Although 6G, or the 6th generation standard for wireless communication technologies, is in development, it is considered to be the key enabler for the TI. Not only will the technology provide high data rates to various applications that can be drivers for use cases of IEEE 1918.1 but it will also introduce intelligence into all the network layers. Since AI is the core functionality of some TI use cases that require the prediction of some reaction from the slave domain, 6G can be seen as the network domain of choice for the TI. In this chapter, we will explain the role of 6G as a new key enabler for the TI and its applications.

4.1. Introduction

The continuous improvement of network technology and the digitization of the economy are key drivers of 6G networks. Automated networks need paradigm changes in communication network management, where communication and computation should eventually merge. This will lead to an increase in user satisfaction, and will ultimately provide new, smaller domains for local and efficient service provisioning (e.g. private networks). Development efforts for 5G concepts, technologies and products converted the fabric of cellular networks with the introduction of eMBB, URLLC and mMTC for more dedicated services.

The novel technology areas in 6G will extend to enhanced mobile access and wireless backhauling with smart and reconfigurable metasurfaces. Here, the wireless channel will be designed to develop system performance and molecular communications. A process that could be achieved using the TI for pairing network components, applying artificial intelligence (AI) for communications/networking and for managing network functionalities and operating improved security techniques for cyberattacks. Energy, health, mobility, transportation and manufacturing are examples of novel domains that future networks will be required to facilitate and support.

However, 5G is still undergoing expansion and may not have reached its full functionality, as outlined by standards, and now is the time to outline the milestones for basic 6G concepts. 6G may become the first network of its kind that will provide for humanity in emergency situations. With 6G, decisions for providing services will be intelligently presented with the use of a number of technology and data enablers: smart transportation, behavioral analysis stored in big data and health monitoring. The blockchain-based networks will redesign the architectures for network domains through the adoption of shared, immutable records of data transactions between varying parties.

Blockchains will develop the efficiency of trust in the communication area with a more robust system of authentication, helping to eliminate the risks of cyberattacks. In addition to this, the adoption of AI to administer the network lifecycle will lead to a substantial decrease in power consumption over network segments. This will reduce the influence of mobile networking over the environment and cut emissions, resulting in greener technologies. And it will show that communication engineering is not only for advancing standards of living but also for helping to preserve the world (David et al. 2020).

The future presents many technical challenges that the current 5G standard cannot meet; those problems will be solved by the next generation, i.e. 6G. High data rate is important but even more important is providing security (Al Mousa et al. 2020). To meet the needs of services and applications planned for post 5G and 6G networks, explorations are progressing towards the integration of architectures with the aim of supporting the variety of new computation-heavy and latency-sensitive applications in the context of the Tactile Internet. 5G deployments have restrictions in terms of integration of new applications. To deal with this problem, next-generation 6G systems worked on the convergence of technologies. 6G will present new system paradigms (e.g. human-in-the-loop communications and human-centric services) (Pérez et al. 2020). The mobile communication technique aims to provide ubiquitous connections between devices such as phones, laptops and buildings for the Internet of Things. 6G should provide for needs across many aspects such as latency mobility and connection density. For this reason, the high dimensions and abilities of 6G will help in the establishment of a connectivity ecosystem in the TI that focuses on building a remote and real-time interactive system; this will provide the ability to interact with real and virtual objects using wireless techniques (Gholipoor et al. 2020a; Jia et al. 2020).

The focus of the TI on ultra-low latency communications with high availability, reliability and security is regarded as a remarkable enabler of mission-critical IoT services. Tactile IoT presents a big change from conventional data-delivery networks to technology-transfer networks. However, current research on tactile IoT focuses on the point-to-point communication link that is able to connect a single haptic device. To reach the next level, multimodel data from massive distributed or area-based IoT devices with sporadic traffic are needed. This creates a huge challenge over the design of physical layer technologies because of the existence of multiple transmitters and limited radio resources. Thus, grant-free non-orthogonal multiple access (NOMA), which exploits the joint benefit of grantless access mechanism and non-orthogonal signal superposition, has been used to simultaneously realize low latency and high-efficiency massive access in tactile IoT in Ye et al. (2019).

The TI is one of the next-generation wireless network services with end-to-end (E2E) delay as low as 1 ms. This ultra-low E2E delay cannot be provided in the current 5G network architecture, and it is also vital to consider the delay of all parts of the network, including the radio access. However, heterogeneous services with variable requirements are proposed in the next-generation wireless networks, and these services are classified as enhanced mobile broadband (eMBB), massive machine-type communications (mMTC) and ultra-reliable and low latency communications (URLLC). eMBB services need high data rate, and mMTC services require a large number of connections with a low data rate for each connection.

