Channel Estimation For Massive MIMO Wireless Communication

To meet the growing demands for high speed wireless data rates services, massive multi-input multi-output (MIMO) technology has drawn considerable attention. By employing a large number of antennas at base stations to serve multiple user equipments (UEs) simultaneously, massive MIMO can improve spectral efficiency and link reliability substantially. In this thesis, channel estimation technologies for massive MIMO communication are investigated both in single-cell and multi-cell scenarios.Firstly, the concept of factor graph is expounded, as well as a detailed introduction about factorizations and marginal. We analyze factor-graph representations of factorizations, and the derivation of sum-product algorithm is given afterwards. We end by combining these, leading to approximate message passing (AMP) algorithm.Next, AMP channel estimation for massive MIMO system is investigated. Taking advantage of the approximate sparsity of a channel that can be obtained by transforming the received signal into a beam domain, a channel estimation method that based on statistical channel state information (SCSI) is proposed under a single-cell scenario. This proposed algorithm can significantly reduce pilot overhead by replacing orthogonal pilots with random pilots, which can approach the performance of using the pilot generated by shifting Zadoff-Chu sequence. This proposed AMP algorithm achieves better mean squared error (MSE) performance than least square (LS) channel estimation, and its complexity is far below that of minimum mean squared error (MMSE) channel estimation. State evolution algorithm is put forward to analysis the convergence of the proposed AMP algorithm and predict the iterations. Simulations show that this algorithm has better MSE and faster convergence speed, compared with the expectation-maximization Gaussian-mixture approximate message passing (EM-GM-AMP) algorithm proposed in earlier works.Finally, we investigate inter cell interference (ICI) in massive MIMO multi-cell scenario. When dis-cussing with multi-cell scenario, previous works often assume that, UEs in the same cell use a set of mutually orthogonal pilot sequences, while the same set of pilot sequences is reused among different cells. The pro-posed AMP algorithm allows UEs in different cells to use different random pilot sequences, and performs well under pilot contamination. Wyner’s 2-D hexagonal cell arrangement is established in this work, and an analysis is carried out for optimal cooperative cell number. Besides, the proposed AMP algorithm achieves great MSE than LS estimation under this model. However, simulation results show convergence problem in high Signal-to-Noise Ratio (SNR) regions when the pilot length is not enough. We proposed a state evolution analysis under this circumstance, and a sufficient condition for convergence is derived afterwards. Moreover, a MSE prediction of the proposed AMP algorithm is performed by the state evolution algorithm. Results show a relatively close MSE compared with the real one.