Research On Adaptive Modulation And Coding Based On 5G

With the innovation of technology,wireless communication has gradually been moving towards higher transmission rates,lower delays,and greater system capacity.As one of the key technologies of the fifth-generation mobile communication(5G),Adaptive Modulation and Coding(AMC)technology can ensure transaction correct ratio,while enhancing transmission rate as much as possible.In the downlink of 5G system,the traditional AMC technology relies on Signal-to-Noise Ratio(SNR)measured by the users to mapped into Channel Quality Indicator(CQI),which is sent to the e NB.The e NB then exploits CQI to select a suitable Modulation and Coding Scheme(MCS)for downlink transmission.In engineering applications,AMC has two technical difficulties: CQI measurement and MCS selection.In order to solve the problems in these two processes,the main contents of this thesis are as follows:Firstly,in order to further improve accuracy of effective SNR mapping in CQI measurement process of AMC,so as to better complete the effective SNR mapping of the fading channel under Additive White Gaussian Noise(AWGN),an exponential effective SNR mapping(EESM)algorithm with improved adjustment factor optimization is proposed.Based on the traditional EESM algorithm,a new adjustment factor is introduced in the optimization process of adjustment factor,and the calculation formula for the optimization factor is modified.The Block Error Rate(BLER)is taken as the logarithm and then Logarithmic results were averaged by number of channel relizations before calculating the mean square error(MSE),thereby to reduce mapping errors.The simulation proves that the performance of the improved adjustment factor optimization algorithm used in this thesis is significantly enhanced,and the MSE is reduced by nearly 0.2 compared with the traditional scheme under high-order modulation(such as 256QAM).Secondly,in order to improve the accuracy of MCS selection in the actual channel and thereby boost the system throughput,this thesis proposes a MCS selection method based on Deep Reinforcement Learning(DRL).Based on the Reinforcement Learning(RL)algorithm,the state space for training is extended from the original current SNR only to two indicators: SNR and the channel matrix obtained by channel estimation.Considering that there are some feedback delays in the process of channel matrix feedback to the base station,a Gauss-Markov model is used to predict the received channel matrix.Then,by setting the target Deep Q Network(DQN)and training DQN separately,the channel quality of this transmission is inferred to select a more accurate MCS to maximize the throughput.The simulation results prove that the throughput performance of MCS selection algorithm based on DRL proposed in this thesis is superior to the traditional MCS selection algorithm based on Look-up Table(LUT)and RL.When the feedback delay is 4ms,the throughput performance has been improved significantly,reaching up to about 13 Mbps.