5G: The Convergence of Wireless Communications

The ongoing deployment of the fourth generation (4G) of wireless mobile systems has prompted some telecommunication companies to consider further development towards fifth generation (5G) technologies and services. Since the appearance of the first generation (1G) system in 1981, new generations have emerged approximately every 10 years. Wireless communication generations typically refer to non-backwards-compatible standards following requirements specified by the International Telecommunication Union-Radiocommunication Sector (ITU-R). Examples of these specifications are International Mobile Telecommunications 2000 (IMT-2000) for 3G and IMT-Advanced for 4G.

Other standardization bodies like IEEE also develop wireless communication technologies, often for higher data rates albeit shorter transmission ranges in most cases. With these additional standards, it is common that predecessor systems occur on the market a few years before new cellular mobile generations. For instance, commercial mobile Worldwide Interoperability for Microwave Access (WiMAX) networks deployed since 2006 are considered predecessors to 4G. Later, first-release Long-Term Evolution (LTE) systems were marketed as 4G in Scandinavia and the United States in 2009 and 2010, respectively. However, these networks were not strictly compliant with the original IMT-Advanced technical requirements. In contrast, the latest 4G LTE-Advanced (LTE-A) systems will fully comply with the ITU-R specification and support peak downlink data rates of 100 Mbps and 1 Gbps for vehicular and pedestrian mobility, respectively. Future 5G systems are expected to provide significant gains over 4G like higher data rates, much better levels of connectivity, and improved coverage. An overview of the technical solutions considered for the “beyond-4G” cellular system is provided in, e.g., [3]. Various aspects of the next generation networks have been addressed in recent papers, e.g., [1, 14, 40, 43, 58]. Finally, comprehensive vision of the 5G networks and even beyond is presented in [60].

Nevertheless, there is still some debate on the technical characteristics that 5G must possess and the services it will provide. 5G is not a term officially used in any particular specification or in any official document made public by ITU-R or any standardization body. Moreover, some industry representatives have expressed skepticism towards 5G. Others, however, envisage the rollout of 5G for the early 2020s [52]. Enabling technologies for the next generation have started to be researched by consortia comprising key international mobile operators, network infrastructure manufacturers, and academic institutions. For instance, in October 2012, the University of Surrey in the United Kingdom secured £35 million in funding for the 5G Innovation Centre, 5GIC (http://​www.​surrey.​ac.​uk/​ccsr/​business/​5GIC/​), which offers testing facilities to mobile operators developing more energy and spectrum-efficient technologies for data rates beyond 4G capabilities.

In November 2012, the EU Project “Mobile and wireless communication Enablers for the Twenty-twenty Information Society,” METIS 2020 (please see Table 1), started activities toward an international definition of 5G. In February 2013, the European Commission announced €50 million for research to deliver 5G mobile technology by 2020. Some recent EU-funded research projects that address the architecture and functionality for networks beyond 4G are listed in Table 1.

On the regulatory side, the ITU-R Working Party 5D (WP 5D) started two study items in February 2013, aimed at obtaining a better understanding of future technical aspects for mobile communications towards the definition of the next generation: (1) a study on IMT vision for 2020 and beyond, and (2) a study on future technology trends for terrestrial IMT systems. In light of these developments, the European COST Action¹ IC0905 “Techno-Economic Regulatory framework for Radio spectrum Access for cognitive radio/software defined radio,” TERRA (http://​www.​cost-terra.​org/​), commenced activities to identify the role and impact that emerging radio technologies will have on the development of the next generation of mobile communications. This article provides an overview of these selected key technologies and the way they will facilitate the realization of 5G, along with some of the research challenges that must be addressed in the coming years. It has to be noted that the technologies described in the paper do not cover all solutions that could be considered for future wireless systems, and one can find numerous excellent proposals in the vast literature, e.g., [3]. In our work, we address the solutions that are in line with the vision of 5G network we present in the following chapter. The remainder of this paper is organized as follows; as mentioned before—our vision for 5G systems is described in Sect. 2, enhanced modulation schemes for 5G are the topic of Sect. 3, and traffic offloading techniques are presented in Sect. 4 as the alternatives for complementing future cellular networks. The roles of cognitive radio, software-defined radio, and software-defined networking as 5G-enabling technologies are outlined in Sects. 5, 6, and 7, respectively. A discussion on 5G infrastructure is presented in Sect. 8 and we conclude in Sect. 9.

