3.1. Small Cells
Small cells are low power short range transmission systems or BS deployed to cover a small geographical area. The characteristics of small cells are similar to the conventional BSs. These small cells can be deployed in indoor or outdoor regions and it will provide high data rates and low latency. These small cells are classified into three types based on the area covered by the BS, the maximum number of user’s covered, and the maximum power transmitted [
47]. They are femto cells, pico cells, and micro cells, which are shown in
Figure 4. The classification of small cells and their parameters are shown in
Table 4.
5G requires small cells to achieve its key parameters such as enhanced mobile broad band, URLLC, massive connections and enhanced spectral efficiency. Small cells use low power transmitting stations and are very easy to deploy because of their low complexity in hardware installation. These small cells can be mounted on a wall for indoor applications and small towers or lamp posts can be used for outdoor applications. In small cells, the coverage and data rate depends on the operating frequency, transmitted power, and height and tilt angle of an antenna. The radius of the coverage area is always proportional to the transmitted power [
48]. The small cells have many advantages such as high data rates, faster deployment, cost effective, they can work with low powers, and require less installation space. To achieve high data rates and large capacity, the 5G network is integrated with small cells, MIMO, beam-forming, and mmWaves.
The mmWave, i.e., from 30 GHz to 300 GHz, is a much less utilised spectrum and it provides very high speed communication. However, the problem with high frequency mmWave is that it collides with obstacles such as trees, buildings, etc., in the free space, and causes multipath fading, which leads to reduced signal strength. Therefore, mmWaves are the best frequency bands for short distance communication and can be used in small cells. A geographical area without small cells and with small cells is shown in
Figure 5. In
Figure 5a, the mmWave signal is facing multipath fading but in
Figure 5b, with the use of small cells, the multipath fading is avoided and it provides better coverage than compared to the conventional method.
In the literature, researchers deployed the femto, pico, and micro cells and identified the network capacity, data rates, and coverage area [
49,
50,
51,
52,
53,
54,
55,
56,
57]. In Ref. [
49], the authors considered a stochastic geometry theorem for the optimization of small BS density in a ultra dense Hetnets. The optimized small BS density enhances the energy efficiency for a particular number of UEs. In Ref. [
50], the authors deployed a small cell in a moving bus and estimated the coverage area as 300 m in an urban environment. For indoor applications such as residential, hospitals, schools, offices, and shopping malls, the femto cells are preferred because of their easy installation, cost effectiveness, and low power requirements [
51]. The femto cells will improve the system throughput and signal to interference noise ratio (SINR) and also minimizes the latency and the cell edge coverage problems [
52]. The performance of femto cells is affected by the location they are placed and the infrastructure present in the place, such as number of walls, curtains, doors, ceiling, etc. They will obstruct the signals and cause multi path propagation.. This may reduce the SINR and overall coverage area [
53].
The maximum coverage radius of femto cells is up to 50 m; if the area covered by a cell increases, then the femto cells are replaced by the pico cells. The pico cells are used in indoor and outdoor regions and it also increases the coverage radius up to 200 m and the maximum power required is 250 mW [
54]. If this coverage radius is extended from 200 m to 2.5 km, then the pico cells are replaced by the micro cells. A micro cell covers a large area and the power radiated by an antenna is also increased to 5 W. The pico cells are deployed at the cell edge of macro cells to enhance the coverage capacity, throughput, and traffic offload at the macro cell edge [
55]. The coverage area of a macro cell is added with the pico cells to improve the overall system throughput and the received signal power is used to estimate the path loss [
56]. This method will enhance the coverage capacity and cell edge spectral efficiency. A macro cell and pico cell cooperative scheduling scheme is considered to reduce the UL and DL interference due to the macro cell on pico cell UEs [
57]. The problem with these small cells is handovers, i.e., frequent handovers cause major issues in the small cells. Therefore, the open research challenges in small cells are: power demands, coverage radius, deployment and testing, and mobility and handovers. Different techniques have been used in recent research to resolve the coverage and capacity issues in 5G networks with small cells, as summarised in
Table 5.
The future challenges of small cells are:
Power Requirements: For data transmission, the small cells use mmWave frequency bands, which require strong operating powers. A huge number of small cells must be installed in order to cover a vast area with a high population density, which requires more transmission power. Therefore, the power required will increase with the increase of small cells, which increases the deployment cost and energy. Hence, power optimization techniques must be proposed in the future to enhance the network coverage;
Coverage Radius: Small cells coverage radius is less compared to the conventional BSs. The users need more coverage capacity in their working places or living places, where the small cells are to be deployed. In places where there is no need of any coverage, such as ponds, lakes, and forests, the small cells are not required to be deployed. Therefore, the deployment of small cells depends on the population density, available carriers, and topography. Along with these it also depends on small cell parameters such as supporting frequency bands, carrier frequencies, and coverage area size. The coverage radius is one of future challenges in the small cells to provide a better coverage and to avoid any gaps in the coverage;
Mobility and Handovers: Small cells cover a small geographic area. When a small cell is installed in a shopping mall and a person is moving from one floor to another floor and browsing, emailing, calling, or downloading files, then it should not cause any degradation in the connectivity. Therefore, how many UEs it can handle at a time with high data rates is also one of the challenging issues to be considered in the future;
Deployment and Testing: Small cells can utilise the existing infrastructure such as street lamp poles, walls, top of the apartments, etc. The weight of a small cell is one of the important factors when deploying them in indoor or outdoor regions. The correct way to test a small cell is to improve the quality of service and also to satisfy the subscribers demands. The user traffic, mobility and handovers, and overall system load should be considered as the testing parameters. These parameters should be improved by proper testing. So, AI based algorithms need to be proposed to test the small cells automatically. Identifying the correct testing method is also one of the future challenging issues.
3.2. Carrier Aggregation
Carrier aggregation is the process of combining one or more carriers to increase the data rate and network capacity for continuous transmission or reception [
58]. Initially, the unused spectrum bands or carriers are identified and then added with the primary carriers. These combined carriers are referred to as component carriers (CC), and these are only the downlink carriers or a combination of the downlink and uplink carriers but not only the uplink carriers [
59]. CC is divided into two types: primary CC (PCC) and secondary CC (SCC). The PCC is always in active mode, but the SCC can be in sleep or active mode based on the requirements of the user. If the user requires higher data rates, then the SCC is in active mode. Otherwise, it will be in sleep mode, but the PCC will always be in active mode until the handover is required. In CA, the primary serving cell or PCC, carries both the controlling signals and user data. The secondary serving cell or SCC, carries only user data; it does not carry any control information, as shown in
Figure 6 [
60].
In a network, if four carriers are aggregated, then one is PCC and the remaining three are SCCs. CA in 5G is divided into two types: intra band and inter band, and again these two are subdivided into contiguous and non-contiguous ways of carrier aggregation, which are classified as follows [
61].
