1 Introduction

Mobile wireless communication industry has come a long way since advent of first generation—voice-only systems a few decades ago. Previous four generations saw steady evolution in wide range of technologies like, digital modulations, effective frequency reuse, WCDMA, OFDMA, MIMO, HARQ, penetration of packet-based Internet etc. Popularity of smart devices and on demand video is strengthening IP based fourth generation wireless communication. IP data handled by wireless networks is expected to exceed 500 exabytes by 2020 [1]. Penetration of smart devices in everyday life is further motivating set of novel, multimedia applications, like video conferencing, video streaming, smart healthcare and online gaming. Many, unforeseen applications are anticipated to materialize by 2020 [1]. Researchers are also predicting annual download of around 1 terabyte by an average mobile user by 2020 [2]. Moreover, a lot of research work is focused on Vehicle to Vehicle communication (V2V), Vehicle to everything communication (V2X), augmented reality, Internet of Things (IoT), Device to Device (D2D) communications, smart healthcare, smart grids, Machine to Machine (M2M) communications and Financial Technology (FinTech). Expecting legacy network to meet rapid increase in data demands and connectivity would be overambitious. Though, developments like, MIMO, HetNets, small cells, Coordinated Multi-Point (CoMP) transmission and multiple antennas can improve capacity and data rates in LTE cellular networks, they are unlikely to be sustainable in a long run [2, 3]. There is an immediate necessity to investigate next-generation wireless system i.e. 5G, to address not only requirements by the users but also to tap new business opportunities for wireless operators for increase in revenue. Various organizations, worldwide famous wireless operators and vendors are already channelizing their resources to different aspects of 5G wireless systems.

Performance requisites for successful implementation of 5G networks are far beyond what is extended by legacy networks [4]. There are eight major requirements of next generation 5G systems as recognized by industries and research community [1, 5, 6]. They are identified as:

  1. 1.

    1–10 Gbps connections.

  2. 2.

    Very low latency (round trip of around 1 ms.)

  3. 3.

    1000 times increase bandwidth in unit area.

  4. 4.

    Enormous number of connected devices.

  5. 5.

    Perceived \(99.999\%\) availability.

  6. 6.

    Almost \(100\%\) coverage.

  7. 7.

    Reduction in network energy usage by almost \(90\%\).

  8. 8.

    High battery life especially for low power devices.

As a mutual consensus between industry and research community, transition to 5G era would not be an incremental evolution of 4G-LTE [1, 4]. Vision of various market players ([7,8,9,10,11,12,13,14,15,16,17]) to achieve efficient 5G communication is summarized in Fig. 1.

Fig. 1
figure 1

Vision 5G: industrial and research perceptive

Full size image

To meet aforesaid requirements a paradigm shift, in terms of carrier frequencies, bandwidths, network architecture and technologies, is inevitable. Extreme densification of base stations and devices with unprecedented number of antennas summon novel approach for PHY, MAC, and network layers [1, 4]. According to Ramjee Prasad [18], 5G system must integrate various technologies, for instance RAT, evolved versions of LTE, HSPA etc. to fulfill terabit communication requirements. Recently, 3GPP released the technical report for the study items that describe the potential physical layer evolution towards New Radio (NR) for 5G wireless [19]. NR for 5G communications is being designed to operate at frequency bands up to 100 GHz [20]. Higher frequencies are expected to support higher data rates. Thus, NR offers to improve data rates by alleviating critical constraints of bandwidth in global wireless communication. In [19], various aspects of NR, for instance, physical layer channel, waveform, modulation schemes, multiple access scheme, multi-antenna based beamforming, MIMO, physical layer scheduling and the likes are discussed for their requirements and complexity evaluations. NR further defines, that a random access preamble format comprises of one/multiple random access preambles(s) where each preamble consists of one preamble sequence along with cyclic prefix [19]. Moreover, multiple Random Access Channel (RACH) preamble formats are supported in NR. Numerology designated for RACH preamble is allowed to be different and would depend on frequency ranges [19].

In 5G communications, enhanced applications and numerous devices offered by IoT would challenges the connectivity and diversity. It is expected that billions of smart entities would be deployed worldwide by 2020 [21]. Due to umpteen requests from myriad of devices, the legacy mechanism of access would suffer from congestion and overloading in the wireless system, degrading the channel performance [22]. Thus, random access procedure requires redesigning in 5G wireless. It should not only support many devices but also accommodate novel architecture based on high frequency bandwidths. Pertinent 5G random access enablers, like relaying, D2D communications, license assisted access, radio access networks as service and the likes discussed in [23] would be relevant to IoT evolution. Bluetooth and IEEE802.15.4 standard played crucial role in the early stages of IoT evolution [23]. Recently, IEEE802.15.4 physical layer has been considered by an IP-enabled IETF protocol stack. By offering distributed solutions for address assignment and routing, it helps in integration of low-powered wireless networks into the Internet [22]. At the same time, 3GPP has been working toward the support for M2M applications on broadband networks which would ultimately lead to M2M communications in the 5G systems [23]. 5G is believed to be a timely technology that promises lower cost, reduced energy consumption and support for numerous devices. These requirements are highlights of 5G MTC design and would be integral to IoT in years to come [22]. Furthermore, 5G is expected to support different services with variable delay requirements. While some applications are delay tolerant, services like critical communications and driverless cars are delay sensitive. Moreover, popularity of tactile internet (applications at fingertips) is motivating low latency internet connectivity [1]. To enable low latency, the mini-slot with length of one symbol is supported in NR [19].

