Research On MAC Algorithm Based On Reinforcement Learning In IEEE 802.11ax

In recent years,with the rapid development of mobile Internet of Things(Io T)services,people's demand for wireless network traffic and Quality of Service(Qo S)keeps growing.Wireless Local Area Network(WLAN),together with cellular networks,has attracted extensive attention because of its high speed,flexible deployment and low cost.However,as WLAN deployments become more and more intensive,it becomes more and more important to solve the problem of high density network deployment.IEEE 802.11 ax standard provides an uplink multi-user random access mechanism.This mechanism can solve the problems of serious Station(STA)conflict and low channel utilization in highdensity network to some extent,but it cannot realize STA intelligent access channel.In this thesis,based on IEEE 802.11 ax,we introduce reinforcement learning to solve the above problems and provide an intelligent access mechanism.Firstly,in order to solve the problems of serious high-density WLAN conflict,low channel utilization and abnormal throughput performance,this thesis proposes a Media Access Control(MAC)mechanism based on IEEE 802.11 AX Orthogonal Frequency Division Multiple Access(OFDM)technology,which combines Multiple rounds of competition and fast back-off.The Contention Window(CW)is a mechanism for every update cycle where all stations hold the same Contention Window as the contention channel.At the same time,the Access Point(AP)optimizes the competitive window by introducing Q-learning(QL)in different competitive window update cycles.The simulation results show that this mechanism can not only ensure the fairness between stations,improve throughput and reduce delay,but also concentrate the contention window optimization problem on AP to reduce the computing load between stations.In addition,a MAC algorithm based on Deep Q-Network(DQN)was proposed to optimize the contention window for the unsatisfactory performance improvement of the algorithm in high-density WLAN traffic instability scenarios.The algorithm evaluates the complex network system through the neural network,and designs a more flexible contention window optimization strategy and network state evaluation mechanism,which makes the algorithm suitable for more complex network scenarios.At the same time,different application scenarios are designed to analyze the performance of MAC algorithm,such as throughput,delay,fairness and convergence speed.Theoretical analysis and experimental simulation show that the proposed algorithm can achieve better performance in most application scenarios.