Intra-band contiguous method: One or more adjacent carriers of the same band are aggregated to form a single carrier, which is called the intra-band contiguous method of CA;
Intra-band non-contiguous method: One or more carriers separated by some band gap in one frequency band are aggregated to form a single carrier. This is called the intra-band non-contiguous method of CA;
Inter-band non-contiguous method: One or more carriers of different frequency bands are aggregated to form a single carrier. This is called the inter-band non-contiguous method of CA.
Figure 7 [
59] depicts these.
CA is introduced in 3GPP release-10 to enhance the data rates and mitigate the inter-cell interference of the LTE-advanced network [
62]. In the initial stage, it can combine two or more CCs to achieve a maximum bandwidth of 100 MHz, with each CC having a bandwidth of 1.4 MHz, 3 MHz, 5 MHz, 10 MHz, 15 MHz, and 20 MHz [
63]. In both intra and inter bands, the maximum number of UL or DL carriers that can be aggregated is 5. CA uses both FDD and TDD duplexing modes and also uses licensed and unlicensed bands for aggregation. The number of wireless devices increases day by day, and the bandwidth of 100 MHz is not sufficient to support the required data rates. Hence, new releases with increased bandwidths and new applications are proposed by the 3GPP. Release-13 is used by LTE Advanced Pro, and it will support a maximum bandwidth of 640 MHz and aggregate up to 32 UL and DL carriers [
60]. This release allows unlicensed frequency bands in the 5 GHz range to aggregate with licensed bands. Releases 15 and 16 are proposed for 5G New Radio (5G-NR) and provide a maximum bandwidth of 6.4 GHz. Up to 16 CCs are aggregated, and those CCs can select their frequency bands from both the FR1 and FR2 bands. The CA of both LTE and 5G-NR is known as dual connectivity (DC), and it is used by the 5G headset for data transmission to achieve a maximum bandwidth of 6.4 GHz [
64].
The coverage can be enhanced by using carrier aggregation techniques [
65,
66,
67,
68,
69]. The high and low frequency components are aggregated based on cooperation in order to enhance the coverage and capacity in a building [
65]. In Ref. [
66], contiguous and non-contiguous CA methods are deployed, and the authors observe that the proposed methods improve the bandwidth and also enhance the coverage and data rates. In Ref. [
67], aggregation of two CCs is considered, and the performance of CA with independent carriers is compared. The authors observed that the CA achieves a higher data rate compared with the independent carriers. In Ref. [
68], single and multi-flow relay-assisted CA systems were analysed, and it was observed that the relay can be used as a capacity booster for low-frequency CCs. It will improve the capacity and fairness of the CA system. LTE and 5G-NR carriers are aggregated to enhance the coverage, channel capacity, and data speed [
69]. CA has several advantages, including an unutilized spectrum that can be effectively used to improve spectral efficiency, to increase the UL and DL data rates with higher throughput, and improve the system performance. As has been summarised in
Table 6, different techniques have been used in recent research to resolve the coverage and capacity issues in 5G networks with CA.
CA is facing DL, UL, and implementation challenges that need to be addressed in the future:
DL Sensitivity: Instead of designing a duplexer for each carrier component, the interference between DL and UL at the receiver is used to design it. A separate multiplexer or duplexer is required if there is a large frequency separation between the UL and DL frequency bands. Therefore, designing a multiplexer for reducing RF front-end design at large frequency variations is a future challenge;
Harmonic Generations: The harmonics are generated by the use of non-linear components in duplexers, power amplifiers, and transceivers. These harmonics are generated to reduce system complexity. This is one of the open challenges for researchers to design the best harmonic generator;
Filter Design: In California, proper filter design is required to decode the carriers at the receiver without any interference;
Power Amplifier: The intra-band UL CA signals use a higher bandwidth and a high peak-to-average power ratio (PARP) to reduce the maximum power. The maximum power can be reduced by adjusting the resource block configurations. Therefore, to achieve reduced power with high bandwidth and high PAPR, tuned power amplifiers are required. So, the open challenge is the design of tuned power amplifiers for UL CA;
Implementation Issues: The hardware implementation of CA is very critical, as it requires oscillators, radio frequency chains, signal processing techniques, a strong battery life, etc. [
70].
3.3. Device to Device Communication
The way information is exchanged between two UEs is changing as technology evolves. In mobile communication, even though two pieces of user equipment (the source and destination devices) are very close to each other, they need a central entity called BS to establish a connection between them. They are not allowed to establish a direct connection between them without BS, and it increases the data traffic at BS [
71]. Therefore, to address this issue, D2D communication is proposed, and it is one of the most popular coverage improvement techniques on mobile networks [
46]. In D2D communication, two devices can communicate with each other without the use of a base station. The features of D2D communication are traffic offloading, low latency, low power consumption, more spectral efficiency, and a large cellular coverage area [
72].
In D2D communication, the cellular architecture is divided into two tiers. In the first tier, the communication is between the BS and the device in the macro-cell layer, called the macro-cell tier. In the second tier, also called the device tier, only devices will communicate with each other. In D2D communication, the users can share their spectrum resources in the device-tier using licensed spectrum and in the macro-cell-tier using unlicensed spectrum. The device-tier mode of communication is divided into underlay and overlay modes. The D2D users share the same spectrum resources as cellular users in in-band mode, but in macro-cell tier mode, D2D and cellular users share different spectrum resources [
73]. The two-tier network is classified into four types [
74].
Device relaying with an Operator Controlled Link Establishment (DR-OC): In this method, a device will communicate with the BS if there is poor cell coverage or if it is at the cell edge. This method is used to relay the data to other networks, as shown in
Figure 8.
Figure 8.
Illustration of DR-OC communication.
Figure 8.
Illustration of DR-OC communication.
Direct D2D Communication with Operator Controlled Link Establishment (DC-OC): In this method, the BS will establish a link between the source device and the destination device. Once the link is established, the devices will communicate with each other directly without any role of BS, as shown in
Figure 9.
Figure 9.
Illustration of DC-OC communication.
Figure 9.
Illustration of DC-OC communication.
Device Relaying with Device Controlled Link Establishment (DR-DC): In this method, the source and destination devices will communicate with each other using relay transmission data. In this method there is no role of BS, as shown in
Figure 10.
Figure 10.
Illustration of DR-DC communication.
Figure 10.
Illustration of DR-DC communication.
Direct D2D communication with Device Controlled Link Establishment (DC-DC): In this method, the source and the destination devices will communicate with each other by establishing a direct link between them on their own, and there is no role of BS, as shown in
Figure 11.
Figure 11.
Illustration of DC-DC communication.
Figure 11.
Illustration of DC-DC communication.
The successful implementation of D2D communication faces many issues and challenges such as mode selection, device discovery, interference management, and security and mobility management [
75,
76,
77].