According to [18], reducing power consumption is one of the major challenges of future mobile networks. Legacy network design considerations are based on high peak load. However, in real case scenarios coverage decreases with increase in load on the network. Thus, network topology should be designed such that as the load decreases the base station should be able to cover more regions. Furthermore, this can facilitate switching off of some BSs. In 3GPP release 14, the base station transmit power is considered as 43 dBm for urban cell at frequencies above 6 GHz for system level evaluation assumptions. The UE transmit power is taken as 23 dBm at 30 GHz and 21 dBm at 70 GHz [19]. Transmit power provides degree of freedom to wireless networks for the management of energy, interference and connectivity. In NR both closed-loop and open loop type of power control are assisted [19]. The closed-loop power control depends upon network signaling, while the open loop power control is based on pathloss estimation. Since the beginning of the smart-phone era, managing UEs power has been a significant challenge. Power requirement at UE is expected to increase further in 5G communication in view of expected advances, like, beam forming, high order modulations, advanced computations, massive MIMO [24, 25]. Moreover, mobile gaming, ultra high definition streaming, video conferencing, frequent use of multimedia would increase power requirements at the UE [26]. Role of a smartphone as an M2M gateway is also being discussed for various applications like health care and smart cities [27]. M2M system can be adopted in various applications, like smart homes, smart buildings, smart energy management and monitoring of critical infrastructure [28]. However, as the functionalities increase exponentially, so does the computational power of the devices [29]. Thus, improvement of battery life in an important power requirement in 5G networks. Emerging technique discussed in [30] (like, Fog computing, edge computing, computational offloading, distributed content delivery etc.) not only offer storage, latency and computational solutions but also promise power saving. An exemplar scenario where fog Computing is implemented for reliable eHealth is presented in [31].

With many novel techniques, services and applications, 5G would be different from legacy network in many ways. Through this article we offer practical guidelines to sought future wireless networks of highest integrity. The effective designing and implementation of 5G networks can be efficiently brought about on the basis of key crucial concepts viz. ‘Ten Commandments’ discussed subsequently in this paper. Some techniques like MIMO, Cognitive Radio, SDN and CRAN are gaining momentum. However, a comprehensive study is needed for their progressive integration into 5G. Concepts such as non orthogonality, multicarriers, full duplex and H-CRAN are in their infancy but they present a notable candidature for smooth 5G transition. On the other hand, in depth understanding of mm-wave channel characteristics, beamforming and beam training methods will enhance success probability of 5G deployment. In this article we explore spectrum considerations, key ingredients and upcoming technologies that lead to the new 5G era.

The rest of the paper is organized as follows: Sect. 2 states advantages of a paradigm shift to high frequency mm-wave channel. In Sect. 3, we discuss new, directional antenna concept. It also includes advances in beamforming and beam training techniques. Section 4 provides a review of Massive MIMO, crucial for physical layer fundamentals of mm-wave based 5G wireless communications. Subsequently, we present the review of modifications required in MAC layer in Sect. 5. Then, we take a look into Filter Bank-Based Multicarrier (FBMC) technique for 5G wireless in Sect. 6. Section 7 provides details of Cognitive Radio (CR) and its importance in emerging next generation. Concepts of cloud computing and HetNets, highly relevant to 5G wireless, are integrated into H-CRAN and elaborated in Sect. 8. Section 9 provides details of SDN, a software based network management principle. Major research works related to non orthogonality and asynchronism are described in Sect. 10. We point out importance of Full Duplex (FD) a novel research issues in Sect. 11. Finally Sect. 12 concludes our article. Figure 2 depicts the organization of our paper.

Fig. 2
figure 2

Ten 5G enabling technologies

Full size image

2 Millimetre Waves: Appropriate Bandwidth

Spectral efficiency, cell size and bandwidth are key parameters for evaluation of wireless communication channel capacity [32]. Cell sizes are progressively shrinking and physical layer technology is already at the verge of Shannon capacity [1, 3, 33]. It is the system bandwidth that still remains under exploited. Terrestrial wireless operations are largely restricted to the relatively small range of frequencies, ranging from 300 MHz to 3 GHz band, often referred to as “sweet spot” or “beachfront spectrum” [1, 2]. This band offers convenience of reliable propagation over larger area (several KM) in various radio environments [2, 34]. However, the spectrum crunch is already evident and its capability to accommodate novel 5G applications, with exploding mobile traffic and connectivity, seems questionable [34]. Embracing new challenges of high frequency mm-wave band appears to be a sensible choice for next generation wireless communications [35]. This unused, spectrum ranging from 3–300 GHz is attracting growing attention as a key contender for next-generation wireless networks [36]. Spectrum between 81–86 and 59–64 GHz respectively, for peer to peer and unlicensed wireless communications is unveiled by US Federal Communication Commission (FCC). It opens plethora of activities in related research [37].

Over past few decades, mm-wave band has become popular with several military applications, radars, airport communications and radio astronomy. However, hostile propagation qualities like strong path loss, atmospheric and rain absorption, low diffraction around obstacles and restricted penetration through objects discouraged their use in cellular communication [1]. Fortunately, developments in low-power CMOS RF technologies, enables miniaturized antenna placement in small dimensions [36]. Large antenna arrays can attain high gains while keeping the antenna aperture constant thus, diluting the frequency dependency of path loss. For instance, wireless gigabit alliance (WiGig) IEEE 802.11ad products can support high data rate for short-range communications using mm-wave frequency band at around 60 GHz [38]. Adaptive antenna array cast directive beams resulting in limited interference and high spatial degree of freedom [36, 39]. Moreover, as shown in Fig. 3, degradation due to absorption is prominent only at 57–64 and 164–200 GHz of huge 3–300 GHz mm-wave spectrum. The available bandwidths in mm-wave frequencies can easily enable hundreds times more data rate and capacity than all cellular allocations today [34, 36]. Progressive research in mm-wave channel modeling, advances in probabilistic LOS-NLOS models along with the availability of a huge chunk of high frequency mm-wave spectrum is bringing up a new opportunities for the spectrum constrained legacy wireless networks [34, 36, 37]. Researchers in [40] address subscriber gathering and group movement for 5G networks since they believe that operating in mm-wave frequencies, every subscriber would have huge data demand.

From above discussion, we conclude that mm-wave represents a key 5G transition factor. Research efforts are needed to integrate the new spectrum in the most compatible way. However, it has a strong impact on architecture designs requiring radical changes in transceiver components [39].

