Research Projects

5G Networks
Description: The time for planning the next generation of cellular networks is now. Demand for wireless data is growing at such a rate that it is predicted that we will need 1000x today's capacity in just a few years time. The three big technologies required to do this are (i) millimetre wave communications (ii) small cells and (iii) massive MIMO. I have projects in all three areas. Millimetre wave communications offers the prospects of vast bandwidths, but propagation is very different than at conventional microwave frequencies. There is a huge difference between LOS and NLOS conditions and channel variations are much faster. This means that the whole cellular architecture will have to be rethought. Small cells and HetNets also lead to totally new architectures, with increased spatial re-use, but major challenges in interference management and cell association (see HetNets project below). Providing wireless backhaul into small-cell networks is another major challenge. A complementary way to increase spatial re-use is through massive MIMO. Here, the problems are coordinating beam-patterns in a multi-cell network, coordinating channel state information measurement and propagation across multiple cells, and pilot contamination. Interference alignment is an important concept in multi-cell MIMO networks, particularly in wireless backhaul, where channel variations are not onerous. 5G networks are also about integration: integrating different technologies (WiFi, femtocells, mmwave, cellular) and managing offload from one technology to another. This area of 5G really covers many different possible PhD projects, and you will focus on one area or another.

Heterogeneous Networks (HetNets)
Description: Femtocells and WiFi offloading are now becoming very popular, and when we consider cellular networks with a mix of scales, from tiny femtocell and picocells, to microcells and large macroscales, we get the exciting new area of HetNets. Relays can also be added to these networks to increase coverage. Femtocells in particular offer an ability to offload traffic from wireless onto, say, a user's home broadband ISP connection. Users can install a femtocell base station in a plug and play manner, much as we do with WiFi access points. There are major challenges in the design of these networks. Traditional cellular networks are carefully planned, but femtocells are typically not planned and better modelled as random. However, what models should we use? Networks should be optimized yet be self-organizing. Should femtocells be open access or closed access? A macrocell link may require a relatively large amount of transmit power to reach its destination, and so interfere very strongly at a nearby small-cell base stations. Problems in power control, signal processing, cell-site association, and intercell interference avoidance are all different for HetNets and must be revisited. Lots of challenging research problems in this area.

Cognitive Radio
Description: This project will examine ways to re-use radio spectrum more efficiently. We are installing a multi-user MIMO network on campus, which forms the primary network. The project is to design a cognitive radio network (CRN) that efficiently re-uses the spectrum of the primary network. The primary network carries real internet data, and has interesting time, frequency and spatial dynamics. The project involves spectrum sensing, channel state measurement, coordination, and interference avoidance. It concerns dynamic topology discovery and resource allocation.

Cooperative Base Stations
Description: There has been recent world-wide interest, both in academia and in standards, in the area of cooperative base stations. A particular interest for me is in the area of coordinated beamforming. It is well known that a base station equipped with multiple antennas can beamform so as to send to multiple users simultaneously. In coordinated beamforming, base stations in a cellular network share channel state information (CSI) in order to jointly design beamforming vectors, in a way that is better than what they could do on their own - beamforming that takes into account intercell interference. Much of the recent work assumes perfect CSI, but in practice CSI is imperfect and the quality of the CSI may vary from one base station to another (and as a function of the bandwidth of the network backhaul). Distributed solutions must be found to coordinated beamforming problems, and instantaneous CSI should be locally measurable.

Network MIMO and distributed antennas
Description: If the backhaul bandwidth is large enough, base stations can share user data as well as channel state information (CSI) and full network MIMO is possible. Although this idea goes back a long way, it is recently coming alive in practice, and there is a vast array of problems to explore. In conventional cellular networks, much of the focus is on the poor cell edge user who gets the weakest signal and the strongest interference. In network MIMO, how should we define cells, and who is the cell-edge user? There are many questions of an information-theoretic nature which are interesting but difficult to answer. There can be many antennas in relative proximity (microdiversity) as well as many distributed antennas (macrodiversity). With a huge number of antennas, can we open up the frequency spectrum and send at much higher frequencies than before? What are the energy savings from such an approach? How do we control such a complex system?

Compressive Sensing in Communications
Description: There has been a flurry of recent interest in the signal processing community in the area of "compressive sensing" or "sparse signal recovery". It has been known for years that sparsity in a signal reduces the number of samples required to represent it, but recent results on random sampling have initiated a surge of interest in this area. Sparse signals arise in many communications problems. Strong interference in a HetNet scenario may have sparsity in the time-frequency domain. The channel impulse response in underwater acoustic communication is typically very sparse. Much of the recent work in sparse signal recovery focuses on very general signal processing algorithms that work well in a wide variety of applications. This project will look at more particular problems in communication where special structure can be further exploited.

Networking models and algorithms for wireless networks
Description: The project will develop novel networking algorithms and protocols for wireless ad-hoc networks that exploit state of the art advances in physical layer communications. There have been huge advances in our recent understanding of physical layer communications, in areas such as multi-user MIMO, interference channels, cooperative communications, network coding, and multi-terminal information theory. Few of these advances have been incorporated into real-world wireless technology: At the network layer, the designs are usually based on models that trivialize the physical layer. On the other hand, much physical layer research ignores the problems of coordination in a random environment. This project will enrich networking theory and performance analysis using new physical layer models and techniques. Resource allocation algorithms will be designed which are distributed, based on locally available information, and with limited overhead requirements.

Networks with large propagation delays
Description: The effect of propagation delay is rarely considered in information-theoretic work, in contrast to the way it is considered in control-theoretic approaches to, say, congestion control for the Internet. This is perhaps due to the focus in information theory on point-to-point channels. However, recent work in multi-terminal information theory and network coding applies information-theoretic principles in a network setting, and therefore it is reasonable to consider the impact of propagation delays. Very recent work has suggested that large propagation delays can make a big difference to the capacity regions of networks. Applications include underwater communications, where propagation delays are very significant because the speed of sound is very slow. This project will investigate further the performance of networks with significant propagation delays.