Optimal Power Allocation for NOMA-Based Satellite Networks Under Delay QoS Constraints

In this paper, a system capacity maximization problem of power domain non-orthogonal multiple access (NOMA) based satellite networks is proposed to determine the power control scheme. Specifically, under the constraints of both users’ various delay quality-of-service (QoS) requirements and minimum rate limitations, this maximization problem is decomposed into two steps. In the first step, the concept of effective capacity is employed to describe each NOMA user’s performance under certain statistical delay constraints. Based on which, the influences of minimum performance limitations and system key parameters on the power allocation strategy are investigated. Then, with the objective of maximizing system capacity, a deep reinforcement learning (DRL) with varying bounds is proposed, with which the lower and upper bounds of action space can be flexible varied according to reward, to optimize the power allocation scheme in the considered system. As illustrated by numerical simulations, our proposed method has superior performance in further enhancing convergence rate and achievable performance for the proposed networks.