Performance Optimization Of LAA And Wi-Fi Coexistence Network Based On Machine Learning

LTE(LTE in unlicensed spectrum,LTE-U)technology deploys LTE system to the unlicensed spectrum,which can increase the spectrum utilization resources of LTE system.However,the medium access control(MAC)protocol adopted by LTE system will seriously interfere with the Wi-Fi system deployed in the unlicensed spectrum.In order to coexist more friendly with Wi-Fi network,3GPP proposes the licensed assisted access(LAA)technology which adopts the mechanism of listen before talk(LBT).This technology alleviates the interference of LTE system to Wi-Fi system to a certain extent.In order to make LAA network and Wi-Fi network more friendly and peaceful coexistence,a lot of work has been carried out to analyze and optimize the performance of coexisting network.In the first part of this paper,we mainly study the performance of the LAA / Wi-Fi hybrid network which includes many operators.When multiple LAA networks and Wi-Fi networks coexist,the network conflict will become more complex.Firstly,the backoff process of coexistence network is modeled as a two-dimensional Markov chain model,and the channel access probability of nodes is solved.Then,we abstract the channel access of asynchronous network into a superframe structure,and based on this,we derive the throughput of hybrid network.Finally,the correctness of our theoretical analysis is verified by Monte Carlo simulation,and the effects of delay period and slot parameters on throughput performance are evaluated.The experimental results show that the theoretical analysis of this paper is correct.The second main work of this paper is to propose a dynamic adjustment mechanism of LBT competition window based on machine learning.This mechanism uses Light GBM model to predict the feedback of hybrid automatic repeat request(HARQ),and intervenes the size of competition window in advance to improve the system throughput.The simulation results show that the proposed dynamic adjustment mechanism can significantly improve the throughput of the LAA / Wi-Fi hybrid network.