One central topic of debate is what the architecture of 5G systems will look like. The evolution from 1G to 4G was characterized mainly by a shift in the channel access method, i.e., FDMA \(\rightarrow \) TDMA \(\rightarrow \) CDMA \(\rightarrow \) OFDMA, in conjunction with improved modulation and coding schemes. However, it is likely that 5G will not be a single air interface based on a single radio access technology (RAT) on the model of the previous generations. As modern communication systems approach the Shannon limit [19] and a wider range of devices demand wireless connectivity, some industry representatives envisage 5G as a “network of networks,” i.e., a heterogeneous system comprising a variety of air interfaces, protocols, frequency bands, access node classes, and network types [20]. In that vein, 5G is not really about the deployment of a completely new single-technology solution (something that will replace LTE, for example), but how existing technologies and spectrum-usage schemes can be better combined. If 5G reflects this prognosis, one of the major challenges will be the seamless integration of evolved versions of currently existing wireless technologies and complementary new communication networks.

Solid spectrum access methods and RRA algorithms will be crucial for efficient spectrum management in evolved RATs for 5G. The requirements regarding the permitted power levels transmitted in and outside the nominal band in the form of spectrum masks and acceptable interference power induced into neighboring systems have an impact on radio resource management (RRM). Such issues are of particular importance in the context of carrier-aggregation-like approaches (considered for LTE/LTE-A systems [44]), or in scenarios where the coexistence of various wireless systems in the same geographical area is allowed. The latter case has been widely discussed in the context of cognitive radio networks [39, 55, 65]. Based on the obvious observation that better spectrum utilization can be achieved when signals are closer to each other in the frequency domain, future 5G systems must attempt to minimize the allowed level of out-of-band (OOB) and spurious emissions. Moreover, in non-contiguous multicarrier transmission schemes where a small portion of subcarriers can be switched-off, e.g., non-contiguous orthogonal frequency-division multiplexing (NC-OFDM) [49, 72], the need for efficient and significant reduction of the power emitted in narrow spectrum gaps is particularly important. Such a scheme can be seen as a special case of CA for narrower sub-bands. For instance, let us consider the requirements defined by the Office of Communications in the United Kingdom [57] for the so-called White Space Devices (WSD). Although the values provided therein are strictly devoted to cognitive-radio-based solutions, it is wise to analyze these data and project them into the guidelines for 5G systems, where the reduction of the interference observed by neighboring users must be as high as possible. In [57] it is stated that the out-of-band equivalent isotropically radiated power (EIRP) spectral density, \(P_{\mathrm{OOB}}\), expressed in terms of dBm per 100 kHz, must satisfy the following relation:where \(P_{\mathrm{IB}}\) is the in-block EIRP spectral density and A represents the adjacent frequency leakage ratio (AFLR), which varies from −74 to −54 dB, depending on the WSD class. These values have been adopted by the European Conference of Postal and Telecommunications Administrations-Electronic Communication Committee (CEPT ECC) and released in a dedicated report for WSD requirements, ECC Report 185 [18], and are proposed in the draft standard for WSD by the European Telecommunications Standards Institute (ETSI) [22]. Similar values have also been identified by the Federal Communications Commission (FCC) [26], where the fixed adjacent channel emission limit is −72.8 dB in a 100 kHz bandwidth, measured relative to the total in-band power over a 6 MHz bandwidth. One can notice that the requirements put on WSD are higher than the corresponding parameters in technical standards for current cellular systems [21]. Therefore, it can be stated that effective solutions that will guarantee a strong reduction of the power radiated outside the nominal frequency band will be applied not only for cognitive radio systems, but can—or rather should—be considered for the future 5G networks. Such an approach will lead to a lower interference inducement into adjacent channels and, in consequence, to a better utilization of the radio resources.

However, establishing such challenging restrictions on the interference induced to the adjacent band will, on the one hand, improve the spectral efficiency, but on the other, will entail the application of spectrally efficient modulation schemes and sophisticated transceiver front-ends. This is due to the fact that there are two main sources of unwanted power emission outside the nominal band: (1) the modulation process resulting in high sidelobes in OFDM systems, and (2) nonlinear phenomena appearing in the terminal front-end, e.g., the influence of high power amplifiers that possess non-linear processing characteristics; other electronic components should also be considered [31]. The latter strongly depends on the physical modules implemented in the device front-end. It has been shown that quite effective hardware solutions are already available [37], i.e., the level of OOB power, measured very close to the nominal band, can be reduced to the level of −100 dBm per 100 kHz, even for signals occupying the frequency band of 30 MHz. Therefore, it is more critical to focus on the first source of unwanted OOB emission and strive to improve the modulation processing.

