Resource Allocation for UAV-Assisted NOMA Systems With Dual Connectivity

In this letter, we consider a non-orthogonal multiple access (NOMA)-based uplink cellular system assisted by an unmanned aerial vehicle (UAV) through dual connectivity. To balance efficiency and fairness, we aim to maximize the sum weighted bit rate by jointly optimizing the three-dimensional UAV placement and power allocation. Considering the non-convexity of the original problem, we propose a hybrid offline-online algorithm: In the offline training, when the UAV placement is given, the property of fractional programming is utilized to obtain a sub-optimal power allocation solution; Based on the results of power allocation, the deep deterministic policy gradient algorithm is leveraged to learn the optimal UAV trajectory policy that can be deployed in an online fashion. Finally, simulations are conducted to show the efficiency of the proposed algorithm.