GAN Model Based Low Papr Sequences And Their Application In 5G NR

In recent years,with the development of fifth-generation(5G)mobile communication,users have put forward higher requirements for communication quality.In 5G scenarios,the scarce low-frequency 5G resources significantly limit the application of eMBB.On the other hand,the 5G New Radio(NR)mm-wave spectrum will seriously deteriorate the uplink coverage performance.NR requires a low-latency physical uplink control channel(PUCCH)for ultra-reliable and low-latency communication(URLLC),which will also limit uplink coverage performance.Due to the poor wireless channel and low signal-to-noise ratio(SNR),users at the edge of cells often cannot obtain high-quality communications.5G NR usually uses sounding reference signals(SRS)for resource scheduling,link adaptation and beam management.Studies have shown that the sequences with low peak-to-average ratio(PAPR)characteristics can improve user terminal(UE)power efficiency and improve the uplink coverage performance of communication systems.By considering mainly PUCCH and SRS in 5G NR,this thesis will investigate the impact of different transmission sequences on the uplink coverage performance,and design required low-PAPR sequences to enhance the performance of uplink physical channels based on generative adversarial networks(GAN).This thesis firstly introduces the content of PUCCH and SRS in 3GPP 38.211 standard,including the sequence function and design requirements,and focuses on the analysis and generation of the Zadoff-Chu(ZC)sequence characteristics using MATLAB.Secondly,a lowPAPR sequence generation method based on the GAN model is presented.This method utilizes the low-PAPR features of Golay sequences of different lengths to screen the PAPR and the time-domain peak cross-correlation,resulting favourable optimized sequence sets.The results show that,compared with the ZC sequences used in the 3 GPP standard,the new sequences obtained have lower PAPR with a slight loss of the peak cross-correlation in the time domain.Finally,the author builds a simulation platform for the relevant parts of the 5G NR uplink,and simulates the performance of false alarm rate and missing detection rate of PUCCH.Then the channel estimation performance by employing various SRS,are conducted.The results show that the related performance is improved slightly by applying the low-PAPR sequences generated by the aforementioned GAN model,compared with the 3 GPP standard ZC sequences.