It is evident that the application of sophisticated PHY techniques can result in a better protection of the transmitted data, as well as the reduction of OOB emission levels, thereby ensuring a better exploitation of the spectrum and creating new degrees of freedom for its management. In order to guarantee a high spectral efficiency together with a relatively low complexity of the receiver structure, OFDM has been applied to many wireless systems, including 4G LTE. Nevertheless, one of the main drawbacks of OFDM is a high OOB power emission due to high sidelobes of the sinc-like transfer function of the transmit pulse. This undesirable phenomenon can be mitigated by extensive filtering and the use of sidelobe suppression algorithms, but even after applying such techniques, the amount of power radiated outside the nominal frequency band may remain high. To help address this issue, one technique involves extending the pulse duration to allow for smoother transitions between consecutive symbols. This is illustrated in Fig. 1 where the rectangular pulse is replaced by a square root raised cosine pulse with the time support of \(T_s + 2T_p\) [67]. However, the cost associated with the extension of the duration of the time-domain symbol (the presence of a symbol prefix and postfix), is the reduction of the overall system efficiency (e.g., the rate expressed in bit/s/Hz/J). Another approach is to perform advanced filtering. However, it has to be taken into account that such signal processing improves the OOB reduction at the expense of some filtering distortions; an interesting analysis of the trade-off between interference and filtering distortions in OFDM systems can be found in [53]. Furthermore, the efficiency and the cost of various approaches are discussed in [24, 71]. These observations lead to the conclusion that new modulation processing technique should be applied; such techniques should guarantee a similar efficiency and robustness as OFDM, while outperforming the existing solutions in terms of OOB reduction. In this context, new modulation schemes have been promoted for mobile systems beyond 4G [30, 51], and cognitive radio solutions, e.g., the IEEE DySPAN-SC P1900.7 standard. Since rectangular (or almost rectangular) pulses, though simple in implementation and analysis, do not guarantee a low OOB emission, alternative solutions involve shaping the transmit and received pulses to result in a better energy concentration in the time-frequency plane. Let us first generalize multicarrier transmission by providing a generic signal description, and then indicate some of the most promising solutions that might be considered for the next generation of cellular networks. We denote g(t) and \(\gamma (t)\) as the pair of transmit and receive pulses. Thus, the generic transmitted multicarrier signal s(t) and the estimation \(\hat{c}_{m,n}\) of the user data, \(c_{m,n}\) (e.g., Quadrature Amplitude Modulation (QAM) symbols), can be defined as:andwhere M is the number of subcarriers, T is the time distance between consecutive pulses, F represents the distance between two neighboring pulses in the frequency domain, and \(g_{m,n}(t)\) is the translated and modulated version of the original pulse, \(g(t)\). Thus, the generic time-frequency representation of the transmit frame consisting of \(M\cdot N_{\mathrm{tot}}\) pulses can be illustrated as in Fig. 2. In the OFDM case, the shape of both the transmit and received pulses is rectangular, the distance between the symbols in the time domain is set to \(T_S\), and the distance between the adjacent subcarriers is reciprocal to the orthogonality time of the OFDM symbol. Relaxing these constraints, one can conclude that in the design process of the transmit-received pulse, various parameters have to be considered, such as the time and frequency support of the pulse, overlapping of pulses, and distance between pulses in both dimensions. Beside those parameters, also other criteria can be considered, such as the minimization of the minimum mean-square error (MMSE) of the assumed channel model [42], and others [9, 27, 46]. In the context of the reduction of OOB emission, the steepest possible power attenuation in the frequency domain is crucial, while maintaining the orthogonality or biorthogonality between pulses [27]. Intensive research has been done over the last decade in the area of efficient pulse shaping, and many newly proposed schemes offer clear advantages over purely rectangular shapes. Currently, among the various proposals that have been investigated, a class of pulses strictly devoted to filter bank-based multi-carrier (FBMC) systems seems quite promising [62]. Various pulse shapes have been considered for this modulation scheme [7, 23, 62, 64]. It has been shown that FBMC systems with isotropic orthogonal transfer algorithm (IOTA) filters can achieve an OOB attenuation at the level of −60 dB relative to the transmit power level in the nominal band. Moreover, good time-frequency localization of the pulses allows for removing the cyclic prefix typically applied in OFDM in order to combat the inter-symbol interference (ISI) problem. As a consequence, a higher spectral efficiency with a very low interference power induced to neighboring systems can be achieved at the expense of a slightly higher transceiver structure complexity. The FBMC transceiver could be implemented very effectively by means of inverse fast Fourier transform (IFFT) blocks followed by polyphase filters. However, despite its benefits, FBMC suffers from various other problems that still require attention, e.g., sensitivity to phase-noise or so-called intrinsic interference and its influence on signal detection [76]. A detailed analysis of the advantages and disadvantages of this solution has been presented in [62], and in extensive reports published within the framework of the PHYDYAS (http://​www.​ict-phydyas.​org/​) and EMPhAtiC (http://​www.​ict-emphatic.​eu/​) projects. In Table 3, we briefly summarize the main features of OFDM and FBMC signals in terms of their potential application in 5G systems.

