Theses: Transmission Strategies for Interfering Networks with Finite Rate and Outdated Channel Feedback
The emergence of very capable mobile terminals, e.g. smartphones or tablets, has dramatically increased the demand of wireless data traffic in recent years. Current growth
forecasts elucidate that wireless communication standards will not be able to afford future traffic demands, thus many research efforts have been oriented towards increasing
the efficiency of wireless networks.
Wireless communications introduce many issues not present in wired systems, e.g. multipath effects or interference. Some of these issues may be tackled by the use of multiple
antennas, i.e. MIMO technologies. This solution allows increasing not only the reliability and robustness of the communications, i.e. the diversity gain, but also its efficiency, i.e.
the multiplexing gain or degrees of freedom (DoF). The DoF describe the slope of channel capacity at very high signal-to-noise-ratio (SNR) regime. For a point-to-point (P2P)
communication, assuming that the wireless channel response is Rayleigh distributed (typical in urban environments with no dominant line of sight propagation), it is known that
the DoF are equal to the minimum between the number of antennas at the transmitter and the receiver. Consequently, the throughput of a MIMO system may be scaled in a
promising way, considerably boosting the efficiency of the system.
However, the DoF behavior for multi-user networks is still an open problem in general. This thesis focuses on the problem of characterizing the DoF of such networks, where a
new problem comes into play: the interference. The most trivial way of tackling it is by means of orthogonalization of the transmission, i.e. the signals intended to each user
are transmitted along different time slots. Although completely avoiding the problem of interference, with this solution resources are exploited by each user only a fraction of time
inversely proportional to the number of users, i.e. each resource block is used only by one user.
As an alternative, the spatial dimensions provided by MIMO technologies may be exploited for orthogonalization,thus obtaining promising gains with reuse of time resources.
Nevertheless, this approach implies sacrificing spatial dimensions in pursuit of managing the interference, i.e. instead of delivering more data symbols. Consequently, it is only
advisable in case of an excess of antennas at the terminals. Since having a massive number of antennas may not be practical, especially at the user terminal, the interference
alignment (IA) principle stands as a candidate for such dimensionally-limited systems.
Introduced for the interference channel, with K transmitter-receiver pairs, a large number of applications of this concept have flooded the literature in the last years. In a
nutshell, the IA approach allows having interference at the receiver under the constraint that interference terms are observed overlapped, which is achieved by smartly designing
the transmitted signals. This way the number of dimensions occupied by interference can be reduced, thus increasing the number of symbols per user that can be delivered.
Following this idea, one of the first works on IA showed the surprising result that in an interference channel with an arbitrary number of transmitter-receivers pairs, “each user gets half of the cake”, where the cake refers to the number of degrees of freedom of the P2P channel.
At the beginning, the IA concept was exploited for scenarios where full channel state information is assumed to be available at the transmitter side (full CSIT), i.e. the information
is acquired instantaneously, and with perfect quality. The first part of this thesis studies this case and completes the DoF characterization of the 3-user MIMO interference
channel for some antenna configurations when channel coefficients are constant. For this case, the current approach in the literature, developed for time-varying channels, cannot
reach the optimal DoF. In this regard, the first part of the dissertation describes a linear precoding strategy able to attain them regardless the channel dynamics.
In practice, CSIT should be obtained from channel feedback after a pilot-based training period, thus incurring delays and errors. Although the latter issue has been extensively
studied, the requirement of having instantaneous CSIT is one of the critical aspects of IA.
This is of special interest for implementation on networks with high channel dynamics, appearing on scenarios with high mobility, where the delay associated to channel feedback
may be larger than the channel coherence time. In this regard, IA concepts were extended to networks with only past information of the channels, in a new framework known
as delayed CSIT. This alternative form of IA is denoted as retrospective interference alignment, since the transmission is carried out in multiple phases, and signals may be
aligned along space and the different phases. The second part of the thesis deepens into the DoF of two network topologies: the interference channel and interference broadcast
channel, for which we propose linear precoding strategies where design employs only delayed CSIT. Moreover, for the interference channel we derive DoF-delay trade-offs,
which are relevant as most strategies based on delayed CSIT entail long communication delays.
Finally, the last part of this thesis is focused on implementation issues. First, we define a common formulation relating the feedback procedure and settings to the feedback quality.
Then, we study the performance of one scheme proposed in the second part (TDMAgroups scheme, or TG) for the downlink of a wireless scenario in terms of mean and
outage rate, and with finite feedback parameters. After that, the effect of finite feedback quality is studied, by deriving its achievable DoF for the case of imperfect delayed CSIT.
The last contribution of the third part studies the net DoF, whereby the DoF are studied as a function of the coherence time (or user mobility), and taking into account all issues
related to channel acquisition at both the transmitter and receiver side: consumption of resources for feedback transmission, consumption of resources for channel training, and
feedback delays. Consequently, they represent the most accurate metric for analyzing the performance of transmission protocols. In this regard, we build two protocols upon the
TG scheme, and evaluate its net DoF performance as a function of user mobility