Reinforcement Learning Based Joint Resource Allocation and User Fairness Optimization in mmWave-NOMA HetNets

In this paper, we propose heterogeneous network (HetNet) with mmWave and hybrid non-orthogonal multiple access (NOMA) and orthogonal multiple access (OMA) transmission, where uplink (UL) of macrocell users (MCUs) and downlink (DL) of small cell users (SCUs) share the same resource block (RB) to increase the network capacity. Also, an imperfect successive interference cancelation (I-SIC) decoding is considered due to hardware impairment of the real-world NOMA systems. We formulate the problem of joint power and resource block (RB) allocation to maximize rate-fairness and sum-rate by considering minimum quality of service (QoS) requirements. Reinforcement learning (RL) named Q-learning (QL) and deep Q-learning network (DQN) algorithms are employed to solve these problems. The results indicate that the proposed QL and DQN algorithms increase the fairness and sum-rate while low number of iterations is required.