Another promising modulation scheme is NC-OFDM. In contrast to traditional OFDM signals, NC-OFDM enables a subset(s) of subcarriers to be switched off, allowing other signals to be transmitted in the vacated frequency bands. A low interference between neighboring signals can be achieved very efficiently by applying cancellation carriers (CCs) or windowing algorithms [49]. In such a case, the OOB emission can be minimized up to −60 dB in very narrow frequency gaps, where the power spectral density (PSD) of NC-OFDM signals with 320 subcarriers is presented. If implemented, it is conceivable that the above modulation schemes would serve as a supporting technology for RRA algorithms in 5G networks.

We will now outline the research activities focusing on generalized multicarrier schemes, where the constraints assumed in OFDM and even in FBMC schemes, are relaxed. The pair of the transmit and receive pulses, as well as the shape of the transmit frame, are designed in such a way that some specific predefined criteria are fulfilled. In [28, 38, 47, 59, 61, 63], the advantages of advanced waveforms are highlighted. In the context of 5G, which envisages the coexistence of various systems in close vicinity, together with the need for a better utilization of the frequency resources, the necessity of applying advanced waveforms is crucial. Although various proposals have been presented in the vast literature, FBMC and NC-OFDM schemes seem to be the most promising solutions at the current stage of development. The application of these waveforms will guarantee the fulfillment of the challenging OOB requirements (mentioned at the beginning of this section) that should be defined for 5G mobile terminals, as it has been done for WSD. At the same time, the spectral efficiency of these new modulation schemes will be kept at a very high level usually, outperforming the efficiency of today’s OFDM solutions. Such an observation is confirmed not only by the research community, but also by the recent standardization activities. For two-tier heterogeneous networks (HetNets), like macrocell/femtocell layouts, the recently proposed Vandermonde-subspace frequency division multiplexing (VFDM) [10] overlay spectrum sharing technique has the potential to provide a spectral efficiency increase of up to 1 bit/s/Hz in HetNets equipped with cognitive radio capabilities. Thus, new sophisticated modulation schemes are the key evolution technologies toward 5G networks.

Smartphones, tablets, and other mobile broadband devices generate unprecedentedly large amounts of traffic. Mobile operators are facing great challenges to serve such a tremendous traffic growth with the current cellular infrastructure. It is evident that mobile networks beyond 4G will need to implement sophisticated traffic offloading strategies [2] beside the traditional network scaling of deploying more base stations per area. Traffic offloading consists in using complementary radio access networks to deliver data originally intended for mobile cellular systems, thereby decreasing the congestion on each individual radio link and respective backbone connection. Traffic offloading encompasses multiple solutions, which can be classified as overlay (sharing the same spectrum with the cellular access network) and non-overlay solutions; some of them are discussed below.

Cognitive radio (CR) has long been considered an enabling technology for the next generation of mobile communications [6]. The CR paradigm proposes an opportunistic utilization of the underused parts of licensed frequency bands, i.e., spectrum holes [68], by unlicensed (secondary) users and/or the efficient sharing of the licensed-exempt spectrum. For this sake, cognitive mobile terminals must acquire accurate real-time knowledge on transmission opportunities through RF spectrum scanning to identify the unoccupied radio channels/bands within the time-frequency resources table. Standards like IEEE 802.11af, IEEE 802.16h (cognitive WiMAX), and IEEE 802.22 are aimed at using CR techniques to enable sharing of the TVWS spectrum on a non-interfering basis with improved coexistence mechanisms.

