Derechos:
Atribución-NoComercial-SinDerivadas 3.0 España
Resumen:
Managing the interference is the main challenge in cellular networks. Multiple-Input
Multiple-Output (MIMO) schemes have emerged as a means of achieving high-capacity
in wireless communications. The most efficient MIMO techniques are based on managing
the iManaging the interference is the main challenge in cellular networks. Multiple-Input
Multiple-Output (MIMO) schemes have emerged as a means of achieving high-capacity
in wireless communications. The most efficient MIMO techniques are based on managing
the interference instead of avoiding it by employing orthogonal resource allocation
schemes. These transmission schemes require the knowledge of the Channel State Information at the Transmitter (CSIT) to achieve the optimal Degrees of Freedom (DoF),
also known as multiplexing gain. Providing an accurate CSIT in cellular environments
involves high-capacity backhaul links and accurate synchronization, which imply the use
of a large amount of network resources. Recently, a Blind Interference Alignment (BIA)
scheme was devised as a means of achieving a growth in DoF regarding the amount of
users served without the need for CSIT in the Multiple-Input Single-Output (MISO)
Broadcast Channel (BC). It is demonstrated that BIA achieves the optimal DoF in
the BC without CSIT. However, the implementation of BIA in cellular networks is not
straightforward. This dissertation investigates the DoF and the corresponding sum-rate
of cellular networks in absence of CSIT and their achievability by using BIA schemes.
First, this dissertation derives the DoF-region of homogenous cellular networks with
partial connectivity. Assuming that all the Base Stations (BSs) cooperate in order to
transmit to all users in the network, we proposed an extension of the BIA scheme for the
MISO BC where the set of BSs transmits as in a network MIMO. It is shown that the
cooperation between BSs results futile because of the lack of full connectivity in cellular
networks. After that, this dissertation presents several transmission schemes based on
the network topology. By differentiating between users that can treat this interference
optimally as noise and those who need to manage the interference from neighbouring
BSs, a network BIA scheme is devised to achieve the optimal DoF in homogeneous
cellular networks. Second, the use of BIA schemes is analyzed for heterogeneous cellular networks. It is demonstrated that the previous BIA schemes based on the network topology result nonoptimal in DoF because of the particular features of the heterogenous cellular networks. More specifically, assuming a macro-femto network, cooperation between both tiers leads to a penalty for macro users while femto users do not exploit the particular topology of this kind of network. In this dissertation, the optimal linear DoF (lDoF) in a two-tier network are derived subject to optimality in DoF for the upper tier. It is demonstrated
that, without CSIT or any cooperation between tiers, the lower tier can achieve nonzero
DoF while the upper tier attains the optimal DoF by transmitting independently of
the lower tier deployment. After that, a cognitive BIA scheme that achieves this outer
bound is devised for macro-femto cellular networks.
The third part of this dissertation is focused on the implementation of BIA in practical
scenarios. It is shown that transmission at limited SNR and coherence time are the
main hurdles to overcome for practical implementations of BIA. With aim of managing
both constraints, the use of BIA together with orthogonal approaches is proposed in this
work. An improvement on the inherent noise increase of BIA and the required coherence
time is achieved at expenses of losing DoF. Therefore, there exists a trade-off between
multiplexing gain, sum-rate at finite SNR and coherence time in practical scenarios. The
optimal resource allocation for orthogonal transmission is obtained after solving a very
specific optimization problem. To complete the characterization of the performance of
BIA in realistic scenarios a experimental evaluation based on a hardware implementation
is presented at the end of this work. It is shown that BIA outperforms the sum-rate
of schemes based on CSIT such as LZFB because of the hardware impairments and the
costs of providing CSIT in a realist implementation.[+][-]