1. Introduction

The term IoT has been considered as an expanding technique applied in various applications and functions, from smart environments and houses to personal healthcare and others [35]. It is described as a smart concept for the internet relating everything to the Internet and data organization and information exchange [36]. Large-scale IoT intelligent systems have become more efficient and effective by using the properties of “symmetry” and “asymmetry”. This can help in a range of IoT applications, for example, in water quality analytics, bee colony status monitoring, accurate agriculture, data communication balancing, smart traffic management, spatiotemporal predicting, and intelligent engineering. Several studies are currently working on IoT technologies to sustain their necessity in platforms developing technology [37]. Although there are diverse definitions and explanations for understanding IoT, it has a subsequent edge associated with the assimilation of the physical world with the virtual one of the internet [38].

The paradigm of IoT is simplified as any-time, any-place, and any-one connected [39]. The implementation of this technology makes things and people closer and everyday life easier [40]. The purpose of IoT is to ensure a connection between devices, where each provides information and data. These devices are generally personal objects that are frequently carried, including smartphones, vehicles, healthcare devices, and office connected devices [41]. Moreover, Radio-Frequency Identification (RFID) is considered to be one of the first applications that saw the light and has played a crucial role in numerous technologies, such as sensors, smart objects, and actuators [42]. However, Machine-to-Machine communication (M2M) [43] and Vehicle-to-Vehicle communication (V2V) [44] represent the actual applications showing the significant advantages of IoT [45,46].

2. IoT Applications

Figure 3 represents a complete taxonomy of IoT in the significant fields of application. The principal areas of application are focused on health care, the environment, smart cities, commercial, industrial, and infrastructural fields [85,86].

Figure 3. Taxonomy of IoT applications.

The applications and use of IoT in the different domains are what drive and explain the development of this new trend, leading to the acceptance of IoT by the new world [87]. The study of IoT applications improves the understanding and enhancement of IoT technology, and thus, the design of new systems for newly developed cases [88]. The concept of IoT can be summarized as generating daily information from an object and transferring it to another one. Therefore, enabling communication between objects makes the range of IoT applications extensive, variable, and unlimited [89,90].

Table 3. IoT healthcare applications.

Table 4. IoT environmental applications.

Table 5. IoT smart city applications.

Table 6. IoT commercial applications.

Table 7. IoT industrial applications.

Table 8. IoT infrastructural applications.

3. Future Research Challenges

For future research directions, we present some of the challenges in current 5G networks that need optimum solutions for designing 6G networks. The future research challenges for the studied topics are summarized as follows:

A. Carrier Aggregation

An optimal SS framework to allocate multiple resources efficiently among users is crucial for our future SS systems [237]. Multiple CCs across the available spectrum can be utilized to create a wider bandwidth channel to increase the network data throughput and overall capacity [238]. Moreover, an application-aware resource allocation scheme is needed for the users of HetNet to achieve fragmented spectrum allocations and aggregate licensed and unlicensed carrier spectra [239]. A Clear Channel Assessment (CCA) may be performed in response to the uplink grant to determine the availability of an unlicensed spectrum [240]. Moreover, the Licensed Assisted Access (LAA) method is the latest approach presented by 3GPP that can exploit high-spectrum bandwidth to address the limitation of the current 5G network [241]. Various Machine Learning (ML)-based resource allocation techniques can also be applied. For instance, a deep learning method can be used to overcome the resource allocation management of BS by performing fractional spectrum access proactively and selecting the channel dynamically [242]. Lack of dynamic control of wireless network resources leads to unbalanced spectrum loads and introduces capacity bottleneck. Therefore, a solution similar to extended Dynamic Spectrum Access (eDSA) is needed to provide quality load balancing in available spectrum bands, traffic allocation, and capacity enhancement through the aggregation of current resources [243]. Moreover, some AI-based solutions for resource management must be proposed; for example, Evolutionary Programming (EP) algorithm [244].

B. Cognitive Radio

CR for spectrum utilization offers the opportunity for flexible spectrum access in the current wireless systems [245]. Spectrum sensing involves the classification of a part of the spectrum or a frequency band as either “occupied” or “unoccupied” [246]. Several types of CR-based schemes are presented recently (e.g., matched filter, energy, and cyclostationary feature detection) [247]. When more accurate information about the primary user is needed, then the best-matched filter is required to perform optimal detection [248]. The cyclostationary approach can also be utilized by using cyclostationary elements of the available spectrum [249]. Moreover, the implementation of cooperative sensing must be performed in a distributed manner; that is, SUs receive information from the neighbors and make a choice on an individual basis [250]. Another way to use the free spectrum efficiently is the utilization of the dynamic genetic algorithm for PUs and SUs [251]. A new promising approach is to utilize ML techniques with CR to improve the spectral and energy efficiency of the network [252]. The handover between the PUs and SUs during resource sharing is a critical task that needs some dynamic handover schemes to achieve high QoS [253]. Moreover, various AI-based approaches are required for effective resource management in CR networks [254]. Although this requires different optimization parameters for different environments, real-time processing can be achieved by combining CR with AI into the Multi-Agent System (MAS), and real-time processing can be achieved [255].