Fig. 3
figure 3

Mm-wave spectrum availability in 3–300 GHz

Full size image

3 Directional Transceivers

Shorter radio wave lengths of mm-waves, enable implementation of large number of miniature antenna arrays in a limited space. By controlling amplitude and phase angle of antennas in an array, electromagnetic waves can be focused in the desired direction creating beams [41]. This shift of air interface from omnidirectional to a directional transmission is achieved by using adaptive beamforming approach, further resulting in commencement of Spatial Division Multiple Access (SDMA) [42]. We defer the details of SDMA and other related technologies till next section. Based on similar technique, authors in [18] present a Beam Division Multiple Access (BDMA). In BDMA, base station allocates dedicated beam to each user. The direction and width of downlink beams are ascertained based on moving speeds and position of users. Several beams at different angles are used to serve data simultaneously if users are at different angles. However, users share same beams when they are at the same angle with the base station. Base station has flexibility in terms of number of beams, width of beams and direction of beams. This access technique increases the capacity of the system [18].

Table 1 Key parameters for design directional air interface
Full size table

Effective beamforming algorithms applied to various configurations of antenna arrays and sub arrays steer the beam in desired direction. Beamforming is attained by designated beamforming weights applied to analog or digital front end [43]. Digital beamforming is characterized by coefficient multiplications at every RF chain, over modulated baseband signals. The operation is performed before/after Fast Fourier Transformation (FFT) at the transmitter/receiver. Beamforming coefficients in analog beamforming are applied to alter RF signal in time domain. Analog beamforming is a beneficial and simple method but suffers from inflexibility. On the other hand, digital beamforming endeavours improved performance at heightened complexity and expenses [39]. Hybrid architecture, helps coalesce advantages of both analog and digital beamforming by providing sharp beams with phase shifters at analog domain along with the flexibility of digital beamforming [39, 44]. Communication in 5G wireless with MIMO and hybrid RF beamforming architecture is proposed by the team of Nokia Solutions and Networks to curtail expenses and energy [45]. Dallas Technology Lab (Samsung Electronics) propose use of broader beams for control channels and sharp beams for data transfer [46]. Beam broadening without increase in antenna spacing is achieved by dividing of antenna array into several logical sub arrays [46]. To achieve balance between interference minimization and the desired signal maximization, a novel beamforming algorithm is presented in [47]. Sum-rate maximization in virtual cell network is reformulated into an equivalent joint precoder design and power scaling problem [47]. Beamforming with overlapped virtual cells enhances capacity while improving edge user performance [47].

Fig. 4
figure 4

Antenna panel

Full size image

Beamforming achieved by steerable and directional antenna arrays, justifies use of mm-waves for future 5G deployment. The hardware complexities of antenna modelling for 5G communications is detailed in [48] by 3GPP. A uniform rectangular panel array is considered for the base station antenna as shown in Fig. 4. The antenna panels are placed uniformly in vertical and horizontal directions with respective spacing given as \(d_{gV}\) and \(d_{gH}\). The array comprises of \(N_{g}\) and \(M_{g}\) number of panels in a row and column respectively. Each panel further consists of N number of columns and M number of antenna elements in each column (with same polarization ). The antenna panel can either be dual polarized (\(P=2\)) or single polarized (\(P=1\)). Thus the tuple (\(N_{g}, M_{g}, M, N\)) is used to describe the array antenna in 5G network [48]. Furthermore, since beam forming is integral to emerging 5G communications, NR is expected to supports beam specific power control as baseline [19]. Beamforming introduces beam alignment problems in complex communication protocol design. As shown in Fig. 5, communication between transmitter and the receiver antennas is not possible if the associated beams do not point towards each other. Thus, an efficient “beamforming training protocol” for searching the best beam angle pair is imperative [49]. Iterative antenna training using pseudo noise sequences marked early work in the field of mm-wave antenna pointing protocols. Further, protocols using narrowband pilot signals for rapid determination of antenna pointing directions and multipath angular spreads are presented in [50]. Researchers from Samsung Electronics in [51] propose, transmit precoding and receive combining method called Singular Value Decomposition (SVD) that is based on training antenna coefficients in an iterative approach. SVD training method promises an effective way to bring down computational complexities in mm-wave communications with only a few RF chains and multiple number of antennas [51]. For efficient handling of beamforming architecture, identification of the beam errors is emphasized by researchers at Stanford University [52]. Probing tests to identify and correct beam errors using only few network resource is proposed [52]. In beam coding technique proposed in [49], a unique signature code is assigned to angle of each beam. Quick identification of the best angle pair is accomplished by concurrent steering of coded beam angles in a training packet [49]. This robust approach is promising for Non Line of Sight (NLOS) environment, intrinsic to emerging mm-wave communication [49].

Fig. 5
figure 5

Antenna training: beam steering for link alignment

Full size image

To ensure existence of mobile device within the antenna beam, an antenna tracking system, similar to directional satellite communication, appears encouraging [53]. Acquiring acceptable received signal by moving the axis in small steps is the underlying principal behind antenna tracking. However, the accuracy of measurement depends on size of each step and SINR [53]. Motion sensors (present in smart phones and tablets) are used to predict user movement to maintain beam alignment [4]. Devices like, accelerometer, gyroscope and magnetometer are already embedded in most modern devices, thus promising realistic approach and economic viability [4]. Moreover, fused data from virtual sensors (combination of sensors) can easily assist in identifying next possible beam-pairs [4]. According to Rappaport and team benefits of both spatial multiplexing and beamforming can be simultaneously derived through a proper hardware architecture [54]. MIMO techniques for beamforming and spatial multiplexing are explored in [54] for selecting the best communication technique for mm-wave propagation environment. Multi-user MIMO provides high benefits than conventional point-to-point MIMO and Massive MIMO brings higher improvements by focusing energy into smaller regions [55]. Massive MIMO promises solution to many traditional research problems but at the same time uncovers entirely new issues [55] and therefore, it requires further deliberation. (Key design parameters for beamforming and directional air interface like types of antennas [56], type of communication environment [57, 58], transmission power [59] and beamforming methods [60] are listed in Table 1.)

4 Multiplicity of Antennas: Massive MIMO

Equipping transmit side with large excessive antennas was first proposed by Marzetta in 2006. Since then scaling up of MIMO with huge number of antennas as demonstrated in Fig. 6 has brought around remarkable advantages [61]. Moreover, massive MIMO at a base station is easy to employ by using low-power and inexpensive components [55]. Massive MIMO significantly improves system performance, data rate, spectral and energy efficiency [6, 62, 63]. Linear signal processing and coherent superposition of wave fronts, are the fundamental principles behind massive MIMO enabled BSs [55, 62]. Emitted wavefronts are designed to add constructively at the intended location [55]. Multiple antennas facilitate accommodation of more information data with an additional degree of freedom (supplementary to time and frequency dimensions) [63]. Thus, spatial multiplexing capacity increases by several magnitudes at Massive MIMO enabled BSs [64]. Moreover, simple linear precoding and detection methods easily mitigate effects of noise-fast fading and intracell interference in massive MIMO enabled systems [63].

