write a research paper after reading the article bellow

profileabby morgan
TM355_TMA-5GFILE_Fall2021_7cea89ac40698089b656553e142a2f1b.pdf

Received December 4, 2019, accepted January 7, 2020, date of publication January 13, 2020, date of current version January 24, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.2966271

5G Technology: Towards Dynamic Spectrum Sharing Using Cognitive Radio Networks W. S. H. M. W. AHMAD 1, N. A. M. RADZI 1,2, (Senior Member, IEEE), F. S. SAMIDI1, A. ISMAIL 1,2, (Member, IEEE), F. ABDULLAH 1,2, (Senior Member, IEEE), M. Z. JAMALUDIN1,2, (Senior Member, IEEE), AND M. N. ZAKARIA3 1Institute of Power Engineering, Universiti Tenaga Nasional, Kajang 43000, Malaysia 2Electrical and Electronics Engineering Department, College of Engineering, Universiti Tenaga Nasional, Kajang 43000, Malaysia 3Architecture and Governance, Tenaga Nasional Berhad Information and Communication Technology (TNB ICT), Kuala Lumpur 59200, Malaysia

Corresponding author: N. A. M. Radzi ([email protected])

This work was supported in part by UNITEN R & D Sdn Bhd through Tenaga Nasional Berhad Seed Fund under Grant U-TC-RD-19-04, and in part by the Universiti Tenaga Nasional BOLD2025 under Grant 10436494/B/2019019.

ABSTRACT The explosive popularity of small-cell and Internet of Everything devices has tremendously increased traffic loads. This increase has revolutionised the current network into 5G technology, which demands increased capacity, high data rate and ultra-low latency. Two of the research focus areas for meeting these demands are exploring the spectrum resource and maximising the utilisation of its bands. However, the scarcity of the spectrum resource creates a serious challenge in achieving an efficient management scheme. This work aims to conduct an in-depth survey on recent spectrum sharing (SS) technologies towards 5G development and recent 5G-enabling technologies. SS techniques are classified, and SS surveys and related studies on SS techniques relevant to 5G networks are reviewed. The surveys and studies are categorised into one of the main SS techniques on the basis of network architecture, spectrum allocation behaviour and spectrum access method. Moreover, a detailed survey on cognitive radio (CR) technology in SS related to 5G implementation is performed. For a complete survey, discussions are conducted on the issues and challenges in the current implementation of SS and CR, and the means to support efficient 5G advancement are provided.

INDEX TERMS 5G, new radio, spectrum sharing, spectrum efficiency, cognitive radio, enabling technologies.

I. INTRODUCTION 5G is the next-generation mobile communication technol- ogy designed to provide greater capacity and higher data speeds than the previous generation Long Term Evolution (LTE). 5G technology, which is expected to be realised by 2020 [1], [2], promises ultra-low latency and ultra- high reliability, thus enabling innovative services across different industry sectors [3]. 5G standards are currently under development and will include the evolution of exist- ing LTE and 5G New Radio (NR) technologies. Several 5G application services have been identified according to International Telecommunication Union (ITU) stan- dards. These services include enhanced Mobile Broadband (eMBB), massive Machine Type Communication (mMTC), Ultra-Reliable Low-Latency Communication (URLLC) and

The associate editor coordinating the review of this manuscript and

approving it for publication was Ke Guan .

fixed fibre-like wireless access. Each user is expected to experience a minimum of 100 Mbps data rate with a peak data rate of 20 Gbps [2], [3]. For data rates higher than 100 Mbps, Fixed Wireless Access (FWA) or fibre-like wireless can be used. This broadband wireless access is beneficial for residential customers and enterprises using pre-5G or 5G access technologies, including full- dimensional Multiple-Input Multiple-Output (FD-MIMO), massive MIMO and millimeter wave (mmWave) radio access technologies [3], [4]. In ITU’s radiocommunication Inter- national Mobile Telecommunications (IMT) 2020 vision, at least 800 MHz of contiguous spectrum per 5G network is required for very-high-capacity 5G networks, such as hotspot and FWA connectivity.

In recent years, applications, such as virtual reality, aug- mented reality and cloud-based services, have emerged and become an integral part of the new generation’s lifestyle. By 2030, the vision of connecting 50 billion devices is

14460 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 1. 5G application services.

expected to be realised as part of the Internet of Things (IoT) evolution. Sensors, actuators, electronic appliances, street lighting and other devices will be wirelessly connected to the Internet and one another via device-to-device (D2D) communication, which is also known as massive Machine Type Communication (mMTC). Other advancements will demand URLLC; these advancements include connected and autonomous cars, aerial vehicles, remote control of robots in extreme hazardous conditions, industry automation as part of Industry Revolution 4.0, remote surgery and smart grid applications. With the realisation of 5G technology, ultra-fast, ultra-reliable and ultra-low latency application services can be achieved, as compactly illustrated in Figure 1. 5G will benefit numerous industry sectors and acceler-

ate many applications, such as IoT and Mobile Edge Com- puting (MEC). According to a study by Rimal et al. [5], highly localised services are required in Radio Access Net- works (RANs) that is in close proximity to mobile sub- scribers. This requirement has led to the emergence of the MEC concept, in which cloud services are delivered directly from the network edge. Authors have also discussed potential service scenarios for MEC, such as edge video orchestration and distributed caching, backhaul optimisation, vehicle-to-vehicle/roadside communication and IoT services. Other have identified the design challenges of MEC-enabled networks, namely, network integration and coordination, distributed resource management, coexistence of human-to- human and MEC traffic, cloud and cloudlet interoperabil- ity, reliability and mobility. Furthermore, the possibility of integrating Fiber Wireless (FiWi) access networks to offer MEC capabilities has been explained with different design scenarios from the architectural perspective. The scenarios are as follows: (1) MEC over FiWi networks, (2) MEC over Ethernet-based FiWi networks, (3) MEC over 4G LTE-based

FiWi networks and (4) coexistence of MEC and centralised- RAN (C-RAN) in FiWi LTE-A Heterogeneous Networks (HetNets). In the same study, the authors provided a per- formance analysis of MEC over Ethernet-based FiWi in terms of delay, response time, efficiency and battery life to demonstrate the feasibility of MEC over FiWi. The results showed the significant benefits of MEC over FiWi networks, with efficient human-to-human/MEC coexistence without network performance degradation.

Numerous researchers have exerted a substantial amount of work to enable applications, such as MEC. An overview of the current status of 5G industry standards, architecture, spectrum allocation, use cases, relevant scenarios and state of the art system advances was presented in [3], [6]–[11]. [12] and [13] studied the opportunities and challenges faced by existing solutions in implementing mmWave for 5G, including the characteristics of and standards in enabling mmWave and related applications (small cell (SC) access, cellular access and wireless backhaul). The authors pre- sented a thorough discussion of the directions of mmWave in the future mainly in terms of physical layer technol- ogy (MIMO at mmWave and full-duplex (FD)), software- defined architecture, control mechanisms, network state measurement and HetNet. A comprehensive analysis of var- ious types of 5G potential waveforms was recently per- formed by Samal et al. [14]. They discussed issues related to single-carrier modulation schemes suitable for mmWave in 5G wireless communication systems and enumerated the limitations of Orthogonal Frequency-Division Multiplexing (OFDM), Orthogonal Frequency-Division Multiple Access (OFDMA) and single-carrier frequency-division multiple access (SC-FDMA) spectrum access schemes.

A survey of the current state-of-the-art research in 5G-IoT was presented by Lie et al. [15], with focus on the key enabling technologies, the main research trends and the challenges. Despite extensive research effort on 5G-IoT, the technical challenges identified by the authors involved the 5G-IoT architecture, namely, scalabil- ity and network management (NM), interoperability and het- erogeneity, security assurance and privacy concerns. The challenges in wireless Software Defined Network (SDN) for 5G data networking, Network Function Virtualisa- tion (NFV), D2D communication with efficient spectral resource and interference management, large-scale (with limited resource) and heterogeneous environment of IoT application deployment and multiple access FD transmission remain to be addressed for the successful commencement of 5G-IoT.

Considering recent research, this study is expected to provide an overview of the most recent advancements in 5G-enabling technologies. Our latest study has shown that one of the most urgent issues in future implementation is the scarcity of the radio spectrum [16]. Currently, wireless connectivity is prevalent worldwide and makes full use of the available spectrum band. Unlicensed bands are often congested due to overuse [17], and licensed bands are always

VOLUME 8, 2020 14461

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

underutilised. The Federal Communication Commission indicated that only 5.2% of bands below 3 GHz are utilised at a given time or location. This situation presents an oppor- tunity to solve spectrum scarcity by using the spectrum sharing (SS) concept within a band to maximise spectrum utilisation. An in-depth survey of previous reviews on SS by other authors is conducted, followed by a discussion of the most recent methods in SS. CR is one of the methods with promising implementation. The analysis performed in this study highlights the issues and challenges in current CR implementation to provide insights into potential research directions.

The main contributions of this study are summarised as follows:

• A complete list of the most important 5G-enabling tech- nologies and a brief overview of each technology are provided.

• Different SS techniques are classified. • Related SS surveys conducted from 2014 to 2019 are reviewed. The focus and contributions of each study are summarised and presented in a table.

• Related studies on SS techniques relevant to 5G net- works are reviewed. The studies are categorised into network architecture, spectrum allocation behaviour and spectrum access method, which is also one of the main SS techniques. Related SS works focusing on energy- efficient (EE) improvements are also discussed. The related works are summarised, compiled and presented in a table.

• CR technology in SS and other applications related to 5G implementation are reviewed. CR in SS can be consid- ered a potential technology that can propel 5G networks into the future.

• The issues and challenges in the current implementation of SS and the improvement of current methods to sup- port 5G advancement and efficiency are discussed.

The rest of this paper is organised as follows. Section 2 dis- cusses current 5G-enabling technologies in literature. Section 3 focuses on SS methods, including surveys and recent works, and summarises them in a tabular form. Section 4 discusses the evolution of CR technology related to SS and 5G applications. Section 5 discusses the issues, challenges and future research directions of SS and CR. Section 6 concludes the work. Table 1 lists the acronyms and notations used in the paper.

II. 5G ENABLING TECHNOLOGIES The deployment of 5G systems has sparked countless aca- demic studies and greatly benefited society. A few impor- tant enablers empower 5G technology in communication networks, as illustrated in Figure 2. Each enabler has its own features, and the combinations of enablers define 5G technology.

TABLE 1. Definitions of acronyms and notations.

14462 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 2. Important 5G enablers.

A. HETEROGENEOUS NETWORK Currently, the emergence of various devices using wireless and IoT technologies is increasing remarkably. Networks that combine different cell types and access technologies are called HetNets. An example of the HetNet architecture is shown in Figure 3. Various types of devices are connected to a femtocell base station (BS) covering multiple small subsets of macro BS. Given that the network is flexible, further research on implementing this network is important to avoid interference and fulfil the Quality of Service (QoS) promise to end users [18]. Therefore, researchers are eval- uating the performance of FiWi-enhanced LTE Advanced (LTE-A) HetNets. For example, Beyranvand et al. [19] inves- tigated the temporal and spatial probability, delay, maximum aggregate throughput and offloading efficiency of FiWi con- nectivity. They proposed an algorithm to improve the max- imum aggregate throughput performance of FiWi and used the algorithm to enhance LTE-A HetNets as an initial step towards achieving 5G.

A study [21] conducted a comprehensive review of the challenges, technologies and potential use cases of HetNets, with focus on implementing mmWave and massive MIMO in 5G networks. Several challenges that are crucial for effec- tive deployment, such as issues in network planning, traffic management and radio resource management, were reported. The solutions to these issues were presented in several papers. For instance, in [22], the authors improved the call ses- sion control function server by adding traffic prediction, bandwidth negotiation and connection admission control to improve traffic management and increase the accuracy of traffic forecasting without sacrificing the prediction accuracy of the system. For ideal large-data streaming in 5G HetNets, an improved version of traditional traffic prediction was pro- posed by [23]. Another study proposed the development of

a linear predictor that uses compressed sensing by adopting support vector classification. The proposed predictor has a simple structure, and its results are promising. The predictor’s performance is better than that of the traditional load predic- tion method. With regard to HetNet network planning, [24] presented a fast handover technique that uses a wireless link signature based on the user location as the handover authen- tication data. The techniques are time-varying, unpredictable and secured with physical encryption to guarantee a distinct and safe handover.

B. MASSIVE MIMO Another important 5G enabler is massive MIMO, where data rates are increasing with reduced interference by using the beamforming technique to focus signals on one another [2]. The use of massive MIMO provides low latency and achieves EE communication, which is suitable for 5G develop- ment [25]. Massive MIMO is implemented by adopting large- scale and advanced antenna arrays whose width and tilt can be controlled vertically and horizontally. An example of massive MIMO implementation is shown in Figure 4. A uniform planar array, which can be rectangular, hexag- onal or circular, is used [26]. Enabling massive MIMO requires regulatory masks to support its statistical nature, and spectrum regulation management must be enhanced to consider time, spatial and direction domains. A recent work addressed Non-Orthogonal Multiple Access (NOMA) in various forms of MIMO-NOMA transmission protocols, designed a cooperative NOMA and identified the relationship between two 5G technologies (NOMA and CR) [27]. The security issues in 5G networks, specifically in the physi- cal layer of massive MIMO, HetNet and mmWave, were also discussed in [27]. A comprehensive survey was con- ducted by [28], who addressed physical layer technologies, including massive MIMO, new channel model estimation, directional antenna design, beamforming algorithms, Media Access Control (MAC) layer protocols and multiplexing schemes.

Beamforming is a spatial filtering technique that aims to enhance spectral and energy efficiency and increase system security. Hybrid beamforming is a method that combines digital and analogue beamforming. Fully digital beamform- ing requires a complete radio frequency chain behind each antenna and is impractical for mmWave. Sun et al. [29] explored the multi-cell, multi-user, multi-stream commu- nication in mmWave homogeneous networks. The follow- ing hybrid beamforming techniques were proposed and compared: (1) leakage-suppressing and signal-maximising precoding, (2) signal-to-leakage-plus-noise-ratio-based pre- coding, (3) generalised maximum-ratio precoding and (4) feasibility of zero-forcing precoding. The highest spectral efficiency was obtained with the second method. The digital predistortion issue in massive MIMO transmitters for lin- earising the radio frequency hybrid beamforming array was addressed by [30]. The solution involved the linearisation of the main beam direction for its combined far-field signal,

VOLUME 8, 2020 14463

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 3. Example of HetNet architecture [20].

FIGURE 4. Uniform Planar Array configurations, from left to right: rectangular, hexagonal or circular [26].

which can effectively broaden the linear angle range. Hybrid beamforming for single users was applied in mmWave mas- sive MIMO by [31] by using the dual-stage approach based on singular value decomposition and zero-forcing. An issue was identified in the singular value decomposition algorithm, that is, too small or too large channel matrix eigenvalues limit the frequency-selective channels. Adaptive beamforming is a versatile approach of detecting and estimating the signal of interest at the output sensor array via data-adaptive spatial or spatiotemporal and interference cancellation. Research on this method in moving vehicles was conducted by [32] by using a predictor antenna with a 64-element massive MIMO for complex OFDM downlink channels. The obtained accuracy was close to the ideal beamforming gain for non- line-of-sight channels. Further investigation of real-time pre- diction with realistic time-frame structures in time division duplex (TDD) systems was suggested. Pei etal. [33] emulated the line-of-sight channel accurately via pre-faded synthesis and analysed over-the-air 5G cellular communication with

adaptive beamforming using a sectored multi-probe anechoic chamber. A recent study by [34] tackled the interference issue by using the adaptive beamforming algorithm to miti- gate interference. The algorithm achieves coherence between different beams, and each beam is suitable for a specific terminal. The beamwidth of the main lobe is narrow, and massive MIMO systems can manage the good sectorisation between user equipment without interference. The network capacity and data rate can also be increased.

