Throughput Maximization in Multi-UAV NOMA Networks Based on Deep Reinforcement Learning

Abstract

Unmanned aerial vehicles (UAVs) can be launched as aerial base stations (ABSs) to help enhance the coverage and quality of communication. Additionally, non-orthogonal multiple access (NOMA) can support massive connectivity for a network to improve spectrum efficiency. In this article, a downlink multi-UAV NOMA communication system is studied to operate as an alternative solution for demands like natural disasters and unexpected traffic in a timely fashion. We intend to maximize the sum system throughput, in which trajectory design and power allocation of all the UAVs are jointly optimized. In order to tackle this complicated non-convex problem, we propose a multi-UAV deep deterministic policy gradient (DDPG) based algorithm for throughput maximization (MDDPG-TM). Simulation results validate that the proposed algorithm outperforms the benchmarks.