Average Sum Rate Optimization in RIS-Assisted NOMA Satellite Network: A Deep Reinforcement Learning Approach

Satellite-ground communication network will be an important network architecture in the coming six generation (6G). In this letter, we formulate a reconfigurable intelligent surface (RIS) assisted non-orthogonal multiple access (NOMA) satellite network to improve the poor wireless channel gain caused by long-distance transmission between the satellite and the ground. To maximize the system average sum rate, we joint optimize the phase-shift matrix of RIS, user pairing and power allocation. Specifically, we propose a novel algorithm named PPO-DQN which merging the proximal policy optimization (PPO) and deep Q-network (DQN) algorithm to effectively solve the optimization problem with both continuous variables and discrete variables. Finally, the effectiveness of the proposed algorithm is verified through various simulations.