For effective future base station design, directional antenna and distributed antenna arrays are proposed for Massive MIMO deployment schemes in [55]. For large number of base station antennas, overheads are caused by exchange in form of Channel State Information (CSI) and coordination among different cells. This hampers the system performance, especially with limited-capacity backhaul links. Effective multicasting along with Massive MIMO, in non-cooperative cellular networks, is investigated for estimation of CSI in [65]. Researchers in [66] propose a novel Massive MIMO design comprising of electromagnetic lens with large antenna array in an effort to derive two fold benefits, better energy focus and spatial rejection of the interference. Energy efficiency and spectral efficiency are compared for Massive MIMO and Small Cell Networks (SCN) in [67]. BSs with low traffic enter sleep mode in SCN. Therefore, energy efficiency of SCN is higher than Massive MIMO. However, with higher base station sleep threshold achievable spectral efficiency is reduced [67]. Massive MIMO design with optimized number of cells in SCN presents a suitable candidature for improving energy aswell as spectral efficiency. Thus, an effective merger of the two technologies are expected to provide enhanced results [62].

Fig. 6
figure 6

Massive MIMO and beamforming

Full size image

IEEE 802.11ac compatible wireless products (already available off-the-shelf) can deliver a \(4 \times 4\) multi-user MIMO (MU-MIMO) for downlink [38]. Distributed phased array MIMO architecture for 5G UE hardware design is presented in [38]. The authors provide answers to various technical constraints (for instance, high path loss, human blockage, self-heating issues etc) in designing of mobile hardware. Unlike existing system architectures, these challenges are prominent to 5G cellular user equipment [38]. A systematic framework for receiver architecture for 5G wireless is also proposed in [68]. Flexible, high performance and low power hardware are the desirable for modern smartphones [68]. An energy efficient architecture design that reduces hardware complexity of time domain Decision Feedback Equalizer (DFE) while manifesting 2.15 Gbps for 16 QAM is proposed in [69] for 5G communications. This is critical since in 5G gigabit applications the volume of computations would be higher and DFE would grow with feedback filter order [69]. Several efficient DFE architectures have also been proposed for IEEE 802.11b, utilizing fixed coefficient filter [69]. We believe untapped mm-wave frequency bands along with highly directional beamforming antennas and massive MIMO design promise a robust backbone for 5G networks. The above discussion lays emphasis on need of progressive research activities for novel physical layer components. It also drives attention towards developments in MAC layer for a unified framework.

5 SDMA and Advanced MAC Protocols

Directional transmissions using beamforming techniques reduce interference and increase spatial multiplexing. Thus, effective capabilities of spatial division multiple access (SDMA) appears convincing for emerging cellular applications [34]. To realize SDMA, BSs should be able to simultaneously transmit and receive several beams in multiple directions [43]. This is attained by digital baseband beamforming applied to multiple transmitter RF chains [43]. However, it is essential to train beamforming coefficients beforehand to achieve desired spatial beam patterns [70]. Successful antenna training protocol is proposed in [70], for estimation of optimal beamforming vectors without the detailed knowledge of the channel by Pengfei Xia’s team, at Samsung Electronics. Another important requirement for efficient implementation of SDMA is the evaluation of the angle as well as direction of arrival for RF links between base station and user equipment [71]. In [50], the Angle of Arrival (AOA) distributions are analysed for different transmitter locations. Moreover, information of AoA is also promising for identification of alternate paths in NLOS scenario [56].

To exploit spatial features relevant for 5G wireless networks it is necessary to update MAC protocols and modify multiplexing techniques. Some such progressive directions are listed in Fig. 7. Directional Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) is proposed for Markov Chain model based multiuser MAC protocols [72]. Channel Time Allocation (CTA) composed of super frames exploit the spatial reuse by implementing many local links that communicate concurrently [73]. Research work is also focused on comparison between RTS/CTS transmissions in both omni and directional mode [41, 73]. Directional Network Allocation Vector (DNAV) is proposed in [74], for maintaining a table that records tracking directions where the node must initiate a transmission (or should not initiate a transmission). DNAV integrated directional MAC protocol, assumes upper layer to be aware of its neighbors [74]. Problems of deafness, under-utilization, hidden node and dead-lock, hinders effective implementation of DNAV. Multihop MAC (MMAC) protocol, performs better than directional MAC by including Direction-Direction (DD) and Direction-Omni (DO) neighbor establishment methods [74]. A DD neighbor node can receive directional transmission from neighboring directional nodes only when it is beam aligned with directional neighbor. However, even in an omni mode the DO neighbor node is able to receive from neighboring directional nodes [74].

Fig. 7
figure 7

Key features—advanced MAC protocols

Full size image

To keep up with advancing protocols many new multiple access techniques are proposed by researchers. Though SDMA appears a natural fit, further sophistication will help achieve stringent 5G requirements. Generalized Frequency Division Multiplexing (GFDM) proposed in [75, 76] offers higher flexibility in multiplexing. The block structure enables low latency specification of 5G systems [75]. Interleave Division Multiple Access (IDMA), a special case of CDMA considers spread sequence specific to the user. Specific interleaves are implemented for user segregation [77]. Fusion of QAM symbol mapping and spreading is proposed in Sparse Code Multiple Access (SCMA) scheme. In SCMA, multi dimensional codeword can be directly mapped over incoming bits (using codebook) [78]. OFDM remains preferred choice by some researchers as it is effective and a little less sensitive to offsets due to time [79]. However, OFDM suffers from high spectral leakage, high peak to average power ratio and strict orthogonality requirements [80]. Filter Bank Multi-Carrier (FBMC) is believed to be one of the major candidate to overcome the challenges of 5G system [81]. It yields better gains than existing techniques at a reasonable increase in complexity [80] and is further elaborated in the next section.