3.3.1. Mode Selection
In D2D communication, two users can directly communicate with each other. Mode selection is the process of choosing the right mode, i.e., D2D or cellular mode, for communication between two UEs, which will improve the system’s performance, throughput, spectral efficiency, and reduce latency. The mode selection issues and proposed solutions are discussed in this section [
78,
79,
80,
81,
82,
83,
84,
85,
86,
87]. In Ref. [
78], a joint optimization algorithm has been proposed for mode selection and power allocation in D2D underlay communication. This will improve the total sum rate and overall system performance using relay-based D2D. A degree of freedom-based mode selection algorithm and an interference alignment algorithm for interference management are proposed in Ref. [
79]. These two algorithms will reduce interference and enhance the system’s performance at high SNR. These two algorithms will be suitable for large MIMO systems and small cell 5G networks. A distributed coalition formation algorithm has been proposed to achieve high system performance under link allocation and mode selection [
80]. An underlying D2D multi-hop relay-aided scheme has been proposed to improve the coverage capacity [
81]. A joint mode selection, relay selection, and resource allocation optimization algorithm are proposed to improve the throughput and SINR [
82]. A greedy heuristic algorithm is proposed for mode selection based on QoS [
83]. This method will improve the total sum rate and SINR without affecting the signalling overhead.
In Ref. [
84], authors proposed an optimal mode selection and resource allocation algorithm for D2D communication in order to encourage mutual cooperation among D2D users, which also increases the throughput and efficiency. In Ref. [
85], a partial CSI algorithm with a low overhead was proposed. Along with this, rate adaptation, mode selection, and user scheduling algorithms are also used to reduce intercell interference and cross-link interference. The analytical expressions for mode selection in underlay mode are derived from a single antenna at the BS. In Ref. [
86], an interlay mode selection model is proposed for NOMA systems. This will use SIC decoding and power domain multiplexing to reduce interference between the users. A joint mode selection and resource allocation scheme is proposed to improve the sum rate and achieve better SIC decoding outputs. In Ref. [
87], an optimal mode selection algorithm for a full-duplex CR has been proposed to improve the spectral efficiency. The crucial zone is defined to identify the operating mode of users, i.e., if the receiver is inside the guard zone, then the operating mode is considered to be a half-duplex mode. If the transmitter and receiver are outside the guard zone, then the operating mode is considered full duplex mode. Different mode selection, power allocation, and resource allocation algorithms have been used in recent research to resolve the coverage and capacity issues in 5G networks, as summarised in
Table 7.
The future research challenges of mode selection in D2D communication are:
Dynamic mode selection: The papers so far discussed in this section are mostly focused on the static mode selection schemes of D2D communication. Mode selection depends on the spectrum and resource availability. Therefore, an optimal mode selection scheme needs to be proposed based on the resources available;
Mode selection overhead: The mode selection depends on the channel estimation, signalling, and CSI, which leads to system overhead. Therefore, the amount of overhead needs to be minimised by proposing an appropriate algorithm to improve the throughput and device lifetime in D2D communications.
3.3.2. Device Discovery
In D2D communication, before any two user equipments (UE) communicate with each other, a secured device discovery process is required. In this process, the first step is to identify whether there is any device that is near another. It is possible if one device will send a discovery signal to identify the presence of other devices that are near it, and then they can form a pair by sharing the link establishment data between the devices and BS. If the two devices are discovered by each other and a secured link is established, then they can transfer their data without revealing their identities. The D2D discovery procedure is classified mainly into centralised and distributed methods [
88]. A central entity known as a BS or an access point will allow all other devices to discover one another in the centralised method. After discovery, if any device wants to connect with another device, it should send the purpose of the connection to the BS. Initially, the BS will verify the channel conditions, interference control policy, and transmission power before establishing a connection. BS will also verify the SINR, path gain and the connection possibility between the two devices. Finally, it will confirm the connection between the two devices.
In the distributed method, one device can send a device discovery signal to any other device without informing the BS. In this method, the connecting devices have to verify all the channel conditions, path gain, SINR, connection possibility, and so on. This method will cause some interference, security, and synchronisation issues. In the literature, many device discovery algorithms are proposed for both centralised and distributed methods [
89,
90,
91,
92,
93,
94,
95,
96].
To detect a large number of UEs and reduce latency and collision probability in device discovery, Ref. [
89] proposed a direct discovery scheme based on a random backoff procedure. In Ref. [
90], a VANET-based discovery scheme has been proposed to reduce the latency in the discovery process and to enhance throughput and energy efficiency. In Ref. [
91], an adaptive D2D discovery scheme has been proposed to reduce the energy consumption required for device discovery by reducing the discovery signalling messages in the network. This scheme will improve energy efficiency and throughput. In Ref. [
92], a socially aware peer discovery scheme was proposed to identify malicious and trusted users. This scheme will increase the probability of identifying trusted users and reduce the number of malicious users. Therefore, the throughput increases with the increase in trusted users in D2D communication. In Ref. [
93], an energy-efficient device discovery scheme has been proposed for device discovery over static and random backoff patterns. This method will increase the overall number of devices discovered for D2D communication. A mathematical model for discovering neighbourhood devices was developed in 1994 [
94]. In this method, devices are discovered when they are moving, static, or out of the network.
To improve spectrum efficiency and reduce device discovery delays, a public safety full-duplex device discovery scheme was proposed [
95]. In this method, there is a possibility of switching the device discovery mode from full duplex to half duplex and vice versa based on the QoS requirements. The priority in device discovery, i.e., public safety, has the highest priority as compared to random access devices, which will enhance the discovery time. In Ref. [
96], a device discovery and localization scheme using UAVs is proposed for public safety systems. The MUSIC algorithm is used to estimate the accuracy, throughput, and packet error rate. Different device discovery algorithms have been used in recent research to resolve the coverage and capacity issues in 5G networks, as summarised in
Table 8.
The future challenges that need to be addressed in device discovery are:
Initial device discovery signal: In D2D communication, initially the device will send a device discovery signal to detect the neighbouring devices. Any malicious user can also send the discovery signal to identify trusted users and create some security issues. Therefore, to avoid security issues due to malicious users, the initial signal generation parameters such as signal data size and control data have to be secured;
Multi cell device discovery: The proposed research till now has focused on single cell device discovery models. If the user is at the cell edge or dynamically changing his location from one cell region to another cell region, then multi-cell device discovery is required. Therefore, detection of multi-cell devices is one of the future challenges in D2D communication systems;
Non Repudiation and Traceability: Non repudiation prevents D2D users from being denied transmission and reception of messages. Traceability is mandatory to detect the source of false messages. This is also one of the future challenges to providing a secured data transfer;
Availability and Efficiency: In D2D communication systems, the availability of a device largely depends on the degree of cooperation of devices. Efficiency is the ability of a communication system to implement and operate economically. Therefore, achieving higher efficiency in device discovery is one of the future challenges in D2D communications.
