Deep Learning Based Dynamic Uplink Power Control for NOMA Ultra-Dense Network System

Abstract

As the development of blockchain and 5G, all kinds of intelligent devices have increasingly higher requirements on data rate and computing power, a large number of base stations are used, optimizing service cost and equipment energy consumption have become new challenges. It is certain that blockchain will be an important technology for the successful development of 5G network. As a new research direction, non-orthogonal multiple access technology (NOMA) combined with ultra-dense network (UDN) can effectively improve system capacity and reduce service cost. In this paper, we study a dynamic energy efficiency (EE) optimization problem under uplink NOMA communication in UDN. In order to ensure the real-time requirement of user equipment, a markov decision process (MDP) model is constructed by quantifying resources in access points (APs) and user equipments (UEs). On this basis, we propose a Deep Q-Network (DQN) based dynamic uplink power control algorithm to maximize the EE. According to different uplink channel gains in different base stations, UE transmission power is controlled through the center node. Through emulation and comparison with traditional Q-learning algorithm, experimental results show that DQN algorithm can effectively improve the EE of the system.