Research On Channel Estimation And Feedback Algorithms For Massive MIMO Systems

With the rapid development of the mobile networks,the demand for higher data rate of wireless communication systems is becoming more and more urgent.The Massive MIMO(Multiple Input Multiple Output)technology can increase the system throughput significantly depending on massive antennas equipped at the base station.Hence,the research about the Massive MIMO technology becomes one of the hot topics in the field of wireless communications.In order to improve spectrum efficiency and reduce interference,the base station needs to pre-code the transmitted data.This requires the base station to know the exact downlink channel information in advance.However,the number of antennas at the base station may reach hundreds in Massive MIMO system.It is difficult for users to perfectly complete the channel estimation in Massive MIMO system with conventional channel estimation algorithms.In addition,the users also need to feed back the huge amount information of the estimated channel to the base station.Thus,our research focus on how to estimate the channel based on compressed sensing technology by exploiting the channel sparsity feature.In order to reduce the load of feedback,this dissertation also studies compression and decompression techniques.The main contributions of this dissertation are listed as follows.1)An Expected Pruning Matching Pursuit(EPMP)algorithm and a Weighted Matching Pursuit(WMP)algorithm are proposed to improve channel estimation accuracy.In the conventional MP(Matching Pursuit)algorithm,once an element is selected into the support set,it cannot be abandoned during subsequent process,resulting in the degradation of the estimation performance.To address this issue,the compressed sensing based EPMP algorithm is proposed in this dissertation.Those possible errors can be corrected by checking the elements of the support set repeatedly.At each sparsity level of the channel,the positions of the atoms(the columns of the sensing matrix)which have a larger inner product value with the current residual signal are selected into the support set to form an expanded support set.Then,the expanded support set is checked and the wrong selected elements are removed to generate the best support set.The estimated channels and their relative probabilities of occurrence corresponding to these support sets are calculated.Finally,the mathematical expectation of the channel is calculated and treated as the final estimation of the channel.The simulation results show that the EPMP algorithm can reduce the complexity without degrading the channel estimation accuracy.The range of mean square error of the estimated channel with EPMP algorithm is also analyzed.The analytical expressions of the upper and lower bounds are given in the dissertation.The conventional MP algorithm indistinguishably deals with the estimated values obtained in different iterations.However,these estimated values have different importance for the sparse signal.Assigning the same importance to different estimated values cannot provide satisfactory estimation accuracy.To solve this problem,the WMP algorithm is proposed in this dissertation.According to the ratio of the true value and the error both included in the estimated values,different estimated values obtained in each iteration are assigned different weights.The weights are dynamically adjusted to match the change of the true value and the error with the iterations.Experimental results show that appropriate weights can improve the channel estimation accuracy performance.2)A downlink channel feedback scheme is designed to reduce the load of feedback.The pre-coding process at base station requires the downlink channel information which is sent from the mobile user in FDD mode.In order to reduce the load of feedback,a channel feedback scheme is proposed in this dissertation.The downlink channel is first compressed and then fed back to the base station.The base station reconstructs the downlink channel by decompressing the compressed data.To ensure the accuracy of the reconstructed channel,a KSVD(K-Singular Value Decomposition)dictionary is designed by exploiting the correlation between antennas at the base station.Aided by the KSVD dictionary,the channel can be represented with a more sparse form.Thus,better channel estimation accuracy is obtained at base station.The reconstructed downlink channel is utilized by the base station to perform matched filtering precoding.The asymptotic performance of the sum rate in single-cell and multi-cell are analyzed when the base station antenna tends to infinity.3)A pilot interval adaptive adjustment based channel estimation algorithm is proposed for saving pilot resources.The large number of antennas and sub-channels in massive MIMO-OFDM systems require large amount of pilots for channel estimation.The pilot intervals in the time domain and frequency domain are uniform and fixed in conventional channel estimation algorithms.They cannot be adaptively adjusted according to the varying channel parameters.In this dissertation,an Adaptive Orthogonal Matching Pursuit(AOMP)algorithm with varying pilot interval is proposed.In the uplink,the base station estimates the speed of user and adjusts the downlink pilot interval in time domain accordingly.In the downlink,the user estimates the maximum channel response delay and feeds it back to the base station to adjust the downlink pilot interval in frequency domain.In addition,the joint sparsity of the channels between different base station antennas and the same mobile user is utilized to improve the channel estimation accuracy performance.Compared with the channel estimation algorithm with fixed pilot interval,the AOMP algorithm can save pilot resources while preserving estimation accuracy.4)A modified channel estimation algorithm is proposed for the virtual sparse channel to improve the estimation accuracy performance.In general,high-order MIMO systems channel do not exhibit sparse feature.The compressed sensing technology cannot be directly used to estimate the channel.Hence,this dissertation introduces a sparse virtual channel model of high order MIMO system and proposes a compressed sensing based algorithm to estimate channel.The virtual channel is firstly estimated by exploiting the sparse feature of high-order MIMO system channel in the virtual angle domain,and then converted to the actual physical channel.A pilot matrix with simple structure is designed to reduce the complexity of the proposed algorithm.The parameters of the virtual channel are properly choosen to weaken correlation between atoms.The pilot sequence length is adjusted dynamically according to the channel state change to ensure stable estimation accuracy performance.Experimental results verify that the proposed algorithm has ability to provide more accurate estimation.