Research On Channel Estimation For Massive MIMO System Based On Beam-Domain Decomposition

Massive MIMO as one of the key technologies of 5G communication system,compared with the traditional MIMO technology,the spectrum efficiency and transmission rate can be significantly improved in massive MIMO system.However,the premise of these performance gains is that the channel state information(CSI)need to be accurate for both the uplink and the downlink.Due to the movement of mobile terminals and the influence of environmental factors,the difficulty of massive MIMO channel estimation is increased,especially in some high-speed mobile scenarios.For the TDD massive MIMO system,when the number of users accessed is large,the pilot overhead of the system will be greatly increased.For the FDD massive MIMO system,the feedback overhead of downlink training is proportional to the number of antennas at the BS.Therefore,how to reduce the pilot overhead and algorithm complexity is a big problem for massive MIMO applications while ensuring CSI is as accurate as possible.In this thesis,the above mentioned problemes are studied.Firstly,a massive MIMO system is briefly introduced in this thesis.Two kinds of the wireless fading channel models are described,including large-scale fading model and small-scale fading model.Two channel estimation algorithms of massive MIMO system are analyzed in detail,which are LS channel estimation algorithm and MMSE channel estimation algorithm.Secondly,a channel estimation algorithm for massive MIMO system based on beam-domain decomposition and SVD is proposed in this thesis.The beam-domain decomposition is used to transform the massive MIMO channel matrix from the angle domain to the beam-domain,and a beam allocation criterion is designed to reduce the pilot overhead.The obtained channel auto-correlation matrix is executed SVD operation to reduce the the complexity of algorithm.After that,the closed expression of channel estimation error is derived.Moreover,a low-rank LMMSE channel estimation algorithm based on beam-domain decomposition is proposed.Compared with the LMMSE channel estimation algorithm using exponential correlation model,the proposed scheme has better channel estimation performance.Then,aiming at the fast channel tracking of massive MIMO system in high-speed train scenario,a beam-domain channel tracking scheme for the massive MIMO system is proposed in this thesis.In order to simplify the system model and reduce the pilot overhead,massive MIMO channel matrix is also transformed from angle domain to beam-domain.Owing to the time-varying characteristics of HST channel,the time-dependent model is introduced into the beam-domain channel modeling,and the transfer equation and state equation of the system are derived.According to the state equation and the transfer equation,Kalman filtering is used for beam-domain channel tracking.A two-stage beam-domain channel tracking scheme is proposed,which achieves a compromise between the channel tracking accuracy and the pilot overhead.The simulation results show that,compared with the proposed angle-domain channel tracking algorithm,the proposed scheme has less pilot overhead and better channel tracking performance when the number of antennas at the BS is less than 100.