6 Filter Bank Multi-Carrier (FBMC)

Future wireless radio networks require flexible spectrum allocation to implement dynamic access spectrum management and cognitive radio [82]. Filter Bank-Based Multicarrier (FBMC) technique can approach the theoretical capacity limits in communications, offers high spectrum resolution and provides independent sub-channels [82]. FBMC improves access flexibility and encourages investigations into alternative multicarrier waveforms for spectrum agility and better adjacent channel leakage performance [83]. Moreover, diverse and dense wireless connectivity along with M2M and IoT applications generate sporadic traffic [83, 84]. Reduced signalling overheads on the cellular network would enhance network performance and improve user experience [84]. Filter bank multicarrier are natively non orthogonal and do not require elaborate synchronization thus, promising a viable solution [1].

Fig. 8
figure 8

Block diagram for OFDM and FBMC

Full size image

To get high frequency resolution FFT, FBMC Scheme is completed by additional processing performed by a polyphase network, similar to OFDM [85], as shown in Fig. 8. Data rates similar or higher than those of OFDM are enabled by Offset Quadrature Amplitude Modulation (OQAM). Moreover, orthogonality is limited to neighbouring sub-channels only. Thus, dissolution by a single empty sub-channel can render groups of sub-channels as independent. This spectral independence property presents itself as a key enabler for future wireless networks. Major benefits of FBMC include (1) ease in implementation of frequency division multiplexing, (2) uplink spectrum assignment to users do not require ranging and distant synchronization, and (3) mobile system coverage is improved by mixed multicarrier and single carrier schemes along with enhanced flexibility. Though, FBMC promises flexibility and reduction in synchronization overheads, major research effort is necessary for full exploitation [85]. Efforts by PHYDYAS: PHYsical Layer for Dynamic Access Spectrum and cognitive radio at the European level are focused to deliver the best methods and the most efficient algorithms [85].

Frequency Spreading Filter Bank Multicarrier (FS-FBMC), where input data flow is spread over a number of carriers in frequency domain is proposed for multicarrier transmission in [86]. Effective frequency spreading in the transmitter along with despreading and equalization operations are important to achieve perfect sub-channel equalization [86]. Another major benefit of FBMC is improved spectral efficiency which is achieved by dropping of redundant cyclic prefix (CP) and an upgraded control of out-of-band emission [87]. Moreover, multiple users are easily accommodated in an FDMA pattern. An effective linear Minimum Mean Square Error (MMSE) equalization concept for the equalization of OQAM multicarrier systems is proposed in [87]. Though complexity overhead of FBMC against OFDM is high, savings in transmit power is sufficiently substantial [87]. Longer equalizers and slightly increased computational complexity can reduce the number of subcarriers [87]. This further improves the robustness against carrier frequency offsets and timevariant channels while simultaneously alleviating the Peak to Average Power Ratio (PAPR) problem of OFDM [87]. Moreover, in OFDM, severe degradation is observed in system performance due to timing errors between BSs [88]. The result is loss of orthogonality among all subcarriers. However, better frequency localization of prototype filter makes FBMC waveform less sensitive to base station timing errors between different cells [88]. For operation in mm-wave band, authors in [89] integrate FBMC-OQAM technique with the baseband processing. The feasibility is then verified for VLSI architecture based on IEEE 802.15.3c/802.11ad standard. An efficient polyphase network (PPN) architecture with pipeline and 8 times parallelism is proposed to lower complexity than the state-of-art designs, where the memory reordering architecture allows the lower memory expenses at the receiver end.

A comprehensive survey on multicarrier communications with emphasis on filter design, lattice structures, and evaluation based on practical implementation aspects is presented in [80]. Cognitive radio (CR) based FBMC systems are analyzed in [90], and are found to have more capacity than CR based OFDM. Musbah Shaat and Faouzi Bader of CTTC, Spain believe multicarrier communication (MC) systems to be one of the prime candidates for evolving CR systems. The CR technology promises increase in spectrum utilization and has ability to fill the spectrum holes left by primary users. The CR concept is expanded in subsequent section.

7 Cognitive Radio Capabilities

According to Federal Communication Commission (FCC) licensed spectra are scarcely utilized continuously athwart space and time [91]. Cognitive radio (CR) technology is proposed to enable effective utilization of spectrum [91]. In a CR network, as shown in Fig. 9, intelligent CR users/devices are capable of sensing and harnessing available unused bands originally allocated to the primary users/devices [91, 92]. The transmission parameters are opportunistically adjusted and thus, presents CR as a candidate technology for spectral efficiency improvement [91]. CR would help in furnishing important information like free channel, occupied channel, modulation scheme, position of UE, type of data that is to be transmitted and more generally the awareness of the environment. This kind of knowledge would facilitate in meeting the required QoS [18]. An integrated energy and spectrum harvesting exemplar for 5G wireless systems is elaborated in [91]. Spectrum sensing in CR networks is usually based on energy detection techniques and is affected by the spatio-temporal characteristics of radio channel [93]. The mm-wave freqencies especially above 10GHz suffer from tropospheric attenuation, more specifically by rain fading [93]. Researchers from National Technical University of Athens, have proposed mm-wave CR technique to evaluate and analyse threshold in the energy detector under rain fading [93]. Moreover, highly flexible communication to support multiple RATs in order to exploit available spectrum across the various frequency bands is essential [4]. Flexible spectrum management and opportunistic spectrum access is advocated in [94]. Novel user-centric architecture along with flexible medium access and signalling protocols promise solution to flexible communication in multiple RAT [4].

Fig. 9
figure 9

cognitive radio architecture

Full size image

In emerging 5G systems, different complications can be potentially resolved by CR operations [92]. These are listed in [92] as: (1) smooth migration from 4G to 5G systems to support non-LTE standard compliant techniques, (2) relaxed date rate and latency demands in machine-type communication (MTC), (3) enabling of D2D communications along with uninterrupted primary systems, and (4) load balancing by averting high data rate transmissions through secondary links. GFDM along with CR waveform is used to access frequency holes in an LTE legacy system is proposed by Martin Danneberg and team at Vodafone Mobile Communications Systems in [92]. D2D is gaining momentum as one of the prominent technology within 5G [95]. Moreover, there is potential for vehicular coordination to derive benefit from D2D advances [95]. The approach is further extended in [95], by exploring CR along with geo-location database for offloading vehicular users. A novel CR based resource allocation policy is employed for D2D supported Vehicle to Vehicle (V2V) communications in [95]. Cognitive capabilities can also provide backhaul link diversity leading to reduced energy consumption of the backhaul network under low traffic scenarios [96]. CR assisted backhaul link selection and resource assignment operations optimized over QoS constraints are investigated in [96]. An end-to-end architecture for 5G era with cognitive and cloud optimized network evolution is formulated in [97]. While cognitive network autonomously adapt itself, cloud architecture enables automated management and on demand assignment of virtual resources [97].