3.3.3. Interference Management
In a cellular network, the same spectral resources are used by both cellular users and D2D users, which will cause interference for the cellular users. The most commonly occurring types of interference in D2D communication are co-tier and cross-tier interference. In co-tier, the interference occurs between the same tier network elements, i.e., between one D2D user (DU) and the neighbouring DU, because they are located very near each other. This is due to assigning the same frequency channels and resource blocks to multiple DUEs. This co-tier interference can be overcome by proper assignment of spectrum bands and user pairing. Cross-tier interference occurs between the device network elements, i.e., DUE and CUE, which are of different tiers. This is between one CUE and one DUE or one CUE and multiple DUEs. The basic reason for cross-tier interference is sharing the same frequency channels or resource blocks with one or more D2D users.
To manage the interference between D2D users, centralised, distributed, and semi-distributed interference-controlled approaches are proposed. In a centralised method, a central entity collects the channel state information (CSI), channel quality, and interference range from all the users in the network and then manages the interference between CU and DU [
97]. The main issue with this method is the significant amount of signalling overhead required for exchanging CSI and feedback, which grows exponentially with the number of DUs. This method is useful if the network size is small. In the second method, there is no need for a centralised entity to collect complete data from all the DUs. Each user will collect the data using their own methods. Therefore, the signalling overhead will be reduced and will be used for large-sized networks. The only disadvantage is the difficulty in coordinating the users. To overcome the pros and cons of centralised and distributed methods, a hybrid interference control approach is popularly used. The interference between cellular users and D2D users can be overcome by using interference management schemes [
98,
99,
100,
101,
102,
103,
104,
105,
106,
107].
An advanced coding and decoding scheme is proposed for interference cancellation (IC) in the CUE and DUE and to enhance the network capacity. A two-way decode and forward relay mode-based scheme is proposed in D2D communications, and M-antenna BS is used in cellular user communication to reduce interference [
98]. IC and beam-forming concepts are also used to reduce the interference between the CUE and DUE in both asymmetric and symmetric cases. In Ref. [
99], a multiple interference cancellation scheme (MIC) is used instead of SIC in NOMA. The MIC optimises power consumption and provides better system performance compared to the SIC receiver. In Ref. [
100], the authors proposed an innovative method to overcome the interference problem by using a MIMO with zero forcing and a minimum mean square error SIC scheme to improve the performance of the NOMA system. In this method, a practical interference cancellation scheme reduces the bit error rate and improves the system performance by predicting the interference signals. In Ref. [
101], the performance of cellular networks using interference cancellation and beam-forming with M-antennas at the BS is analysed. From the results, it is observed that the interference between the D2D users and cellular users is decreased and the SNR is enhanced. In Ref. [
102], a guard zone-based interference mitigation algorithm is proposed based on the SIC at the BS. This method will improve the average throughput of the UE and the transmission probability.
To avoid interference between the DUEs and CUEs, interference avoidance techniques are proposed. In Ref. [
103], a multi-hop D2D resource allocation scheme was proposed to avoid interference. From the simulation results, it is observed that the system’s performance in terms of throughput has increased. In Ref. [
104], a massive MIMO system operating in FDD mode has been proposed to reduce the feedback overhead. In Ref. [
105], the authors proposed a relay-based D2D system to improve the system’s capacity and performance by avoiding interference. The relay can use the untapped resources of the CUE to enhance the spectrum’s efficiency. In Ref. [
106], a fractional frequency reuse and a blank sub-frame concept were used to avoid interference between the CUEs and DUEs. The numerical results show that the interference level of the DUEs decreased as compared to the conventional methods. This method increases the system throughput and efficiency. In Ref. [
107], a dynamic distance-based algorithm was proposed to reduce the interference. Using this algorithm, outage probability, SINR, and cell density are estimated for D2D-enabled hetnets. Different interference management algorithms have been used in recent research to resolve the coverage and capacity issues in 5G networks, as summarised in
Table 9.
The future challenges in the interference management that need to be addressed are
D2D in mmWave Communication: Multi-cell D2D communication is possible with the small cell concept and mmWave in 5G networks. The mmWave improves the data rates, network capacity, and latency but different interferences occur due to the penetration losses of mmWaves. Therefore, the interference mitigation algorithms for 5G mmWave need to be addressed to provide multicell D2D communication;
Cell Densification and Offloading: The small cell and macro cell concepts of 5G are integrated with D2D communication to provide multi-cell D2D transmissions. Multi-cell D2D is used to improve the spectral efficiency and overall system performance of D2D communication. Therefore, the resource allocation and interference control algorithms for multicell D2D communication are challenging future issues.
3.3.4. Security and Privacy
In D2D communication, device discovery was done using a centralised, distributed, and semi-distributed scheme. The centralised and distributed schemes are paired in a semi-distributed scheme to identify neighbouring users, which causes some security issues in both cellular networks and ad-hoc wireless networks. These issues will affect the availability, confidentiality, and authenticity of the devices on the network. Therefore, to overcome these issues, some privacy- and security-based algorithms are proposed [
108,
109,
110,
111].
In Ref. [
108], a user sub-channel matching algorithm has been proposed to address the spoofing and interference between the CUEs and DUEs. The proposed algorithm will reduce the interference and improve the overall sum rate. In Ref. [
109], an optimization algorithm has been proposed to reduce the interference due to eavesdroppers. This proposed method will reduce interference and enhance the physical layer security of CUs and throughput, but it will protect only one CU from multiple eavesdroppers. In Ref. [
110], two privacy-preserving authentication protocols are proposed to provide a secured link between the devices in D2D communication. These methods will improve the efficiency and security by reducing the interference between the DUs. In Ref. [
111], a secure key generation algorithm has been proposed to secure the DUs from malicious users, which leads to improved information confidentiality between the DUs. Only a few authors proposed algorithms to solve the security and privacy issues. Different security and privacy algorithms have been used in recent research to resolve the coverage and capacity issues in 5G networks, as summarised in
Table 10.
The future challenges relating to privacy and security issues are as follows:
Lack of Standardisation: To provide a secured connection between the DUs, there are no standard rules for data transmission, the amount of data to be transferred, feedback signalling, security policies, and so on. Therefore, the standardisation of these parameters needs to be addressed;
Security Threats: A lot of security threats are happening in D2D communications. They are denial of service, man in the middle attack, impersonation attack, session hijacking, interference attack, data leakage, malware attack, free riding attack, data modification, privacy violation, forge attack, active attack on controlled data, and so on. All these need to be addressed in the future;
Privacy and Security: D2D communication allows DUs to communicate directly with one another. However, identifying the trusted users and malicious users is a challenging issue. Therefore, efficient algorithms need to be proposed to identify the trusted users and malicious users before providing a secured connection;
Data Confidentiality and Integrity: Data confidentiality protects the data being revealed from unauthorised users as well as preserves the user content and secures the privacy. Integrity ensures that data is not altered during the transmission. This is also one of the future challenging issues in providing an accurate data transmission;
Session Key Agreement: This session key agreement is used to ensure the security of data transfer over air interface. It is mandatory to keep the session key secret to avoid any type of security issues;
Non Repudiation and Traceability: Non repudiation prevents D2D users from being denied transmission and the reception of messages. Traceability is mandatory to detect the source of false messages in a D2D communication system.