A highly potential 5G network infrastructure for com- munication known as cell-free massive MIMO was intro- duced by [35]. It entails joint signal processing from many distributed access points (APs) and can offer similar QoS to all user equipment despite its low complexity. An illus- tration of how cell-free and cellular differs is shown in Figure 5. Each AP in cellular massive MIMO is serving an exclusive subset of the users. While cell-free massive MIMO has many distributed APs that are jointly serving the users with the signal encoding/decoding taking place in a CPU. Björnson and Sanguinetti [36] performed the first comparison of cell-free and cellular massive MIMOs. Cell- free massive MIMO was implemented with four levels of receiver cooperation from fully centralised to fully dis- tributed with multi-antenna APs. Cell-free massive MIMO exhibited higher spectral efficiency for all user equipment compared with cellular massive MIMO. Through mini- mum mean-squared error processing, the received and esti- mated signals are sent to a central processing unit instead of being preprocessed at APs. Ullah et al. [37] proposed

14464 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 5. Comparison between cellular and cell-free massive MIMO systems [39].

training size optimisation for cell-free massive MIMO sys- tems that is effective when the coherence is low or the number of users is very large. The method can achieve a higher downlink rate than the conventional pilot length method. Zhang et al. [38] conducted a comprehensive survey on cell-free massive MIMO systems by exploiting chan- nel hardening and favourable propagation conditions. The exploitation aimed to reduce the transmission energy and inter-cell interference in a centralised or distributed man- ner. The study showed that cell-free systems provide better coverage than conventional cellular systems and uncoordi- nated small cells. Chen and Björnson [39] investigated a cell- free massive MIMO network with channel hardening and favourable propagation from a stochastic geometry perspec- tive. They found that hardening can be obtained efficiently by deploying multiple antennas per AP, and having a few multi- antenna APs is better than having many single-antenna APs. Users that are spatially well-separated communicate with dif- ferent APs subsets and thus exhibit favourable propagation. However, [39] suggested not to rely on both properties when designing and analysing cell-free networks; instead, resource allocation schemes and achievable rate expressions that work well without the two properties should be used.

The 5G network is expected to serve a massive num- ber of users and support instantaneous demand variations at different times and events. Therefore, NOMA has been explored as one of the promising solutions to this problem. An example of work in NOMA is the study of [40], who elaborated on the working principle of the uplink NOMA framework. The data of two users are transmitted simultane- ously, and frequency resources and Successive Interference Cancellation (SIC) are used for reliable data transmission. The authors also presented the challenges in implementing a massive number of IoT devices in 5G cellular networks, such as BS and traffic estimation, channel estimation, interference management, power allocation and management of device synchronisation. For NOMA-based multi-user MIMO, [41] presented a joint interference alignment with the power allo- cation framework to overcome the immense increase in traffic in the data network. The framework utilises the sum rate

to provide QoS and guarantee an effective SIC constraint whilst managing the power to reduce interference within a cluster. Furthermore, a NOMA two-way relaying method was developed by [42] by using Karush–Kuhn–Tucker conditions with dual composition techniques. The relay divides single- source data into two parallel parts, which are then transmitted using amplify-and-forward relay. The method can be further expanded to multiple users. The use of NOMA also aims to provide control over the complexity of data processing and signal overhead in 5G networks. In the study of [43], compressed sensing for NOMA for mMTC was presented. The proposed schemes use low coherent spreading, which can improve spectral efficiency and reduce latency by enabling user data detection and joint activity without activity infor- mation from the user.

C. ULTRA-LEAN DESIGN Future 5G technology is expected to enable an ultra-lean design [2], [44] where ‘always-on’ signals are reduced to a bare minimum to achieve an EE network at a low opera- tional cost, as shown in Figure 6. The implementation of this ultra-lean design can reduce network transmissions without affecting user data delivery [45]. This feature is critical for very dense local areas to reduce the overall interference level for end-user performance at low-to-medium loads and is essential to high-frequency bands where networks are yet to be deployed. Ultra-lean design also needs special attention in terms of backward compatibility for low-frequency bands because a large number of terminals are already deployed. Moreover, the implementation of ultra-lean design in Ultra- Dense Network (UDN) has exhibited significant improve- ment in enhancing mobility support, increasing throughput and saving energy as experimentally confirmed by [46], who provided future research insights into the deployment of 5G.

D. ULTRA-RELIABLE LOW-LATENCY COMMUNICATION Another important promise of 5G is the ultra-low latency enabler [2] that can reduce processing delays and transmis- sion time intervals and widen the bandwidth of radio resource blocks in which a specific amount of data is transmitted.

VOLUME 8, 2020 14465

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 6. Ultra-lean design for 5G data transmission [47].

This feature can also avoid queuing delays at the radio trans- mitter whilst the direct communication link (i.e. D2D) pro- vides low-latency transmission for devices in close proximity. To achieve this vision, the physical channel structure must be designed for fast decoding at the receiver, and the MAC has to enable immediate access. Collision risks must be minimised by providing dimensioned instant-access resource alloca- tions. Several recent studies performed detailed analyses to achieve this low latency requirement. Lauridsen et al. [48] measured the performance of current LTE network imple- mentations and compared it with the initial LTE requirements. They also identified key performance indicators for consider- ation in designing and standardising 5G technology, includ- ing an analysis of critical connected mobility parameters, such as user and control plane latency, handover execution time and coverage. In 2018, Moradi et al. [49] proposed a scalable architecture by combining SDN and NFV for a customisable and low-latency 5G core network. The system enabled the creation of custom services for user equipment with low latency and efficient signalling.

E. ENABLERS FOR mMTC 5G technology also needs enablers for mMTC [2] to achieve ‘zero-overhead’ communication by simplifying the connec- tivity states of devices and providing channel access with minimal signalling. Maximising the devices’ sleep opportu- nities is also useful in reducing energy consumption. Doing so would lead to long battery life, and devices can operate for years with small batteries. Optical transmission modes are also needed to provide connectivity at low rates. This feature can be attained by providing a ‘spectrum-compatible’ interface for the best coexistence with legacy radio technol- ogy. Jovović et al. [50] presented an overview of a next- generation mobile network that integrates all machines and devices via the IoT concept enabled by the deployment of mMTC in 5G network. The main advantages of 5G machine communication is the increased data speed transmission and capacity of up to 1 Gbps, with latency as low as 1 ms. mMTC is also targeted to support emerging services and applications based on the IoT concept. With the implementation of 5G in the future, mMTC will not only promote the IoT concept but will also provide an open opportunity to explore the implementation of the Internet of Everything (IoE).

F. NETWORK MANAGEMENT Managing a network is one of the crucial issues in any communication technology. Given the multitude of exist- ing services and the additional services to be offered in the new economic sector using 5G technology, intelligent means of managing the network efficiently are urgently needed. Autonomous NMs that have self-awareness, self- configuration, self-optimisation and self-healing proper- ties are necessary for enhanced cost and energy savings. Pulcini et al. [51] demonstrated the use of specific proce- dures based on carrier Ethernet for the reliable management of HetNets dedicated to 5G networks using SDN configured for the implementation of energy saving tasks and QoS con- trol. Self-Organising Network (SON) management in 5G was identified by Moysen and Giupponi [52] as the key driver of improvements in operation, administration and management activities with minimal human intervention. The authors also reviewed the basic concepts and taxonomy of SON, NM and machine learning (ML). SON reduces the installation and management costs of 5G by simplifying operational tasks due to its capability of configuring, optimising and healing itself. The autonomous management vision using SON is expected to be extended to end-to-end networks to satisfy 5G NM requirements.

The planning of ultra-dense wireless networks [53] and the characteristics of typical UDNs that appear in a 5G network were explained by [54]. The study included a comparison with traditional cellular networks based on the definition by ITU. The key issues in applying UDNs in 5G were identi- fied as follows: network architecture and protocol procedure enhancements, interference avoidance and inter-cell coordi- nation, EE and super SON. To address one of these UDN key issues, a survey was conducted on how ML solutions can benefit 5G SON management with an end-to-end per- spective [52]. The key improvement of SON is its capability to configure, optimise and heal itself. The evolution of SON in 3GPP was also presented together with detailed implemen- tations of SON on different architectures. The elements that contributed to SON’s evolution were discussed thoroughly, including self-configuration, self-optimisation, self-healing, self-coordination, minimisation of drive tests, core networks and SON in virtualised and softwarised 5G architecture. An overview of ML-based NM’s relevant literature based on these elements was also presented in a comprehensive manner. Moreover, a high-level classification of different NM problems in managing the SON network was meticulously performed. For each class of problem, the authors identified the ML tools that can be used.

The multiple access and multi-service features envisioned in 5G can be realised by slicing a single physical network into multiple isolated logical networks. Foukas et al. [55] presented a review of 5G network slicing in different aspects, namely, virtualisation of radio resources (dedicated or shared resources), granularity of network functions (coarse or fine grained) and service description (human-readable format or set of functions and network components). The challenges

14466 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

that need to be addressed before fully realising the vision of network slicing based on multi-service softwarised 5G mobile network architecture are RAN virtualisation, service composition with fine-grained network functions and end-to- end slice orchestration and management.

The security aspect of 5G wireless communication net- works was explored by [56], specifically in HetNets (spa- tial modelling, mobile association and device connection), massive MIMO systems (low power consumption, TDD operation, artificial noise, antenna correlation, confiden- tial broadcasting and hardware impairments) and mmWave communication (large bandwidth, short range transmission, directionality and large antenna arrays). The authors sug- gested appropriate ways of safeguarding the said technologies by using physical layer security. Another work conducted a survey of existing authentication and privacy-preserving schemes for 4G and 5G cellular networks [57]. The authors classified the schemes into seven types: (1) handover authen- tication with privacy, (2) mutual authentication with privacy, (3) radio frequency identification authentication with privacy, (4) deniable authentication with privacy, (5) authentication with mutual anonymity, (6) authentication and key agree- ment with privacy and (7) three-factor authentication with privacy. The authors also presented a thorough survey of the techniques for threat models (attacks against privacy, integrity, availability and authentication), countermeasures (using cryptography methods, human factors and intrusion detection methods) and security analysis (informal and for- mal techniques). The opportunities and challenges for secu- rity designers in planning future 5G networks were likewise identified [56]. The concept of 5G cellular network was studied comprehensively by [58].

G. CONVERGENCE OF ACCESS AND BACKHAUL The convergence of access and backhaul [2] can simplify the deployment of wireless connectivity between radio network nodes. Convergence is an alternative to relying on optical fibre alone. It is achieved with the exploration of high- frequency bands, extensive beamforming and low-latency transmission. The combination of these methods can provide high bandwidth that is comparable to an optical link. Conver- gence can be accomplished with the dynamic split of spec- trum resources between access and backhaul for increased efficiency by using the same radio interface for both links. The links can also be used for operational and maintenance systems. Convergence backhaul sharing towards realising 5G was studied by [19] for FiWi-enhanced LTE-A HetNets. Meanwhile, in [59], the authors addressed the convergence of access and virtual backhaul networks. It was implemented with an SDN NFV orchestrator to serve mobile network operator capacity requests automatically by computing and allocating virtual backhaul tenants.

Six major requirements and challenges in preparing 5G backhaul solutions were enumerated by [60]; these six are capacity, availability, cost, long-distance reach requirement, ultra-low latency and UDN. The issues in backhaul access

that are foreseen to arise in the 5G environment were dis- cussed by [61]. Backhaul will be used to connect the core net- work with ultra-dense and heavy traffic cells, thus exposing it to extreme requirements in terms of latency, capacity, energy, cost efficiency and availability. The authors discussed 5G backhaul with evolving and disrupting 5G features, the evo- lution to a 5G cellular backhaul network in terms of the requirements, the tailored solutions and trade-off between centralised RANs and the fronthaul cost from the perspective of joint backhaul/RAN. The advantages and shortcomings of the backhaul solution, the unsolved challenges and the consolidated 5G backhaul vision were also presented. The enabler and challenges of 5G RAN slicing for vertical spec- trum were explained in detail by [62].

With regard to channel assignment in backhaul/access in SC, a comparative observation was performed by [18] to determine the optimal location for backhaul channels by using the maximum received signal power (max-RSP) and the minimum received signal power (min-RSP). The results showed that the min-RSP model outperforms the max-RSP model, implying that backhaul channels are more suitable for use as the cell edge of secondary BS. These channels allow numerous high-density channels in the network to be a base study for future backhaul implementation in 5G networks. Furthermore, a focused study on mobility management for joint access backhaul was conducted by [25], who studied the previous development of backhaul scenarios and listed differ- ent approaches in designing joint-access backhaul solutions. The effects of the future backhaul/fronthaul condition on 5G mobility management were also investigated. Intelligence is required in deploying 5G backhaul solutions. Several other aspects, such as adaptive and dynamic adoption and alloca- tion strategies, were also explored by [61] as future research directions.

H. mmWave For resolving spectrum scarcity issues to realise 5G archi- tecture, mmWave is one of the solutions explored in many studies. mmWave is an ultra-high-frequency band that ranges from 30 GHz to 300 GHz [12]. It provides extremely high data rates, ultra-high capacity, very large bandwidth and very low latency for new services and the economy sector that can benefit from 5G. Authors have discussed the characteristics of mmWave communication, such as wireless channel mea- surement, directivity and sensitivity to blockage, standardis- ation of mmWave, challenges and existing solutions in terms of integrated circuits and system design, interference man- agement and spatial reuse, anti-blockage and dynamics due to user mobility and mmWave applications, such as SC access, cellular access and wireless backhaul. Investigations on 5G SS were conducted by [3] for the spectrum below 6 GHz and within the mmWave range. For the application of 5G access fronthaul, the mmWave signal frequency used was 60 GHz which was generated using a dual-wavelength fibre laser [63]. The same authors presented a radio over a fibre system for the transmission of OFDM with 5 GHz bandwidth.

VOLUME 8, 2020 14467

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

Recently, Samal et al. [14] conducted a study on single- carrier modulation schemes that are suitable for mmWave and provided suggestions on using modern signalling schemes for smart grid communication technologies. An entire chapter in the paper of [64] described mmWave in terms of its applications, frequency spectrum, characteristics and stan- dard channel models for 5G networks. They also discussed EE in networks operating at mmWave frequencies, antenna technology, cognitive radio network (CRN) and mmWave technology and the optimisation and projects related to mmWave.

The millimetre band of the unlicensed 59 GHz to 64 GHz is designated for future research on 5G to achieve increased data capacity, throughput and low latency for future dense network traffic. However, using mmWave as the data carrier entails limitations, especially in utilising the 60 GHz band. Generally, mmWave has a short wavelength that varies from 1 mm for 300 GHz to 10 mm for 30 GHz. Given the short wavelength characteristics, mmWave can only travel in less than a kilometre and can be easily blocked by any obstacle along the path. However, this characteristic is useful for object imaging and short-distance networking. For example, the 60 GHz frequency band is used in IEEE 802.11ad as a wireless networking standard [65] for Wireless Fidelity (WiFi) in close areas [66]. As mentioned by [68], mmWave has high oxygen absorption and atmospheric attenuation, which further shorten the coverage and strength of the waves. In addition, mmWave travelling in rain or highly humid con- ditions experiences signal loss and distortion. For countries with high humidity and heavy rainfall all year long, such as Malaysia [67], implementing mmWave, especially at 60 GHz, is a challenge due to the unsuitable weather [68]. However, the advantages of using mmWave in dense network traffic are too valuable to ignore. Flament [68] addressed the problem of frequency scattering that affects mmWave’s throughput performance. The proposed solutions involved using multiple techniques, such as beamforming and leveraging the reflec- tion and refraction of the radio path for D2D communication. Another improvement was proposed by [69]. The proposed solution improves microstrip antenna design with an inserted slot to increase the frequency gain in utilising the 60 GHz band for a point-to-point 5G communication system. Other possibilities are available for implementing the 60 GHz band or utilising other suitable bands for an alternative 5G car- rier frequency in unfavourable climate conditions, and they require further exploration.

I. SPECTRUM SHARING AND FLEXIBILITY SS and flexibility are amongst the most important issues to be addressed in overcoming spectrum scarcity and utilising the spectrum band in 5G technology. SS is a means to optimise spectrum usage by enabling the sharing of the same frequency band amongst multiple users of different priorities without impeding one another. Transmission for NR uplinks and downlinks occurs at new (increased frequency) bands for 5G. LTE/NR uplink SS is one of the 5G enablers because it can

TABLE 2. 5G frequency spectrum [4].

co-exist with 5G and share the same low-frequency bands. 5G NR uplink can also exploit the low-frequency bands used by operators for LTE.

In terms of spectrum flexibility, a new regulatory frame- work is needed for a fully flexible radio interface design and the dynamic use of paired and unpaired frequencies [2]. This will allow for smooth adaptation to service requirements, such that the downlink spectrum resource is usable for uplink transmission and vice versa. Future systems are also expected to provide a high degree of spectrum flexibility for high- frequency bands to allow the use of an unlicensed spectrum in boosting the capacity [2] for critical control signalling and mobility handling, preferably in combination with a licensed spectrum.

