Low-Complexity Channel Estimation In Massive MIMO System

Massive MIMO(Multi- input Multi-output) technology is the latest research direction in the present wireless communication domain, and strongly promotes development of next generation 5G mobile communication technologies. Massive MIMO technology improve the spectrum efficiency, energy effect and the robustness of the system, etc through allocating hundreds of low-power antennas at the base station, which attracts more and more attention by the majority of scholars. But when we conduct channel estimation, because of involving the process of the inversion to channel covariance matrix, at the time that using increasing the number of the antenna at the base station to break the interference bottleneck between the cells and improve the communication quality, the high computational complexity in channel estimation also become a inevitable problem to large-scale MIMO technology.First of all, this thesis expends research around the massive MIMO technology and makes a comparison the respective advantages and disadvantages between the TDD communication model and the FDD communication model and their applicability. We introduce the massive MIMO system model. With the system model, we recommend three kinds of frequently- used channel estimation algorithm in detail and deduce the estimation results of each algorithm according to the algorithm principle.Secondly, we research a MMSE channel estimation algorithm based on the Kronecker model, and analyze the computational complexity problem emphatically when the MMSE estimator is used in the massive MIMO system. To solve the high complexity problem of the MMSE channel estimation algorithm, we present a diagonalizable channel estimation method. We separately ana lyze the gradualness of the estimation error of the diagonalizable channel estimation method in noise- limited scenario and pilot contamination scenario when the transmitting power of the pilot enlarges progressively, and make the simulation to the diagonalizable channel estimation algorithm. The research results shows that the diagonalizable channel estimation method enables the complexity of the channel estimation is reduced from the original cube level to linear level, and this method is more suitable for noise- limited scenario.Then, to solve the high complexity problem of the MMSE channel estimation algorithm in the massive MIMO system, we propose using the SOR to reduce the computational complexity of the channel estimation by transforming the process o f inversing the covariance matrix to the process of solving the linear equation and analyze the effect of the choice of the relaxation factor to the complexity. The Studies show that the channel estimation based on SOR iterative method can make the high complexity issue of the MMSE channel estimation be resolved and the estimation performance of the SOR estimator can achieve the MMSE estimator ’s absolutely with the increase of the iterations. The simulation revealed that the method is more appropriate in the pilot scenario.Finally, we propose a low-complexity channel estimation based on the conjugate gradient method, and this method has global convergence, avoids the SOR estimator’s convergence problem. In order to speed up the convergence rate of the conjugate gradient method, we design a preconditioned matrix by splitting the coefficient matrix of the linear equations, and propose a new preconditioned conjugate gradient method. The Studies show that the channel estimation based on conjugate gradient and preconditioned conjugate gradient all can make the high complexity problem of the MMSE channel estimation be resolved and their estimation performance all can achieve the MMSE estimator’s absolutely with the increase of the iterations. The simulation revealed that the method is more appropriate in the pilot scenario.