3.4. Non Orthogonal Multiple Access
To enhance the coverage and capacity, the cellular networks use many access techniques, such as frequency division multiple access (FDMA), time division multiple access (TDMA), code division multiple access (CDMA), and orthogonal frequency division multiple access (OFDMA). The concept of orthogonality is used in these accessing techniques to reduce the interference between users, but it also reduces the number of users multiplexed to access the spectrum.
Therefore, the concept of NOMA has been proposed in 3GPP long-term evaluation release 15 as the latest multiple access technology, and it is also implemented in 5G [
112]. Initially, NOMA was used with a single cell to improve spectral efficiency. The spectral efficiency is increased by increasing the number of multiplexed users accessing the single channel with different channel gains [
113]. NOMA is basically divided into two types: power domain NOMA (PDNOMA) and code domain NOMA (CDNOMA). In PDNOMA, multiplexing is based on the transmitted power, and in CDNOMA, multiplexing is based on code [
114,
115,
116]. Along with CDNOMA and PDNOMA, there are some other NOMA techniques such as signature NOMA (S-NOMA) and compressing-based NOMA (CS-NOMA) [
117]. NOMA works with MIMO, cognitive radio (CR), HetNets, millimetre waves, mobile edge computing, visible light communication, vehicles for any communication, etc. This will provide high spectral efficiency, high data rates, and massive connectivity while decreasing inter-cell and intra-cell interference [
118]. In NOMA, fairness in power allocation is obtained by assigning less power to strong channel gain users and more power to weak users. This power efficiency will increase the overall throughput and efficiency. NOMA can be implemented with a single cell or with multiple cells. The authors in Ref. [
119] implemented the NOMA on the multi-cell and identified some issues that occurred due to the multi-cell NOMA.
By utilising the idea of high- and low-gain users, NOMA increases the throughput and spectrum efficiency by eliminating interference. It is called successive interference cancellation (SIC). A SIC receiver is one of the most important blocks in NOMA transmission, which is shown in
Figure 12 [
114]. The quality detection of the strongest user who has a strong signal is identified by subtracting the other user’s signal. Similarly, the weak user detects its own signal by subtracting the others’ signal, assuming noise or interference, which will improve the system’s performance by minimising the propagation effects. However, while implementing the SIC in NOMA, security issues will arise. These issues are classified as resource allocation issues, signalling issues, and security issues, which are discussed in the following subsections.
3.4.1. Resource Allocation Issues
In NOMA, one resource block (RB) is shared by multiple users, and the SIC receiver is used to decode the user information at the receiver end based on the user’s channel gains. Interference between users can be avoided by choosing proper power allocation algorithms. Otherwise, resource allocation issues such as user pairing and power allocation (PA) will arise. In user pairing, the users with less power are allocated with more channel gain, and the users with more power are allocated with less channel gain to make the channel fair for all the users at the transmitter end. At the receiver end, the SIC receiver is used to decode the same. In this method, if the number of users increases, then the complexity of decoding also increases at the receiver end. This is one of the major problems with user pairing. Along with this, there is another problem, i.e., if the users with high and low gain are transformed to mid gain, then mid gain users may be paired or may not, which leads to reduced channel capacity.
To overcome the user pairing issues, optimization techniques, game theory, and machine learning algorithms are proposed in the literature. The authors proposed an optimization method while pairing two users [
120,
121,
122,
123]. To optimise the user pairing, the channel gain should not be less than the predefined threshold. A strong channel pairing algorithm (CSS-PA) was proposed to increase the system capacity and fairness in user pairing [
120]. In Ref. [
121], the authors used a new pairing concept, i.e., the highest channel gain users are paired with the next highest gain users, and so on. Game theory for multiple user pairing and machine learning algorithms for user pairing are proposed to reduce the channel capacity [
122,
123]. Different user pairing and resource allocation algorithms have been proposed in recent research to resolve the coverage and capacity issues in 5G networks, as summarised in
Table 11.
3.4.2. Power Allocation Issues
In the power allocation scheme, users who have more channel gain are given less power, while users who have less channel gain are given more power. In NOMA, the users with various PA are multiplexed in one RB, which will increase the system’s performance, but inefficient PA methods cause interference and also reduce the system’s performance. Therefore, proper selection of power allocation methods plays an important role and depends on many parameters such as CSI, QoS, total system power, and channel conditions.
Therefore, many optimization techniques are proposed for PA fairness [
124,
125,
126,
127,
128,
129,
130]. The PA is optimised by maximising the maximum and average sum rates using physical layer QoS [
124,
125]. The optimal PA was proposed by minimising the power consumption [
126,
127,
128]. Different PA techniques have been proposed for maximising the capacity rate and total power [
129,
130]. PA based on expenditure or cost is explained using graph theory [
131,
132,
133,
134,
135,
136,
137]. In Ref. [
131], the authors proposed a simulated annealing to optimize the PA and to maximise the throughput. In Ref. [
132], two user concepts were considered. In this method, power is allocated to two users or only one user, whichever has a higher power gain. In Ref. [
133], the authors proposed a price-based PA optimization algorithm for two users based on their QoS. The data rate is maximised using an optimization algorithm [
134]. In Ref. [
135], the authors proposed an auction-based game design by bidding on the transmitting power. The spectrum scarcity issues in both licensed and unlicensed bands for NOMA were discussed in Refs. [
136,
137].
The joint PA and user pairing problems are solved by using an optimization algorithm [
138,
139,
140,
141,
142,
143,
144,
145,
146]. To identify the best multiplexed users in NOMA, a greedy algorithm, and to optimise PA, an optimization algorithm was proposed in Ref. [
138]. A joint PA and user pairing optimization has been used to maximise the transmitted power and reduce the computational complexity of the NOMA system [
139]. In Ref. [
140], the authors proposed a proportional fair scheduling algorithm to reduce the complexity of a single carrier NOMA system. A dynamic PA scheme was used to optimise the PA in both UL and DL and also to maximise the system throughput [
141]. The proposed fractional error factor method will increase the energy efficiency and also maximise the performance gain of both the UL and DL of UE [
142]. A swarm optimization algorithm has been proposed to find the outage rate and average data rate of all the user pairs [
143].
The spectral efficiency and error probabilities are improved by using a sort-based user pairing method [
144]. This method was used to maximise the system capacity and reduce the interference. The user pairing and PA were dynamically varied to identify the system capacity of an individual and paired users [
145]. The system capacity for pairing users was better than for individual users. In Ref. [
146], the bit error rate is considered along with the individual and paired user’s capacities. As summarised in
Table 12, various PA algorithms and joint user pairing and PA algorithms have been proposed in recent research to address the coverage and capacity issues in 5G networks.
