Throughput maximization of EH-CRN-NOMA based on PPO

The cognitive radio system with non-orthogonal multiple access and energy harvesting (EH-CRN-NOMA) is an important research aspect in 6G network. To maximize the total throughput of multiple secondary users (SUs) in the EH-CRN-NOMA a novel deep reinforcement learning algorithm (DRL) based on proximal policy optimization (PPO) is proposed in this paper. Firstly, we simplify the optimization variables through mathematical, so as to reduce the burden of agents. Furtherly, we use a new decoding sequence for better successive interference cancellation (SIC) in the base station (BS). Finally, experimental results show that the proposed algorithm achieved faster convergence and higher throughput. At the same time, it simplified the action space of DRL, but the time-consuming performance is poor.