Spectrum allocation for 5G is categorised into three main bands, namely, low, high and very high, as summarised in Table 2. The spectrum at frequencies below 1 GHz, partic- ularly at 700 MHz [3], [4], enables 5G coverage in wide areas and deep indoor coverage. The spectrum at high frequencies with relatively large bandwidths below 6 GHz (at 3.4 GHz to 3.8 GHz) [3], [4] provides the necessary capacity to sup- port numerous connected devices and ensure high speed for concurrently connected devices. This spectrum delivers the best compromise between capacity and coverage. At very high frequencies above 24 GHz (e.g. 24.25 GHz to 27.5 GHz) with very large bandwidths, the spectrum provides ultra-high capacity and very low latency [3], [4]. The cells at these frequencies have a small coverage (from 50 m to 200 m). The build-out of 5G networks in mmWave bands will initially be focused on areas with high traffic demand or specific loca- tions or premises requiring services with extremely high data rates (in Gbps). This ‘pioneer’ mmWave band also provides ultra-high capacity for innovative new services, thus enabling new business models and sectors of the economy to benefit from 5G [3], [4]. An overview of the 5G spectrum and its uses is illustrated in Figure 7.

The C-band of the spectrum, which ranges from 3300 MHz to 5000 MHz, is designated as the primary frequency band to introduce 5G in the year 2020. The channel bandwidth provided for 5G must be at least 100 MHz per network to meet the requirements. The implementation is very cost-efficient

14468 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 7. Frequency spectrum for 5G and its uses [4].

because the channel capacity can be enhanced without net- work densification costs. 5G’s use of the C-band can be realised by the adoption of massive MIMO for its acceptable complexity and capability to boost peak, average and cell edge throughput. The low frequency used for mobile can also be exploited by combining the 3300 MH to 3800 MHz fre- quency as one of the 5G features in 3GPP standards through the use of the co-existing LTE/NR uplink.

Amongst all 5G-enabling technologies presented, this work focuses on the potential of fully utilising the limited spectrum band by reviewing recent SS studies dedicated to 5G technology. The reason for such focus is the promising solution for licensed SS. The main advantages are improve- ments in spectrum utilisation and increment in capacity according to the different types of services required.

III. TOWARDS SPECTRUM SHARING IN 5G NETWORKS SS has been identified as a potential solution to spec- trum scarcity but was initially avoided by network providers/mobile operators in the past due to market competi- tion. However, sharing activities have been observed between mobile operators in recent years mainly due to the limited infrastructure. With SS, capacity demands have increased substantially despite the decrement in average revenue per user [70]. Another advantage of infrastructure sharing is the multitude of costs related to network deployment and maintenance, which leads to low capital expenditure. This is especially true when traffic is low in remote areas. For dynamic and asymmetric traffic, multi-operator SS is ben- eficial when set in a cooperative manner. However, creating a trustworthy relation between operators is the main issue in the aspects of sharing fairness, data transparency and service quality agreements.

A. CLASSIFICATION OF SPECTRUM SHARING TECHNIQUES Figure 8 illustrates SS techniques, which can be divided into three main approaches: network architecture, spec- trum allocation behaviour and spectrum access method.

Network architecture can be either centralised (infrastructure- oriented) or distributed (infrastructure-less) [71]. A com- parative analysis of both approaches with different per- formance evaluation parameters and the associated chal- lenges were presented in [72]. Spectrum allocation behaviour can either be in a cooperative or non-cooperative manner. Dludla et al. [73] discussed both behaviours, specifically in Licensed Shared Access (LSA) (non-cooperative) and Television White Space (TVWS) (cooperative). Meanwhile, the spectrum access method is divided into three techniques: (1) dynamic exclusive that consists of a commons model, shared use and exclusive use [17]; (2) open access; and (3) hierarchical model that consists of overlay, underlay and interweave. The commons model and open access are tech- nologies that can enable simultaneous spectrum usage but with limitations, as discussed in [74]. The commons model uses self-regulation, which limits the number of users. The users are in charge of co-managing spectrum usage. This model also has a regulator that assigns and enforces rights but is not responsible for managing the spectrum. The open access model has unrestricted access to the spectrum and can be used by any service or application. No usage rights are enforced, but users must comply with certain rules defined by National Regulatory Authorities (NRAs). Fees may be applied for NRAs to be borne by the users.

The hierarchical model is also known as vertical SS (ITU-R 2014) or primary–secondary sharing [75]. The differ- ences between vertical and horizontal spectrum access were explained by [76] with different sharing scenarios associated with licensed bands from the mobile operators’ perspec- tive. The overlay model, which is based on CR technology, is divided into two other models: three-tier and database- assisted models. The database-assisted model has several other methods associated with it, including TVWS, LSA and Spectrum Access System (SAS).

In the underlay and overlay mechanisms of the hierarchical model, secondary users (SUs)) can transmit simultaneously with primary users (PUs) as long as performance degradation does not occur in PUs. However, the underlay model has

VOLUME 8, 2020 14469

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 8. Classification of SS techniques.

a predefined noise threshold to protect PUs from potential interference, and SUs has to obey it by tuning its power level or must liaise with PU to avoid performance degradation. The interweave mechanism is an opportunistic access model where SU can dynamically access the licensed spectrum on an opportunistic basis whenever PU is not using the spectrum. It can be either in a single domain or in a combination of temporal, frequency and spatial domains [76]. SU transmits over spectrum holes in the licensed band in a specific geo- graphical or time domain. Spectrum holes are the temporary space–time–frequency that is not in use by any licensed or unlicensed user and changes over time [77]. The interweave model also requires information on the user’s activity from the spectrum. Spectrum occupancy is periodically monitored and detected by the system. Minimal interference occurs when the system communicates opportunistically over avail- able spectrum holes.

Several SS licenses can be applied for 5G use, such as exclusive use (individual licenses), license-exempt rules (unlicensed or commons), LSA and authorised shared access, citizen broadband radio service with SAS, pluralistic licens- ing and Licensed Assisted Access (LAA) co-primary shared access. An overview of these license schemes was presented by [70], with legal regulations (in USA, Europe and else- where) and trial activities for SS schemes (LSA, LAA and SAS) as well as the characteristics of SS from inter-operator and virtualised network architecture perspective.

B. PREVIOUS SURVEY ON SPECTRUM SHARING An early survey on the convergence of the current network to a heterogeneous mobile network for 5G communication was conducted by [78] in consideration of the communication

of machine-to-machine (M2M) application. The convergence of mobile networks combines device convergence, protocol convergence and service convergence, which are expected to improve QoS. In the future development of M2M technology, the first step is to enhance current OFDMA and MIMO anten- nas with improvements in end-user device performance. This task can be done by introducing a new network architecture that can support cellular and M2M interconnection. Further- more, an improved traffic balancing model is also needed for QoS control policies. Interference is one of the challenges in using M2M and cellular in the same spectrum. Therefore, Coordinated Multipoint (CoMP) for interference avoidance and M2M integration for HetNet are suggested for future research.

A critical insight into the challenges in expanding future research on CR in 5G was provided by [79]. The enabling devices are specifically found in the CR terminal and expected to be developed in software-defined devices, and the outcome will allow researchers to study spectrum-related matters more efficiently. The authors specified the important aspects of cognitive terminal challenges, such as adaptive prediction algorithms, seamless connectivity, fast and reliable reconfigurable hardware and dynamic spectrum allocation. For example, with dynamic spectrum allocation, the spectrum can be utilised more efficiently without reducing the trans- mission quality. However, future research on the implemen- tation of adaptive power and spectrum allocation in MAC is needed. Furthermore, the ability to exchange information is also required to enable high QoS amongst the heterogeneous devices in a 5G network. Problems in security, interference and 5G terminal locations were also critically reviewed in the paper.

14470 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

The survey on emerging technologies for 5G architecture conducted by [7] focused on massive MIMO, D2D and net- work architecture. The 5G network architecture was sug- gested to improve the current network architecture, such as OFDMA, which can still be used for a long time. However, with the current cellular architecture, users inside a building experience high signal penetration loss and reduced data rate and spectral efficiency. Arrays of antennas utilising massive MIMO technology are the enabler for this challenge. SS and shared horizontal and vertical spectrum will be used to obtain increased spectrum capacity for the usage of various devices in the future. To enable SS for the 5G mobile broadband system, two SS techniques were identified: distributed and centralised solutions. The distributed solution is highly effi- cient when used in a local framework, and the centralised solution is useful for systems with higher granularity than a radio system. This solution divides users on orthogonal resources with incomplete data but has benefits in certainty, control and reliability.

The well-known fundamental challenge in SS is to reliably identify when and where spectrum holes exist. This task is achieved by implementing spectrum sensing and spectrum prediction in the system. The work of [80] verified the fun- damentals of predictable and unforeseeable underlying real- world Radio Spectrum State (RSS). The authors provided technical guidance on how to apply the predicted RSS in future wireless communication design. The authors also sug- gested several applications of spectrum prediction in SS to meet 5G requirements, such as cost-efficient wideband car- rier aggregation, dynamic frequency selection and predictive interference mitigation.

A comprehensive survey on the licensed SS mechanism for mobile operators was conducted by [76]. The authors classified the authorisation regimes into three categories, namely, individual authorisation, light licensing and general authorisation. Their focus was on individual authorisation, for which state-of-the-art co-primary shared access (shared use between mobile operators) and licensed/authorised shared access/spectrum access sharing (shared use of license bands between incumbent and mobile operators) were presented. The gaps were analysed, and the requirements needed for efficient sharing algorithms were presented. The paper con- cluded that the main challenge in SS is the need for efficient and cost-effective sharing schemes. Although other issues exist, such as interference, the authors believed that they do not exert a considerable effect on deployment due to enhanced management approaches. Other challenges in the SS scheme include integrating multiple layer protocols to per- form interference management, multi-band resource schedul- ing and accurate sensing with reduced signalling overhead. The scheduling algorithm for SS must be able to minimise delay by having on-demand and highly dynamic features. However, a trade-off between complexity and practicality must be considered.

In [81], the integration issue of five SS techniques was reviewed together with their potential challenges.

The techniques were CR, D2D communication, in-band FD communication, NOMA and LTE on unlicensed spectrum (LTE-U). The authors stated that if the spectrum in the 5G network is to be used exclusively, 76 GHz of spectrum resources are needed, but SS of only 25% of the total spec- trum resource is required. The main challenge addressed by CR in the 5G network is the requirement for an EE spectrum sensing scheme because 5G uses ultra-high-frequency bands, thus forcing users to scan a wider bandwidth than that at low-frequency bands. This task appears to be a challenge considering the battery limitation of terminals. A possible workaround for this is to scan a small band instead. However, this workaround imposes a different challenge in which the system must have the cognitive ability to learn from the scan history in order to accurately reduce the frequency range to the correct band. With the implementation of 5G, numerous nodes are expected to be connected to the network. The increased number of nodes in the 5G network is unsuitable for a centralised scheme due to the signalling overhead cost. Another challenge to consider is distributed interference mit- igation schemes. Spectrum scarcity increases with increasing nodes.

Although SS is beneficial, the challenges of its adoption must be clearly addressed by operators and competitors to ensure benefits to all parties involved. Furqan et al. [70] presented an overview of SS schemes considered for 5G networks and discussed SS techniques, legal regulations and trial activities in detail. Furthermore, a basis of spectrum management for SS was discussed. The study also addressed the challenges of SS in virtualised networks. Amongst the challenges is ensuring service differentiation, which is easy to achieve with exclusive spectrum access. However, as the spectrum is being shared, a more detailed concession is needed amongst competitors. Hence, fair rules and regula- tions must be adopted. Moreover, by sharing the spectrum, operators need to expose crucial information to competitors, such as traffic condition and bandwidth utilisation. Therefore, a regulator is needed to develop mutual trust with real-time updates to estimate the remaining resources accurately. With the implementation of SS, new network services or control functions are required to receive input and implement spec- trum management and sharing decisions. Given that operators are normally bound by a long-term contract on 5G, a study on operators’ willingness to share the spectrum is necessary. Furthermore, control of operators must be considered because several operators might opt for partial SS rather than full SS to prevent exposing internal network information. The place- ment of responsibility between sharing operators is another crucial issue in 5G.

SS classification can be divided into three parts: accord- ing to the spectrum access behaviour (cooperative or non-cooperative), network architecture (centralised or decen- tralised) and spectrum access methods (overlay or under- lay) [72]. Several of the challenges were addressed by [70]. These challenges included spectrum management and spec- trum assignment, metrics to quantify spectrum usage,

VOLUME 8, 2020 14471

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

interference coexistence and management, security and enforcement, protocols and standards and regulatory, pol- icy and economic problems. The challenges that were not addressed by Furqan et al. are listed below. • Radio software and hardware: Design of hardware and software for reconfiguration purposes, development of smart radio architectures and radio hardware to support operations in the mmWave band and design of sus- tainable low-powered devices and hardware that offer improved direction-finding, geolocation, etc.

• Experimentation, testing and standardisation: Develop- ment of test cases, virtual test beds to assess and model coexistence methods in large-scale environment and benchmark for designing test beds

The SS that is implemented in 4G uses a ‘one-fits-all’ architecture, which is unsuitable for 5G due to the vari- ous services offered, such as URLLC, eMBB and mMTC. Although numerous wireless nodes running these services are connected to the physical network, they have different requirements to be fulfilled. If spectrum resource allocation is not handled properly, users may experience deterioration in QoS, reduced data rates, unreliable transmission and high latency. A potential solution that was proposed by [82] is an SS scheme that uses the reservation-based sharing policy. This scheme is applicable to a 5G tenant-based cellular net- work with careful planning. The challenge presented in this study was the pursuit of a learning technique that has not yet been utilised for SS in multi-tenant cellular networks. The work was criticised for its wide use of stochastic learning, which is unsuitable for practical applications (such as in a power grid that requires high-speed protection).

The studies discussed in this section are compiled and presented in Table 3 chronologically.

C. RELATED WORK ON SPECTRUM SHARING TECHNIQUES This section presents related work on SS techniques that are relevant to 5G networks. The techniques are categorised into main SS approaches based on network architecture, spectrum allocation behaviour and spectrum access method. Related SS studies on EE improvement are also discussed.

1) NETWORK ARCHITECTURE An early work on SS based on network architecture was con- ducted by [84], who studied centralised SS using harmonised SDN-enabled approach (HSA). The system was utilised for synergistic SS by taking spectrum availability as the input. Incorrect decisions that are influenced by inconsistent QoS can be reduced using this method. This approach includes distributed sensing devices, BS and an SDN controller in the system. The centralised management minimises the SS processing by the BS. The SDN controller processes the information based on defined policies to allow PU or SU to enter and leave the network.

The spatial focusing effect was studied by [85] based on the trend of using massive MIMO and time reversal

wideband in 5G networks. The spatial white space SS con- cept was designed to enable SS for concurrent multi-users without orthogonal resource allocation. Through Signal-to- Interference-plus-Noise Ratio (SINR) analysis, centralised and distributed protocols were proposed to study the con- gruency of the wide time reversal band and massive MIMO system. The main design consideration was the downlink network path for multiple users with a spatial focusing effect from the multi-path channel. The outcome showed the similar performance of the distributed and optimal centralised proto- cols with low complexity in the time reversal wideband and massive MIMO system.

Zhang et al. [86] proposed an interference graph and a game theoretic approach to achieve joint optimisation for decentralised SS for reducing complexity and maximising user satisfaction. Concurrent Best Response Iterative (CBSI) and Spatial Adaptive Play Iterative (SAPI) algorithms were used to converge Nash Equilibrium (NE) for either global or local optimisation. NE was used to measure user satis- faction across the network corresponding to the global or local maximiser. The global optimal solution was obtained using the SAPI algorithm with a large learning parameter and arbitrarily high probability. The authors also suggested an improvement of service experience through user allocation according to small BSs and spectrum band suitability.

