Energy efficient and low complexity techniques for the next generation millimeter wave hybrid MIMO systems
View/ Open
Date
31/01/2020Author
Kaushik, Aryan
Metadata
Abstract
The fifth generation (and beyond) wireless communication systems require increased
capacity, high data rates, improved coverage and reduced energy consumption.
This can be potentially provided by unused available spectrum such
as the Millimeter Wave (MmWave) frequency spectrum above 30 GHz. The high
bandwidths for mmWave communication compared to sub-6 GHz microwave frequency
bands must be traded off against increased path loss, which can be compensated
using large-scale antenna arrays such as the Multiple-Input Multiple-
Output (MIMO) systems. The analog/digital Hybrid Beamforming (HBF) architectures
for mmWave MIMO systems reduce the hardware complexity and power
consumption using fewer Radio Frequency (RF) chains and support multi-stream
communication with high Spectral Efficiency (SE). Such systems can also be
optimized to achieve high Energy Efficiency (EE) gains with low complexity but
this has not been widely studied in the literature. This PhD project focussed on
designing energy efficient and low complexity communication techniques for next
generation mmWave hybrid MIMO systems.
Firstly, a novel architecture with a framework that dynamically activates the
optimal number of RF chains was designed. Fractional programming was used
to solve an EE maximization problem and the Dinkelbach Method (DM) based
framework was exploited to optimize the number of active RF chains and the data
streams. The DM is an iterative and parametric algorithm where a sequence of
easier problems converge to the global solution. The HBF matrices were designed
using a codebook-based fast approximation solution called gradient pursuit which
was introduced as a cost-effective and fast approximation algorithm. This work
maximizes EE by exploiting the structure of RF chains with full resolution
sampling unlike existing baseline approaches that use fixed RF chains and aim
only for high SE.
Secondly, an efficient sparse mmWave channel estimation algorithm was developed
with low resolution Analog-to-Digital Converters (ADCs) at the receiver.
The sparsity of the mmWave channel was exploited and the estimation problem
was tackled using compressed sensing through the Stein's unbiased risk estimate
based parametric denoiser. The Expectation-maximization density estimation
was used to avoid the need to specify the channel statistics. Furthermore, an
energy efficient mmWave hybrid MIMO system was developed with Digital-to-
Analog Converters (DACs) at the transmitter where the best subset of the active
RF chains and the DAC resolution were selected. A novel technique based on the
DM and subset selection optimization was implemented for EE maximization.
This work exploits the low resolution sampling at the converting units and provides
more efficient solutions in terms of EE and channel estimation than existing
baselines in the literature.
Thirdly, the DAC and ADC bit resolutions and the HBF matrices were jointly
optimized for EE maximization. The flexibility in choosing the bit resolution
for each DAC and ADC was considered and they were optimized on a frame-by-frame
basis unlike the existing approaches, based on the fixed resolution sampling.
A novel decomposition of the HBF matrices to three parts was introduced to
represent the analog beamformer matrix, the DAC/ADC bit resolution matrix and
the baseband beamformer matrix. The alternating direction method of multipliers
was used to solve this matrix factorization problem as it has been successfully
applied to other non-convex matrix factorization problems in the literature. This
work considers EE maximization with low resolution sampling at both the DACs
and the ADCs simultaneously, and jointly optimizes the HBF and DAC/ADC bit
resolution matrices, unlike the existing baselines that use fixed bit resolution or
otherwise optimize either DAC/ADC bit resolution or HBF matrices.