3.4.3. Signalling Issues
With the increase in wireless users in 5G, effective spectrum utilisation plays an important role. NOMA has been proposed to improve spectrum efficiency by allocating multiple users to a single RB. However, the allocation of multiple users to one RB will cause signalling issues such as CSI imperfections, latency, SIC receiver complexity, and interference management. In the literature, many authors use optimization techniques as well as analytical techniques to address all of these issues.
SIC Issues: The concept of SIC in NOMA was introduced to decode the signals of paired users at the receiver [
147]. The authors used a PD-NOMA for multiplexing the users into one RB, and SIC was used to decode the users at the received PD-NOMA for multiplexing the users into one RB, and SIC was used to decode the users at the receiver. The SIC hardware implementation is very complex at high powers, but theoretical implementation is possible using Moore’s law. The performance gain of NOMA has been affected by SIC characteristics such as receiver complexity, decoding order, imperfect channel estimation, and so on [
148]. The SIC receiver has been affected by the decoding of users. If the decoding order of the users does not match the multiplexed sequence, then it will cause imperfect SIC implementation. In this case, a re-transmission request is forwarded to the BS, and the whole process has been repeated again to get the perfect SIC results from the receiver. However, it takes more time to decode the data at the receiver end [
149].
Therefore, optimization algorithms have been proposed to avoid SIC imperfections [
150,
151,
152,
153,
154]. In Ref. [
150], the authors proposed an efficient PA algorithm for multi-carrier NOMA (MC-NOMA). The optimal SIC decoding value has been determined by the outage threshold of the carrier-to-noise ratio. Two algorithms, i.e., complementary geometric programming (CGP) and arithmetic geometric mean approximation (AGMA), have been proposed to optimise the total power [
151]. A two-phase algorithm has been proposed to solve the non-convex resource allocation problems in NOMA [
152]. Using a matching theory algorithm, the user scheduling is optimised first, and then the branch and bound technique is used to optimise PA. In Ref. [
153], an efficient successive convex approximation (SCA) algorithm is proposed to optimise the sum rate. The sum rate and system utility are maximised by the SNR and the number of iterations in the MIMO-NOMA system. The iterative PA algorithm provides a maximum weighted sum rate in the NOMA system [
154].
The analytical techniques used to overcome signalling issues such as outage probability and CSI of NOMA in UL and DL are addressed by the authors in Refs. [
155,
156,
157]. In Ref. [
155], the authors proposed a dynamic SIC receiver concept based on the users’ received power, and the outage probability is estimated theoretically. SIC analytically estimates the statistical CSI to improve the outage probability [
156]. In Ref. [
157], a new decoding algorithm is used to decode the signal information at the SIC receiver, and its outage performance is verified analytically.
CSI Issues: In literature, most of the authors assumed a perfect CSI at the receiver. However, in NOMA, the concepts of SIC decoding, user pairing, and PA depend on imperfect CSI. This leads to computational complexity, overhead signalling, and delayed feedback. To overcome these issues, optimization and analytical methods are identified, and with these methods, the outage probabilities are increased by assuming imperfect CSI and also reducing the number of feedback bits [
158,
159,
160,
161,
162]. The minimum outage probability is obtained from a minimum feedback rate. An optimal power allocation scheme is used to improve the total sum rate of a NOMA system [
158]. The outage probability is minimised by optimising the dynamic PA using the one-bit CSI feedback method [
159]. Imperfect CSI and perfect CSI models are used to estimate the system throughput and bandwidth [
160]. The outage probability based on channel feedback is estimated analytically [
161,
162]. As summarised in
Table 13, various optimization and analytical algorithms have been proposed in recent research to resolve signalling issues and improve coverage and capacity in 5G networks.
3.4.4. Security Issues
In NOMA, multiple user messages are superimposed on a single RB. The SIC receiver is used to decode the overlaid signals on the receiver. In the SIC receiver, the strong users are allocated with lower powers and the weak users are allocated with stronger powers for channel fairness. At the receiver, strong users have to subtract the low-power signals to get their own signals. However, it causes some security issues, which have to be considered in NOMA. These security issues can be overcome by using encryption and decryption schemes at the sender and receiver, respectively. This process increases the latency and processing requirements. However, the basic idea of NOMA in 5G is to decrease the latency and improve the data rates, spectral efficiency, and bandwidth.
The NOMA implementation causes some security issues, such as the implementation of SIC, transmitted power, and outage probability. These security issues mainly occur at the physical layer. To overcome these security issues, optimal and analytical solutions are proposed in the literature.
Optimization Methods: The authors in Refs [
163,
164,
165,
166,
167] proposed optimization techniques to overcome the security issues in NOMA technique. An optimization algorithm is used to maximise the security sum rate (SSR) in a secured physical layer of a NOMA system. The authors in Ref. [
163] proposed an optimal PA and optimal power splitting algorithms to maximise the security sum rate. The proposed algorithms significantly improve the system performance by assuming a uniform PA and a fixed power splitting scheme. In Ref. [
164], the authors proposed two mobility models, such as a random way point (RWP) and random direction (RD), to observe the security performance of the physical layer. The average security sum rate of NOMA users was estimated and analysed. From the numerical results, it is observed that the RWP achieves a high safety rate than all other mobility models. In Ref. [
165], the security sum rate is maximised by using a two-stage optimization technique. In the first stage, the SINR of a particular user is fixed to its maximum value, and then it is used to maximise the security sum rate by using a one-dimensional search algorithm. The optimal beamforming technique is used to improve the safety rate by adjusting SINR between signal strength and interference. In Ref. [
166], the authors considered the security outage probability (SOP) to maximise the NOMA security performance. Design parameters such as PA, decoding order, and data rate are estimated even though the CSI is not perfectly known at the transmitter. Pairing a strong user with an unstructured weak user improves the physical layer performance [
167]. The optimization algorithm minimises the pairing outage probability. The mathematical expressions for outage probability and secured outage probability are derived based on the proposed optimization algorithm;
Analytical Methods: The physical layer security is estimated using analytical methods along with optimization methods. The numerical expressions for SOP are derived using analytical methods in Refs [
168,
169,
170]. NOMA users and eavesdroppers are randomly placed, and a protected area is generated around the source of the eavesdroppers with an imperfect SIC. The security performance of the physical layer is improved by increasing the eavesdroppers’ execution area. This is extended for multiple antenna transmissions and estimates the numerical expressions for diversity order, which will reduce the SOP of multiple antenna transmissions [
168]. In Ref. [
169], the authors assumed a new concept where BS and users are operating in half-duplex mode and eavesdroppers are in full-duplex mode. To interrupt the NOMA transmissions, the eavesdroppers are performing active jamming and passive eavesdropping. A novel transmission outage probability scheme is proposed to improve the outage transmission probability. The analytical expressions for SOP and security diversity orders are derived to improve the security performance. The security performance of an overlay CR-NOMA system was described in Ref. [
170]. In this method, we consider the primary user (PU) as a secured user and the secondary user (SU) as an eavesdropper. This proposed system provides guarantees of QoS for the PUs and reduces the interference between the users. The analytical expressions for connection outage probability, SOP, and security throughput of the PU’s are derived by assuming the Nakagami-m fading channel. By reducing the number of SUs, security performance can be improved. As summarised in
Table 14, various optimization and analytical algorithms have been proposed in recent research to resolve security issues and improve coverage and capacity in 5G networks.
