Q-learning based trajectory design for UAV communications network with fairly non-orthogonal multiple access

Benefit to the advantages of flexibility and low-cost deployment, employing unmanned aerial vehicle (UAV) as the aerial base station (ABS) constitutes a promising technology to support the multi-user access. However, facing the challenges of both mobile ABS and ever-increasing users, it is crucial to design the UAV trajectory in the dynamic environment, in order to providing communications services for multiple mobile users with fair consideration. Therefore, in this paper, we propose a fairly non-orthogonal multiple access (FNOMA) scheme for UAV communications network, which enables the ground mobile users to be accessed to the ABS by sharing the same spectrum resources with high fairness. For the sake of optimizing the attained system throughput, a novel greedy genetic algorithm assisted Q-learning (GGA-Q) method is conceived, where the UAV trajectory is elaborately designed by jointly considering dynamic user grouping and power allocation. Our simulation results indicate that the proposed UAV trajectory planning algorithm based on FNOMA scheme can significantly improve the fairness level while enhancing system throughput.