Artificial Intelligence-Based Internet of Things for Industry 5.0

Abstract

In the Industry 5.0 paradigm, systems based on artificial intelligence are an important component of the Internet of Things. Industry 5.0 demonstrated the important link between intelligent systems and people in most applications through precision manufacturing automation and critical thinking. In addition, Industry 5.0 brings with it several valid tools that help organizations operate cheaply and change immediately without capital investment. In recent years, smart devices, wireless communication, and sensor nodes have advanced greatly, transforming Internet of Things (IoT) ecosystems. With IoT devices, users can receive information even in rural areas and generate unbounded reports. Also, as previously mentioned, they meticulously guide people with intelligent judgments through communication technology. Many connected devices collect significant amounts of detected raw data when they require pre-processing. Although, it hardly becomes valuable for IoT devices and sufficient resources require Edge computing. AI-based algorithms are essential tools for data inference in Edge computing. In addition, observed data collected by IoT applications is usually unstructured and needs further analysis, where AI-based models help extract relevant information. Furthermore, malicious files are possible when data are transferred from one device to another. Therefore, this chapter looks at Industry 5.0, IoT architecture, and AI-based IoT; we analyze the IoT network and its specifications; communication is possible thanks to technologies.

1. Introduction

Nowadays, IoT devices, wireless communication, mobile computing, smart sensors, and communication protocols are the catchline in industrial academia. Generally, the Internet of Things works by implanting portable short-range transceivers into eclectic devices and everyday objects, enabling new methods of communication between people and objects and objects. Therefore, the Internet of Things would bring a new dimension to information and communication. IoT devices are networked devices with intelligent communication mechanisms. According to a recent study, the number of IoT devices, for example, embedded devices, sensors, laptops, and smart devices will exceed 60 billion in 2025 [1]. Normally, the competence and development of the Internet of Things is very similar to the current society, where people and devices are practically integrated into information systems through wireless sensor technology. The integration and primary purpose of IoT is to share information that enables intelligent environments to recognize and recoup. Nowadays, the IoT platform offers curb and observes services for new equipment to improve their work productivity.

The Internet of Things (IoT) was expounded and utilized by famous scientist Kevin Ashton in the early 2000s. Kevin Ashton and According to explanations, the Internet of Things is a formation of material things from the actual world that connect to the Internet across smart sensors. Ashton also developed RFID technology, which was widely used in non-human transportation tracking services. For example, IEEE and IoT are a framework for new devices joined through the Internet and enable communication in the real and fake world of M2M communication [2]. According to the definition of Internet Architecture Boards (IAB), the Internet of Things is the networking of intelligent objects, which means that many devices communicate sensibly with the presence of Internet Protocol, which is not directly used by people, but which are part of IoT objects, for example, buildings, vehicles, or environment [3].

As aforementioned, many IoT systems are enhancing progressively dynamic, complex, and adaptable, and therefore, organizing such an IoT model is difficult. Additionally, such IoT models and assistance need to improve efficiency. Artificial intelligence (AI) has recently made enormous progress in several fields using changes in computing technology [4]. Machine learning (ML) is another unique AI technology and subset applied to IoT for better services. Both AI and ML considered critical components for IoT and enable intelligent network management and operations. The IoT domain can also profit from AI and ML support. The use of AI and ML-based models in the Internet of Things offers enormous potential for deep analytical and profound advances in well-organized smart real-world devices [3, 5].

Before learning about IoT technical research trends, everyone should look at and understand how IoT works and affects our daily lives. Here, IoT device integration standards represent advanced challenges, while authentic connectivity to everything. IoT device integration challenges are considered important IoT issues because they are essential for the further development of IoT projects [5, 6, 7]. Today, many standard managers, organizations, researchers, and industries are making efforts to expand, modernize, and put the IoT in the right direction. However, there is a lack of a broad context with unified ethics within a single Internet of Things.

2. Industry 5.0 paradigm

When it comes to the twenty-first century, most fields are going digital. At the same time, we recognize that companies have difficulties with the digitization of their business when they combine artificial intelligence, IoT, and 4.0 technologies. In addition to the mentioned technologies, the next step of the industrial revolution in the coming days seems to be called Industry 5.0 [2, 3, 4, 8, 9, 10, 11, 12, 13, 14]. The term Industry 5.0 has been known since the beginning of 2015; however, it was called the fifth industrial revolution, which created a huge impact in various fields, especially in daily business activities, due to increased industrial and technological improvement and the changing integration of human processes [15, 16, 17, 18].