The CR technology is now at an advanced stage and real-world trials and evaluations are being carried out. As an example, an experimental trial of CR and the dynamic spectrum access technology using the TVWS spectrum is being carried out in Ireland. This Irish research testbed is assessing the feasibility of taking an opportunistic spectrum usage approach in the 700 MHz band to establish an 18.3 km TVWS link connecting the Trinity College Dublin in Dublin city center to Intel Labs Europe (Ireland) based in Leixlip, Co. Kildare. Demonstrating the spectrum agility and the ability to avoid interference to incumbents in order to provide a robust link suitable for the support of new services is the key objective of this trial [56]. In future 5G systems, CR will be a key component to ensure interference coordination in 3GPP-compliant overlay networks as well as in different traffic offloading techniques. The 3GPP LTE standard has taken provisions to facilitate interference mitigation in overlay networks. Enhanced Intercell Interference Coordination (eICIC) was introduced in 3GPP Release 10 LTE-A to tackle interference issues in HetNets. In this approach, Almost-Blank-Subframes (ABSs) are introduced, which are mostly control channel frames with a very low power. A macrocell can configure ABSs, so that mobile terminals in small cells can send their data within ABSs to avoid interference with overlay macrocells. The use of CR can improve this approach resulting in a more efficient utilization of the radio resources with an even better interference coordination for two-tier HetNets, which ultimately leads to an enhanced system capacity and coverage [13]. CA for HetNets developed in compliance with 3GPP Release 12 LTE-A will introduce new CR-wise techniques for efficient spectrum allocation in heterogeneous wideband fading environments.

The future 5G network infrastructure must have a sufficiently wide flexibility in order to cover different radio frequency (RF) parameter settings in a dynamic and adaptive way to allow efficient spectrum management. Reconfigurable platforms based on software-defined radio (SDR) will facilitate the dynamic air interface reconfiguration of the network nodes by software modifications, reflecting current traffic demands. The flexibility of the RF chains must be further reflected in the baseband processing capabilities, where the down-converted RF signals are to be processed. As 5G systems will need to exploit underutilized frequency bands to avoid the foreseen spectrum crunch, CR implementation on SDR platforms should consider the cooperation and interoperation of multiple radio technologies, e.g., through common radio resource management (CRRM). This implies that a reconfigurable platform must be capable to operate at different power levels, channel bandwidths and frequencies, modulation and coding schemes, modifying transmission parameters and characteristics according to the particular constraints of the radio technology standards in use; these constraints include unwanted emission in the operating band, AFLR, or intermodulation products. Promising SDR development initiatives include the GNU Radio (http://​gnuradio.​org/​), an open-source software development toolkit for SDR implementation on various programmable platforms like the Universal Software Radio Peripheral (USRP) boards (http://​www.​ettus.​com/​), and the Open Base Transceiver Station (OpenBTS), which was recently used to demonstrate the implementation of a software-based Global System for Mobile Communications (GSM) base station on the Raspberry PI hardware platform (http://​openbts.​org/​).

Looking at recent developments for LTE-A, the CA feature [70] with its maximum configuration of \(5\times 20\) MHz radio channels is a good starting point for future research on the incorporation of cognition, to further improve the flexibility of radio resources utilization. Nevertheless, the RF chains bandwidth and their tuning are considered here as a bottleneck, especially when a variety of different frequency bands are used for cellular networks in different markets. The multi-standard radio base station (MSR-BS) addresses these issues in the recent releases of 3GPP networks, enabling a single radio unit in the base station to simultaneously operate with multiple RATs at the same time without harming each other with unwanted RF emissions.

