During the last years, mobile data traffic has skyrocketed, paralleling the devel- opment of the wired Internet traffic at the beginning of this millennium. Since 2008, the total global mobile data volume has more than doubled every year and is expected to grow at a similar rate in the future [1]. The main drivers of this massive growth are smartphones, tablet PCs, and laptops, whose promi- nent usage is about data rather than voice. Soon, there will be as many wireless devices as humans on earth. As a result of this development, current networks reach their capacity limits, especially in highly populated metropolitan areas. Congestion problems arise in the wireless and the backhaul networks alike.
At the same time, there is a growing concern about the possible effects of information and communication technology (ICT) on global carbon emissions [2,18]. Although ICT’s contribution to the global emissions is and will remain a rather small percentage of the global figures (with 1.25 % in 2002 and around 2.5 % in 2020), the general trend of a 10 % yearly increase in ICT-related carbon emissions is alarming. This means that, despite significant progresses in energy- efficient technologies, the growth in data traffic will outpace our ability to reduce or even maintain the overall energy consumption and related emissions. Thus, more network capacity on the one hand and less energy consumption on the other are two seemingly contradictory future requirements on ICT. This begs the question how mobile operators can satisfy the future traffic demands, both economically and ecologically.
Let us now have a look at several different ways how the capacity of wireless networks could be generally increased:
More spectrum
Essentially all commercial mobile communication systems are operated on fre- quency bands in the range from 300 MHz–3 GHz. This is due to the favorable radio propagation conditions at these frequencies. As a consequence, this part of the spectrum is almost entirely allocated and new spectral resources can only be found at higher frequencies. For this reason, there is a growing inter- est in millimeter-wave (mmWave) communication systems [3] which exploit the spectrum from 3–300 GHz. Today, mainly short-range indoor and fixed-wireless
communication systems are operated in these bands. Owing to high reflection and penetration losses, the communication at very short wavelength is more or less limited to line-of-sight (LOS) conditions. Providing indoor coverage from outdoor base stations (BSs) seems impossible. Nevertheless, there are current research activities which assess the feasibility of mmWaves for mobile commu- nications [4]. However, this research is still in its infancy.
Better coding and modulation schemes
With low density parity-check (LDPC) [5] and Turbo [6] codes, orthogonal frequency-division multiplexing (OFDM), link adaptation, and hybrid auto- matic repeat request (HARQ), modern mobile communication systems like LTE [7] or WiMAX [8] operate at spectral efficiencies close to the theoretical limits. Thus, revolutionary breakthroughs in coding or modulation are not to be ex- pected in the near future. Current research targets, for example, non-orthogonal modulation schemes which do not require a cyclic prefix, such as Isotropic Or- thogonal Transform Algorithm (IOTA)-OFDM [9], or techniques like Vander- monde frequency-division multiplexing (VFDM) [10] which exploit the cyclic prefix of OFDM to allow for interference-free underlay networks. Also rate-less or Fountain codes [11], which generalize in some sense the concept of HARQ, as well as Polar codes [12] promise further improvements in spectral efficiency. However, despite these efforts, advances in coding or modulation are unlikely to be the main capacity driver of future mobile communication systems.
More antennas
Since the seminal papers [13, 14] in the mid–1990s, multiple-input multiple- output (MIMO) wireless communications have attracted enormous interest from researchers and industry alike. By now, the benefits of multiple antennas at the transmitters and/or receivers are well understood [15]. MIMO techniques can provide power gains, improve the link reliability, and increase the throughput by multiplexing several independent data stream on the same time-frequency resource. Simple MIMO schemes are already an integral part of modern mobile communication standards [7, 8] which support today up to eight antennas at the BSs and the user terminals (UTs). Large-scale MIMO systems where BSs are equipped with hundreds of antennas are the subject of numerous ongoing research projects [16, 17, 18]. In theory, large-scale or “massive” MIMO can significantly reduce transmit powers while achieving high spectral efficiencies. Although not directly MIMO techniques, one needs to consider also higher- order cell-sectorization [19] and 3D-beamforming [20,21] as effective means to increase the spectral efficiency. These techniques aim at increasing the spatial reuse by focusing the transmitted energy towards the intended UTs and by reducing interference to other UTs at the same time.
