Energy Efficiency Optimization Of Networked Large-scale Antenna Array Based On Reinforcement Learning

With the rapid development of the mobile internet,the surge in data traffic and business types,mobile data traffic will experience explosive growth.Some experts estimate that the total demand for data capacity in the next ten years will rise ten times today.In order to address the challenge of the new era,the concept of the Fifth Generation Mobile Network(5G)was proposed.User-centric Massive MIMO(UC-MMIMO)is one of the key technologies in 5G.As a member of large-scale networked antenna array,due to the dense access point antenna arrangement structure,UC-MMIMO can effectively improve the total system capacity.And the distributed deployment also enable UC-MMIMO to satisfy the "ubiquitous"communications philosophy.So,UC-MMIMO has become the hot topic in communication research field.UC-MMIMO still has some trouble.Firstly,the deployment of numerous access point antennas makes the traditional antenna selection algorithms are no longer suitable for Massive MIMO systems,because the traditional algorithms rely on channel state information(CSI)for antenna selection,however,it is difficult to get enough CSI in Massive MIMO.Secondly,plenty of access point antennas will generate huge energy consumption when they are working,and the deployment cost and the maintenance cost will also increase significantly.Therefore,a efficacious antenna selection algorithm and energy efficiency optimization algorithm need to be designed for UC-MMIMO.In recent years,the development of artificial intelligence(AI)and deep learning(DL)technologies has provided solutions for the problems faced by various industries and has achieved fruitful results.As an important branch of AI technology,reinforcement learning excels in many problems.It is of great research significance to introduce reinforcement learning methods in communication systems to solve the antenna selection problem and the energy efficiency optimization problem of UC-MMIMO systems.Focusing on the above two issues,two studies in this paper are carried out,as follows:1.For the antenna selection problem,this paper introduces reinforcement learning method into the antenna selection problem,and proposes an antenna selection method based on reinforcement learning.This method consists of two modules——the antenna selection module and the antenna adjustment module.Through simulation,this antenna selection algorithm based on reinforcement learning has a lower algorithm complexity and less dependence on CSI compare with two traditional antenna selection methods.2.For the energy efficiency optimization problem,because of the differences between Massive MIMO systems and traditional MIMO systems,this article considers the calculation energy and bandwidth coefficients of base stations into the energy efficiency optimization problem,and uses reinforcement learning to solve energy efficiency problem.Through simulation,the power allocation algorithm based on reinforcement learning has higher energy efficiency than the average power allocation algorithm.Combining this part of work with the work of the antenna selection problem in this paper,it is found that the power allocation algorithm based on reinforcement learning has a large capacity for users the promotion also helps.