2) SPECTRUM ALLOCATION BEHAVIOUR A survey on the technical and economical views of coopera- tive and cognitive advanced SS and the differentiation classes of multi-level SS was conducted by Yang et al. [87]. The survey focused on spectrum trading and leasing, spectrum mobility, relaying, routing and harvesting. The SS schemes from the survey can enhance spectral and energy efficiency. The authors proposed a structure of 5G HetNets with the assumption that macro-cell eNodeB and SC eNodeB can use a cognitive application that can determine the capacity achieved and the SINR of the user equipment with its range. A poor SINR equipment is offloaded from MeNB. The cog- nitive SeNB also processes the requirement of offloading a user equipment to access the underutilised spectrum. The pro- posed advanced spectrum flow scheme was implemented on 5G HetNets, and it improved energy and spectral efficiency.

A hybrid SS system was introduced by [88]. The system is called Size-Negotiable Auction Mechanism (SNAM), and it combines auction and negotiation for multiple users. The SNAM system allows users to bid for the spectrum in per unit space and coverage areas as they prefer. To correctly model the interference, a mixed interference graph that can quantify up to five levels of interference was constructed to prove SNAM’s rationality. The system can accommodate many BSs and provide fair spectrum competition in the case of small firms competing with large ones. It can observe the different sizes of large to small radius ratios for interference-free areas. Overall, the proposed mechanism performs better than the undirected graph method in terms of spatial efficiency, seller revenue and buyer satisfaction.

14472 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

TABLE 3. Summary of previous survey/review papers related to SS with their focus and contributions.

A paid sharing approach based on fluid models was pro- posed by Rattaro et al. [89] to encourage the SS behaviour of PUs in CR networks. The idea is that SUs must pay for spectrum utilisation, but priority is strictly given to PUs over SUs. Reimbursement is provided to affected SUs, but a cost is implied for the PUs’ service provider. The approach was implemented by characterising the behaviour of the system using a fluid approximation stochastic model that considers pre-emptive situations with reimbursement, admission con- trol decisions and multi-resource allocation. The proposed model can also benefit other economic scenarios by mod- elling the dynamic control queuing system with the said pre- emptive situations.

A study on unlicensed SS that allows an unlicensed band to coexist with WiFi for 5G was conducted by Bairagi et al. [90]. The proposed SS scheme is called LTE

over an unlicensed band and improves users’ Quality of Expe- rience (QoE) whilst securing the necessary wireless system, such as WiFi. The proposed game-theoretic approach is a Vir- tual Coalition Formation Game (VCFG) that is divided into two sub-problems, where time sharing is solved by using the Kalai–Smorodinsky bargaining solution and resource allo- cation is implemented by the Q-learning algorithm. QoE is measured by the mean opinion score for selected applications, which are web browsing, video streaming and file download- ing, because the requirements differ for every application. The proposed method can reduce unsatisfied users, achieve good fairness and manage WiFi systems properly.

3) SPECTRUM ACCESS METHOD SAS is an emerging SS model that is currently gaining attention [70], [73], [76], [88], [91]–[93]. In the USA,

VOLUME 8, 2020 14473

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

SAS can be used by operators to access the 3.5 GHz military radar band for commercial use. This opportunity presented by SAS was exploited by [91] by proposing a scalable SAS framework that can manage the mMTC uplink interference to the incumbent with a low overhead. The database-driven dynamic spectrum access promoted by CR technology has issues in location privacy, where SUs must reveal their loca- tion in the process of querying and locating spectrum oppor- tunities for them to use. This issue is encountered in white space geo-location spectrum databases for locating the avail- able spectrum in TVWS. A multi-server Private Information Retrieval (PIR) was discussed by [94] as an enabler for private access to spectrum databases to protect the location information of SUs in database-driven SS. A database-driven CR network was designed to contain multiple synchronised spectrum databases that share the same content but are oper- ated by different service providers. The authors concluded that information-theoretic location privacy can be efficiently provided to SUs via multi-server PIR.

The analytic hierarchy process approach of SS was dis- cussed by [95] by addressing a solution to interference and performance improvement in 5G NR. The method is a mul- tilevel hierarchical structure consisting of objectives, criteria and alternatives. It can find solutions to complex problems and was derived from a set of pair-wise comparison matri- ces. The method enables the calculation of user priority to access the spectrum without any conflict of interest amongst other SUs. It has efficient spectrum utilisation even with the increased number of users in the systems.

An efficient network sharing scheme that uses a two-time- scale hierarchical model for wireless network virtualisation resource management was studied by Jiang et al. [96]. The spectrum is sliced into sectors for PUs, SUs and tertiary users. Each slice has its own subchannel and power for every time slot. The PU’s slice is allocated with its own subchannel and suitable power for guaranteed QoS, and the SU’s slice is assessed in a static manner. The tertiary user’s slice is integrated, and the idle spectrum band is assessed using a listen-before-talk scheme. The system is optimised with the Lyapunov algorithm for large-time-period frequency allocation, the time-slot subchannel and power scheduling algorithms. The scheme focuses on packet delay, data rate and IoT throughput for the three types of users. Simulation of the proposed scheme showed effectiveness in terms of delay and power allocation (PA) optimisation, but fairness amongst users has not been investigated yet.

The dynamical advance access SS method for public and dedicated telecommunication operators was introduced by Lin et al. [97] to ensure the QoE of SU. The system was modelled based on a finite-state Markov chain for the anal- ysis of the state transition model. The system introduces an enhanced mobility management entity that is responsible for SS management. The BS of both operators is set so they can only execute tasks from their own users whilst all service requests and dynamic spectrum access are performed by the central BS. The results showed that the queuing time and

proposed architecture are better than those of the shortest queue length access and QoE-driven access methods because the design considers the variation in channel condition and service state, resulting in better QoE performance.

A new hybrid spectrum access method was introduced by [98] to combine exclusive access and pooled spectrum access. This hybrid design aggregates the low-frequency car- rier for exclusive access and the high-frequency carrier for spectrum pooling. The method is reminiscent of the LAA structure in the Third Generation Partnership Programme (3GPP) system that aggregates the licensed carrier with the unlicensed spectrum. The proposed system was compared with exclusive and pooled systems for benchmarking. The results showed that the proposed approach provides bet- ter performance to the average user and a slightly better throughput compared with fully licensed and fully pooled approaches, respectively.

One of the promising multiple-access techniques, NOMA, was recently recognised to encourage SS with a significant improvement in the spectral efficiency of mobile communi- cation networks [99]. The technique is based on the power domain, unlike the previous generation of mobile networks that rely on the time/frequency/code domain. The technique also allows SS for multiple users in the same time-frequency resource block, where the set of users served by a BS via NOMA is known as a user cluster. Many researchers have demonstrated that NOMA is compatible with numer- ous 5G techniques. Latest innovations and applications of NOMA have been documented in the IEEE Access Special Section Editorial [100] to bridge the gap between theory and practice in designing NOMA for 5G. A study on the EE of PA was conducted by [101] for a MIMO–NOMA system with multiple users in a cluster and considers the QoS of all users. The authors proposed an optimal PA strategy to solve the EE maximisation problem when feasible. Moreover, a low- complexity user admission protocol was proposed where users are admitted in ascending order according to their power requirement (one by one) to satisfy their QoS requirements. The recent work of [102] presented a large cellular Poisson network that employs NOMA in the downlink and addresses the challenges in NOMA for multiple users with the use of SIC for decoding at each user. Three models for NOMA user clustering were proposed to highlight the importance of choosing network parameters and constraints and ordering to balance the cell sum rate and fairness requirements. The authors also showed that efficient interference-aware user clustering is important in improving performance in terms of maximising the cell sum rate. The cell sum rate can be maximised by allocating all resources (power in the NOMA network) to the best user [103].

Three key performance metrics have been identified in the development of a sustainable 5G system; these three are spectrum utilisation, energy consumption and cost efficiency. SS in 5G scenarios has been studied for many applications, such as improvement of EE and multiple spectrum access. The energy issue for the SS of cognitive SCs was studied

14474 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

by [104] in relation to the Energy-aware Dynamic Spectrum Sharing (EDSS) algorithm. They contributed to literature on the selection of the most appropriate mode in the downlink for all individual SCs with theoretical upper bound performance. A known problem in EE maximisation for underlay and over- lay SS is the non-convex fractional problem of the resource allocation system. Considering the scenario models of CR and D2D communication, Zappone et al. [105] proposed an optimal resource allocation policy for the SU through a suit- able reformulation of the original non-convex fractional prob- lem in the underlay model. For the overlay model, the same authors proposed two algorithms but with trade-off between complexity and optimality claims. An SS approach that con- forms with green communication and uses the zonal approach was presented by [106] for primary transmitter (PT)-forming D2D communication links with multiple primary receivers (PRs). The method has a low energy consumption at the PT with optimised transmission power within a zone that adequately meets the target rates demanded by PRs. The method was implemented using an iterative approach. It can improve the EE of the system and enhance the battery lifetime of the PT.

Zhou et al. [107] conducted a study on SS that focused on vehicle communication, specifically D2D communication. In their research, vehicle communication using SS was done to achieve an immersive experience with dedicated short- range communication. The architecture was explored for licensed and unlicensed spectra for improved efficiency and flexible usage. The proposed design also solves the vehicular- D2D (v-D2D) underlay shared spectrum resource problems of inconsistent TVWS spectrum due to location and time. The proposed architecture consists of four components, which are heterogeneous vehicular network, radio resource shar- ing cloud, the Internet and channel access. Moving forward, the research focused on mobility for vehicle communication and SS management for software-defined solutions.

For a more intelligent SS that supports the requirements of 5G wireless networks (high spectrum efficiency, massive connectivity, low latency and improved fairness), the inte- gration of NOMA and CR was investigated by [108]. State- of-the-art cognitive NOMA architectures, namely, underlay and overlay NOMA and CR-NOMA, were also discussed. NOMA principles exploit power domain multiplexing at transmitters for signal combination and SIC at receivers for signal detection. By integrating CR and NOMA, spectrum efficiency can be enhanced from different perspectives, such as improved spectrum efficiency, massive connectivity, low latency and enhanced fairness. The authors proposed coop- erative relying strategies in cognitive NOMA to improve reception reliability in addressing inter-network and intra- network interferences that cause major degradation in the performance of cognitive NOMA networks. The method can also be used to improve the cost of installing relays and has the potential to decrease outage probabilities significantly.

SS and spectrum aggregation utilising enhanced CRNs (E-CRNs) for sharing licensed TVWS spectrum, the LTE

TDD band and unlicensed spectrum bands were studied by [93]. The E-CRNs framework was embedded with Dynamic Spectrum Management (DSM) and the water-filled algorithm for the spectrum access method. The proposed E-CRNs system model implements imperfect spectrum sens- ing that guarantees PU priority and allows the SU interfer- ence during transmission to be controlled rather than perfect spectrum sensing which contributes several limitations. Fur- thermore, dynamic spectrum aggregation and spectrum lean management were introduced to reduce the interference to the WiFi network and maximise the system output, respectively. The numerical results proved that high system QoS perfor- mance, which contributes to good outage probability, sum rate and spectrum efficiency, is achieved. Holland [109] stud- ied the parameter performance of the TVWS band, especially during the aggregation of TVWS resources. They assessed the experiment with two conditions, which are mobile broadband downlink and indoor broadband provisioning. A study on link performance test was also conducted to identify the strength of using TVWS for indoor environment. The experi- ment was modelled using Carlson RuralConnect WSD, which was coded using the coded OFDM waveform and multiple modulation techniques that can be selected by the devices to evaluate the TVWS band’s availability and capacity.

In addressing spectrum scarcity using CR technology, traf- fic congestion can only be alleviated over the air, and the strain in the case of limited backhaul cannot be reduced. The heavy burden at the backhaul can be relieved using fog-computing-based wireless caching, which can be incor- porated into CR paradigms for improved performance. The primary transmission can be protected from unintended sec- ondary interference using an opportunistic spectrum access, namely, the guard zone and interference cancellation-based opportunistic spectrum access proposed by [110]. The method can enhance the spectrum accessibility of SUs.

Table 4 summarises the aforementioned SS techniques with their focus, contributions and key features and sorts them according to their SS approach, SS technique and year of publication.

IV. PROMISES OF COGNITIVE RADIO CR technology is one of the promising technologies that enable open SS for 5G. This emerging technology can satisfy the strict spectrum requirement of 5G networks. It has cogni- tive potential, is reconfigurable and its transmission parame- ters can be adjusted according to environment characteristics. The functions of CR include spectrum sensing, management, mobility and sharing. Spectrum sensing is a key function to detect the unused spectrum (spectrum hole). Spectrum hole is the temporary space-time-frequency that exists, is not used by any PU or SU and changes according to the time and place [77]. Spectrum management is utilised to determine the best channel for establishing communication based on user necessities. Spectrum mobility moves the spectrum with a low priority to the next vacant channel whenever PU is not in range. SS distributes the spectrum amongst SUs accordingly.

VOLUME 8, 2020 14475

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

TABLE 4. Summary of previous SS techniques with their focus, contributions and key features.

The handover procedure for CRN was described in detail by [71] as follows. Upon the arrival of PU, SU has to vacate

the occupied channel and is forcibly terminated. A method called fraction guard channel assignment performs this task,

14476 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

TABLE 4. (Continued.) Summary of previous SS techniques with their focus, contributions and key features.

resulting in throughput increment of unlicensed users. How- ever, the value cannot be attuned effectively. Four metrics were identified by [111] to present the performance of short- term and long-term spectrum handover. The metrics are link maintenance probability, number of spectrum handover, switching delay and non-completion probability. The com- plete CR cycle is shown in Figure 9. A survey on cognitive and cooperative SS schemes was

performed by [87], with focus on enhancing spectral and

energy efficiency in a cost-effective manner. The approaches to SS were discussed from two perspectives: economic mar- keting perspective, which emphasises spectrum trading, spec- trum leasing and multi-tier spectrum trading and leasing, and cross-layer technical implementation perspective, which discusses spectrum mobility, spectrum relaying, spectrum routing and spectrum harvesting. The buying and selling process and exchanging the rights to the radio spectrum are collectively known as spectrum trading, which also enhances

VOLUME 8, 2020 14477

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

FIGURE 9. CR functions’ cycle (adapted from [71]).

the utilisation of the radio spectrum. SS based on game theory is one of the market-driven spectrum trading schemes. Spectrum leasing is the act of leasing parts of the bandwidth by the PU who owns a given licensed spectrum bandwidth to SUs in exchange for economic revenue or technical coop- eration. Multi-tier spectrum trading and leasing were pro- posed in [87]. They have different tiers (upper, medium and low) to represent long, medium and short terms, respectively. In CRN, spectrum mobility enables the suspension of SU’s transmission to vacate the channel in order for PU to reclaim its licensed channel that has been temporarily leased to SU. The main aim is to guarantee the QoS and performance of PUs and SUs during the switching process. The ongoing data transmission of the SU must be transferred from the current channel to an alternative free channel via a process known as spectrum handoff. In spectrum mobility, PU has a higher priority than SU. Thus, SU must immediately vacate the licensed channel once it causes interference. The SU may experience spectrum mobility due to the degradation of link quality. Spectrum relaying can be enabled with relay-assisted protocol for spectrum mobility in CRN, where each SU has more than one connection path to the BS through dynamic spectrum relaying. In wireless communication, spectrum routing is a series of channel switching decisions by TVWS sharing subscribers when they foresee channel availabilities. Meanwhile, spectrum harvesting is a new idea for CRN, where SS occurs between PUs and cognitive users as the SUs. To maximise spectrum utilisation, a new service provider dedicated for SUs was introduced; it harvests the available spectrum bands with CR capabilities.

Spectrum access for dynamic sharing models is categorised as common, shared use and exclusive use. For CRs, the most promising model for satisfying the spectrum needs of users is the exclusive-use model with spectrum trading solutions, such as game theory, market equilibrium, classical optimisa- tion and their hybrid. The overview of the said trading model together with its features, limitations, feasibility and stability of the pricing solution was presented by [17].

Service provisioning challenges, such as providing channel access opportunities for new service requests and guarantee- ing continuous connections for ongoing flows until service completion, exist in wireless networks. These challenges can be addressed with the Dynamic Channel Reservation (DCR) algorithm and Dynamic Spectrum Access (DSA) scheme

with three access privilege variations to explore the advan- tages of channel reservation in performance improvement in error-prone channels [112]. DCR is used to reserve a dynamically adjustable number of channels for uninterrupted services to maintain service retainability for current users. Alternatively, it is also used to enhance channel availabil- ity for new users. The DSA scheme has a DCR algorithm embedded in it to enable spectrum access for PUs and SUs based on their licensed shared access. In addition, the DSA scheme is utilised to investigate the performance of CRN in homogeneous and heterogeneous channel failures by using the continuous-time Markov chain model.

