Study On Channel Estimation Methods Of Massive Multiuser MIMO Uplink Systems

Multiple-input multiple-output (MIMO) technology can improve speed and relia-bility of information transmission through spatial multiplexing and antenna diversity without extra bandwidth and power consumption. However, with the growing populari-ty of smart terminals, future mobile communication systems need to meet higher re-quirements:(1) serveing more users on the same frequency resources; (2) meeting high-er transmission rates and reliability; (3) possessing lower power consumption. To this end, Marzetta proposed massive MIMO technology in 2010 which can effectively meet the above requirements. Channel estimation is essential in massive MIMO system to guarantee system signal detection and collaboration transmission between the base sta-tion. Hence, this paper focuses on the channel estimation method of massive MIMO system.First, the thesis analyzes the causes of pilot contamination and its effect on massive MIMO channel estimation. In the case of a multi-cell pilot multiplexing, massive MIMO system’s performance is mainly limited to pilot contamination. The traditional channel estimation method (such as MMSE, LS), when the number of users is very large, more affected by the pilot pollution, and possess the lower frequency spectrum utilization. Therefore, it is necessary to research a new and effective channel estimation method. Next, this thesis estimates the channel by using eigenvalue decomposition (EVD) on the matrix that utilizes the pairwisely orthogonality of the channels vectors between the users and the BS in massive MIMO systems, without requiring any specific structure of the transmitted signals. The channel vector corresponding to each user can be estimated via using the covariance matrix of the received signals, but up to a re-maining scalar multiplicative ambiguity which can be resolved by sending a short pilot sequence. In order to further reduce the estimation error, the iterative least square pro-jection algorithm (ILSP) can be applied to EVD-based estimation method. The simula-tion results show that even in the case of a short pilot sequence, the error of EVD-based estimation method is far smaller than the conventional channel estimation method. Fi-nally, this thesis studies the channel estimation problem of massive multiuser MIMO system with finite scattering path model that closer to the actual scene. Since the de-grees of freedom of the channel matrix is much smaller than its large number of free parameters for physical finite scattering channel model, thus, this thesis analyses an ef-fective channel estimation method which particularly applicable to this scene, i.e., the CS-based low-rank matrix approximation. In this method, the problem of low rank ma-trix estimation is converted into the constrained optimization problem and solved by quadratic SDP. The simulation results show that in the case meeting the same estimation error, this method needs less pilot sequences and lower pilot transmission power than LS, what is more, this method not requires any knowledge about the statistical distribu-tion of the transmission channel.