Research On Key Technologies Of Channel Estimation In Massive MIMO Systems

The massive multiple-input multiple-output (MIMO) has been proposed as one of the key technolo-gies for the next generation wireless communication systems. There are lots of performance superiorities in massive MIMO systems. However, in practice, these superiorities is achieved only when the uplink and downlink channel state information (CSI) are known perfectly by both the base station (BS) and user-s. In frequency division duplex (FDD) massive MIMO systems, a large percentage of precious downlink resources is reserved by training sequence for channel estimation due to the large number of antennas. In this thesis, the downlink channel estimation and limited feedback in FDD massive MIMO systems are studied.Firstly, the development of traditional MIMO systems and the principle of massive MIMO systems are introduced. The channel estimation and the limited feedback in FDD massive MIMO systems are summarized. Generally, the CSI is acquired based on the training sequences in massive MIMO system-s. Some channel estimation schemes, such as Least Square (LS) method and Linear Minimum Mean Square Error (LMMSE) method are studied. Moreover, the fundamental principle of channel estimation in massive MIMO systems is analyzed. There are some problems need to be focused urgently, such as pilot contamination and reciprocity calibration.Secondly, a design method of optimal training sequence is studied based on the closed-loop down-link training scheme using Kalman filter in FDD massive MIMO systems with Uniform planar array (UPA) antenna structure. To design efficient training sequence, the channel statistics for FDD massive MIMO systems are exploited, which is derived from dynamic channel models and the UPA channel spatial correlation models. The correlated time-variation in the channel is exploited by adopting the Gauss-Markov channel model. The performance of channel estimation is enhanced by using the optimal Kalman filtering and prediction that exploits the corresponding training sequences received consecutive-ly. Moreover, the power allocation of training sequences for channel estimation is studied. Simulation results show that the proposed optimal training sequence based on the closed-loop training scheme out-performs the existing ones.Finally, the limited feedback techniques in FDD massive MIMO systems are studied. The vector quantization (VQ) schemes are used frequently in traditional MIMO systems. In this chapter, the VQ schemes based on random vector quantization (RVQ), unitary matrix and discrete fourier transformation (DFT) codebooks are analyzed. Limited feedback codebooks that adapt to spatially correlated channels are studied at the same time. The limited feedback based on trellis-coded quantization (TCQ) is studied at last. In this scheme, the encoding and decoding of convolution code with constellation are used for limited feedback. It is shown that the complexity of this scheme depends on the convolution code, and it is proportional to the number of transmit antennas. Simulation results show that the near-optimal performance can be achieved by the TCQ scheme with moderate complexity and feedback overhead.