CR is known to create opportunities for SUs to use spec- trum holes or white spaces that are not being used by PUs, but the main challenge is to recognise without fail the specific time and location of spectrum hole existence [80]. Previously, CR-based SS used opportunistic PU-SU access in unlicensed bands (such as TVWS) only, but the 5G requirement demands for sharing of both licensed and unlicensed bands. A new E-CRN based on SS and spectrum aggregation for 5G was introduced by Zhang et al. [93]. They exploited the licensed spectrum of PU networks, including TVWS and LTE TDD bands, and the unlicensed spectrum from industrial, scientific and medical bands. The framework for E-CRN includes DSM for licensed SS and unlicensed spectrum aggregation. The method can control harmful interference but with trade-off between sharing and aggregation efficiency. Thus, spectrum- lean management was introduced to achieve DSM. In addi- tion, a water-filling algorithm was proposed to access the available spectrum dynamically and assign system traffic offloading to the shared and aggregated spectrum bands. The authors in [93] also studied the co-existence of E-CRN and WiFi.

The CR in the 5G environment has the following charac- teristics: interoperability, context awareness, learning ability, self-optimisation, dynamic spectrum management, adaptive decoding and self-healing. It can be realised by using a CR terminal with several components, such as a software-defined device, a geo-locator, a learning system, a policy database, sensors, optimisation algorithms and the cognitive engine itself [79]. CR technology was initially implemented using half-duplex (HD) radio due to the hardware and software complexities of the superior FD communication [113]. After five decades, this FD CR technology gained serious attention with the recent advances in signal processing, machine-to- machine communication and deep learning methods. A sur- vey on SS in CR networks was conducted by [83] in four key steps: spectrum sensing, allocation, access and handoff; the key enabling technologies related to 5G were presented. The technologies were FD spectrum sensing, spectrum database- based spectrum sensing, compressive spectrum sensing and carrier aggregation-based spectrum allocation. Spectrum pre- diction and spectrum sensing can be used to reliably iden- tify when and where spectrum holes exist. Several of the applications of spectrum prediction in 5G SS include cost- efficient wideband carrier aggregation, dynamic frequency

14478 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

selection and predictive interference mitigation [80]. The theory of interplay between predictable and unforeseeable underlying real-world RSS was studied by [80]. Technical guidance on how to apply the predicted RSS to the design of future wireless communication was provided.

In terms of energy, the HD technique consumes relatively minimal energy due to its low complexity. Although the FD technique can enhance the gain in throughput [114], it is uncertain whether FD can outperform HD in EE. A com- parative study was conducted by Li et al. [115] in 2018 to analyse both techniques. The authors proposed an adaptive SS scheme that utilises the HD and FD of CR technol- ogy. Their analysis showed that the adaptive scheme has a higher EE than either techniques alone. In 2019, the authors developed a polarisation-enabled FD hybrid SS scheme [116] that guarantees a sufficiently small collision ratio for PUs and a larger throughput improvement compared with tra- ditional schemes. Another study on EE was conducted by Park and Hwang [117] for cognitive femtocells and picocells. They proposed an optimal EE PA using sensing-based SS for uplink multiple cognitive femto users operating in multiuser MIMO mode at the same frequency bands. Data transmission to a femto BS is performed in the presence of a pico BS and multiple pico users. The design adopts a user-wise EE with Pareto optimisation. An additional power constraint is introduced to convexify the non-convexity optimisation prob- lem without losing global optimality. Each femto user’s EE is quasi concave.

To improve EE in the CR network, [118] and [119] pro- posed an energy-harvesting-aided spectrum sensing and data transmission scheme by turning the harmful interference from PU into favourable factors for improving sensing per- formance and providing power to SUs. The scheme was developed by combining the individual and cooperative sens- ing of multi-users corresponding to the PUs’s strong and weak signals. A study on the trade-off of spectral EE in CR was conducted by [120] for link and system levels at various Signal-to-Noise Ratio (SNR) values. The amount of energy needed to achieve benchmarked CR spectral effi- ciency at the link level was investigated for low and high SNRs. Furthermore, the interference constraint was stud- ied to protect PR. The results showed that at a low SNR, constraining the energy level at average power provides a better EE than that at peak power. However, neither power constraint poses an impact on EE at a high SNR. At the system level, extreme value theories were implemented to study the CR EE for indoor environments, specifically in the uplink CR-cellular network. The result showed high spectral efficiency for a low PR, but the spectral efficiency rapidly drops as the number of PR increases. However, this situation can be compensated for by reducing the threshold of the inter- ference, which can be achieved by increasing the secondary receiver within the range of the secondary transmitter.

Wang et al. [121] considered the potential benefits of using massive MIMO in underlay CRNs. They derived the distribution of SINR for downlink transmission in secondary

network (SN) by considering the downlink multi-user MIMO transmission in the PU network. The measure was for the performance analysis of any finite number of antennas. An asymptotic analysis at the primary and secondary BSs for massive antenna arrays was also presented, particularly on the effect of a large number of PUs and imperfect channel state information on SN. Their analysis showed that the usage of large antenna arrays improves the power efficiency in SS networks. The application is for SS in multi-user MIMO primary and multiple-input single-output SN using the com- bination of primary and secondary BSs with single-antenna PU and SU, respectively. A comprehensive system analysis was performed to mathematically evaluate the lower bound of the average achievable rate.

A study by [113] presented the challenges in CR technol- ogy, including enhancing spectral efficiency and multi-user interference starting from HD and FD and how to implement them efficiently in 5G networks. The authors also addressed several recent works, specifically on cognitive and FD net- working, that focused on security, minimising delay, resource optimisation using the PA algorithm and asynchronous CR with the new MAC protocol. Physical layer security was addressed by [122] by proposing a cognitive security con- cept but with several challenges, such as condition detection, correct statement identification and security mechanism for many resources. Khan etal. [123] discussed the requirements, design challenges and review of routing and MAC protocols for CR-based smart grid systems. The mutual interference for spectrum users in the same band can be reduced by using a cognitive radar with spectrum sensing and transmission notching abilities. Ravenscroft et al. [124] conducted a study on the case where another spectrum user moves in the fre- quency during the radar’s coherent processing interval, and the radar emission is inherently robust to sidelobes that other- wise arise for spectral notching. The interference considered was in-band OFDM signals that hop around the band, and the fast spectrum sensing algorithm determined where notches are required. The efficacy factors for the proposed algorithm are the shape of the notch, maintaining the notch depth when generating the final emitted waveform, transmitter distortion and Doppler smearing due to notch hopping.

In CRN, a reliable spectrum detection depends on the accu- racy of PU location. A deep sensing algorithm was studied as a joint estimation for applications that require spectrum and location awareness. The emission state of a PU is iden- tified by estimating its evolving positions [125]. For mobile PUs, effective spectrum sensing can be realised by tracking their locations incessantly, and the location information can be utilised in cognitive performance optimisations. Another study was carried out to propose a wideband SS and compress the sensing method based on frequency-domain cyclic prefix autocorrelation. The study provided an accurate detection of OFDM-based primaries in wideband CR systems [126]. The method can overcome the noise uncertainty and frequency- selective multi-path channel effects in realistic wide- band communication scenarios and can utilise the spectral

VOLUME 8, 2020 14479

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

domain sparsity. In addition, the method can be combined with other wideband spectrum sensing approaches, such as sub-band energy detection. A wideband sensing platform can also run parallel sensing processes for different frequency channels and diverse types of primaries.

Spectrum decision and prediction were studied by [127] on the basis of a context-aware cloud-based spectrum monitor- ing system that utilises the benefit of the crowd-sensed data of static and mobile users with the cooperation of multiple SUs. The aim was to overcome the problem faced by SUs in finding idle frequency bands by using stand-alone spec- trum sensing that is prone to fading, shadowing and noise uncertainty effects and leads to undetected PU activity. The crowd-sensed data contained spectrum data and contextual information, such as time, location and building complexity. Current and future channel statuses were predicted using supervised learning methods, such as k-nearest neighbour, support vector machine and artificial neural network.

CR is very promising for sharing the licensed spectrum and enhancing system throughput. System robustness against errors can be improved by increasing the number of decoding blocks, but this will also increase the delay. By utilising the SS model, the authors in [128] proposed a minimisation of the outage probability over block-fading channels for optimal resource allocation under the PU outage constraint. The aim was to address the challenge of finding a trade-off between tuning the decoding blocks’ length and sharing the spectrum. The algorithm is based on alternating optimisation, which uses a problem with a verified strict quasi-convex structure. Additional sensing information can be obtained, and system performance is analysed under the CR constraint with or without sensing information. Another outage analysis was performed by [129] to investigate the outage performance of cellular underlay SS by exploiting multi-antenna and multi- user diversity using the transmit antenna selection algorithm and opportunistic scheduling. The authors derived the outage performance for a PU simultaneously with that for an SU PA.

On the architecture side, Lorenzo et al. [130] proposed a novel collaborative cognitive dynamic network architec- ture that incorporates cognitive capabilities to exploit the underutilised spectrum in a flexible and intelligent manner. The architecture leverages the density and heterogeneity of wireless devices to provide ubiquitous Internet connectivity. The design principles are aligned with the 5G functional- ity requirements, which are energy and spectrum efficiency, scalability, dynamic reconfigurability, support for multi-hop communication, infrastructure sharing and multi-operator cooperation. The work also illustrated the benefits of the cognitive dynamic network architecture in tackling user asso- ciations for data and spectrum trading with low complexity and enabling self-organising capabilities.

The transmission of video data for future multimedia traffic was studied in [131] by using CR to meet the critical spectrum demands. Data were transmitted over underlay cellular CRN with cooperative diversity. To ensure the QoS of primary transmissions and prevent packet loss due to PU interference

and channel fading, the authors used RaptorQ code as the application layer forward error correction scheme. Addition- ally, for the efficient utilisation of the licensed channel and robust video transmission, the authors employed different time-sharing ratios ranging between direct and best indirect links.

For an environment-friendly network, Sboui et al. [132] proposed low complexity profit maximisation algorithms for CR cellular networks consisting of collaborated primary network (PN) and SN, with a certain threshold constraint for carbon dioxide (CO2) emission that is controlled by a regulator. The optimisation was based on decentralised and centralised approaches, in which renewable energy availabil- ity and roaming cost are important parameters for network profits. In the collaboration CRN, PN and SN maximise their own profits whilst meeting the PN’s QoS and the imposed total carbon dioxide emission constraint. The PN switches off several of its BSs to reduce carbon footprints, and the corresponding users are roamed to the SN’s BSs. In return, the SN receives certain roaming cost and is free to exploit the PN’s spectrum. The spectrum is managed dynamically between the two networks either by sharing or leasing, depending on the profit per CO2 emission of the PN’s BSs. CR creates endless possibilities in sharing a limited spectrum band more effectively, with the aim of improving EE and spectrum efficiency, optimising power utilisation and eliminating interference with promising integration with FD technology.

Table 5 summarises all related CR studies on SS together with their focus, contributions and key features. The stud- ies are sorted according to their CR functions and year of publication.

V. ISSUES, CHALLENGES AND FUTURE RESEARCH DIRECTIONS Network system and architecture are developing rapidly. With 5G deployment, promising network improvements in terms of speed for future IoE-enabled applications, problems and solutions must be identified accordingly. To support the development and requirement of 5G, challenges and issues are presented in this section.

A. SPECTRUM SHARING Dense areas are created due to the emergence of hetero- geneous devices and the need for high-capacity data trans- mission. Jo et al. [78] conducted a detailed survey on the challenges of and solutions to converging the current network to a heterogeneous mobile network for performance improve- ments. Three types of converged mobile networks exist: device, protocol and service convergence. These converged mobile networks offer high spectrum efficiency and coverage in CoMP M2M HetNet. These future networks will also have integration issues in CR, D2D communication, in-band FD communication, NOMA and LTE-U, as addressed in [81]. For example, CR needs to scan a wider bandwidth, which makes the EE spectrum sensing scheme crucial in 5G networks.

14480 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

TABLE 5. Summary of previous CR work in SS with their focus, contributions and key features.

Open research challenges in database-driven CRNs for SUs’ location privacy were acknowledged by [94]. The challenges are private collection of SUs’ usage data, partial replication of

database content, coexistence and interference amongst SUs, multi-SU coordinated queries, protecting the privacy of PUs and spectrum access policy enforcement.

VOLUME 8, 2020 14481

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

TABLE 5. (Continued.) Summary of previous CR work in SS with their focus, contributions and key features.

For SS techniques, several gaps must be bridged to realise efficient SS in the 5G environment. A dense network leads to multi-user interference, security problems [122] and service deterioration because of the high competition for limited shared resources. Furthermore, the need for large bandwidth by heterogeneous devices leads to the lack of spectrum avail- ability. Despite the availability of numerous SS techniques, accurate spectrum sensing is required to reduce the interfer- ence. Determining the most efficient design implementation of sharing algorithms is another issue, as identified by [76]. For SS in a virtualised network, several key challenges were enumerated by [70], and these included addressing service differentiation, sharing of information, new network

functions, long-term contracts, management and control and responsibility assignment. Nonetheless, improvements in the hardware level are still needed to create a device that can overcome propagation losses and interference.

Dynamic SS is open to several policy-domain challenges. Interdisciplinary research is required for spectrum usage and access to manage the sharing of limited spectrum resources efficiently in the fields of market- and non-market-based mechanisms. Other open problems in SS include spectrum management and spectrum assignment/allocation, metrics to quantify spectrum usage, interference coexistence and management, security and enforcement, radio software and hardware, protocols and standards, experimentation, testing

14482 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

and standardisation and regulatory, policy and economic issues [72].

The challenges in the network architecture of SS in terms of spectrum sensing, decision-making and admission control were explained by [84]. The authors stated that interfer- ence occurs when SUs start their transmission whilst PU transmission is ongoing in the spectrum because of the lack of sensing accuracy. This is a serious issue in hybrid SS, especially in systems that utilise licensed bands. Moreover, propagation losses and interference occur from searching the hidden terminals of PUs and missed transmission opportuni- ties. Software-defined research on SS design has also been conducted to reduce device-level limitations in producing inaccurate results and reducing fluctuation in signal qual- ity [84]. Another means of alleviating the interference from the architecture perspective is to place emphasis on physi- cal layer design in order to strengthen the spatial focusing effect from mmWave and massive MIMO. The MAC layer design will focus on how to utilise the focusing effect to accommodate more users. This will lead to other issues in the MAC layer, and re-investigation is required when the spatial focusing effect is considered [85]. The issues are admission control, handover and security. Zhang et al. [86] proposed a formulation for an optimisation problem with the objective of maximising users’ satisfaction degree across the network. In the future, when EE is improved by employing energy harvesting technologies at secondary BSs, an issue might be encountered in terms of users’ QoE due to the intermittent arrival of renewable energy. Therefore, the authors suggested that the energy status of secondary BSs must be considered during SS.

SS using spectrum allocation behaviour also has its issues. The deployments of SC will be trending in future 5G for cost-effective spectrum and energy-efficient solutions. The challenges in SCs include serious interference and high energy consumption. However, SS amongst SCs and macro- cells can be explored either in inter or intra-cooperation. Yang et al. [87] suggested a cooperation method using coop- erative capacity offload with spectrum leasing, cooperative power coordination for SC range expansion and cooperative relay using spectrum trading and virtual currency. For SS in mobile network operators, the challenges include incomplete information, economic properties and spatial efficiency [88]. EE network design is one of the 5G development issues in delivering high QoS to all PUs and SUs. As stated by [104], restricted transmission power can affect the performance of SS, especially in terms of data rate and transmission efficiency. Therefore, future research with focus on design- ing solutions to the high complexity and huge signalling overhead of the EDSS and UDN of SC is needed [89]. Rattaro et al. [89] presented a paid SS system to promote the SS concept to PUs, and the incorporation of dynamic pricing features was proposed as future advancement. How- ever, the sensing accuracy and interference induced during the process must be solved.

