Research On Channel Estimation And Detection Technologies For Massive MIMO Systems

Massive MIMO (Multiple-input Multiple-output) can achieve a significant performance imporvement in terms of reliability, spectral efficiency, energy efficiency, and channel capacity, and it has been widely recognized as one of the key technologies for the future fifth generation mobile communication systems. Both the forward precoding and backward detection at the BS (Base Station) require CSI (Channel State Information), so, the quality of the channel estimation result will directly affect the performance the whole communication systems. Meanwhile, the computational complexity will be a key factor that determines whether the signal processing algorithms are feasible for massive MIMO systems, where the BS will be equipped with a large number of antennas. Therefore, it is crucial important for massive MIMO systems about how to achieve high accuracy result about channel estimation and signal detection with low complexity. In this thesis, we mainly discuss the channel estimation technologies and signal detection technologies for massive MIMO systems.Pilot-based linear channel estimation algorithms require users to send a large number of pilots, which posses a low spectral efficiency and the performances of such algorithms are seriously limited by pilot contamination. Firstly, a recursive method to compute the ambiguity matrix is derived in this thesis. Then, the FRRH (Fast Recursive Row-Householder) subspace tracking algorithm is applied to reduce the computations in the estimation of signal subspace of the autocorrelation matrix of the receiving signals. Finally, a FRRH-based semi-blind channel estimation algorithm is given in this thesis. The proposed algorithm can significantly reduce the computations of semi-blind channel estimation algorithms when comparing with the EVD-based or SVD-based semi-blind channel estimation algorithms. Simulation results show that the proposed FRRH-based semi-blind channel estimation algorithm achieves good convergence performance, and it can converge to a const value with very little receiving samples. Also, it can obtain better estimation performance than the LS (Least-square) channel estimation algorithm and the EVD-based channel estimation algorithm, and the nearly similar performance when comparing with the SVD-based channel estimation algorithm. Meanwhile, the proposed algorithm can effectively relieve the effect of pilot contamination on the performance of channel estimation.In terms of signal detection technologies for massive MIMO systems, this thesis mainly discuss three linear detection algorithms such as MRC (Maximum Ratio Combining), ZF (Zero Forcing), MMSE (Minimum Mean Square Error), and present the lower bound of the achievable uplink rate for each user in each cell corresponding to each detection algorithm. In the simulations, the BER (Bit Error Rate) performances of the three linear detection algorithms are compared under the single-cell model and the multi-cell model respectively. The results show that, in the single-cell model, the BER of the three linear detection algorithms decrease significantly as the number of antennas at the BS becomes larger, the performances the three linear detection algorithms nearly are the same, and the ideal BER can be achieved at a low SNR. In the multi-cell model, the ZF detection algorithm obtains the similar BER performance with the MMSE detection algorithm, and both of them can achieve lower BER than MRC detection algorithm. While, the MRC detection algorithm can still achieve approximately BER performance with the ZF and MMSE detection algorithms, when the number of antennas at the BS is large and the SNR value is high.