Deep Reinforcement Learning Based Dynamic User Access and Decode Order Selection for Uplink NOMA System With Imperfect SIC

摘要

Perfect successive interference cancellation (pSIC) is often assumed in power domain Non-orthogonal multiple access (NOMA) systems. Besides, signals are generally decoded in the order of received signal strength from the strongest to the weakest (DS2W). However, pSIC is impractical in the real NOMA system, and choosing the order of DS2W for the imperfect SIC NOMA systems may lead to decoding failure. In this letter, the imperfect SIC is considered, and a hierarchical deep reinforcement learning-based user access and decoding order selection (HDRUD) algorithm is proposed for the considered uplink NOMA. The simulation shows that the algorithm is significantly better than traditional access and decoding schemes, especially when the SIC is imperfect or the quality of service (QoS) requirements of users are different.