3.4.5. NOMA—Future Challenges and Research Issues
NOMA can be included with the other 5G enabling technologies such as MIMO, CR, HetNets, millimetre wave, and so on. This helps in achieving the 5G key parameters such as improved spectral efficiency, low latency, high data rates, extended coverage capacity, and mass connectivity. However, the inclusion of NOMA with 5G and beyond 5G leads to some issues and challenges such as resource allocation, pairing, decoding, security, and signalling. These issues can be overcome by using some technologies and techniques such as optimization and analytical techniques, game theory, and machine learning algorithms. There are still some challenges and issues with the implementation of NOMA that need to be addressed. The major concerns that need to be addressed in the future are:
User Pairing in NOMA: In the NOMA implementation, multiple users are multiplexed into one RB, and the SIC receiver will decode the paired users at the receiver. We have discussed pairings of two or three users in previous sections, but not multiple users. Nowadays, the demand for connected devices such as IoT, V2V, and massive machine types is increasing, which requires them to pair a large number of users to one RB. Therefore, NOMA requires some new pairing techniques to fulfil the requirements of future wireless networks. Along with this, there is another challenge to be addressed, which is the practical implementation of user pairing in NOMA systems;
Receiver Complexity in SIC implementation: In NOMA, the users are multiplexed at the transmitter side using pairing techniques and decoded at the receiver using a SIC receiver. As the number of users increases, the allocation of transmission power to pair the users becomes very complex, and at the same time, decoding the strong user to the next strong user and up to a weak user is a very difficult task at the receiver end. This process increases the latency and also introduces interference with the increased number of multiplexed users. Therefore, reducing the latency and interference and providing an efficient and dynamic SIC receiver for the NOMA system is a future challenge;
Multi-cell NOMA system: In NOMA systems, most of the researchers considered a single cell, and very few addressed the multi-cell concept because the multi-cell system causes inter-cell interference. The interference will affect the performance of the weak users who are at the cell edge. Therefore, to solve the interference issues along with the pairing and decoding issues, a small cell concept of 5G is included in the multi-cell NOMA system. This is the future challenge in the multi-cell NOMA systems to be addressed;
Mobility in NOMA: NOMA is used to enable 5G and beyond 5G technologies. In IoT, V2V, and M2M communications, mobility is the key parameter. Most of the research in NOMA systems is based on static systems. The PA, pairing, and SIC receiver algorithms are proposed based on the static behaviour of the users. However, in future communications, dynamic PA, pairing, and SIC algorithms will be required. As the user moves from one location to another, the channel gains vary with respect to the user’s location. Therefore, proposing dynamic algorithms for NOMA systems is one of the challenges of the future;
CSI in NOMA: Most of the research is carried out assuming perfect CSI in NOMA systems, but fewer users consider imperfect CSI. The CSI plays an important role in user pairing and decoding for users with SIC receivers in the NOMA system. To increase the system performance and spectrum efficiency of a dynamic system, the estimation of CSI is very important. For a dynamic channel, the estimation of CSI using machine learning and game theory algorithms is one of the important future challenges in NOMA systems.
3.5. Multiple Input and Multiple Output
The key parameters of 5G can be enhanced with the use of MIMO technology. Exploiting the bandwidth and spectral efficiency features of wireless networks will help increase the throughput. The present research is concentrated on methods that increase spectral efficiency because increasing the bandwidth has the disadvantage of decreasing the SNR for the same transmitted power. Utilizing numerous antennas at the transceivers is a well-known method of enhancing spectral efficiency [
171]. To provide high-speed transmission with a minimal quality of service, MIMO communication systems have been proposed based on the usage of an antenna array at the transmitter and receiver [
172].
Foschini [
173] and Telatar [
174] observed a sharp, linear rise in the channel capacity with the increasing number of antennas in the MIMO systems. MIMO systems use two unique dimensions, i.e., diversity and capacity of the radio link, to enhance the system’s performance. Diversity enhances the communication channel’s dependability by utilising multiple antenna links, and capacity can be increased by employing multiple antennas and multiplexing techniques, which maximise the amount of information transmitted through the channel. However, the increase in data traffic these days increases the need for spectrum for 5G wireless networks and beyond. Therefore, to enhance the data rates and the channel capacity, a mm-wave spectrum with a huge number of small radiating elements at the BS, called massive MIMO (M-MIMO) technology, has been proposed [
175].
Massive MIMO is an advancement in MIMO technology that increases the spectral efficiency, energy efficiency, data rates and throughput by using hundreds or even thousands of antennas connected to a BS, which is shown in
Figure 13 [
29]. M-MIMO is a significant technique used to enhance the throughput, spectrum efficiency, data rates, capacity and overall system performance [
176,
177,
178,
179,
180,
181]. In Ref. [
176], the authors introduced a M-MIMO technology for 5G and beyond. The system performance, throughput and data rates are enhanced through the proposed technology and they achieved a maximum data rate of 100 Gbps. M-MIMO, in conjunction with the planar antenna array, is used to enhance the system capacity. However, the planar antenna array reduces the coverage radius to only tens of kilometres. So, to enhance the coverage radius up to 100 km, a cylindrical M-MIMO system has been proposed [
177]. The proposed system enhanced the coverage radius along with the system capacity by 2.1 times compared to the conventional system with a planar array.
The beam multiple access technique was used to enhance the system capacity of a M-MIMO system [
178]. This will reduce the bit error rate and enhance the system capacity by 10 times and energy efficiency by 100 times compared to conventional methods. In Ref. [
179], the authors proposed a cell-free MIMO UL receiver with zero forcing, MMSE, and maximum ratio combining detectors to enhance the spectral efficiency. When compared to conventional methods, the zero forcing and MMSE methods improve spectral efficiency. In Ref. [
180], the authors considered a triangular lattice structure of antennas within the frequency bands of 24.25 GHz and 29.5 GHz for M-MIMO systems. This structure achieves higher spectral efficiency by maintaining a minimum distance between the antennas. In Ref. [
181], the authors estimated the optimal antennas by using the trade-off between energy efficiency and spectral efficiency in a M-MIMO system. This will reduce the cost of the transmitted power and the total energy consumption. In Refs. [
182,
183], the authors discussed MIMO, its opportunities, recent trends, and advantages in 5G networks and beyond 5G.
