Array Processing and Precoding Design for Next Generation of Wireless Communication Systems

Abstract

The next generation of wireless communication systems promises to provide a better user experience in terms of high data rates, coverage, reliability, and energy efficiency. One of the competing candidates is the viable combination of millimeter-wave (mmWave) with the introduction of a large number of antennas. On one side, the use of mmWave will facilitate the deployment of a large number of antennas but on the other side will impose a challenge of energy-efficient hardware implementation. Therefore, in addition to spectral efficiency, energy efficiency will be an important design goal. Introducing a large number of antennas at the base station (BS) will also complicate the channel parameter estimation. The channel parameter estimation must be obtained with high-resolution at the user equipment (UE), because these parameters need to be quantized before being sent back to the BS for precoding. If the channel parameters are not estimated with high accuracy, the BS will receive the erroneous parameters with additional quantization errors, resulting in deterioration of performance. This thesis presents an energy-efficient solution to overcome the challenge of hardware implementation due to the introduction of a large number of antennas by introducing Butler matrix (BM) in the analog domain using partially connected analog phase shifting (PCAPS) approach. The deployment of BM improves the hardware implementation but makes the channel parameter estimation and hybrid precoding more challenging. To cater to these problems, maximum likelihood (ML) estimator is initially derived for frequency flat fading channels, while a two-stage approach is designed for one-dimensional parameter estimation assuming frequency selective channels. The first stage is accomplished by proposing parameter estimation based on a DFT grid (PREIDG) algorithm to find the coarse estimates, which is used to initialize the space alternating generalized expectation-maximization (SAGE) algorithm to get ML estimates of the parameters. Furthermore, the problem is extended to two-dimensional parameter estimation, which is solved by the two-stage algorithm. In the first stage a modified PREIDG is proposed to perform coarse estimation which is used to obtain the high-resolution estimates of the parameters using the SAGE algorithm in the second stage. The performance of the parameters estimation algorithms is assessed by deriving Cramer-Rao lower ´ bound (CRLB). Finally, the analog and baseband algorithm is obtained using hybrid beamforming (HBF)-weighted minimum mean square error (WMMSE) method.