For the spectrum access method, recent studies on NOMA were discussed by [100]. NOMA has several challenges, which were addressed by [102]. These challenges include the following: (1) determining the number of UEs to be served by a BS; (2) determining the UE clustering; (3) organizing the UEs within a cluster to follow the link quality measure; (4) achieving the objective of the cluster (either to prioritise individual UE performance, total cluster performance or the middle ground between them) and (5) allocating resources for UEs in a cluster. The resource allocation scheme based on the multi time-scale hierarchical model was studied by [96], who proposed to consider the fairness amongst users in each slice in future work. mmWave has a very wide spectrum, but its features raise broad questions. mmWave issues and several initial answers were discussed by [98], such as ‘how the mmWave should be utilised amongst multiple operators’, ‘to what extent should the spectrum be shared’ and ‘how does the optimal SS arrangement vary with different frequency bands’. The same authors also suggested a study on char- acterisation for user allocation strategies in the interaction of dynamic traffic and interference, which will lead to a time-varying throughput performance. In time-varying traf- fic, users dynamically come and go according to statistics. In an SS system, spectrum handoff introduces extra handoff delay that has a great impact on 5G low-latency commu- nication. Therefore, spectrum handoff probability and delay must be reduced. Another challenge in improving the learning efficiency of dynamic spectrum management implementation thus arises [97]. This includes learning the radio environment, such as learning the traffic patterns of PUs and the interfer- ence levels, for practical 5G SS systems.

B. COGNITIVE RADIO The challenges for 5G cognitive terminals were highlighted in [79] to ensure the success of CRN in providing EE and high-speed connectivity to end users. Several key characteris- tics were discussed, and these included interoperability, con- text awareness, learning ability, self-optimisation, dynamic spectrum management, adaptive decoding and self-healing behaviour.

Novel architectures for cognitive networks that are intended to further increase spectrum efficiency also have their issues. The architecture based on D2D experiences inter- ference when multiple D2D pairs share the same resources. In cognitive HetNets consisting of microcell and SCs, inter- cell interference can be reduced with cognitive capabilities for a spectrum-efficient network. However, a remarkable increase in overall energy consumption and infrastructure cost will occur after solving the capacity demand issue. The architecture that uses sleep mode techniques can solve the said problems, but an issue in the infeasibility of the techniques arises due to fluctuations in traffic demand over space, time and frequency. Another method to increase the coverage and spectrum efficiency in CRN is by enabling multi-hop communication. The solution can be implemented

VOLUME 8, 2020 14483

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

by providing backup channels for improving link reliability. However, the higher the number of hops is, the higher the switching delay is to the backup channels. The collaborative cognitive dynamic network architecture that was proposed to flexibly and intelligently exploit underutilised spectrum bands also has many challenges in terms of reconfigurable user equipment, economic models and security design [130].

Spectrum sensing as one of the CR elements has vari- ous techniques, as discussed in [16]. Sensing through the estimation approach estimates the interference in real data, but estimation errors can occur even with stationary radio frequency interference. On the one hand, underestimating (missed detection) or overestimating (dales alarm) interfer- ence leads to SINR degradation. On the other hand, correctly estimating radio frequency interference may degrade SINR when the bandwidth is varied from pulse to pulse [124]. Crowd sensing has elicited attention in recent years despite its issues, such as manifestation of abnormal data in crowd sensors [133], data incompleteness and inaccuracy [134], need for scalable radio-frequency spectrum monitoring, low- power and low-cost sensors [135], fading, shadowing and noise uncertainty effects [127]. The suggested future direc- tions for the crowd sensing method include finding ways to process the spectrum, streamlining contextual data, improv- ing the spectrum decision quality and implementing the sensing application in various possible platforms other than Android [127].

The spectrum allocation issues in methods based on graph, auction, game and carrier aggregation were elabo- rated in [83]. Other problems that cannot be overcome by fundamental protocols were identified in [71], and these include disproportionate and unequal spectrum utilisation caused by network entry, initialisation and hidden incumbent problem. In dynamic spectrum reservation, channel alloca- tion for CRNs is prone to failures from the perspective of SNs, thus making users experience a less predictable QoS. Hence, challenges exist in ensuring the requirements of CR users for the reliability and availability of related services [112].

In [83], issues that are important for CR development towards 5G, such as common control channel, energy har- vesting, NOMA and CR-based aeronautical communication, were addressed together with the importance of improving the spectrum and EE. The challenges in implementing cognitive NOMA as a spectrum access technique, including interfer- ence management, imperfect channel state information, EE, multi-carrier cognitive NOMA, cognitive MIMO-NOMA, relay selection/user scheduling, physical layer security and integration of FD technology, were discussed in [108]. Cog- nitive NOMA is a promising technique for efficient spec- trum utilisation, and NOMA and CR are interference-limited. However, power domain multiplexing of NOMA causes the coexistence of inter-network interference between PNs and SNs and intra-network (co-channel) interference, resulting in the degradation of reception reliability [108]. The challenges in the exclusive use of the spectrum access trading model, viability assessment, CR-based cloud computing, security

issues, optimal reservation contract design, embedding good- will factor, alliance of sellers, mobility management, auto- mated intelligent trading and social networking were covered in [17].

Spectrum handover based on packet-switched handover does not require spectrum sensing in the procedure, thus reducing handover interruption. However, during the spec- trum handover process, the number of unavailable channels increases, and a channel must be verified to reduce the packet loss probability and bandwidth fragment ratio [71]. Synchro- nising all CR functions (SS, sensing, allocation, access and handover) needs efficient spectrum management functions. The challenge in managing dynamic spectra in CRN lies in integrating the said functions in many layers of a pro- tocol stack to allow SUs to communicate reliably over the environments.

Enabling FD for CRN enhances the gain of throughput for PUs and SUs, but EE SS remains an issue. The trade-offs between EE and throughput and between EE and spectrum sensing accuracy were studied in [136], where SUs acted as FD relays for PUs. The energy consumption of SUs is greatly affected by the PUs’ behaviours with constant change in their activities. Li et al. concluded that EE can hardly be achieved with a single and fixed choice of SS scheme [115]. In [116], FD SS for CRN was proposed from another perspective, that is, the PU-oriented method. This approach leads to the obstruction of SUs’ opportunities to access the spectrum due to the self-interference of PUs. Then, the authors proposed a polarisation-based method to enhance SUs’ transmission chances by using the signal waves’ intrinsic property in electromagnetic field transmission. However, the polarisation learning accuracy requires further improvement.

Outage analysis for cognitive SS was studied to minimise the outage probability for users. However, only a few studies have been conducted on the outage analysis of two-way relaying networks for cellular scenarios in the presence of PU interference [129]. Hence, the overall outage probability of the SN system must be minimised by solving the relay loca- tion optimisation problem. To minimise delay whilst sharing a spectrum, [128] addressed the challenge in optimal resource allocation in outage analysis to determine the trade-off in tuning the length of decoding blocks.

For general 5G improvements, network synchronisation is recommended for future research. The efficient use of spectrum resources must be guaranteed. Potential inter- operator network synchronisation must be facilitated to avoid guard bands between operator assignments. This feature is beneficial for the efficient deployment of 5G NR networks in unpaired assignments and the alignment of uplink/downlink transmissions for slot and frame synchronisation.

VI. CONCLUSIONS Realising a technology with the vast growth of devices con- nected together in a UDN sharing the same spectrum band is one of the major challenges in 5G. The urgency to address this important issue is the motivation of this study. Here, a detailed

14484 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

survey on SS techniques and CR technology was conducted. Recent 5G-enabling technologies, such as HetNets, SS and flexibility, massive MIMO, ultra-lean design, URLLC, con- vergence of access and backhaul, mMTC, mmWave and NM, were also explained. Classification of SS techniques was discussed together with the survey on SS and its techniques relevant to 5G networks. The focus, contributions and key features of each technique were highlighted. The CR tech- nology in SS and other applications related to 5G were also reviewed. Future research directions in SS and CR were outlined together with their issues and challenges.

The main 5G development issues are related to EE and interference. Multi-user interference in dense networks is a serious issue, especially in licensed bands. Propagation losses and the interference that fluctuates in signal quality must be addressed at the hardware level. The interference of multiple D2D pairs in shared resources is another issue. The current solution can reduce the interference but at the cost of energy consumption and infrastructure. Multi-hop communication increases spectrum efficiency but has a high switching delay. Interference management is also needed for cognitive NOMA to handle inter-network interference, which degrades reception reliability. This comprehensive survey is expected to aid researchers in addressing important issues in achieving successful 5G applications.

REFERENCES [1] J. G. Andrews, S. Buzzi, W. Choi, S. V. Hanly, A. Lozano, A. C. K. Soong,

and J. C. Zhang, ‘‘What will 5G be?’’ IEEE J. Sel. Areas Commun., vol. 32, no. 6, pp. 1065–1082, Jun. 2014.

[2] E. Dahlman, G. Mildh, S. Parkvall, J. Peisa, J. Sachs, Y. Selén, and J. Sköld, ‘‘5G wireless access: Requirements and realization,’’ IEEE Commun. Mag., vol. 52, no. 12, pp. 42–47, Dec. 2014.

[3] M. Nekovee, Opportunities and Enabling Technologies for 5G and Beyond-5G Spectrum Sharing. Singapore: Springer, 2018, pp. 1–15.

[4] Update on 5G Spectrum in the UK, Ofcom, London, U.K., 2017. [5] B. P. Rimal, D. P. Van, and M. Maier, ‘‘Mobile edge computing empow-

ered fiber-wireless access networks in the 5G era,’’ IEEE Commun. Mag., vol. 55, no. 2, pp. 192–200, Feb. 2017.

[6] A. Gohil, H. Modi, and S. K. Patel, ‘‘5G technology of mobile communi- cation: A survey,’’ in Proc. Int. Conf. Intell. Syst. Signal Process. (ISSP), Mar. 2013, pp. 288–292.

[7] A. Gupta and R. K. Jha, ‘‘A Survey of 5G network: Architecture and emerging technologies,’’ IEEE Access, vol. 3, pp. 1206–1232, 2015.

[8] R. N. Mitra and D. P. Agrawal, ‘‘5G mobile technology: A survey,’’ ICT Express, vol. 1, no. 3, pp. 132–137, Dec. 2015.

[9] N. Panwar, S. Sharma, and A. K. Singh, ‘‘A survey on 5G: The next gen- eration of mobile communication,’’ Phys. Commun., vol. 18, pp. 64–84, Mar. 2016.

[10] D. Soldani, Y. J. Guo, B. Barani, P. Mogensen, I. C. , and S. K. Das, ‘‘5G for ultra-reliable low-latency communications,’’ IEEENetw., vol. 32, no. 2, pp. 6–7, 2018.

[11] I. Noman and W. A. W. Abdul, 5G Networks: A Holistic View of Enabling Technologies and Research Challenges. Hershey, PA, USA: IGI Global, 2019, pp. 37–70.

[12] Y. Niu, Y. Li, D. Jin, L. Su, and A. V. Vasilakos, ‘‘A survey of millimeter wave communications (mmWave) for 5G: Opportunities and challenges,’’ Wireless Netw, vol. 21, no. 8, pp. 2657–2676, Nov. 2015.

[13] S. A. Busari, S. Mumtaz, S. Al-Rubaye, and J. Rodriguez, ‘‘5G millimeter-wave mobile broadband: Performance and challenges,’’ IEEE Commun. Mag., vol. 56, no. 6, pp. 137–143, Jun. 2018.

[14] U. C. Samal, B. Appasani, and D. K. Mohanta, 5G Communication Networks and Modulation Schemes for Next-Generation Smart Grids. Singapore: Springer, 2019, pp. 361–399.

[15] S. Li, L. Da Xu, and S. Zhao, ‘‘5G Internet of Things: A survey,’’ J. Ind. Inf. Integr., vol. 10, pp. 1–9, Jun. 2018.

[16] Y. Arjoune and N. Kaabouch, ‘‘A comprehensive survey on spectrum sensing in cognitive radio networks: Recent advances, new challenges, and future research directions,’’ Sensors, vol. 19, no. 1, p. 126, Jan. 2019.

[17] M. R. Hassan, G. C. Karmakar, J. Kamruzzaman, and B. Srinivasan, ‘‘Exclusive use spectrum access trading models in cognitive radio networks: A survey,’’ IEEE Commun. Surveys Tuts., vol. 19, no. 4, pp. 2192–2231, 4th Quart., 2017.

[18] U. Siddique, H. Tabassum, E. Hossain, and D. I. Kim, ‘‘Wireless back- hauling of 5G small cells: Challenges and solution approaches,’’ IEEE Wireless Commun., vol. 22, no. 5, pp. 22–31, Oct. 2015.

[19] H. Beyranvand, M. Levesque, M. Maier, J. A. Salehi, C. Verikoukis, and D. Tipper, ‘‘Toward 5G: FiWi enhanced LTE-A HetNets with reliable low-latency fiber backhaul sharing and WiFi offloading,’’ IEEE/ACM Trans. Netw., vol. 25, no. 2, pp. 690–707, Apr. 2017.

[20] M. Feng, L. Guomin, and G. Wenrong, ‘‘Heterogeneous network resource allocation optimization based on improved bat algorithm,’’ in Proc. Int. Conf. Sensor Netw. Signal Process. (SNSP), Oct. 2018, pp. 55–59.

[21] T. E. Bogale and L. B. Le, ‘‘Massive MIMO and mmWave for 5G wireless HetNet: Potential benefits and challenges,’’ IEEE Veh. Technol. Mag., vol. 11, no. 1, pp. 64–75, Mar. 2016.

[22] S. Wu, W. Mao, C. Liu, and T. Tang, ‘‘Dynamic traffic prediction with adaptive sampling for 5G HetNet IoT applications,’’ Wireless Commun. Mobile Comput., vol. 2019, pp. 1–11, Jun. 2019.

[23] S. Wu, W. Mao, T. Hong, C. Liu, and M. Kadoch, ‘‘Compressed sensing based traffic prediction for 5G HetNet IoT video streaming,’’ in Proc. 15th Int. Wireless Commun. Mobile Comput. Conf. (IWCMC), Jun. 2019, pp. 1901–1906.

[24] J. Yang, X. Ji, K. Huang, Y. Chen, X. Xu, and M. Yi, ‘‘Unified and fast handover authentication based on link signatures in 5G SDN-based HetNet,’’ IET Commun., vol. 13, no. 2, pp. 144–152, Jan. 2019.

[25] R. I. Rony, A. Jain, E. Lopez-Aguilera, E. Garcia-Villegas, and I. Demirkol, ‘‘Joint access-backhaul perspective on mobility management in 5G networks,’’ in Proc. IEEEConf.StandardsCommun.Netw.(CSCN), Sep. 2017, pp. 115–120.

[26] W. Tan, S. D. Assimonis, M. Matthaiou, Y. Han, X. Li, and S. Jin, ‘‘Anal- ysis of different planar antenna arrays for mmWave massive MIMO sys- tems,’’ in Proc. IEEE 85th Veh. Technol. Conf. (VTC Spring), Jun. 2017, pp. 1–5.

[27] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, C.-L. I, and H. V. Poor, ‘‘Application of non-orthogonal multiple access in LTE and 5G networks,’’ IEEE Commun. Mag., vol. 55, no. 2, pp. 185–191, Feb. 2017.

[28] M. Agiwal, A. Roy, and N. Saxena, ‘‘Next generation 5G wireless net- works: A comprehensive survey,’’ IEEE Commun. Surveys Tuts., vol. 18, no. 3, pp. 1617–1655, 2016.

[29] S. Sun, T. S. Rappaport, and M. Shaft, ‘‘Hybrid beamforming for 5G millimeter-wave multi-cell networks,’’ in Proc. IEEE INFOCOM IEEE Conf. Comput. Commun. Workshops (INFOCOM WKSHPS), Apr. 2018, pp. 589–596.

[30] X. Liu, Q. Zhang, W. Chen, H. Feng, L. Chen, F. M. Ghannouchi, and Z. Feng, ‘‘Beam-oriented digital predistortion for 5G massive MIMO hybrid beamforming transmitters,’’ IEEE Trans. Microw. Theory Techn., vol. 66, no. 7, pp. 3419–3432, Jul. 2018.

[31] X. Zhao, E. Lukashova, F. Kaltenberger, and S. Wagner, ‘‘Practical hybrid beamforming schemes in massive MIMO 5G NR systems,’’ in Proc. WSA 23rd Int. ITG Workshop Smart Antennas, Apr. 2019, pp. 1–8.

[32] D. Phan-Huy, S. Wesemann, J. Bjoersell, and M. Sternad, ‘‘Adaptive massive MIMO for fast moving connected vehicles: It will work with pre- dictor antennas!’’ in Proc. WSA 22nd Int. ITG Workshop Smart Antennas, 2018, pp. 1–8.

