Font Size: a A A

Research On Limited SRS Resources With Massive Antennas

Posted on:2021-05-13Degree:DoctorType:Dissertation
Country:ChinaCandidate:T Y ZengFull Text:PDF
GTID:1368330605981235Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
With the development of communication technologies and the demands of people,massive antenna technology,namely massive multiple-input multiple-output(MIMO)is more and more preferred by academia and industries.When massive antennas are equipped by the base stations,the Multi-User(MU)technologies can be adopted to obtain the spatial gain and the great improvement of Spectral Efficiency(SE).However,the gain depends on the acquisition of Channel State Information(CSI).Given that the current services are dominated by downlink data services,the acquisition of downlink CSI is crucial.In practice,even though Time Duplex Division(TDD)systems have facilitated the acquisition of downlink CSI,due to the fact that the resources of Sounding Reference Signals(SRS)are limited,the downlink CSI still cannot meet the requirements.Especially when a lot of users are to be scheduled and the moving speed varies apparently,the insufficient CSI caused by the limitation of SRS resources will degrade the system performance severely.This thesis is motivated by alleviating the problem,and investigates from three aspects,namely channel estimation,MU grouping and pilot allocation.The main contributions of this thesis are as follows.(1)The improvement of channel estimation performance based on Denoising Convolutional Neuron Network(DnCNN).DnCNN has achieved great success in image denoising of computer vision,it has the potential to handle the denoising problem of wireless signals if it is adopted properly.Since enhancing the ability of acquiring CSI(i.e.channel estimation)for base station is the direct solution for SRS-limited problem,it is focused on the single user and multi-user channel estimation based on DnCNN in the first part.For single user channel estimation,it is proposed the feature map establishment for CSI,provided the recommended DnCNN configurations and proposed fast training scheme to satisfy the practical latency constraint.In this part,some practical problems are fully evaluated,such as the data pollution caused by error,the applicability for different cases and the bound of performance,etc.Simulation results show that the proposed scheme has good robustness and applicability.For multi-user channel estimation,it is proposed improved DnCNN model.To guarantee the sufficient features contained in datasets,based on the current design of SRS sequences,the SRS sequences are modified,and the matched adopting method is proposed.Finally,the channel restoration is provided to separate the channels of multi-users.The factors such as antenna groups and Signal to Noise Ratio(SNR)are studied in this part.Besides,the theoretic and operation time for time complexity are analyzed as well.Results show that the proposed scheme remarkably enhances the capacity of channel estimation and has decent robustness.(2)The MU grouping and scheduling scheme based on CSI Reference Signals(CSI-RS)measurement.Screening the outdated users to expand the available CSI set for base station is one possible solution for SRS-limited problem.In this part,it is focused on the screening outdated CSI and the measurement of outdated CSI is proposed based on CSI-RS.The outdated degree for CSI is defined and derived based on the current applications of CSI-RS,and the measurement method for outdated users is provided as well.Then according to the Line of Sight(LoS)conditions,two screening approaches for outdated users are proposed in terms of different user number and movement speed.Finally,the MU scheduling scheme is proposed on the premise of balancing the fairness and system throughput with measurement results.Simulation results show that,compared to other schemes,the proposed scheme can effectively avoid scheduling the outdated users with large CSI errors and therefore guarantees data transmission efficiency.Besides,it is confirmed that the LoS conditions can be valid basis to alleviate SRS-limited problem.(3)The high-precision LoS identification based on Convolutional Neuron Network(CNN).As mentioned above,LoS information can be adopted to assess the outdated situation of CSI for users.However,the precision of existing methods is not satisfactory,and the acquisition process does not match the resource allocation stage.Thus,it is necessary to propose novel identification method that can facilitate the acquisition of LoS conditions for resource allocation.According to the receive signals,the design method for training samples and the CNN settings are provided.By discussing the influence of different factors such as antenna number,extracted taps,multi-user number,inter-antenna separation and etc,the recommended configurations are provided.(4)The pilot allocation schemes which can alleviate the CSI outdating and energy consuming.Compared with the topics focused by existing pilot allocation schemes,in this part it is innovatively proposed to adopt pilot allocation method to handle SRS-limited problem.Two algorithms are included in this part.The first method aims to alleviate the cost caused by correlation.First,the metric that evaluates the degradation of system SE is defined.Then the mathematic relationship between the correlation and proposed metric is derived theoretically,which is the basis of proposed pilot allocation scheme.In the second part,the research aims to solve the waste problem of SRS resources caused by the inappropriate allocation.The evaluation metric of CSI precision is first provided.Then it is verified that the linear relationship between the metric and time in the short interval.Based on this conclusion,the measurement method for user-specific sounding period and the pilot allocation scheme are provided.These two schemes can enhance system SE and have meaning for practical cases.
Keywords/Search Tags:limited SRS resources, massive antennas, channel estimation, pilot allocation, MU grouping
PDF Full Text Request
Related items