We can conclude from aforesaid discussion that their is an emerging diverse ecosystem which requires unification. Alignment of different concepts and technologies which can be materialized by software driven control. Thus, to simplify automation of diverse connectivity, we should focus on amalgamation of heterogeneous networks with virtualization through cloud computing. Heterogeneous-CRAN offers unification of virtualization and physical layer developments and is elaborated in our next sections.

8 The Evolution of H-CRAN Architecture

To achieve vision 5G, a breakthrough is required to resolve computationally intensive operations, adapted to the evolved air interfaces [98]. Hence, evolution in baseband, signal processing and RF technology is necessary [98]. Moreover, efficient working of ultra dense radio nodes advocate improvements in integrated access and heterogeneous convergence [98]. While Cloud radio access networks (CRANs) curtail capital and operating expenditures by enabling high transmission bit rate at improved energy efficiency (EE), Heterogeneous Networks (HetNets) guarantee backward compatibility by deploying low power nodes (LPNs) along with high power macro/micro base station [99]. H-CRAN architecture derives benefit of large-scale cooperative signal processing and networking by embedding cloud computing technology into HetNets. Integration of the two concepts, promise a substantial improvement in spectral and energy efficiencies beyond existing HetNets and CRANs [98].

Cloud architecture comprises of remote radio heads RRHs as soft relays responsible for compressing and then forwarding signals from UEs to BaseBand Unit (BBU). BBUs from multiple remote sites are centralized at the BBU pool connected to different RRHs through wired/wireless fronthaul links [100, 101]. Work in [102] demonstrates 5G-enabled CRAN prototype that contains several VBSs software instances. The hardware for VBS cluster (VBSC) prototype comprises of dual-core Intel X5355 processors over a Linux kernel. A high-speed optical fiber (10 Gb/s) is employed for the fronthaul communication between RRHs and VBSC. Several VBSCs are interconnected by means of 10 Gb/s X2 interface. Using hardware abstraction layer application programming interface, authors in [102] provide different RRH interfaces to each VBS instance. China Mobile Research Institute presented a step by step procedure n-32,n-23for CRAN design using 8−12 macro sites with the ring range up to 40 km [103].

Power requirements in 5G should not only focus on base station but also on other power consumption elements of the network for instance virtual base station and UEs [104]. Virtual base station (VBS) is key element in BBU. Computational resources vary in VBS based on processing power of UEs a VBS is required to assist [104]. Since CRAN allocates the baseband resources for computation dynamically, the power consumption of VBSs fluctuate. Thus, flexibility would be crucial in power matrix of 5G wireless. Authors in [104] evaluate DRX models for VBS based 5G networks. Visualization of energy flow and real time monitoring of power consumption would be effective for equalization of power load in future networks [25]. Energy efficiency is also an important challenges in IoT landscape with millions of devices [105]. 5G networks are expected to connect myriad number of devices under low power consumption constraint especially in scenarios where reliable connection is required (eg wireless sensors) [106]. Authors in [105] present concept of mini clouds that host different types of machines to address power requirement in IoT. Another promising technology is the use of alternate sources like solar cells, wind energy etc. However, reduction in cost of installation would be defining factor for the commercial use of alternate technologies [25]. Furthermore, virtulaized BBU pools offer cost reduction, scalability, resource savings, incorporation of various services and shortens field trial time requirements [107]. Improved statistical multiplexing gains and energy efficient operations are realized by CRAN architecture [107]. Effective computing capabilities of cloud can easily address all complicated control processes [108]. Moreover, a seamless merger of cloud applications and wireless networks is possible by an SDN enabled programmable interfaces [109].

According to recent paper by Huawei, cloud-native architecture would be the foundation to 5G innovation [110]. Furthermore, at IEEE 5G SUMMIT (2017) [111] the split between central unit (CU) and distributed unit (DU) at the base station was highlighted for the centralized deployment. It is recognized in [111] that the CU and DU separation can not be more than 200 km. In [112], CRAN and next generation architecture is presented. Implementation of CRAN and its deployment would be governed by split between CU and DU. For instance CU may comprise of PDCP (Packet Data Convergence Protocol) layer while functionalities of MAC, PHY, and RLC(Radio link control) layers are at the DU [112]. White paper by China mobile research institute [113] also delineates 5G -CRAN architecture. In [113], DU are focused on PHY-level collaboration. The fronthaul is expected to provide high capacity, low cost, low latency and low jitters [113].

On the other hand, HetNets are typically composed by deployment of low power nodes besides the high power macrocells [101, 114, 115]. Work in [116] considers HetNets with carrier aggregation. In [116] the macro cells operating at 800 MHz (for desired coverage) is placed along small cells operating at 2.6 GHz (for throughput enhancement at hotspots). Such coexistence calls for a coordinated operation between legacy macro and small cells, in order to reduce mutual interference [117]. Time synchronization is crucial among radio transceivers to collaboratively transmit among several BSs and to align received signals [118]. It becomes more pronounced when macrocell users hand over to a small cell or vice versa. To mitigate such problems, cooperative distributed algorithms is proposed in [118]. Research works focused on multi-tier networks and interference management are addressed by researchers from industries in [119]. Proficient coupling between multiple RATs offers to improve coverage in HetNets [120]. Efficient RAT handover decisions and optimized partitioning of common resources promise mitigation of unnecessary signalling overhead caused by inappropriate RAT [121]. Radio resource management schemes (cross-tier and co-tier), frequency scheduling algorithms and frequency reuse techniques further improve HetNets performance [122].