The TI covers numerous applications, such as remote surgery, remote monitoring, distance education and remote driving. Because of the importance of E2E delay in TI services, it is crucial to pay attention to all delays in the network to provide the E2E delay requirement. Delays in an E2E connection are queuing delays, transmission delays and network processing delays. The required radio resources in the network are power and bandwidth. They are limited and it is necessary to distribute them to users according to their service requirements, such as data rate and delay. Also, the queuing delay and the transmission delay in the radio access are vital for delay-sensitive services, and they are influenced by the resource allocation scheme. Radio resource allocation (R-RA) has an important role in guaranteeing the QoS, and the TI is highly sensitive to delay; thus, R-RA becomes vital in this case (Gholipoor et al. 2020b).

4.2. The architecture of 6G

Since the 1980s, developments within information communication technology (ICT) have tended to result in next generation revolutionary technologies emerging in 10-year cycles, as shown in Figure 4.1. The continuous improvement of information and communication technology has played a crucial role in the perpetual development of the information system and the prosperity of society. Prior to 4G, mobile communication was concentrated on people-oriented individual consumer markets. 5G has achieved remarkable technological breakthroughs with better transmission speeds, ultra-low latency, reduced power consumption and a huge number of connections. After the integration of 5G and artificial intelligence, the progress made by next-generation information technologies - i.e. big data and edge computing - has promoted industry improvement in areas such as healthcare, manufacturing and transportation. The 6G network will assure that everything will be linked deeply, intelligently and seamlessly, to adapt to the deep integration of IoT in different industries (Lu and Zheng 2020).

4.2.2. Space network

A huge number of satellites are distributed in the space network. As for the different orbital altitudes of communication satellites, they can be divided into low Earth orbit (LEO) at 500–2,000 km, medium Earth orbit (MEO) at 2,000–36,000 km and high geostationary Earth orbit (GEO) at 36,000 km. In addition, communications between satellites in high orbits and those between high orbits and UAVs can use visible light lasers. The broadcast and multicast technologies can be used to improve the network capacity and decrease traffic burden in the multi-layered satellite communication system. Various services such as emergency rescue, Earth observation and navigation are ensured by the global coverage of the space network.

4.3. 6G channel measurements and characteristics

6G wireless channels exist at multiple frequency bands and in multiple scenarios, as illustrated in Figure 4.3. Moreover, the channel sounders and characteristics for each individual channel differ greatly. A comprehensive survey of variable types of 6G channels is given by grouping them under all spectra and global coverage scenarios.

4.3.4. Satellite channel

Satellite communication has attracted significant attention in wireless communication systems. It is considered a novel solution to ensure global coverage due to its feasible services and lower cost. Satellite communication orbits are considered to be in geosynchronous orbit and non-geostationary orbit. The circular geosynchronous Earth orbit (GEO) is 35,786 km above Earth’s equator and traces the direction of Earth’s rotation. Non-geostationary orbits are classified as low Earth orbit (LEO), medium Earth orbit (MEO) and high Earth orbit (HEO). The classification depends on the distance of satellites from Earth. The applied frequency bands are the Ku (12–18 GHz), K (18–26.5 GHz), Ka (26.5–40 GHz) and V (40–75 GHz) bands. In general, the satellite communication channel is affected by weather dynamics, including rain, cloud, fog and snow. At frequency bands above 10 GHz, rain is the major source of attenuation. Moreover, the satellite communication channel presents large Doppler frequency shift, Doppler spread, frequency dependence, large coverage range, long communication distance and so on. The channel is viewed as LOS transmission, and multipath effects can be ignored because the distance is extremely long. High transmitted power and high antenna gains are required to fight against the high path loss caused by the long-distance and high-frequency bands (Wang et al. 2020).

Mobile communication standards have been improved for a new era of 5G and 6G. 6G will integrate with terrestrial mobile communications, high, medium and low Earth orbit satellite communications, and short-distance wireless communications. Moreover, 6G will combine communication, calculation, navigation, conception and intelligence. It will provide three-dimensional global coverage in space, on Earth and at sea with high-data-rate broadband communications with the use of intelligent mobility management.

The main aim of 6G is to provide seamless communication, anytime and anywhere. 6G will integrate with networks, terminals, frequencies and technologies, enabling wider innovation for ICT markets and applications. 6G will construct a universal mobile communication network. Research on 5G-based satellite communication will provide a basis for future 6G to integrate high, medium and low Earth orbit satellite communications and terrestrial communications.

In 6G, integrations cover the following six aspects:

• – standard integration, a single standard supports both satellite mobile communication and terrestrial mobile communication;
• – terminal integration, a UE has a unified identity for access. It is controlled uniformly by the network without distinguishing between the satellite network and the terrestrial network;
• – network architecture integration, composed of a unified network architecture and a control management mechanism;
• – platform integration, both satellite and terrestrial equipment use the same cloud platform architecture;
• – frequency integration, terrestrial communication and satellite communication share spectrum by spectrum sensing, coordination, sharing and interference resistance;
• – resource management integration, wireless resource management will be controlled and allocated uniformly, using terrestrial communication, satellite communication or both, for instance (Chen et al. 2020a).