Industry 5.0 places a strong emphasis on human-machine collaboration, which implies that the fifth industrial revolution will be even more fascinating as human-machine interaction enables the introduction of human-machine interfaces. Elevating Industry 4.0 to a new level is the primary objective of Industry 5.0. It introduces the idea of collaborative robots, or cobots, for this. Cobots can satisfy the demands of businesses that generate separate items as well as those of today with successful integration [19, 20, 21, 22, 23, 24]. Industry 5.0, distinguished from other similar sectors in industry and medicine, is characterized by advanced manufacturing, software tools, the Internet of Things, and robotics that leverage technological advancements. It offers the customer the opportunity to experience massive customization in different groups and cooperation around the world. Technological innovations are not considered the basis of organizational revolution, and customer goals are needed. To meet the customer’s goals, Industry 5.0 follows several principles: (1) mass customization – offers real price and convenience to customize different products or services for customers; (2) customer orientation – focus on customer goals and try to solve obstacles to business expansion transformation; (3) green computing – emphasis also on environmental conditions; (4) cyber-physical systems – makes intelligent systems of people serving customers, receiving maximum benefits from people with machine intelligence [16, 17, 18, 20, 21, 22, 23, 24].

Visualization tools play a crucial role in the creation of management and customization policies for authentic products and product descriptions. Microsensor Data Interoperability – sensor nodes are now widely used in distributed intelligence systems, autonomous robots, and smart houses, among other applications. These perceptive sensor nodes identify and gather actual unprocessed data, which will be a vital resource for the upcoming industrial revolution. Nonetheless, there are still unanswered research concerns about energy optimization, quick and easy customization, selecting a local agent for data processing, and developing highly modeled distributed intelligence in Industry 4.0 and the Internet of Things Virtual Reality with Digital Twins. Thanks to the continued growth of big data and artificial intelligence cobots, even more realistic digital twins can be created. This enables industry experts to reduce waste in the process flow and system design. Therefore, with advanced imaging techniques, the digital duplicate significantly increases the performance of all sectors. Customers and sales orders with production orders and extra material will find it easier with real-time tracking devices, which enhance real-time production monitoring. Virtual training can be helpful when a trainer or trainee is in separate places but needs to learn a specific job in a simulated or virtual environment. For both parties, this kind of training drastically cuts expenses and time. Sensible self-driving systems – the autonomous leading production lines of the manufacturing sector can benefit greatly from the application of AI models. State-of-the-art AI-enabled ML and DL models successfully change intelligent systems and solutions that aid decision-making. Table 1 depicts the transition from Industry 1.0 to Industry 5.0 along with other details.

Phase	Period	Description	Identification by	Key point
Industry 1.0	1780	Industrial manufacture based on stream and water machines	Mechanization; Water and stream	First mechanicalloom
Industry 2.0	1870	Mass production with electrical energy	Electrification Division of labor Mass production	First assembly line
Industry 3.0	1970	Automation with electronic and IT system	Automation; Electronics; IT systems	The first programmable logic controller
Industry 4.0	2011	The connected device, data analytics, and computerized machinery programs to automate the industry production	Globalization; Digitalization; IoT, Robotics, Big data, Cloud computing	Cyber-physical systems
Industry 5.0*	Future	Cooperation among human intelligence with a machine to improve products and services	Personalization Robotics and AI Sustainability	Human-robot coworking Bio-economy

Table 1.

From Industry 1.0 to Industry 5.0.

*

Top enable technologies for Industry 5.0: Additive manufacturing, multi-agent systems, Smart manufacturing, Digital Eco-system, Collaborative Robotics, Internet of Everything, Mixed reality, Industrial Blockchain, Drones, 5G and Beyond.

3. Issues and constraints with industry 5.0

Industry 5.0 solves most of the production problems associated with removing people from various operations. However, it must include new preview skills, as people can add innovative manufacturing skills in the coming days. There are many capabilities under development, and some of these are listed as follows:

1. Finding out how an autonomous system may integrate ethical standards is vital before adding advanced abilities to industrial management.

2. Ensuring and validating ethical behavior inside the autonomous system concept is crucial.

3. The phenomena of overproduction can be influenced by application function transparency as well as quick and skilled production.

4. Ultimately, there should be intelligible answers for moral behavior in self-governing systems. Experts in a particular field encounter difficulties with implementation and adaptation.

5. By fine-tuning and validating, significant issues between technology, specialists, investors, society, and businesses are avoided.

3.1 Elements of IoT

As we indicated in the opening, knowing the IoT’s fundamental components helps to visualize and comprehend the concept of IoT more fully than just its functionality [23].

Figure 1 lists the six primary components of the Internet of Things.

Figure 1.

Internet of Things Elements.

3.1.1 Recognition

The idea of identification is crucial to any network used for data transfer or communication. In an IoT architecture, precise identification is essential for naming and associating services with their respective demands. In an IoT system, it can be challenging to find the address object ID and the associated IP address. The address provides the item or device’s present address within the network domain, while the ID provides the name and specifics of the object. Because identification patterns are not unique and objects might utilize public IP addresses on the network, the differentiation of object identification becomes authoritative. As a result, the suggested models need to get over the aforementioned challenges and accurately identify every item in the network.

3.1.2 Sensing

An IoT installation’s goal is to gather data from a certain region or area that is arranged according to sensors. Sensors, mobile devices/sensors, and actuators used primarily for sensing are examples of things or equipment that capture actual data from the surrounding atmosphere and send it back to a database or cloud for additional processing. Single-board computers (SBC) like the Arduino Yun and Raspberry PI, for instance, readily make use of the understanding of the Internet of Things devices when paired with sensors and inbuilt TCP/IP and security features. Usually, these gadgets are linked to a centralized management platform that offers crucial client data.

