocation scheme to minimize the energy consumption of multi-UAV network. By exploiting actor-critic, the proposed scheme learns to determine UAV trajectory and transmit power. To deal with the signaling overhead in multi-UAV network, local agent in each UAV distributively decides UAV trajectory and transmit power using local information. Numerical results show that the proposed scheme can significantly improve the energy efficiency of multi-UAV network while ensuring the QoS requirement of network.