As illustrated in Figure 4.5, the 6G network will have a multi-layer architecture including user equipment (UE), satellite stations, terrestrial stations and core networks (CN). The satellite stations may include GEO, MEO and LEO. In addition, the terrestrial stations may have a macrocell base station, a microcell base station and a picocell base station. In the general architecture, both the satellite stations and terrestrial stations are seen as accessing nodes to communicate with UEs under the unified control and management of the CN. Also, satellite communication is compatible with 5G, developing 5G-based technologies and reusing 5G key technologies. However, satellite communication will be integrated within 6G, propagating at high, medium and low orbits and working with terrestrial mobile communication (Chen et al. 2020a).

4.4. 6G cellular Internet of Things

According to International Data Corporation (IDC) predictions, by 2025 there will be 41.6 billion connected IoT devices, producing 79.4 zettabytes of data. Therefore, it is essential to form the sixth-generation (6G) cellular IoT network to meet the higher demands, such as wider coverage, increased capacities and ubiquitous connectivity. To provide real-time processing of mass data from terminal devices, 6G cellular IoT has to ensure accurate computation and efficient communication for a number of devices, which are considered to be two basic tasks of 6G cellular IoT. Besides, it is not unimportant to carry out the two tasks with limited wireless resources. For computation, the traditional way of transmitting and then computing is not appropriate for massive data aggregation in 6G cellular IoT because of the ultra-high latency and the low spectrum efficiency. To solve this problem, a novel solution called over-the-air computation (AirComp) is presented in Figure 4.3 that can compute the target functions including a summation structure over wireless multiple-access channels (MACs). AirComp decomposes the structure of the targeted function and then uses the superposition property of wireless MACs to achieve the sum result of pre-processing data by concurrent transmission. Eventually, the targeted function result can be achieved by post-processing the arrived signal at the base station (BS). In 6G cellular IoT, AirComp can integrate with multiple-input and multiple-output (MIMO) techniques to spatially multiplex multi-function computation and reduce computation errors by using spatial beamforming. For communication, conventional orthogonal multiple access (OMA) schemes cannot provide massive access due to the restricted radio spectrum. Non-orthogonal multiple access (NOMA) into cellular IoT is achieved to realize seamless access to a massive number of devices. Massive NOMA is exposed to co-channel interference that disrupts the quality of communication signals. Spatial beamforming is made use of to fight against co-channel interference and improve system performance. Since the BS of 6G cellular IoT will be equipped with a large-scale antenna array, there are ultra-high spatial degrees of freedom to reduce co-channel interference. To realize exact computation and efficient communication simultaneously, IoT devices should have sufficient energy. Besides, energy supply for massive IoT is an important task. Because of the high cost and environmental strain, frequent battery replacement for massive IoT is prohibitive. Thus, it is useful to adopt the wireless power transfer (WPT) technique to realize one-to-many charging by taking advantage of the open nature of the wireless broadcast channel (Qi et al. 2020).

4.5. Energy self-sustainability (ESS) in 6G

Following commercial deployment of 5G worldwide, studies of 6G mobile communication networks have increased. The major key performance indicator for 6G is its massive connectivity for small devices to enable the so-called Internet of Everything (IoE), a scale up from Internet of Things (IoT). Most of these IoE devices will be either battery-powered or battery-less, so the key challenge is how to prolong the lifetime of these IoE devices. Two aspects need to be addressed: energy efficiency to reduce the energy consumed by IoE devices, and energy self-supply to create new energy supplies for IoE devices. 6G will ensure energy self-sustainability (ESS) to massive IoE devices. 6G technologies such as THz and reflective/reconfigurable intelligent surfaces (RISs) have considerable potential to fulfill this vision of energy self-sustainability. THz frequencies are higher than mmWave and ensure better directionality in 5G, which is more efficient for wireless energy transfer (WET). RIS is applied in close proximity to end devices providing not only improved communication but also energy transfer to IoE devices, using either active or passive transmission. RIS may also be managed to provide on-demand WET (Yang et al. 2020b).

The main aim of 6G IoT networks is to ensure significantly higher data rates and extremely low latency. However, because of the scarce spectrum bands and ever-growing number of IoT devices (IoDs) deployed, 6G IoT networks face two vital problems, i.e. energy limitations and severe signal attenuation. Simultaneous wireless information and power transfer (SWIPT) and cooperative relaying are effective methods for solving both of these problems. 6G relies on massive collectivity where a large number of IoDs are connected for ubiquitous information exchange. However, most are powered by batteries with a restricted operating life. Energy harvesting (EH) is a convenient method for providing energy for 6G IoT networks. IoDs with EH technology can harvest energy from the environment, such as thermal, wind and solar, which means energy self-sustainability (ESS) can be achieved. Nevertheless, these energy sources may be unreliable because of the inherent unpredictability of the environment itself, which is a disadvantage of EH technology. Wireless power transfer (WPT) can ensure more reliable energy supply for IoDs compared to the EH obtained from radio frequency (RF) signals or from magnetic induction. Also, a new protocol is proposed for user cooperation in wireless powered networks to improve the energy efficiency. A carrier-sense multiple access with collision avoidance (CSMA/CA) protocol is proposed for WPT-enabled wireless networks to develop the throughput.