[33] H. Pei, X. Chen, M. Zhang, and A. Zhang, ‘‘Over-the-air testing of 5G millimeter-wave system with adaptive beamforming,’’ in IEEEMTT-SInt. Microw. Symp. Dig., May 2019, pp. 1–3.

[34] H. Manai, L. B. H. Slama, and R. Bouallegue, ‘‘Interference management by adaptive beamforming algorithm in massive MIMO networks,’’ in Proc. 15th Int. Wireless Commun. Mobile Comput. Conf. (IWCMC), Jun. 2019, pp. 49–54.

[35] H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, and T. L. Marzetta, ‘‘Cell-Free Massive MIMO: Uniformly great service for everyone,’’ in Proc. IEEE 16th Int. Workshop Signal Process. Adv. Wireless Com- mun. (SPAWC), Jun. 2015, pp. 201–205.

VOLUME 8, 2020 14485

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

[36] E. Bjornson and L. Sanguinetti, ‘‘Cell-free versus cellular massive MIMO: What processing is needed for cell-free to win?’’ in Proc. IEEE 20th Int. Workshop Signal Process. Adv. Wireless Commun. (SPAWC), Jul. 2019, pp. 1–5.

[37] S. M. S. Ullah, W. A. Mahyiddin, N. A. Zakaria, T. A. Latef, K. A. Noordin, and K. Dimyati, ‘‘Training size optimization with reduced complexity in cell-free massive MIMO system,’’ Wireless Netw, vol. 25, no. 4, pp. 1983–1994, May 2019.

[38] J. Zhang, S. Chen, Y. Lin, J. Zheng, B. Ai, and L. Hanzo, ‘‘Cell-free massive MIMO: A new next-generation paradigm,’’ IEEE Access, vol. 7, pp. 99878–99888, 2019.

[39] Z. Chen and E. Bjornson, ‘‘Channel hardening and favorable propaga- tion in cell-free massive MIMO with stochastic geometry,’’ IEEE Trans. Commun., vol. 66, no. 11, pp. 5205–5219, Nov. 2018.

[40] M. Shirvanimoghaddam, M. Dohler, and S. J. Johnson, ‘‘Massive non- orthogonal multiple access for cellular IoT: Potentials and limitations,’’ IEEE Commun. Mag., vol. 55, no. 9, pp. 55–61, 2017.

[41] M. Rihan, L. Huang, and P. Zhang, ‘‘Joint interference alignment and power allocation for noma-based multi-user MIMO systems,’’ EURASIP J. Wireless Commun. Netw., vol. 2018, no. 1, pp. 1–13, 2018.

[42] S. Wang, S. Cao, and R. Ruby, ‘‘Optimal power allocation in noma-based two-path successive AF relay systems,’’ EURASIP J. Wireless Commun. Netw., vol. 2018, no. 1, p. 273, 2018.

[43] K. He, Y. Li, C. Yin, and Y. Zhang, ‘‘A novel compressed sensing- based non-orthogonal multiple access scheme for massive MTC in 5G systems,’’ EURASIP J. Wireless Commun. Netw., vol. 2018, no. 1, 2018.

[44] A. Babaei. (2017). New radio for 5G: The future of mobile broadband. Ofinno Technologies, Report. [Online]. Available: https://ofinno.com/ wp-content/uploads/2017/05/Ofinno_New-Radio-for-5G_WP.pdf

[45] K. Balachandran, ‘‘5G radio access technology,’’ presented at the Slides for 5G Summit, 2016. [Online]. Available: http://www.5gsummit. org/seattle/docs/slides/Kumar-5GSummit-Radio-Access-revA3.pdf

[46] A. Prasad, A. Bhamri, and P. Lunden, ‘‘Enhanced mobility and energy efficiency in 5G ultra-dense networks with lean carrier design,’’ in Proc. IEEE Conf. Standards Commun. Netw. (CSCN), Oct. 2016, pp. 1–5.

[47] Ofinno. (2019). New Radio for 5G. [Online]. Available: https://ofinno. com/technology/new-radio-for-5g/

[48] M. Lauridsen, L. C. Gimenez, I. Rodriguez, T. B. Sorensen, and P. Mogensen, ‘‘From LTE to 5G for connected mobility,’’ IEEE Commun. Mag., vol. 55, no. 3, pp. 156–162, Mar. 2017.

[49] M. Moradi, Y. Lin, Z. M. Mao, S. Sen, and O. Spatscheck, ‘‘Soft- Box: A customizable, low-latency, and scalable 5G core network archi- tecture,’’ IEEE J. Sel. Areas Commun., vol. 36, no. 3, pp. 438–456, Mar. 2018.

[50] I. Jovović, I. Forenbacher, and M. Periša, ‘‘Massive machine-type com- munications: An overview and perspectives towards 5G,’’ in Proc.3rdInt. Virtual Res. Conf. Tech. Disciplines, vol. 3, Nov. 2015.

[51] L. Pulcini, P. Grazioso, A. Valenti, F. Matera, D. Del Buono, and V. Attanasio, ‘‘Software Defined Networks over Carrier Ethernet for 5G: Tests from a GMPLS test bed,’’ in Proc. 18th Italian Nat. Conf. Photonic Technol. (Fotonica), Jun. 2016, pp. 1–4.

[52] J. Moysen and L. Giupponi, ‘‘From 4G to 5G: Self-organized network management meets machine learning,’’ Comput. Commun., vol. 129, pp. 248–268, Sep. 2018.

[53] A. Al-Dulaimi, S. Al-Rubaye, J. Cosmas, and A. Anpalagan, ‘‘Planning of ultra-dense wireless networks,’’ IEEE Netw., vol. 31, no. 2, pp. 90–96, Mar. 2017.

[54] S. Chen, F. Qin, B. Hu, X. Li, Z. Chen, and J. Liu, 5G Requirement and UDN. Cham, Switzerland: Springer, 2018, pp. 5–9.

[55] X. Foukas, G. Patounas, A. Elmokashfi, and M. K. Marina, ‘‘Network slicing in 5G: Survey and challenges,’’ IEEE Commun. Mag., vol. 55, no. 5, pp. 94–100, May 2017.

[56] N. Yang, L. Wang, G. Geraci, M. Elkashlan, J. Yuan, and M. D. Renzo, ‘‘Safeguarding 5G wireless communication networks using physical layer security,’’ IEEE Commun. Mag., vol. 53, no. 4, pp. 20–27, Apr. 2015.

[57] M. A. Ferrag, L. Maglaras, A. Argyriou, D. Kosmanos, and H. Janicke, ‘‘Security for 4G and 5G cellular networks: A survey of existing authentication and privacy-preserving schemes,’’ J. Netw. Comput. Appl., vol. 101, pp. 55–82, Jan. 2018.

[58] V. Kumar, S. Yadav, D. N. Sandeep, S. B. Dhok, R. K. Barik, and H. Dubey, ‘‘5G cellular: Concept, research work and enabling tech- nologies,’’ in Advances in Data and Information Sciences. Singapore: Springer, 2019, pp. 327–338.

[59] R. Martínez, A. Mayoral, R. Vilalta, R. Casellas, R. Muñoz, S. Pachnicke, T. Szyrkowiec, and A. Autenrieth, ‘‘Integrated SDN/NFV orchestration for the dynamic deployment of mobile virtual backhaul networks over a multilayer (packet/optical) aggregation infrastructure,’’ J. Opt. Commun. Netw., vol. 9, no. 2, p. A135, Feb. 2017.

[60] M. M. Ahamed and S. Faruque, 5G Backhaul: Requirements, Challenges, andEmergingTechnologies. London, U.K.: IntechOpen, 2018, pp. 43–58, doi: 10.5772/intechopen.78615.

[61] M. Jaber, M. A. Imran, R. Tafazolli, and A. Tukmanov, ‘‘5G backhaul challenges and emerging research directions: A survey,’’ IEEE Access, vol. 4, pp. 1743–1766, 2016.

[62] S. E. Elayoubi, S. B. Jemaa, Z. Altman, and A. Galindo-Serrano, ‘‘5G RAN slicing for verticals: Enablers and challenges,’’ IEEECommun. Mag., vol. 57, no. 1, pp. 28–34, Jan. 2019.

[63] S. E. Alavi, M. R. K. Soltanian, I. S. Amiri, M. Khalily, A. S. M. Supáat, and H. Ahmad, ‘‘Towards 5G: A photonic based millimeter wave sig- nal generation for applying in 5g access fronthaul,’’ Sci. Rep., vol. 6, Jan. 2016, Art. no. 19891.

[64] J. Parikh and A. Basu, Millimeter Waves: Technological Component for Next-Generation Mobile Networks. Cham, Switzerland: Springer, 2019, pp. 167–186.

[65] O. Jo, S. Chang, C. Kweon, J. Oh, and K. Cheun, ‘‘60GHz wireless communication for future Wi-Fi,’’ ICT Express, vol. 1, no. 1, pp. 30–33, Jun. 2015.

[66] J. Yinon, Counterterrorist Detection Techniques of Explosives. Amsterdam, The Netherlands: Elsevier, 2011.

[67] C. A. Lockard, Z. Ahmad, O. J. R. Bee, and T. Leinbach. (2019). Malaysia—Land—Climate. Encyclopedia Britannica, Inc. [Online]. Available: https://www.britannica.com/place/Malaysia/Climate

[68] M. Flament, ‘‘Propagation and interference issues in a 60 GHz mobile network,’’ in Proc. 2nd Pers. Comput. Commun. Workshop, Jan. 2000, pp. 59–64.

[69] A. Aishah, M. R. C. Beson, S. Azemi, and S. Junid, ‘‘60 GHz milimeter- wave antennas for point-to-point 5G communication system,’’ MATEC Web Conf., vol. 140, 2017, Art. no. 01006.

[70] F. Ahmed, A. Kliks, L. Goratti, and S. N. Khan, Towards Spectrum Sharing in Virtualized Networks: A Survey and an Outlook. Cham, Switzerland: Springer, 2019, pp. 1–28.

[71] H. Anandakumar and K. Umamaheswari, ‘‘Cooperative spectrum han- dovers in cognitive radio networks,’’ in Cognitive Radio, Mobile Com- munications and Wireless Networks. Cham, Switzerland: Springer, 2019, pp. 47–63.

[72] S. Mishra, S. S. Singh, and B. S. P. Mishra, ‘‘A comparative analysis of centralized and distributed spectrum sharing techniques in cogni- tive radio,’’ in Computational Intelligence in Sensor Networks. Berlin, Germany: Springer, 2019, pp. 455–472.

[73] G. Dludla, L. Mfupe, and F. Mekuria, ‘‘Overview of spectrum shar- ing models: A path towards 5G spectrum toolboxes,’’ in Information and Communication Technology for Development for Africa. Cham, Switzerland: Springer, 2018, pp. 308–319, doi: 10.1007/978-3-319- 95153-9_28.

[74] F. Beltrán and M. Massaro, ‘‘Spectrum management for 5G: Assignment methods for spectrum sharing,’’ in Proc. 29th Eur. Regional ITS Conf., Aug. 2018.

[75] M. Mustonen, ‘‘Analysis of recent spectrum sharing concepts in policy making: Dissertation,’’ Ph.D. dissertation, Univ. Oulu, Oulu, Finland, 2017.

[76] R. H. Tehrani, S. Vahid, D. Triantafyllopoulou, H. Lee, and K. Moessner, ‘‘Licensed spectrum sharing schemes for mobile operators: A survey and outlook,’’ IEEE Commun. Surveys Tuts., vol. 18, no. 4, pp. 2591–2623, 2016.

[77] S. M. Baby and M. James, ‘‘A comparative study on various spectrum sharing techniques,’’ Procedia Technol., vol. 25, pp. 613–620, 2016.

[78] M. Jo, T. Maksymyuk, R. L. Batista, T. F. Maciel, A. L. F. De Almeida, and M. Klymash, ‘‘A survey of converging solutions for heterogeneous mobile networks,’’ IEEE Wireless Commun., vol. 21, no. 6, pp. 54–62, Dec. 2014.

[79] R. Jonathan, Cognitive Radio for 5G Wireless Networks. Hoboken, NJ, USA: Wiley, 2014, p. 1.

[80] G. Ding, J. Wang, Q. Wu, Y.-D. Yao, R. Li, H. Zhang, and Y. Zou, ‘‘On the limits of predictability in real-world radio spectrum state dynamics: From entropy theory to 5G spectrum sharing,’’ IEEE Commun. Mag., vol. 53, no. 7, pp. 178–183, Jul. 2015.

[81] L. Zhang, M. Xiao, G. Wu, M. Alam, Y.-C. Liang, and S. Li, ‘‘A survey of advanced techniques for spectrum sharing in 5G networks,’’ IEEE Wireless Commun., vol. 24, no. 5, pp. 44–51, Oct. 2017.

14486 VOLUME 8, 2020

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

[82] O. Al-Khatib, W. Hardjawana, and B. Vucetic, ‘‘Spectrum sharing in multi-tenant 5G cellular networks: Modeling and planning,’’ IEEE Access, vol. 7, pp. 1602–1616, 2019.

[83] F. Hu, B. Chen, and K. Zhu, ‘‘Full spectrum sharing in cognitive radio networks toward 5G: A survey,’’ IEEE Access, vol. 6, pp. 15754–15776, 2018.

[84] A. M. Akhtar, X. Wang, and L. Hanzo, ‘‘Synergistic spectrum sharing in 5G HetNets: A harmonized SDN-enabled approach,’’ IEEE Commun. Mag., vol. 54, no. 1, pp. 40–47, Jan. 2016.

[85] C. Jiang, B. Wang, Y. Han, Z.-H. Wu, and K. J. R. Liu, ‘‘Exploring spatial focusing effect for spectrum sharing and network association,’’ IEEE Trans. Wireless Commun., vol. 16, no. 7, pp. 4216–4231, Jul. 2017.

[86] N. Zhang, S. Zhang, J. Zheng, X. Fang, J. W. Mark, and X. Shen, ‘‘QoE driven decentralized spectrum sharing in 5G networks: Potential game approach,’’ IEEE Trans. Veh. Technol., vol. 66, no. 9, pp. 7797–7808, Sep. 2017.

[87] C. Yang, J. Li, M. Guizani, A. Anpalagan, and M. Elkashlan, ‘‘Advanced spectrum sharing in 5G cognitive heterogeneous networks,’’ IEEE Wire- less Commun., vol. 23, no. 2, pp. 94–101, Apr. 2016.

[88] H. Wang, D. N. Nguyen, E. Dutkiewicz, G. Fang, and M. D. Mueck, ‘‘Negotiable auction based on mixed graph: A novel spectrum shar- ing framework,’’ IEEE Trans. Cogn. Commun. Netw., vol. 3, no. 3, pp. 390–403, Sep. 2017.

[89] C. Rattaro, P. Bermolen, and P. Belzarena, ‘‘Multi-resource allocation: Analysis of a paid spectrum sharing approach based on fluid mod- els,’’ IEEE Trans. Cogn. Commun. Netw., vol. 4, no. 3, pp. 607–617, Sep. 2018.

[90] A. K. Bairagi, S. F. Abedin, N. H. Tran, D. Niyato, and C. S. Hong, ‘‘QoE- enabled unlicensed spectrum sharing in 5G: A game-theoretic approach,’’ IEEE Access, vol. 6, pp. 50538–50554, 2018.

[91] B. A. Jayawickrama, Y. He, E. Dutkiewicz, and M. D. Mueck, ‘‘Scalable spectrum access system for massive machine type communication,’’ IEEE Netw., vol. 32, no. 3, pp. 154–160, May 2018.

[92] P. Kryszkiewicz, A. Kliks, Ł. Kułacz, H. Bogucka, G. P. Koudouridis, and M. Dryjański, ‘‘Context-based spectrum sharing in 5G wireless networks based on radio environment maps,’’ Wireless Commun. Mobile Comput., vol. 2018, pp. 1–15, Nov. 2018.

[93] W. Zhang, C.-X. Wang, X. Ge, and Y. Chen, ‘‘Enhanced 5G cognitive radio networks based on spectrum sharing and spectrum aggregation,’’ IEEE Trans. Commun., vol. 66, no. 12, pp. 6304–6316, Dec. 2018.

[94] M. Grissa, B. Hamdaoui, and A. A. Yavuz, ‘‘Unleashing the power of multi-server PIR for enabling private access to spectrum databases,’’ IEEE Commun. Mag., vol. 56, no. 12, pp. 171–177, Dec. 2018.