Fig. 10
figure 10

Heterogeneous-Cloud Radio Access Network (H-CRAN)

Full size image

Researchers believe that communication between diverse cells is one of the most prominent challenge in HetNets. Combination of HetNets and CRAN, H-CRAN is gaining popularity as a key technology for 5G communications [98, 101]. Control of radio resources across different radio technologies for enhanced coverage, seamless combination of multiple RATs and link reliability calls for joint management of virtual radio network [33]. Lagrange dual decomposition method is used in [99] to address resource allocation issue in H-CRAN in an energy efficient way. A new entity ‘Node C’ is proposed in [98], for software-defined H-CRAN design. ‘Node C’ is an evolution of Node B (base station) and possesses networking functionalities for the RRHs. An exhaustive survey on H-CRAN is presented in [98], that includes system components, application architecture, major technologies, large scale spatial signal processing and fronthaul optimization. H-CRAN presents itself as an advanced wireless access network paradigm, with cloud enabled centralized large-scale cooperative processing to suppress co-channel interference in diverse wireless environment [98] as shown in Fig. 10. Furthermore, one of the effective methods to achieve power saving requirements is to dynamically switch off the BSs when there is lower traffic load and only few BSs can meet the requirements [123]. This would be especially valuable for optimization of power saving in HetNets where different tiers of small cells and large cells coexist. Thus, power consumption is popular optimization objective that requires investigating of switch-off strategies for cloudified BSs in H-CRANs, for greener 5G networks [123]. Moreover, for a complete automation this will go hand in hand with abstraction of control and user plane, i.e. the Software Defined Networks (SDN) [97].

9 Agility and Resilience by SDN

Wireless networks are now rapidly evolving with changes in architecture, air interface and deployment schemes. To accommodate exploding wireless traffic, cell shrinking is already under way. Hence, emerging dense networks are confronted with challenges of configuration as well as maintenance of multiple routers and servers. Software Defined Network (SDN) offers to establish agility and resilience by splitting of control and data planes, creating framework of intelligent programmable networks [1, 124, 125]. Moreover, underlying network infrastructure is abstracted from the application with logically centralized network intelligence [1]. Increase in user plane capacity is no longer constraint by control plane resources. The split ensures enhanced data delivery at the desired locations, without incurring control plane overheads [125]. Software components in SDN help to accomplish the decoupling of control and data planes. These software components manage the control plane thus, effectively reducing hardware requirements. Further, open interfaces, like Open Flow can establish interactions between the two planes for seamless operation [126].

Complete automated administration is achieved by SDN stepping over OSI layers for a remodelled network. Controllers assign policy to participating routers for monitoring which reduces redundant interfaces [108]. Researches at NEC Laboratories America believe that control/data plane separation of SDN realized with radio access networks (RAN) manifests itself as a SON solution [126]. Control plane coordination is achieved at a coarse granularity by SON algorithms optimize over RAN , while fine granular data plane remains unaffected [126]. However, for significant gains, improvement in data plane cooperation by sophisticated RAN optimization is also necessary. Coordinated Multi Point (CoMP) transmission provides candidature for cooperative data transmission from multiple base station at the granularity of fine time scale [126].

Concept of SDN can be applied to core network by use of Softcell [127], MobileFlow [128] and SoftMoW [129]. The network functions are replaced with SDN controllers and switches that provide interconnection between RAN and external packet networks [130]. SoftCell architecture allows fine-grained packet classification at the access switches. These switches are placed next to the base stations so that the state and bandwidth requirements are easily handled [127]. Work in [128] provide testbed implementation of mobileflow approach using real Huawei eNodeB, commercial off-the-shelf (COTS) LAN switches, Huawei Network Traffic Emulator, multiple KVM virtual machines etc. Controller is the key component in the SoftMoW architecture. Modular approach is used for controller design using network operating system, operator applications and the recursive abstraction application (RecA). To cover the users traffic in the core network, SDN is applied with extensions to the OpenFlow protocol to implement GPRS tunneling protocol in [131].

We believe that software driven steering and control will not only help simplify network but will also address low latency requirements. However, to achieve tactile sense of human body capable of differentiating around 1 ms latency [75], even more sophistications are expected. A new PHY paradigm for efficient and scalable air interface supporting diverse 5G requirements with focus on Non-Orthogonal, Asynchronous Waveforms is proposed in [75]. In alignment with this new dimension, the principles of non orthogonality, asynchronism and waveforms are further elaborated in our next section.

10 Non Orthogonality and Novel Waveforms

Applications like M2M, IoT, FinTech, real time health monitoring, smart grids and vehicular Internet are driving transition to 5G networks. The rigid paradigm of synchronism and orthogonality applicable to legacy networks prevents efficient scalability [75]. The disruptive design of non orthogonal waveforms offer key candidature to meet the upcoming needs.

Fig. 11
figure 11

Non-orthogonal asynchronous multiplexing technique

Full size image

While synchronization employ a common clock for all sender operations and processing, orthogonality ensures no cross talk in the receivers waveform detection process [75]. Robustness against multipath and simple execution of Fast Fourier Transform (FFT) algorithms makes Orthogonal Frequency Division Multiplexing (OFDM) widely adopted solution [132]. However, strict synchronization requirements are challenging main scenarios of 5G networks [132]. Sporadic traffic generating devices (M2M, IoT, smart grid devices) are unavoidable in emerging 5G applications. Moreover, uncoordinated interference from highly overlapping HetNeTs and cooperative multipoint (CoMP) signaling are also inevitable with increasing demands. Integrating bulky synchronization procedure designed to meet orthogonal constraints impedes smooth transition to 5G [27]. Robustness to time-frequency misalignment is imposed as an important requirement from 5G [133]. This enables heterogeneous devices and diverse service classes to fit into a single radio frame structure, thus reducing signaling overheads. Universal Filtered Multi-Carrier (UFMC) proposed in [133], presents capabilities of relaxed synchronicity to combat against misalignment overheads. Envisioned applications like internet of vehicles and gaming demand fast response time. Hence, short frame enabled transmission mode with very low air interface latency is crucial designing parameter [134]. Time-frequency efficiency of cyclic prefix based OFDM is less than theoretic maximum. However, subcarriers of FBMC (instead of sinc-pulses of OFDM) promise \(100\%\) time-frequency efficiency in theory. Therefore, FBMC is a key contender for 5G air interface. Analysis in [134] shows FBMC to be non orthogonal with respect to the complex plane [134]. On the other hand, Generalized Frequency Division Multiplexing (GFDM), proposed in [132], offers flexibility to cover CP-OFDM and single carrier frequency domain equalization. The result of this subcarrier filtering is non orthogonality [132]. Thus, to address specific requirements of 5G, orthogonality will have to be forfeit [132]. However, 5G research not only focuses on new waveforms but at the same time re-design of physical and partially MAC layer is already inevitable. Novel physical layer aspects in NR support multiple numerologies, flexible network and channel bandwidth [19]. CP overhead and sub-carrier spacing are used to define numerologies. Physical Resource Block (PRB) is defined in NR such that the number of subcarriers per PRB remains same for all the numerologies. It is to be noted that the number of subcarriers is 12 per PRB [19]. In recent 3GPP report, the maximum channel bandwidth is 400 MHz per NR carrier. 1 ms is fixed as sub frame duration and 10 ms as frame length. Scalable numerology allows subcarrier spacing from 15 to 480 kHz [19].