Different antenna arrays, multiplexing techniques for M-MIMO have been proposed in recent research to enhance the data rates and capacity in 5G networks, as has been summarised in
Table 15.
MIMO—Future Challenges and Research Issues
5G key parameters such as improved spectral efficiency, energy efficiency, high data rates, and enhanced capacity are achieved using the MIMO technology. There are still some challenges and issues in the implementation of M-MIMO that need to be addressed. The key issues that need to be explored in the future are:
Signal Detection: The UL signal detection becomes more complex in M-MIMO systems due to the large number of radiating elements, which decreases the system throughput. Along with this, the superimposition of user-transmitted signals at the BS causes interference, which also reduces the spectral efficiency and throughput. Therefore, to enhance the spectral efficiency and throughput of a M-MIMO UL, intelligent algorithms are required. Designing less complicated and accurate algorithms for UL signal detection is one of the future research challenges;
Channel Estimation: For a perfect CSI, the M-MIMO system performance increases linearly with the number of transmitting and receiving antennas. The BS should know how to use CSI in order to identify and detect the user-transmitted signal at the UL and to precode the signals at the DL. Channel estimation at the DL and UL depends on the FDD and TDD duplexing modes.
FDD Mode Channel Estimation: In this mode, the CSI needs to be estimated for both DL and UL. During the DL, the BS forwards the pilot signals to the user, and the user replies with the estimated CSI to the BS. Similarly, during the UL, the BS estimates the CSI using the orthogonal pilot signals forwarded by the user. The DL channel estimation for a M-MIMO system with a large number of antennas becomes very difficult and is impossible to carry out in real-world applications.
TDD Mode Channel Estimation: The problem during the DL channel estimation in FDD mode is solved by using TDD mode. In this mode, with the advantage of channel reciprocity, the BS estimates the DL channel with the use of UL CSI. During DL, the user will forward the orthogonal pilot signals to the BS, and the BS will estimate the CSI to the user terminal based on these pilot signals;
Pilot Contamination: In M-MIMO systems, the BS requires the user terminal’s channel response in order to estimate the channel. When the user terminal delivers orthogonal pilot signals to the BS, the BS estimates the uplink channel. Additionally, the BS calculates the downlink channel to the user terminal using M-MIMO’s channel reciprocity property. The BS can accurately estimate the channel if the pilot signals in the home cell and nearby cells are orthogonal. However, because there are only a limited number of orthogonal pilot signals available in a given bandwidth and period, adjacent cells must reuse the orthogonal pilot signals, which leads to pilot contamination. Hence, the pilot signal contamination is one of the challenges for future research in M-MIMO systems;
Energy Efficiency: Energy efficiency is defined as the ratio of spectral efficiency to the total transmitted power. M-MIMO systems offer significant energy efficiency by achieving higher spectral efficiency with low power consumption. The increased number of antennas in the M-MIMO system will increase the spectral efficiency, but it will also increase the total power consumption, which reduces the energy efficiency. Numerous studies have been done to develop energy-efficient M-MIMO systems based on the trade-off between energy efficiency and spectral efficiency and also to reduce the power consumption by designing the power amplifiers. Therefore, achieving higher energy efficiency is one of the challenges of future research.
3.6. 5G Optimization Algorithms
Nowadays, cellular networks are becoming increasingly challenged to provide high data rates, low latency, enhanced coverage, and capacity. The user requirements are to be fulfilled with advanced technologies. These technologies might affect network complexity and also cost. Therefore, optimization algorithms are required to reduce the network complexity and to enhance the key parameters of 5G and beyond 5G. The benefits of optimization algorithms include reduction in interference and latency, enhanced SINR throughout the cell, spectral efficiency, throughput, coverage, and capacity [
184]. In the literature, researches have proposed optimization techniques to enhance the key parameters of 5G [
185,
186,
187]. In Ref. [
185], a genetic optimization algorithm is used to identify the location of a mobile station in a cell and the location of the new BSs in the test regions. This method will enhance the coverage capacity and data rates and decrease latency and traffic density. In Ref. [
186], a game theory based optimization algorithm is proposed to enhance energy efficiency and throughput based on link fairness. In Ref. [
187], the authors proposed an innovative scheduling algorithm to reduce the delays and call drop rates in order to enhance the QoS.
The position of an antenna, its tilt angle and height will affect the coverage of a cellular system. Therefore, optimization of antenna parameters will be considered to enhance the coverage of a network [
188,
189,
190]. In a cellular system, directional antennas with higher order sectorization are considered to enhance the coverage. The antenna tilt will improve the coverage and also SINR [
188]. In Ref. [
189], a reinforcement learning (RL) algorithm is used to improve the spectrum efficiency and the sum rate of a network, using a simplified path loss model. In Ref. [
190], a self-tuning algorithm is proposed to adjust the tilt of an antenna. Two fuzzy logic controllers are used for automatic adjustment of the antenna tilt, which will improve the SINR at the cell edge and spectrum efficiency. In Ref. [
191], artificial neural network (ANN) and stochastic learning based optimization algorithms are used to optimise the antenna tilt. Network coverage and capacity are enhanced by using the ANN algorithm. In Ref. [
192], two machine learning algorithms are proposed to reduce the interference and to enhance the coverage by optimizing the transmitted power and antenna tilt.
The random distribution of the traffic across a cell and its uncertainty in the radio propagation needs dynamic antenna tilts and optimization algorithms to enhance the coverage and capacity. Different optimization algorithms have been proposed in recent research to enhance the coverage and capacity in 5G networks, as has been summarised in
Table 16.
5G Optimization—Future Challenges and Research Issues
5G key parameters such as improved spectral efficiency, low latency, high data rates, and extended coverage capacity are achieved using the optimization algorithms. There are still some challenges and issues in the implementation of AI-based optimization algorithms that need to be addressed. The key issues that are to be explored in the future are:
ML and RL Algorithms: New ML and RL based optimization algorithms are required to enhance the 5G network coverage and capacity and also to bridge the gap between intelligent algorithms and 5G technologies;
Scheduling Algorithms: Scheduling algorithms need to be proposed for 5G networks to optimize the throughput and to enhance the capacity of cellular networks;
Optimization of the Antenna Parameter: Optimization of antenna tilt angle, antenna height, and number of antennas is required to enhance the coverage and capacity and to reduce cell edge interference, deployment cost, and complexity.
5G key parameters such as coverage capacity, latency, throughput, spectral efficiency, data rate, outage probability, interference management, and power consumption can be enhanced by using different algorithms and methods in the literature.
Table 17, lists the various key performance indicators and the articles in which they were used for performance evaluation. Statistical analysis of 5G key parameters is shown in
Figure 14.