C. Small Cell

As the use of a high-frequency band in the current 5G network increases, the utilization of small cell deployment is a mandatory approach to serve a higher number of subscribers [256]. However, the existing spectrum allocation algorithms are insufficient to deliver optimum spectrum allocation efficiency in the small cell network [257]. Therefore, an efficient algorithm, such as K-Nearest Neighbor (KNN) learning algorithm, can be used to classify all the small cells according to their geographic locations and interference radius; thus, the spectrum allocation efficiency can be improved [258]. Additionally, for a limited backhaul capacity network, some efficient spectrum allocation solution is required to provide significant performance improvement in throughput enhancement, delay reduction, and energy savings for small-cell networks [259]. The switching between small cells and Wi-Fi remains a challenging task that is being explored in the latest 3GPP releases [260]. Furthermore, in small cell-based HetNet, which consists of multilayers with a shared spectrum, a dynamic spectrum and multicell logarithmic resource allocation algorithm are required [261]. The utilization of various new approaches, such as block-aware power allocation, efficient relay selection, and cooperative caching algorithms, must also be developed to deliver the optimum results for the current 5G network [262,263]. The BS in HetNet is experiencing a seamless switching between different technologies, such as Wi-Fi and cellular. This continuous switching makes the network suffer from negative parameters, such as intercell interference, SNR, fading, and downstream power. Hence, an optimal solution for resource and power allocation using a feed-forward neural network approach can be implemented for the stability of the network [264]. Similarly, traffic offloading is a critical issue in multitier HetNet; therefore, an autonomous traffic offloading technique based on machine learning is required to reduce transmission delay [265]. Moreover, an efficient design for some new AI-based clustering approach can manage the resource framework while enhancing the efficiency and throughput of the small cell [266]. To improve coverage, AI-based optimization approaches are required, especially for software-defined networking controllers [267].

D. High-Spectrum Access

The modeling, as well as the measure of high-spectrum channels, play a vital role in guiding toward the complete knowledge of how this spectrum differs from the currently used spectrum [268]. Limited coverage is another big issue for the mmWave spectrum; therefore, detailed stochastic geometric coverage analysis studies with the realistic channel and antenna radiation models are required [269,270]. Moreover, the use of passive reflectors of different shapes and sizes can help enhance the received power, thus improving signal coverage in the NLOS region [271]. Besides, NLOS is assumed to be more important for a lower 6 GHz band rather than in mmWave communication links. On the contrary, high propagation losses and high absorption in mmWave makes the LOS inevitable [272]. However, only a few research studies have focused on designing the channel models for the NLOS scenario to deliver sufficient results [273,274]. Although existing channel models provide some insights into the propagation characteristics of mmWave in cellular environments, further research is needed to capture the shades of the propagation and fade in the mmWave scenario [275]. The utilization of clustering in narrow-beam antenna [276] and accurate estimation of departure and arrival angles, as well as the time-of-arrival for each observed radio propagation path, can be used to enhance the overall network performance [277]. Various new frequency spectra, such as 60 and 73 GHz bands, can be studied for various propagation environments and compared with the existing frequency band below 6 GHz. Different multifrequency propagation path loss models (in particular, ABG, which is CIF) can be investigated for the evaluation of future high-frequency mmWave networks [278]. Furthermore, some new self-organizing techniques based on ML are required to provide clustering and efficient spectrum allocation for the mmWave system [279]. Moreover, the beam selection for the uplink scenario requires an efficient ML mechanism to deliver a high directional beamforming effect [280]. The AI-based framework can also be used to optimize high-spectrum mmWave compressed sensing for high-speed 5G/6G image transmission [281].

E.  M-MIMO

In mmWave frequency bands, the blockage and path loss phenomena are considerably high. Nonetheless, it can be (partially) surmounted by keeping the structure of antenna array on the basic physical size as it is used in lower frequencies; this can be accomplished by M-MIMO [282]. However, M-MIMO technologies are constructed, implemented, and utilized differently [283]. The main requirements regarding stability, flexibility, and coverage must be investigated for different frequency bands, antenna geometries, and propagation environments [284]. Besides, various current precoding schemes have high-computational complexity and fail to maximize spatial information [224]. Conventional digital beamforming involves the complexity of large antenna arrays in addition to the increased cost of the system, whereas analog beamforming can handle only a single data signal at a time. Therefore, various low-cost and less complex hybrid precoding methods are required to model efficient transmitters and testbeds to mitigate jamming for MIMO-based mmWave systems [285–289]. This goal can be achieved by designing architecture with a combination of analog and digital processing that can be utilized to enable beamforming and spatial multiplexing with minimum complexity in achieving optimal performance [290,291]. An ML algorithm can be utilized to predict various channel characteristics and create a beamforming M-MIMO dataset framework [292,293]. The AI-and M-MIMO-based systems can deliver good QoS performance for high altitude users [294]. Nonetheless, the explainable AI-controlled based architecture would be useful for several current limitations while performing resource allocation, energy optimization, and minimizing interferences [295].

4. Conclusions

Various challenges have been summarized: Such as data privacy and scalability for the healthcare applications, authorization and cost issues for environmental applications, mobility and architecture challenges for smart city applications, cost and implementation difficulties for commercial applications, hardware and production problems for industrial applications, and standardization and trust issues for infrastructural applications. It has stated that various IoT applications still need to be exploited, such as blockchain technology, in order to maintain transaction information, enhance the existing structure performance, or develop next-generation systems. This can help to achieve extra safety, automatic business management, distributed platforms, offline-to-online information authentication, and so on. Moreover, the security and privacy characteristics of IoT are the key factors that can lead to its ability to be developed into a universally implemented technology in the future. However, the self-organizing and accessible nature of IoT makes it susceptible to numerous insider and outsider attackers. This may compromise the users’ security and privacy, enabling access to a user’s private data, financial damage, and eavesdropping. Therefore, more advanced optimized algorithms and protocols are required to secure data privacy. It can be concluded that by designing an energy- and cost-efficient intelligent network with potential business growth for IoT systems, the next generation of development technology can be produced.

This entry is adapted from the peer-reviewed paper 10.3390/sym12101674