Fig. 12
figure 12

Full duplex operation

Full size image

Authors of [75], propose non-orthogonal robustness concept along with required control signaling and several waveform approaches such as UFMC, FBMC, and GFDM, highlighted in Fig. 11. Non and quasi-orthogonal techniques allow spectrum overlap as number of users are no longer limited by the set of orthogonal resources [135]. Moreover, PHY layer abstraction to handle non-orthogonal waveforms and synchronization errors is needed to evaluate the system-level performance. System level simulations of non-orthogonal waveforms is analyzed in [136]. Suitability of non orthogonal waveforms to address manifoldness of emerging services, device classes and transmission set-ups is discussed in [137]. Introducing new waveforms to support heterogeneous traffic will further justify this transition to new disruptive architectural and control paradigm. Another upcoming disruptive challenge demanding attention in transmission mode is the wireless full-duplex transmission which is further extended in the next section.

Table 2 Technologies and Concepts for design and implementation of 5G wireless networks
Full size table

11 Full Duplex Radio Technology

Single frequency Full Duplex (FD) radio technology is identified by Huawie as one of the major contributor for implementing commercial-ready 5G network solutions [14]. Conventionally, crosstalk between transmitting and receiving antenna, fading, impediments due to path loss and self interference discourages concurrent communication on the same frequency channel [138]. Current developments in antenna design and RF circuits along with advances in self-interference cancellation technology has made FD radios a possibility. FD transmission enables a node to transmit and receive signal simultaneously in the same frequency band [9, 121], as depicted in Fig. 12. This not only doubles the capacity, but also improves feedback mechanisms latency while maintaining security in physical layer [121]. Researchers have analysed physical layer network security in the full-duplex relay system and found it to have better secrecy performance than the half duplex scheme [139]. Moreover, network efficiency further enhances as full duplex removes the hidden node problem in contention based networks [121]. Alleviating Self-Interference (SI), caused by a combination of hardware and deployment imperfections, is the major challenge to be resolved in implementation of full duplex [140]. Every FD transmission experiences excessive interference from within and neighboring cells due to parallel scheduling of downlink/uplink on the same resource block [138]. Research work in field of full-duplex enabled transmission is focused on development of advanced self-interference cancellation and/or suppression techniques [141]. Passive and active cancellations are identified as broad SI cancellation categories [140, 142]. Passive technique advocates directional antennas, cross-polarization and absorptive shielding to isolate the transmit and the receive antennas. Antithetically, active technique capitalizes on knowledge of transmit signal (of the node) to cancel the interference by exploiting analog, digital and spatial cancellation [142]. While digital domain negates residue self-interference due to the non-ideal electronic circuity, analog cancellation avoids input of the analog-to-digital converter (ADC) overwhelmed by the self-interference [141]. However, benefits of full duplex strained by tradeoff between spatial reuse and the full-duplex gain in asynchronous contention based environment, are met with skepticism in [143]. To fully exploit the capabilities of full duplex, design of an efficient and novel MAC protocol becomes critical [144]. Asymmetrical duplex MAC protocol proposed in [141]. It recommends dual link among two distinct half duplex clients and full duplex access point based on packet-alignment. Thus, half duplex and full duplex would have to coexist in the same service environment [144]. On similar lines, a MAC protocol that offers both the bidirectional and the unidirectional techniques, with request to send (RTS)/full-duplex clear to send (FCTS), is proposed in [141].

Researchers from NYU Polytechnic School of Engineering, New York believe in careful planning of full duplex transmission, as high interference in both uplink and downlink greatly limits the potential gains. Furthermore, regular scheduling approach when employed for maximum capacity gain cause acute energy efficiency losses [138]. Small cell environment integral to emerging 5G networks is effective for FD operation. Intelligent device scheduling that enables suitable rate/power allocation extracts large capacity gains from the FD operation [138]. Self Interference, crosstalk and scheduling may prevent exact doubling of capacity by FD but improvement in throughput gain are certain. For duplexing, NR tries to maximize the commonalities between different solutions. For instance, supporting different directions of transmission in either parts of the paired spectrum, enabling FDD on paired spectrum and allowing TDD on paired spectrum (while time resources are changing or not changing dynamically) [19]. The direction of transmission includes uplink, downlink, sidelink and backhaul link. We believe FD to have significant potential to contribute to 5G communications. It is a novel communication paradigm and its impact remains to be investigated.

5G is likely to be significantly different from legacy communication generations [18]. Novel ideas in architecture, operation, addressing of possible challenges and mechanisms would drive wireless connectivity towards ubiquitous terabit per second (tbps) capacity [18, 40]. As suggested in [18], 5G must be a combination of advanced and evolved technologies. Table 2 summarizes key contenders and their benefits that are likely to provide solutions for unified 5G era.

12 Conclusion

Emerging 5G wireless communication systems are promising multi-fold increase in data rate, connectivity, applications and services. We have focused upcoming on technologies to address these pressing requirements. This article discussed ten crucial research directions that are likely to form basis for dense and diverse 5G deployment. We hope this article will help in treading along the 5G road.