A DRL Resource Allocation for Downlink NOMA Multi-beam Satellite Communications

With the development of communication technology and the increasing demand, there are more expectations for the research of multi-beam satellite communication system, and the efficient management and allocation of resources on the planet and the improvement of system performance are becoming more and more important. In this paper, the downlink Non-Orthogonal Multiple Access (NOMA) multi-beam satellite is used as the system model, and a greedy principle-based joint user grouping and subchannel joint access algorithm is proposed to solve the problems of user grouping and bandwidth allocation. On this basis, a power resource allocation algorithm based on deep reinforcement learning is proposed. By optimizing the reachability and rate of the system and user fairness under the constraints of the total transmission power and the user's minimum transmission rate, the optimization problem is modeled as Markov decision process, using the Proximal Policy Optimization (PPO) algorithm to achieve optimal allocation. The validity and superiority of the algorithm are verified by simulation, which is of great significance to the research of multi-beam satellite resource allocation.