At the current stage of development, both CR and SDR technologies do not involve the control of the cellular core network. Until now, no coordination of traffic flow is possible at the core network, e.g., a UE cannot receive multiple different traffic flows from different eNBs simultaneously to increase the data rate. The revolutionary concept of software-defined networking (SDN) aims at providing a coordinator that has a global view of the network infrastructure, thereby facilitating a number of networking functionalities. In essence, the SDN concept consists in decoupling the forwarding plane and the control plane [48]. In this approach, the core devices, e.g., routers, do not make decisions on how and where to forward a given data packet; instead, the decision is taken by a central coordinator referred to as the controller. One of the protocols in charge of the communication between the controller and user devices is OpenFlow [54]. Data packets passing through interconnection devices (switches, routers, etc.) are classified into flows to make per-flow forwarding decisions. A flow is defined by a set of matching rules in 12 distinctive fields of a typical Ethernet/IP/UDP header (L2 and L3 addresses, virtual LAN (VLAN) information, ports, etc.). Each time a data packet affiliated to a specific flow enters a device, a counter is updated at the controller. This enables the controller to have a global view of the status of each network component. Therefore, the controller can make the decision to forward the traffic through a less congested route, or use a radio channel that momentarily experiences a good condition to forward the data packet to the end user [11]. SDN aims to support a much better integration of all the existing wireless networks (Wi-Fi, 2–4G, etc.) [74]. It would be possible to perform a seamless handover, not only within the same technology as already exists, but also across h-RATs. This feature will enable the offloading approaches described in Sect. 4. Moreover, SDN will greatly facilitate the management of chaotic deployments of a large number of small cells in LTE systems [34]. The ultimate goal of SDN is to completely abstract the forwarding plane to enable “network programmability” where the infrastructure is seen as a service that can be used by different applications above an intermediate layer referred to as the Network Operating System (NOS). This is the basis for the concept of cloud networks. In the context of 5G, we can refer to this approach as cloud-RAN or RAN as a Service (RANaaS).

5G systems will leverage the availability and variety of the deployed infrastructure to better address the need for capacity, coverage, and high quality of service instead of relying on a single technology, topology, or regulatory approach [29]. In other words, unlike 4G, which involves exclusively-licensed LTE in a predominantly cellular-based topology, 5G will focus on knitting together cellular, local and personal area networks, short range device technologies and topologies. This concept also extends to the regulatory domain; 5G communications will encompass a range of exclusive, light-licensing, license-exempt, and shared spectrum licensing schemes. Through this approach, the flow of data can be more efficiently abstracted from the physical infrastructure and regulatory approach. This approach is likely to increase in dominance over the coming years due to two main reasons: (1) it is unlikely that a single technology/standard will be able to provide a sufficient capacity and coverage to accommodate all market sectors; (2) networks seeking to maximize the utility of existing infrastructure in a bid to increase the return on capital investments will continue to pursue new avenues for offloading, data aggregation, active sharing, and licensed shared access options. As a result, they will position themselves as parts of a blended infrastructure, instead of a standalone network, as in the majority of operator models today. An example of how this blended infrastructure may be implemented is illustrated in Fig. 6. In this example, RF mesh and shared spectrum technologies, e.g., TVWS, complement a core cellular network technology, e.g., LTE, for capacity and high data-rate delivery, and 2–3G cellular technologies for coverage. Wi-Fi will continue to play an important part in home area networks and as part of operator data-offloading strategies. Short range devices, e.g., ZigBee and Z-Wave will dominate the in-building control sector. 5G will better leverage the advantages stemming from a blend of technology and spectrum-usage options. This could lead to usage scenarios in which the communications flow could better suit customers’ priorities, e.g., the requirement for a high QoS and willingness to pay extra for the service, versus low cost communication requirement and acceptance of a lower QoS or higher latency services.

In the quest for higher data rates and lower latencies beyond 4G network capabilities, the next generation of wireless mobile communications has to adopt revolutionary ways of using the radio spectrum. In addition to better ways of exploiting the already allocated frequency bands through enhanced modulation schemes and traditional network scaling, an opportunistic access to the underused spectrum through cognitive radio will further increase the systems’ capacity. Moreover, different offloading techniques will contribute to an “unlimited” access to large amounts of multimedia data anywhere and anytime. The software-defined radio technology will enable reconfigurable platforms for the implementation of the cognitive radio paradigm in underlay cellular networks and traffic offloading solutions, whereas software-defined networking will facilitate the programmable operation of the core network. The evolution towards 5G, as outlined in this paper, opens new research problems. Although we have briefly surveyed the key research projects recently launched in Europe, efforts to develop 5G technologies are being carried out worldwide. Further standardization for cooperation and interoperability between different radio access technologies will be needed, too. With this paper, we intend to motivate the search for innovative solutions to all these challenges.