Research On Limited Feedback Method Based On Partial Reciprocity Of Uplink And Downlink Information

The commercialization of 5G has become the general trend.As one of the key technologies,massive multiple-input multiple-output(Massive MIMO)technology can bring higher capacity and spectrum utilization ratio to the system compared with traditional MIMO technology,and has great potential in user experience.The accuracy of channel state information(CSI)acquisition is a prerequisite for taking full advantage of massive MIMO systems.However,the core idea of a massive MIMO system is to set up a large number of antennas in the base station,which leads to the increase of channel space dimension and also increases the difficulty of CSI acquisition.Moreover,in the scenario where the user moves at a high speed,the training delay and feedback delay caused by the speed and Doppler shift also seriously affect the timeliness of CSI acquisition.Especially in the acquisition of downlink CSI in the frequency division duplex(FDD)system,since the complete channel reciprocity between uplink and downlink does not hold,it is difficult for the base station to directly deduce the downlink channel from the uplink channel.Therefore,FDD downlink CSI acquisition is more difficult than time division duplexing(TDD)downlink CSI acquisition.In response to the above problems,this thesis focuses on the issue of downlink CSI acquisition in FDD massive MIMO systems,and proposes corresponding CSI acquisition methods in consideration of two scenarios,namely coherent channel and fast-changing channel.The main contents and results are as follows:First,in the coherent channel scenario,combined with the spatial reciprocity of the uplink and downlink channels in the FDD system,an adaptive training-feedback method based on a reference frame is proposed.The main idea of this method is to first design a reference frame of matching channel tensor product structure.Based on the spatial reciprocity of the uplink and downlink CSI,the uplink channel information and the singular value decomposition are used to determine the subspace of the downlink channel.Then,these base vectors of this subspace are sent to the user as pilots.Therefore,the user can obtain the coefficients of the downlink channel represented by the set of basis vectors through training.Finally,the user uses the designed precoding codebook to feed back the obtained coefficients to the base station.Combining these basis vectors and the coefficients obtained by feedback,the base station can construct a precoding.Because the pilot designed by this method is generated based on the uplink channel,it is adaptive and can effectively reduce the training overhead and feedback overhead.Simulations also verify that the performance of this method is better than the existing methods,for example,see [14].Second,for the fast-changing channel,using the spatial reciprocity,combined with the idea of principal component analysis(PCA),a PCA-based adaptive training-feedback method is proposed.The main idea of this method is to extract the basis vectors of the channel space through the PCA technology and the singular value decomposition by using the uplink channel information according to the spatial reciprocity and send them as pilots to the user or training.The user then feeds back the coefficients obtained through training to the base station,and the base station can obtain the downlink CSI according to the basis vectors and the coefficients obtained from the feedback.And through simulation verification,even if the user is in a high-speed moving scenario,with a small number of training times,compared with the ideal precoder,this method can still achieve good performance,and its performance loss is acceptable.