RF signals carry both energy and information; thus, simultaneous wireless information and power transfer (SWIPT) technology allows for energy harvesting and information decoding from the same received RF signal. Energy consumption of SWIPT IoT systems is reduced through two resource and task scheduling strategies. Because of the high-frequency bands used in the 6G IoT network, the transmitted signals are likely to be attenuated; thus, energy may not be harvested effectively at IoDs. Cooperative relaying can provide reliable communication and expand the coverage of IoT networks, a novel solution for solving the signal attenuation problem.

Furthermore, two relay destination selection SWIPT schemes for multi-relay cooperative IoT networks are proposed in Lu et al. (2020). Cooperative relaying-based SWIPT can effectively improve energy harvesting efficiency. Orthogonal frequency division multiplexing (OFDM) is mostly used because it can transmit signals efficiently over flexible subcarriers and power allocation. By combining OFDM and the SWIPT technology, a higher transmission rate and more efficient energy transfer in IoT networks can be achieved (Lu et al. 2020).

The main problem of the scalable deployment of the energy self-sustainability (ESS) Internet of Everything (IoE) for sixth-generation (6G) networks is balancing massive connectivity and high spectral efficiency (SE). Cell-free massive multiple-input multiple-output (CF mMIMO) is considered to be a novel solution where multiple wireless access points apply coherent signal processing to jointly serve the users. Besides, massive connectivity and high SE are hard to achieve at the same time because of the limited pilot resource. In 6G, the expectation is to obtain energy self-sustainability (ESS) Internet of Everything (IoE) networks, which handle massive connectivity and distribute large amounts of data traffic, while ensuring a more uniform quality of service (QoS) over the whole wireless network. Cellular massive multiple-input multiple-output (MIMO) is a known component of 5G networks. By inheriting several virtues from cellular massive MIMO, cell-free massive MIMO (CF mMIMO) is thought to potentially meet the needs of ESS IoE networks in 6G, strong macro-diversity and ubiquitous coverage, for instance. The main aim of CF mMIMO is for a large number of wireless access points (APs), which could be deployed in the coverage area and connected to a central processing unit (CPU), jointly serve all user equipment (UE) on the same time-frequency resource under the coordination of the CPU. Besides, an ESS network can be applied by wireless power transfer (WPT), where the ambient and dedicated radio frequency (RF) energy is harvested for the battery-limited UEs. Moreover, time-switching protocol is well known in WPT cellular massive MIMO operating in time division duplex (TDD) mode, where the transmission interval is partitioned into slots for energy harvest and information reception. The main idea behind integrating CF mMIMO with WPT is that each UE can be served by at least one AP with higher channel gain compared with cellular technology, which makes WPT applicable in the dense scenarios (Chen et al. 2020b).

4.6. IoT-integrated ultrasmart city life

The concept of a smart city is a city that improves the quality of life (QoL) of the people in it by optimizing its operations using available infrastructures to monitor, observe, examine and act by connecting the core components that run the city. With 5G, a city can be partially smart, which means major components such as healthcare, monitoring and transportation networks are individually smart. Also, there are many partially smart utilities such as electricity, water and waste. 6G will consider a holistic structure in an integrated way for a smart city. Some use cases about future city life are given below. Super smart home environment: Most of the elements will have a mastermind that will make decisions based on data fusion from a myriad of sensors embedded within them. The IoT infrastructure will be controlled by voice, gestures and other types of sensory communications. Thus, a complete overhaul in the approach to system structure will be required.

Ultra smart transport infrastructure: The automotive and transportation industries are attempting a change, partly because of the connectivity and networking capability presented by 5G systems and beyond. With 6G, the whole transport system will be affected by three factors. In-vehicle sensors and actuators will be intelligent, controlled by mastermind-like AI capabilities, leading to fully autonomous vehicles. Overcoming physical barriers will reduce most travel, and trips will be taken by driverless and fully automated vehicles. With the help of AI and extreme data rate capability, transport infrastructure will be fully autonomous where safety and security of the system will be provided by integrated intelligent sensors and actuators. A large amount of data will be required to be shared between vehicles to update live traffic and real-time hazard information on the roads and for high-definition 3D maps. Since vehicles move at very high speeds, the network requires a short round-trip time for communication. Vehicle-to-everything (V2X) technologies will not improve with 5G. Thus, its whole potential will be realized in 6G with VLC, OAM and emerging terahertz technologies.