More cells
It is well known that cell-size shrinking is the simplest and most effective way to increase wireless throughput [22, 23]. Physics tells us that bringing a radio transmitter and receiver closer together reduces the necessary transmit power to overcome path loss and other phenomena, such as fading and noise. More- over, the area throughput increases theoretically linearly with the cell density.
1.1. Mobile communications: Future challenges
However, the deployment of traditional BSs, such as macro and micro cells, requires huge capital and operational expenditures for cell-site acquisition, net- work planning, backhaul provisioning, operation, and maintenance. Hence there is a growing popularity of femto cells [24], i.e., user-deployed home-BSs utilizing the existing Internet connection as backhaul. Femto cells allow to offload traffic from the macro cells and to provide high-capacity indoor coverage at minimal cost for the network operator. Currently, also “small cells”, i.e., self-organizing operator-deployed outdoor/indoor femto cells, receive considerable interest from academia and industry as a promising means to provide localized high-capacity coverage at low energy-consumption and cost [25,26].
Cooperation and coordination
Cellular networks are first and foremost limited by intercell interference. If the BSs were allowed to cooperate or to coordinate, this interference could be ei- ther exploited or reduced [27]. One distinguishes generally between cooperation (also network MIMO or multi-cell processing), i.e., multiple BSs are connected together via backhaul links and jointly process their data, and coordination, i.e., clusters of BSs jointly decide on precoding/decoding strategies but generally do not share user data. Both techniques, commonly referred to as coordinated multi-point (CoMP), can improve the coverage and throughput, especially for cell-edge UTs, without the need to deploy new cell sites or additional anten- nas. However, many technical challenges need to be overcome before CoMP can be successfully introduced in practice [28]. Moreover, the effective gains of CoMP are less promising once the overhead for the acquisition of channel state information (CSI) is taken into account [29]. Other interesting techniques un- der current research are interference alignment schemes [30], relaying concepts [31], and wireless network clouds [32] where BSs are replaced by fiber-connected remote radio heads and the processing is carried out on centralized server-farms. Cognitive radio
“Most bands in most places are underused most of the time” [33, 34]. This observation stimulated the idea of cognitive radios [35] which are allowed to use parts of the licensed spectrum, given that they can reliably sense if it is currently used or not. In this context, one often speaks about “spectrum holes” in either time, frequency, or space which cognitive radios try to exploit. Despite the heavy research on this topic during the last decade, cognitive radios have not yet been successful in practice. This is mainly due to a lack of reliable sensing algorithms, protocols, hardware, and regulatory constraints. Thus, it is unlikely that cognitive radio will play a major role in next generation mobile communication systems. However, the concept of more “intelligent” or “flexible” radios, which moves away from the classical centralized network architecture to self-organizing networks with intelligent decision making at the nodes, is a promising idea [36].
Other techniques
There are of course many other techniques which provide interesting oppor- tunities for spectral-efficiency improvements. Among them are for example
full-duplex transceivers [37] which could theoretically double the capacity of cur- rent networks, a consequent exploitation of electromagnetic polarization [38,39] which could lead in principle to a threefold capacity increase, and cross-layer approaches like joint source-channel decoding [40] which exploits redundancy and side information at different protocol layers.
From the discussion above, it becomes clear that advanced mobile communi- cations systems are likely to consist of a dense deployment of different types of wireless access points (macro/micro/femto/small cells) with different character- istics (transmit powers, number of antennas, open/closed access, user/operator deployed). Additionally, CoMP techniques will be a desirable feature to miti- gate the increasing interference in such networks. In order to assess which of the possible techniques presented above is the most appropriate with respect to a given goal, it is necessary to provide a fundamental theoretical performance analysis. However, channel models and tools which were developed for the anal- ysis of simple point-to-point links often fail to provide meaningful insights for large, increasingly dense, heterogeneous networks. New theoretical tools for their analysis are needed and the development of these tools is the main goal of this thesis. In what follows, we will focus almost exclusively on the char- acterization of theoretical performance limits. Energy efficiency, although an important parameter, will not be considered in this work.