Optimal Power Allocation for Rate Splitting Communications With Deep Reinforcement Learning

This letter introduces a novel framework to optimize the power allocation for users in Rate Splitting Multiple Access (RSMA) network. In network, messages intended are split into different parts that single common part and respective private parts. mechanism enables RSMA flexibly manage interference thus enhance energy spectral efficiency. Although possessing outstanding advantages, optimizing is very challenging under uncertainty of communication channel transmitter has limited knowledge information. To solve problem, we first develop Markov Decision Process model dynamic channel. The deep reinforcement algorithm then proposed find optimal policy without requiring any prior information simulation results show scheme can outperform baseline schemes terms average sum-rate QoS requirements.