Smart ubiquitous healthcare: The senescent population adds a massive cost to the healthcare system because of the continuation of conventional physical/manual administration and devices with limited communication and networking capabilities as well as restricted agility. The fast and anticipated developments in electronics and nano-bio sensors will improve health monitoring and management. The ubiquitous high-quality coverage of 6G will allow remote healthcare management intended to guarantee care regardless of the position of the patients. Latency and reliability should be guaranteed in order for emergency procedures and medical interventions to be given in a timely and uninterrupted manner. In addition, the improvements in soft and medical robotics, coupled with IXR, will mean remote surgery and medical intervention will be possible. Surgeons with particular expertise will be able to support and supervise robots to perform procedures from anywhere in the world as a result of vital latency and safety requirements being met. Privacy is a vital concern that should be primarily addressed by blockchain or variants of distributed ledger technology (Tariq et al. 2020).

Industry X.0: Efficient integration of robots with automation and warehouse transportation is crucial for industry growth. The concept of Industry X.0 is about improving Industry 4.0 by taking advantage of SMAC (social, mobile, analytics and cloud). Radio environments with a very complex network, composed of many robots, sensors and hardware elements is a problem. 6G will promote the Industry X.0 revolution by providing extreme latency and reliability as well as IoT and built-in AI capability (Tariq et al. 2020).

Pervasive AI: AI and its variants will be located at the core of 6G and will act as its vital enabling technology. Because of the improvements in AI techniques, deep learning and the availability of huge training data, there has been interest in using AI for the design and optimization of wireless networks. AI is expected to play an important role in a number of areas. These can be categorized into three levels, namely: in the device or end user equipment, localized network domain level and overall network level. This will convert the 6G network from a self-organizing regime to a self-sustaining one. At present, the most powerful AI technique is deep learning. It is based on a deep neural network (DNN) that relies on training in a centralized manner. 6G is proceeding towards a more distributed architecture like fog-RAN that provides billions of end-to-end communications anywhere around the world. The distributed cloud structure requires training to be conducted at the network edges. With 6G, AI should associate with game theory to provide a distributed learning mechanism where multiple AI agents can teach and learn from each other by interacting. Collective AI is a related concept that has been presented to deal with the situation where multiple AI agents want to perform the same goal based on local training with no direct communication between the agents. Furthermore, AI and hardware will be co-developed to provide better integration. Progress in formal methods will mean the devices are capable of reprogramming themselves, enabling them to reconfigure their functionalities as required for the purpose (Tariq et al. 2020): DNN and URLLC in 6G.

In sixth-generation networks, URLLC will lay the foundation for the emergence of mission-critical applications with stringent requirements in terms of end-to-end delay and reliability. Studies on URLLC are based on theoretical models and assumptions. The model-based solutions provide useful insights but cannot be implemented in practice. The objective of 6G networks is to enable ultra-reliable low-latency communications (URLLC), the foundation for allowing multiple mission-critical applications with stringent requirements on end-to-end (E2E) delay and reliability; for example, autonomous vehicles, factory automation and virtual/augmented reality (VR/AR). Also, the E2E delay is much longer than the transmission delays in the air interface. Meeting the E2E delay and reliability needs in such highly dynamic wireless networks proposes unprecedented problems in 6G networks. Model-based methods alone cannot solve the problems affecting 6G networks. To gain tractable results, some ideal assumptions and simplifications are unavoidable in model-based methods. As a result, the achieved solutions cannot satisfy the quality-of-service (QoS) requirements in real-world networks. With data-driven deep learning, it is feasible to learn a wide range of policies for wireless networks. To implement deep learning to URLLC, well-established models and theoretical formulas in communications and networking will be useful. Integrating model-based and data-driven methods is a novel approach in 6G networks. A multi-level architecture is an architecture that provides device intelligence, edge intelligence and cloud intelligence at user level, cell level and network level, respectively. Also, deep transfer learning and federated learning are accepted in this architecture.

To improve an architecture that provides deep learning for URLLC, the following features of 6G networks need to be considered. E2E QoS requirement: 5G systems are divided into multiple cascaded building blocks. Consequently, the E2E latency and reliability needs of URLLC are barely satisfied. In 6G networks, the E2E QoS requirement needs to be met by adjusting the whole network according to the stochastic service requests, queue states of buffers, workloads of servers and wireless channels. Scalable and flexible control plane: with a software-defined network, the control plane and user plane are slitted in 5G networks. To provide better scalability and flexibility in 6G networks, the network functions in the control plane can be fully centralized, partially centralized or fully distributed. So, the deep learning algorithms can be centralized or distributed depending on the network functions. Multi-level storage and computing resources: in 6G networks, storage and computing resources will be located at MUs, MEC and central cloud. The central cloud has many resources for offline training, but the communication delay among MUs and the cloud is long. With MEC, it is possible to train DNNs locally, so the response time of the network becomes shorter. By deploying computing resources at MUs, every device can decide with its local information in real time. Such a feature allows us to improve deep learning at various levels (Ye et al. 2019).

Based on these features, a wireless network is considered in Ye et al. (2019). It consists of smart MUs, MEC servers at APs and a central cloud, as given in Figure 4.6. To better show the multi-level architecture, mobility and traffic prediction for each MU, scheduler design at each AP and user association in a multi-AP network are examined.

