Hybrid Precoding Based on Alternating Minimization for Millimeter Wave Systems

Millimeter-wave frequencies, which enable the usage of a greater spectrum than the existing cellular microwave bands, have attracted the researchers' attention significantly. However, the ever-increasing demand for higher data rates in the future 5G and 6G networks requires further improvement of spectral efficiency. Millimeterwave systems take advantage of the decrease in wavelength to employ antenna arrays consisting of a large number of antennas on both transmitter and the receiver. Due to the high hardware cost and high power consumption of the fully digital precoders used in conventional MIMO systems, it is not feasible to employ fully digital baseband precoders at the millimeter-wave MIMO systems. Hybrid analog/digital transceiver architectures, which were made up of digital baseband and analog RF precoders were recently proposed by many researchers to reduce the rather high power consumption. Since analog RF precoders are much cheaper and consume much lower powers, the cost of the system is also significantly reduced with the help of RF precoders while achieving a comparable fully-digital precoder performance. This thesis investigates the performance of the millimeter-wave massive MIMO systems with hybrid analog/digital architecture for several aspects. We propose a novel low-complexity alternating minimization algorithm based on the Barzilai-Borwein (BB) gradient algorithm to maximize the spectral efficiency in single-user millimeterwave systems under Gaussian noise and impulsive noise. It is aimed to minimize the Euclidean distance between the hybrid precoders and fully digital precoder using alternating minimization techniques for both scenarios. In the impulsive noise environment, a novel fuzzy logic-based decoder is also proposed to suppress the effects of impulsive noise. Simulation results demonstrate that the proposed BB method can achieve almost the same spectral efficiency as the competing methods despite its lower computational complexity. Furthermore, the proposed fuzzy logic-based filter successfully suppresses the impulsive noise effects and achieves a better bit error rate performance than the competing methods which also work efficiently in Gaussian noise.