الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 170 |  |  |
| | | from | 170 | | |
Abstract
Multiple-input multiple-output (MIMO) concept has been around
for decades since it can improves capacity and reliability of the wireless
systems. The initial work in MIMO focused on point-to-point MIMO
system, where we have two devices with multiple antennas communicate
together. The direction of research has shifted to multi-user
MIMO (MU-MIMO) which o ers big advantages over point-to-point
MIMO system. In MU-MIMO system, the base station (BS) with
multiple antennas simultaneously serves a set of single-antenna users.
In this way, expensive equipment is only needed at the BS side, while
single-antenna users are relatively cheap. Moreover, in MU-MIMO the
users are typically separated in space by many wavelengths. As a result,
the MU-MIMO system is commonly used in most communication
standards such as 802.11 (WiFi), 802.16 (WiMAX) and LTE. For most
MU-MIMO systems the number of BS antennas is few (i.e. fewer than
10 antennas), which limits the improvement in spectral eciency.
Massive MIMO (also known as ”Very Large MIMO”, ”Hyper MIMO”
and ”Large Scale Antenna Systems”) is a new trend of research in
which the BS is equipped with a very large number of antennas (e.g.,
100 or more) can serve a large number of users sharing the same system
resources (time, frequency and space, etc.). This is possible as the
BS is able to decrease (asymptotically eliminate) the user interference,
as well as the e ect of the small-scale fading and the random additive
noise. Massive MIMO can theoretically o er advantages in terms of
energy and spectral eciencies. In order to achieve these promised
gains, the channel state information (CSI) is required for signal detection
and/or precoding. Training sequences (pilot signals) are used to
acquire these CSI, however, the number of pilot signals is limited as it
depends on coherence time and coherence bandwidth (where channel
i
coherency depends on the propagation environment, the carrier frequency,
and user mobility). In order to serve a large number of users
simultaneously, the pilot sequences are reused through the cellular
system causing the well-known problem of ”pilot contamination”.
In this thesis, we propose two methods to decrease pilot contamination.
1. A structure, i.e. non-random pilot hopping and a weighted moving
average channel estimation is proposed for time-varying massive
MIMO systems. The potential spectral eciency gain is
obtained by combining pilot reuse and a structure pilot hopping
strategy. The proposed algorithm allows the users to experience
a di erent set of contaminating users across time and therefore
make it possible to the BS to average out the interference e ect.
The performance of multi-cell network where the BS performs
matched ltering (MF) or zero forcing (ZF) for data detection
and least square (LS) or minimum mean square error (MMSE)
for channel estimation is analyzed. Using polynomial expansion
approximations and massive MIMO properties, we obtain
achievable rates that are tight even at nite (not very large)
number of antennas. Moreover, the proposed algorithm involves
a weighted moving average that incorporates the past channel
estimation by a correlation factor to improve the channel estimation
of the current one. Due to the massive MIMO properties,
the algorithm is robust to channel time-variations and can be
used with any previously proposed channel estimation method.
Our analytical results are nally illuminated by means of Monte
Carlo simulations that show the e ect of the main system variables
in various cases, e.g. small/large number of antennas, number
of users, Doppler frequencies and the signal-to-noise ratio
(SNR).
2. Increasing the number of antennas at user side (two-antenna
users) and proposing a simple data-aided channel estimation
ii
method for the extra antenna. We show that increasing the
number of antennas at the user side is possible without increasing
the channel estimation overhead linearly with the number
of extra antennas. A semi-blind channel estimation for multiantenna
users is proposed where a pilot sequence is used for
only one antenna whereas the other antenna(s) use a data-aided
channel estimation so there is no overhead in terms of pilot symbols.
The extra antenna may be exploited in di erent schemes:
data multiplexing (i.e. each user transmits two independent signals
from its two antennas), spatial repetition (i.e. the second
antenna transmits the same signal as the rst antenna) and interference
mitigation using interference-alignment (IA) precoding.
It is shown that data multiplexing scheme increases the achievable
rate. However, spatial repetition case has the e ect of virtually
increasing (doubling) the number of antennas at the base
station compared with the conventional single-antenna users, the
third precoding (IA) scheme outperforms the second scheme in
terms of spectral eciency as it e ectively decreases the size of
the interference space. Using polynomial expansion approximations
and new lower bounds, we obtain achievable rates when
matched ltering detection is used in uplink multi-cell MIMO
system. The rate bounds are tight even at nite (not very large)
number of antennas. The simulation results show the gain in
achievable rate and the number of the base station antennas
that can be saved by adding additional antenna at each user
compared to the single-antenna user system. The proposed algorithm
is applicable for more than two antennas.