Device intelligence at the user level: with device intelligence, MUs can make decisions based on local predicted information, such as the state of traffic and mobility. The prediction reliability is vital for decision-making; thus, the prediction error probability needs to be extremely low. Edge intelligence at the cell level: a scheduler at an AP maps the channel state information and queue state information to resource allocation between various MUs. With edge intelligence, DRL can be used to optimize the scheduler. Cloud intelligence at the network level: user association schemes rely on the large-scale channel gains from MUs to APs, as well as the packet arrival rate of every MU. With cloud intelligence, a centralized control plane uses a DNN. Thus, it approximates the optimal user association scheme, which maps the large-scale channel gains and the packet arrival rates of MUs to the user association scheme (Al Mousa et al. 2020).

4.7. AI-enabled 6G networks

The development of 6G networks will be large-scale, multi-layered, highly complex, dynamic and heterogeneous. Furthermore, 6G networks should provide seamless connectivity and guarantee diverse QoS requirements of the huge number of devices. Also, they process a vast amount of data produced from physical environments. AI techniques with powerful analysis ability, learning ability, optimizing ability and intelligent recognition ability that can be used in 6G networks to intelligently carry out performance optimization, knowledge discovery, sophisticated learning, structure organization and complicated decision-making. With AI, an AI-enabled intelligent architecture is presented in Yang et al. (2020a) for 6G networks. It is made up of four layers: the intelligent sensing layer, the data mining and analytics layer, the intelligent control layer and the smart application layer, as shown in Figure 4.7. This four-layer bottom-up architecture can serve as a bridge between the physical world and social world. The physical world is composed of general physical/virtual things, objects, resources and so on. The social world is composed of human demand, social behavior, etc. Some common AI techniques are given below. AI techniques involve multidisciplinary techniques containing machine learning (supervised learning, unsupervised learning and reinforcement learning), deep learning, optimization theory, game theory and meta-heuristics. Supervised learning: supervised learning uses a set of exclusive labeled data to form the learning model, which is divided into classification and regression subfields. Classification examines aims to assign a categorical label to each input sample that contains decision trees (DT), support vector machines (SVM) and K-nearest neighbors (KNN). Regression analysis includes support vector regression (SVR) and Gaussian process regression (DPR) algorithms, and it predicts continuous values based on the input statistical features. Unsupervised learning: the aim of unsupervised learning is to find hidden patterns as well as extract the useful features from unlabeled data. It is divided into clustering and dimension reduction. Clustering seeks to group a set of samples into variable clusters according to their similarities. It contains K-means clustering and hierarchical clustering algorithms. Dimension reduction converts a high-dimensional data space into a low-dimensional space without losing much useful information. Also, principal component analysis (PCA) and isometric mapping (ISOMAP) are two classic dimension reduction algorithms. Reinforcement learning (RL): in RL, every agent learns to map situations to actions. It makes suitable decisions on which actions to take through interacting with the environment, to maximize a long-term reward. Classic RL algorithms contain Markov decision process (MDP), Q-learning, policy learning, actor critic (AC), DRL and multi-armed bandit (MRB).

Deep learning: deep learning is an AI function. It realizes the working of the human brain in understanding the data representations and creating patterns based on artificial neural networks. It is composed of multiple layers of neurons. Also, the learning model can be supervised, semi-supervised and unsupervised. Classic deep learning algorithms contain a deep neural network (DNN), convolutional neural network (CNN), recurrent neural network (RNN) and long short-term memory (LSTM).

Intelligent sensing layer: sensing and detection are the primitive tasks in 6G networks. 6G networks aim to intelligently sense and detect the data from physical environments through many devices such as cameras, sensors, vehicles, drones and smartphones or crowds of people. AI-enabled sensing and detecting can intelligently gather huge amounts of dynamic, diverse and scalable data by directly interfacing the physical environment, including radio frequency utilization identification, environment monitoring, spectrum sensing, intrusion detection, interference detection and so on.

Data mining and analytics layer: a core task that aims to process and analyze the huge amounts of raw data produced from the massive number of devices in 6G networks and provide semantic derivation and knowledge discovery. The data gathered from physical environments may be heterogeneous, nonlinear and high dimensional. Thus, data mining and analytics can be applied in 6G networks to solve the problems of processing the huge amount of data, as well as to examine the data collected towards knowledge discovery. Intelligent control layer: the intelligent control layer is composed of learning, optimization and decision-making. This layer uses the suitable knowledge from lower layers to provide multiple agents such as devices and BSs to smartly learn, optimize and choose the most suitable actions (e.g. power control, spectrum access, routing management and network association), with dual functions to encourage diverse services for social networks. This function is realized by applying AI techniques in 6G networks, where every agent is equipped with an intelligent brain (learning model) to automatically learn to make decisions by itself.

