LAUSR

In this paper, we investigate the joint user pairing and power coefficient allocation for unmanned aerial vehicle (UAV) systems which employ non-orthogonal multiple access (NOMA) to communicate with multiple ground users. Aiming to maximize achievable sum rate and ensure the users' Quality-of-Service (QoS) requirements, we formulate an optimization problem which relies on reinforcement learning (RL) from Multi-Armed Bandit (MAB) framework to propose a solution based on Upper Confidence Bound (UCB) approach. The proposed solution can successfully identify the best action and selects it more often, which leads to maximum system throughput. The attained results show that the proposed scheme finds the best-performing action fast, while the others methods spend a lot of time exploring non-ideal user pairs. As a result, the proposed method accumulates less regret and achieves satisfactory results in terms of system throughput when compared to other user pairing strategies and power allocation (PA) policies.