Joint Optimization Of Base Station Activation And User Admission Control For Large-Scale Wireless Networks

With the rapid development of communication technology,on the one hand,the num-ber of users has been increasing and mobile data has started to explode.On the other hand,in order to ensure the Quality of Service（Qo S）of users,numbers of base stations need to be deployed.The above problem causes the communication network to grow larger,and in this case,efficient maintenance and management of the network becomes a very important issue.In terms of energy consumption,firstly,maintaining the base stations in a normal working environment requires a lot of energy.Secondly,accessing all users may lead to a dramatic increase in energy consumption.Therefore,it is necessary to find suitable methods to manage the base stations and users to reduce the energy consump-tion.The two types of energy consumption can be effectively managed using base station activation and user admission control respectively,but few studies have considered them jointly.In this thesis,we jointly optimize the base station activation and user admission control（JBAUA）.In addition,the actual scenarios of large-scale wireless networks are considered,and design a low-complexity solving algorithms.The main work of this thesis is as follows:（1）We proposed a JBAUA and complete the system modeling.In this thesis,we con-sider a Multiple Input Single Output（MISO）downlink wireless communication network,in which all base stations jointly serve users.The conventional beamforming energy min-imization problem uses the Signal to Interference Plus Noise Ratio（SINR）as a measure of Qo S to optimize the transmit energy.Based on this,the l₀-norm penalty term for base station activation and user admission are introduced into the objective function,and the control strategies with different tendencies are realized by adjusting the weight factors of the two penalty terms.（2）The problem presented in this paper is the Non-deterministic Polynomial（NP）hard,which is very tricky to solve.We use a continuous function to approximate the l₀-norm,and use the Successive Upper-Bound Minimization（SUM）algorithm to approx-imate the original problem as a convex problem.For large-scale wireless network,we design a distributed solution algorithm with lower complexity,which is based on the Al-ternating Direction Method of Multiplication（ADMM）.Compared with the CVX solution,this method can distribute the computation task to each terminal（including base station and user）,which can save a lot of time.（3）We propose a Reinforcement Learning（RL）based online base station activation and user admission strategy,where an agent explores the action space by continuously interacting with the environment and tries to find the optimal solution.We use the Dou-ble Deep Q-Network（DDQN）to avoid the storage and computation cost in Q-learning,and over-estimation of value function in Deep Q-Network（DQN）.Compared with the traditional optimization algorithm,the RL strategy has better performance.From the perspective of green communication,the problem proposed in this thesis greatly reduces the energy consumption,effectively guarantees the Qo S of users.The solution of the problem can efficiently cope with the large-scale wireless network.