Smart application layer: the aims of this layer are to deliver application-specific services to people according to their requirements and also to evaluate the provisioned services before feedbacking the evaluation results to the intelligence process. Intelligent programming and management can be provided by the impetus of AI to ensure more high-level smart applications, such as automated services, smart city, smart industry, smart transportation, smart grid and smart health, and handle global management relevant to all smart-type applications. All the activities of smart devices, terminals and infrastructures in 6G networks are administered by the smart application layer with the AI techniques to realize network self-organization ability (Yang et al. 2020a).

In particular, 6G will improve around ubiquitous AI (artificial intelligence), a hyper-flexible architecture that enables human-like intelligence for every aspect of networking systems. Edge intelligence is one of the important missing components in 5G. It is well known as a vital enabler for 6G to release the full potential of “edge-native AI” and bring network intelligentization. The influence of 6G in data collection, processing, transportation, learning and service delivery will design the progress of network intelligentization, catalyzing the maturity of edge intelligence. Next-generation edge-native AI technologies, in particular should provide the requirements below that are raised by 6G.

Resource-efficient AI: conventional wireless networks concentrate on augmenting the data transportation capability of wireless resources, spectrum and networking infrastructure, for instance. Besides, with more computationally intensive and data-driven AI tasks being applied by 6G, the extra resources needed to perform AI-based processes include data coordination, computing, model training, caching and so on. They should be meticulously evaluated, quantified and optimized. Although there are some communication-efficient AI algorithms such as transfer learning, deep reinforcement learning and federated learning, which demonstrate decreased communication overhead, these algorithms may require a considerable amount of resources compared to most data-centric applications. In addition, these algorithms may be applied to some specific learning tasks. Data-efficient AI: compared to computer vision systems, it is often hard to gather adequate high-quality labeled datasets under every possible wireless environment and networking setup. For this reason, it is of vital importance to shape data-efficient self-learning approaches, which need restricted or no hand-labeled data as input. Moreover, cloud data centers jointly provide the same set of computational tasks. Distributed AI has attracted attention because of the latest popularity of federated learning and its extension-based solutions. Besides, both distributed AI and federated learning continue to evolve. The federated-learning-enabled architecture will play a vital role in the future improvement of distributed AI-based 6G services and applications.

Personalized AI: personalized AI will play an important role in 6G to develop the decision-making abilities of AI algorithms, to assist machines to better understand human users’ preferences and make better human-preferred decisions. Also, there are two types of human-in-the-loop AI approaches that include human intelligence as part of the decision-making process.

Human-oriented performance metrics: instead of concentrating on increasing traditional performance metrics, such as throughput, network capacity and convergence rate, the performance of 6G and AI should be jointly measured and considered by taking into consideration characteristics and potential responses of users. Furthermore, with 6G and mobile services becoming essential within human society, it is vital to improve new metrics, which can assist in providing the social and economic dimensions of 6G and AI convergence (Xiao et al. 2020).

4.8. AI- and ML-based security management in super IoT

5G cellular networks presented a new usage scenario to support massive IoT, i.e. mMTC. In terms of 6G, super IoT has been introduced, which can be enhanced with symbiotic radio and satellite-assisted IoT communications to support a massive number of connected IoT devices and provide extensive coverage. Eventually, more effective energy management mechanisms are expected to support the large scale of IoT systems which can then be used for long periods of time. Moreover, privacy and security issues will arise, especially for IoT systems gathering individual or sensitive information. As illustrated in Figure 4.8, AI and ML techniques are expected to assist 6G networks in making more optimized and adaptive data-driven decisions, in solving communication problems and in meeting the needs of emerging services.

The vast amount of IoT devices and data cause significant problems in terms of privacy preservation and security guarantees. To preserve super IoT systems from various types of threats and attacks, authentication, access control and attack detection are of paramount importance. Conventional privacy and security technologies are barely feasible for super IoT because of the heterogeneity of resources, volume of networks, restricted energy, storage of devices and so on.

By supplying embedded intelligence in IoT devices and systems, AI/ML-based security technologies are used to deal with these security problems. There are many existing AI/ML-based solutions to address authentication, access control and attack detection in super IoT systems. Authentication and access control can assist IoT devices to notice identity-based attacks and hinder unauthorized devices from accessing authorized systems. To develop authentication accuracy, various AI/ML-based approaches can be well implemented based on various scenarios and assumptions. AI/ML technologies can be implemented to examine and notice various kinds of attacks including jamming, spoofing, denial of service (DoS) or distributed DoS (DDoS) attacks, eavesdropping and malware attacks, for instance. The supervised learning includes SVM, KNN, random forest (RF) and DNN, and it can be presented to distinguish these attacks by constructing classification and regression models. Furthermore, unsupervised learning can examine unlabeled data to partition them into several groups; for example, multivariate correlation analysis can assist in detecting DoS and DDoS attacks. Also, RL algorithms have been implemented to assist IoT devices in making decisions on the selection of security protocols against attacks. The applicable algorithms involve Q-learning, DQN, Dyna-Q and so on (Du et al. 2020).