[95] R. Kalidoss, M. Saravanan, and K. Manikannan, ‘‘Analytic hierarchy processes for spectrum sharing in 5G new radio standard,’’ Wireless Pers. Commun., vol. 103, no. 1, pp. 639–655, Nov. 2018.

[96] H. Jiang, T. Wang, and S. Wang, ‘‘Multi-scale hierarchical resource management for wireless network virtualization,’’ IEEE Trans. Cogn. Commun. Netw., vol. 4, no. 4, pp. 919–928, Dec. 2018.

[97] S. Lin, L. Kong, Q. Gao, M. K. Khan, Z. Zhong, X. Jin, and P. Zeng, ‘‘Advanced dynamic channel access strategy in spectrum sharing 5G systems,’’ IEEE Wireless Commun., vol. 24, no. 5, pp. 74–80, Oct. 2017.

[98] M. Rebato, F. Boccardi, M. Mezzavilla, S. Rangan, and M. Zorzi, ‘‘Hybrid spectrum sharing in mmWave cellular networks,’’ IEEE Trans. Cogn. Commun. Netw., vol. 3, no. 2, pp. 155–168, Jun. 2017.

[99] Y. Li and G. A. Aruma Baduge, ‘‘Underlay spectrum-sharing massive MIMO NOMA,’’ IEEE Commun. Lett., vol. 23, no. 1, pp. 116–119, Jan. 2019.

[100] D. B. Da Costa, T. Q. Duong, Z. Ding, H.-M. Wang, K. Tourki, and N. Al-Dhahir, ‘‘IEEE access special section editorial: Non-orthogonal multiple access for 5G systems,’’ IEEE Access, vol. 6, pp. 79280–79284, 2018.

[101] M. Zeng, A. Yadav, O. A. Dobre, and H. V. Poor, ‘‘Energy-efficient power allocation for MIMO-NOMA with multiple users in a cluster,’’ IEEE Access, vol. 6, pp. 5170–5181, 2018.

[102] K. S. Ali, M. Haenggi, H. Elsawy, A. Chaaban, and M.-S. Alouini, ‘‘Downlink non-orthogonal multiple access (NOMA) in Poisson net- works,’’ IEEE Trans. Commun., vol. 67, no. 2, pp. 1613–1628, Feb. 2019.

[103] J. Choi, ‘‘NOMA: Principles and recent results,’’ in Proc. Int. Symp. Wireless Commun. Syst. (ISWCS), Aug. 2017, pp. 349–354.

[104] P. Mach and Z. Becvar, ‘‘Energy-aware dynamic selection of overlay and underlay spectrum sharing for cognitive small cells,’’ IEEE Trans. Veh. Technol., vol. 66, no. 5, pp. 4120–4132, 2017.

[105] A. Zappone, B. Matthiesen, and E. A. Jorswieck, ‘‘Energy efficiency in MIMO underlay and overlay device-to-device communications and cognitive radio systems,’’ IEEE Trans. Signal Process., vol. 65, no. 4, pp. 1026–1041, Feb. 2017.

[106] P. Gandotra, R. K. Jha, and S. Jain, ‘‘Prolonging user battery lifetime using green communication in spectrum sharing networks,’’ IEEE Com- mun. Lett., vol. 22, no. 7, pp. 1490–1493, Jul. 2018.

[107] H. Zhou, W. Xu, Y. Bi, J. Chen, Q. Yu, and X. S. Shen, ‘‘Toward 5G spectrum sharing for immersive-experience-driven vehicular communi- cations,’’ IEEE Wireless Commun., vol. 24, no. 6, pp. 30–37, Dec. 2017.

[108] L. Lv, J. Chen, Q. Ni, Z. Ding, and H. Jiang, ‘‘Cognitive non-orthogonal multiple access with cooperative relaying: A new wireless frontier for 5G spectrum sharing,’’ IEEE Commun. Mag., vol. 56, no. 4, pp. 188–195, Apr. 2018.

[109] O. Holland, ‘‘Some are born with white space, some achieve white space, and some have white space thrust upon them,’’ IEEE Trans. Cogn. Commun. Netw., vol. 2, no. 2, pp. 178–193, Jun. 2016.

[110] C. Jia, X. Song, and X. Meng, ‘‘Fog-aided cognitive radio networks with guard zone and interference cancellation based opportunistic spectrum access,’’ IEEE Access, vol. 6, pp. 59182–59191, 2018.

[111] Q. Zhao and A. Swami, ‘‘A survey of dynamic spectrum access: Signal processing and networking perspectives,’’ in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. (ICASSP), Apr. 2007, pp. IV-1349–IV-1352.

[112] I. A. M. Balapuwaduge, F. Y. Li, and V. Pla, ‘‘Dynamic spectrum reser- vation for CR networks in the presence of channel failures: Channel allo- cation and reliability analysis,’’ IEEE Trans. Wireless Commun., vol. 17, no. 2, pp. 882–898, Feb. 2018.

[113] M. Shikh-Bahaei, Y.-S. Choi, and D. Hong, ‘‘Full-duplex and cognitive radio networking for the emerging 5G systems,’’ Wireless Commun. Mobile Comput., vol. 2018, pp. 1–2, 2018.

[114] X. Liu, D. He, and H. Ding, ‘‘Throughput maximization for UAV- enabled full-duplex relay system in 5G communications,’’ Phys. Com- mun., vol. 32, pp. 104–111, Feb. 2019.

[115] D. Li, J. Cheng, and V. C. M. Leung, ‘‘Adaptive spectrum sharing for half-duplex and full-duplex cognitive radios: From the energy efficiency perspective,’’ IEEE Trans. Commun., vol. 66, no. 11, pp. 5067–5080, Nov. 2018.

[116] D. Li, D. Zhang, and J. Cheng, ‘‘A novel polarization enabled full-duplex hybrid spectrum sharing scheme for cognitive radios,’’ IEEE Commun. Lett., vol. 23, no. 3, pp. 530–533, Mar. 2019.

[117] H. Park and T. Hwang, ‘‘Energy-efficient power control of cognitive Femto users for 5G communications,’’ IEEE J. Sel. Areas Commun., vol. 34, no. 4, pp. 772–785, Apr. 2016.

[118] H. Gao, W. Ejaz, and M. Jo, ‘‘Cooperative wireless energy harvesting and spectrum sharing in 5G networks,’’ IEEE Access, vol. 4, pp. 3647–3658, 2016.

[119] Y. Gao, H. He, Z. Deng, and X. Zhang, ‘‘Cognitive radio network with energy-harvesting based on primary and secondary user signals,’’ IEEE Access, vol. 6, pp. 9081–9090, 2018.

[120] F. Haider, C.-X. Wang, H. Haas, E. Hepsaydir, X. Ge, and D. Yuan, ‘‘Spectral and energy efficiency analysis for cognitive radio networks,’’ IEEE Trans. Wireless Commun., vol. 14, no. 6, pp. 2969–2980, Jun. 2015.

[121] L. Wang, H. Q. Ngo, M. Elkashlan, T. Q. Duong, and K.-K. Wong, ‘‘Massive MIMO in spectrum sharing networks: Achievable rate and power efficiency,’’ IEEE Syst. J., vol. 11, no. 1, pp. 20–31, Mar. 2017.

[122] M. H. Yılmaz, E. Güvenkaya, H. M. Furqan, S. Köse, and H. Arslan, ‘‘Cognitive security of wireless communication systems in the physical layer,’’ Wireless Commun. Mobile Comput., vol. 2017, p. 9, Dec. 2017.

[123] A. A. Khan, M. H. Rehmani, and M. Reisslein, ‘‘Requirements, design challenges, and review of routing and MAC protocols for CR-based smart grid systems,’’ IEEE Commun. Mag., vol. 55, no. 5, pp. 206–215, May 2017.

[124] B. Ravenscroft, J. W. Owen, J. Jakabosky, S. D. Blunt, A. F. Martone, and K. D. Sherbondy, ‘‘Experimental demonstration and analysis of cognitive spectrum sensing and notching for radar,’’ IET Radar, Sonar Navigat., vol. 12, no. 12, pp. 1466–1475, Dec. 2018.

[125] B. Li, S. Li, A. Nallanathan, and C. Zhao, ‘‘Deep sensing for future spectrum and location awareness 5G communications,’’ IEEE J. Sel. Areas Commun., vol. 33, no. 7, pp. 1331–1344, Jul. 2015.

[126] S. Dikmese, Z. Ilyas, P. C. Sofotasios, M. Renfors, and M. Valkama, ‘‘Sparse frequency domain spectrum sensing and sharing based on cyclic prefix autocorrelation,’’ IEEE J. Select. Areas Commun., vol. 35, no. 1, pp. 159–172, Jan. 2017.

VOLUME 8, 2020 14487

W. S. H. M. W. Ahmad et al.: 5G Technology: Towards Dynamic Spectrum Sharing

[127] H. Shirvani and B. S. Ghahfarokhi, Cloud-Based Context-Aware Spec- trum Availability Monitoring and Prediction Using Crowd-Sensing. Cham, Switzerland: Springer, 2019, pp. 29–45.

[128] A. Alabbasi, Z. Rezki, and B. Shihada, ‘‘Outage analysis of spectrum sharing over M-block fading with sensing information,’’ IEEE Trans. Veh. Technol., vol. 66, no. 4, pp. 3071–3087, Apr. 2017.

[129] I. Khan and P. Singh, ‘‘Outage analysis for multiuser underlay cognitive TWRN with antenna selection and user scheduling,’’ AEU-Int. J. Elec- tron. Commun., vol. 98, pp. 89–94, Jan. 2019.

[130] B. Lorenzo, F. J. Gonzalez-Castano, and Y. Fang, ‘‘A novel collabora- tive cognitive dynamic network architecture,’’ IEEE Wireless Commun., vol. 24, no. 1, pp. 74–81, Feb. 2017.

[131] A. Ali, K. S. Kwak, N. H. Tran, Z. Han, D. Niyato, F. Zeshan, M. T. Gul, and D. Y. Suh, ‘‘RaptorQ-based efficient multimedia transmission over cooperative cellular cognitive radio networks,’’ IEEE Trans. Veh. Tech- nol., vol. 67, no. 8, pp. 7275–7289, Aug. 2018.

[132] L. Sboui, H. Ghazzai, Z. Rezki, and M.-S. Alouini, ‘‘On green cognitive radio cellular networks: Dynamic spectrum and operation management,’’ IEEE Access, vol. 4, pp. 4046–4057, 2016.

[133] G. Ding, J. Wang, Q. Wu, L. Zhang, Y. Zou, Y.-D. Yao, and Y. Chen, ‘‘Robust spectrum sensing with crowd sensors,’’ IEEE Trans. Commun., vol. 62, no. 9, pp. 3129–3143, Sep. 2014.

[134] C. Xiang, P. Yang, C. Tian, L. Zhang, H. Lin, F. Xiao, M. Zhang, and Y. Liu, ‘‘CARM: Crowd-sensing accurate outdoor RSS maps with error- prone smartphone measurements,’’ IEEE Trans. Mobile Comput., vol. 15, no. 11, pp. 2669–2681, Nov. 2016.

[135] A. Chakraborty, M. S. Rahman, H. Gupta, and S. R. Das, ‘‘SpecSense: Crowdsensing for efficient querying of spectrum occupancy,’’ in Proc. IEEE INFOCOM IEEE Conf. Comput. Commun., May 2017, pp. 1–9.

[136] M. Costa and A. Ephremides, ‘‘Energy efficiency versus performance in cognitive wireless networks,’’ IEEE J. Sel. Areas Commun., vol. 34, no. 5, pp. 1336–1347, May 2016.

W. S. H. M. W. AHMAD received the B.Eng. degree in electronic engineering (majoring in mul- timedia) and the M.Eng.Sc. and Ph.D. degrees from Multimedia University, Cyberjaya, Malaysia. She has been a Postdoctoral Researcher, since 2017. She is currently with the Innovation and Research Management Centre, Universiti Tenaga Nasional, in a project studying the relationship of spectrum sharing and multiuser energy consump- tion for 5G network. Her current research interest

is in the communication field. Her forte is in the area of medical image analysis, content-based image retrieval, image segmentation, and feature extraction and data mining.

N. A. M. RADZI (Senior Member, IEEE) received the B.E. and M.E. degrees (Hons.) in EE and the Ph.D. degree in engineering from Universiti Tenaga Nasional in 2013, 2010, and 2008, respec- tively. She is currently a Senior Lecturer with the Department of Electrical Electronics Engineering, Universiti Tenaga Nasional. She is also a Profes- sional Engineer and a Chartered Engineer for IET. Her research interests include optical communi- cation and quality of service. She has contributed

50 technical articles in various journals and conferences. She is currently the Treasurer for the IEEE Photonics Society, Malaysia Chapter.

F. S. SAMIDI was born in Selangor, Malaysia, in 1996. He received the B.S. degree in electri- cal and electronics engineering from Universiti Tenaga Nasional (UNITEN), Malaysia, in 2018, where he is currently pursuing the M.S. degree in electrical engineering. He is also a Research Engineer with a Project under UNITEN iRMC. His research interest includes telecommunication field, especially in 5G development.

A. ISMAIL (Member, IEEE) received the B.Eng. degree in electrical and electronics engineering and the M.Eng. degree in electrical engineer- ing from Universiti Tenaga Nasional, Malaysia, in 2008 and 2011, respectively. He is cur- rently a Lecturer with Universiti Tenaga Nasional. His research interest is in optical communications and sensors.

F. ABDULLAH (Senior Member, IEEE) received the B.Eng. degree in electronics from Universiti Tenaga Nasional, Malaysia, in 2001, the M.Sc. degree in network and communication engineering from Universiti Putra Malaysia, in 2004, and the Ph.D. degree in fibre laser sensor from Univer- siti Tenaga Nasional, in 2012. He is currently an Associate Professor with the Department of Elec- tronics and Communication Engineering, Col- lege of Engineering, Universiti Tenaga Nasional.

His research interest is in optical communications, fibre lasers, optical ampli- fiers, and fibre optic sensors. To date, he has published more than 80 journal and conference papers. He has also been an Active Member of the IEEE Photonic Society.

M. Z. JAMALUDIN (Senior Member, IEEE) received the Diploma degree in electrical and elec- tronic engineering from University Technology Mara (UiTM), in 1983, the B.Sc. degree in elec- trical engineering from the University of Miami, FL, USA, in 1986, the M.Sc. degree in electronic (medical system) from the University of Hertford- shire, U.K., in 1994, and the Ph.D. degree in net- work communication engineering from Universiti Putra Malaysia, in 2007. He worked in Motorola

Malaysia Sdn. Bhd, in 1984 and 1986 and UiTM, in 1990. In 1997, he joined GITN Sdn. Bhd, than Digicert Sdn. Bhd to set-up the first Certification Authority, company that issue the Digital Certificate and appointed the Chief Operating Officer, in 2000. In 2001, he joined the Department of Electronics and Communication Engineering, College of Engineering, Universiti Tenaga Nasional. He was main person responsible in setting up the Spin off company under UNITEN that is UNITEN R & D Sdn Bhd (URND) and promoted as the Managing Director of URND, from 2013 to 2017. He is currently a Professor with the Department of Electronics and Communication Engineer- ing and seconded to the Institute of Power Engineering (IPE). His work and interest include, photonics devices, and sensors, optical networks, secured remote data acquisition systems, RF radiation (GSM, Mobile BS), and eth- ernet passive optical networks. He was an Active Researcher with more than RM7.0 million worth of research grants. He has authored or coauthored more than 120 research articles in journals and conference proceedings. He has been an Active Executive Committee Member of the IEEE Photonics Society, International Conference on Photonics (ICP), since 2004 as the Conference Chairmen and the Committee Members, including as the Chairman, from 2007 to 2008, and a member of the IEEE Malaysia for the past 15 years, recently promoted to a member of IET, since 2010.

M. N. ZAKARIA is currently the Network Archi- tect Manager with Tenaga Nasional Berhad Infor- mation and Communication Technology (TNB ICT), Kuala Lumpur, Malaysia. He started more than 13 years ago in the telecommunications sector at Motorola R & D in Penang. Over the last ten years with Network Planning and Network Oper- ation, TNB, he has a strong leadership in deliver- ing a future proof and reliable telecommunication network solution for power utility systems and

applications, from initial conception through all the project initiation.

14488 VOLUME 8, 2020