4.9. Security for 6G

Data privacy and confidentiality will be one of the most significant problems for future 6G networks because of the rising network threats. In IoT networks, threats are composed of computer viruses, DDoS attacks and eavesdroppers. These threats threaten message safety and also copromise Quality of Service (QoS). To protect message safety, various policies have been presented that are focused on the physical layer and the link layer. The link-layer security protection of IoT networks is performed by the data authentication and encryption process. Available encryption and authentication methods have parameter configurations with the key lengths and the adopted algorithm, which result in varying levels of security protection and energy consumption. Available IoT chips support multiple security specifications. The main idea is to overcome the encryption and authentication configuration in the chip initialization process that facilitates the network configuration and fits the resource-constrained IoT chips. Besides, this security configuration may not ensure the service requirements of 6G networks for network QoS, energy efficiency and message safety. The network threats are mostly dynamic which may be far beyond the ensured low-level protection. The security protection causes extra energy consumption. A fixed high-level security configuration may soon run out of batteries, leading to the termination of service provision. It means that the fixed security configuration has the restriction of low energy efficiency. Taking into account the adopted energy harvesting techniques and service requirements in future 6G networks, the proper security configuration needs to be applied to the energy and network threats that can improve the security protection and network performance. Moreover, because of the network threats, the main strategy is to develop the security protection according to the energy. Also, when the harvesting source does not provide enough power, or the IoT devices have a huge sensing workload, it can consider the minimum level of required security protection to maximize working time. When the harvesting power is large, the security protection may be developed for better message safety. Another challenge is the sacrifice of network QoS caused by the extra communication overhead and energy consumption for security protection which is usually neglected. In the 6G IoT, network QoS is vital; thus, the sacrifice due to the security should be decreased (Mao et al. 2020).

4.10. The WEAF Mnecosystem (water, earth, air, fire micro/nanoecosystem) with 6G and Tactile Internet

Self-adapting capabilities, which 6G will push to the edge, urge for a pronounced separation between hardware (HW) and software (SW), this being in full contrast to the consolidated HW-SW co-design philosophy followed up to now. Algorithms will estimate the physical resources they can rely on in order to run in an optimal way. The HW will be required to achieve more symmetry with respect to the SW, in terms of flexibility, function reconfigurability and self-adaptivity/self-evolution. It envisions a prime role for micro/nanotechnologies, embodying materials, electronics and micro/nanosystems, in the scenario of 6G and of the Tactile Internet in Iannacci (2021). The WEAF Mnecosystem is designed using the analogy with the four classical elements in nature. Earth and air represent the classical concepts of HW and SW, respectively. Also, water is the new formulation of HW that, like water, is liquid in terms of functional characteristics. Furthermore, it achieves some features typical of the SW (i.e. air). Fire is the HW devoted to harvest, store and transfer energy, making it available everywhere at the network edge, i.e. when needed, where needed.

There are four identifiable application classes. They are labeled as verticals as they address top-down applications, i.e. from specifications to their physical (HW/SW) implementation. This plot is complemented by horizontals which are research activities, services and methodologies transversal with respect to verticals, yet functional to them and driven by their requirements. The visual representation of horizontals and verticals driven by MEMS within the 6G and TI scenarios is illustrated in Figure 4.9.

Verticals are grouped as follows, including but not restricted to the mentioned (MEMS) devices for 6G and TI applications:

• – Telecommunications: highly reconfigurable and tunable broadband radio frequency (RF) passive components based on RF-MEMS/-NEMS technologies, like low-loss/high-isolation switches, switching matrices, as well as complex multi-state phase shifters, step attenuators, filters, resonators, impedance tuners, etc., for mmWaves (60 to 120 GHz) and THz (above 150 to 300 GHz);
• – Energy conversion and storage: wideband microsystem-based energy harvesting (EH-MEMS) devices can transform environmental energy from mechanical vibrations, focusing at the IW (Indication Weights) range, examining various conversion mechanisms (piezoelectric, electrostatic, electromagnetic);
• – Sensors and actuators: some indicative examples can be sensors (e.g. inertial, proximity, pressure) for gesture recognition and control of environments (domestic and industrial), sensors for BCI (Brain-Computer Interaction) and sensors for healthcare, entertainment and daily living support;

• - Technologies for AI: with the rising trend for massive availability of small data, AI will be locally distributed down to the smaller pieces of HW (e.g. at sensor/component level), i.e. at the network edge.

Horizontals are introduced as follows, and they fall across vertical application macro-domains:

• – Packaging: development/extension of packaging and encapsulation technologies/methodologies for the physical devices spilling out from the above verticals decreasing the impact on their characteristics;
• – Heterogeneous integration: development of integration solutions empowering the realization of hybrid HW sub-systems, for example, multi-source EH platforms (vibration, RF, thermal) interfaced to power management electronics and sensors/transceivers to be powered (Iannacci 2021).