3.1.3 Interaction

The majority of IoT objects typically have enough resources; however because of these resource constraints, enterprise-level objects link to a variety of heterogeneous devices and objects that have noisy, lossy meanings. IEEE, Bluetooth, NFC, RFID, and Wi-Fi standards for specific IoT communications.

3.1.4 Computing

The Internet of Things also faces significant challenges with hardware computing power. The brains of a given device are represented by computer components like microprocessors, microcontrollers, and software-oriented devices. Hardware platforms made for the Internet of Things devices include Arduino, Rasberry PI, UDOO, MULLE, and Gadgeteer. Some additional platforms are as follows: Real-Time-Software – Operating-Systems (RTSOS) – for real-time operation of IoT, Tiny-OS – light for eight operations, and cloud platforms – too much data processing in real-time. Nonetheless, there are defects in some computer components, and research is being done to improve their functionality.

3.1.5 Services

IoT provides a multitude of services. The majority of them are categorized as identity-based services, which includes the majority of real-time devices. Data services: gather unprocessed sensor data in real-time and use it to link the relevant two Internet of Things applications. Ubiquitous services aim to depict collaborative systems that may operate whenever and wherever clients need them. Collaboration-aware services use collected data for data analysis to make judgments. Nevertheless, the previously indicated comfort levels or services were not met; in addition to the difficulties, several issues need to be resolved.

3.1.6 Semantics

The semantic operation of the Internet of Things gathers useful abstract data from many items in an intelligent manner. This is comparable to learning new information, such as locating resources that enhance model performance. Among the well-known semantic technologies utilized in Internet of Things systems are Resource Description Framework (RDF), Web Ontology Language (OWL), Efficient XML Interchange (EXI), and Wide Web Consortium (W3C).

4. Internet of things architecture

Organizations can enhance their work performance or produce better products by integrating IoT and its variations into many industries. These suggestions, however, are serious and challenging to put into practice in the actual world due to the vast differences in device numbers, protocols, and operating environments.

In other words, the problem of creating a consistent architecture for IoT inevitably comes here. Before designing the IoT architecture, it is better to understand the factors that affect IoT behavior, which will facilitate the development of reliable IoT solutions. It also reduces the different resources used to design IoT. It is critical to comprehend the meaning of this concept before delving into the mysteries that offer the distinct framework for this creativity [23, 24, 25, 26, 27, 28]. The IoT architecture is essentially a collection of tools within a sizable core network. Based on communication, networking, detection, and data processing, it was measured as a global network configuration gathered from multiple connected devices. Refer to Figure 2.

Figure 2.

Internet of Things Architecture.

4.1 Perception layer

The Internet of Things (IoT) is a form of worldwide physical networked system that allows objects to be remotely connected and measured. Similar to a bridge connecting the physical and digital worlds, the perception layer is regarded as the foundational stage of IoT systems. In certain instances, it was determined to be a sensor layer. Smart sensors, tags, and actuators are among the smart wireless devices that occupy the majority of the perception layer. The wireless systems are equipped with tags or sensors that are now required to detect and share data among various devices. The methods and sizes of equipment can differ, ranging from tiny cars to transitory vehicles. Sensors: Gather data about the surrounding environment, transform it into an electrical signal, and send it to Internet of Things models. Actuator: Transforms electrical impulses that an Internet of Things system gathers into physical actions.

4.2 Connectivity layer

It manages all device, system, and cloud center communications to create the ideal IoT scheme, which is regarded as the second stage of the IoT scheme. The physical layers’ communication connectivity to cloud centers can be established by software/hardware modules or TCP/UDP. Wi-Fi is a widely used wireless connectivity used for residential IoT setups; Ethernet is used to link fixed IoT devices; NFC is used to transmit data between two devices; and Bluetooth is used to transfer small-size data, not appropriate for huge data files. Message-oriented protocols are used by IoT in certain special cases, depending on the application’s need for data communication. Message Queue Telemetry Transport (MQTT), Data Distribution Service (DDS), Constrained Application Protocol (CoAP), and Advanced Messaging Queuing Protocol (AMQP) are a few of them.

4.3 Edge computing layer

For IoT systems to support the increasing number of connected devices and in-person services, edge computing is essential. To store and preprocess observed data as soon as it is identified and close to the source, edge computing designs are used. Thus, it can help the Internet of Things devices save time and resources. It also lowers system latency, which enhances performance accuracy.

4.4 Layer of processing

Every piece of data from IoT systems is gathered by the processing layer. Preprocessing models are used to supply data for subsequent processes or to employ abstract information for decision-making. Real-time data is a central location for event- and query-based data collection, and it is monitored through APIs and utilized for non-real-time applications. As a result of data abstraction techniques and multidimensional data collection from many devices, only other linked devices can currently understand the data.

4.5 Application layer

Data analysis was performed using software packages to provide meaningful answers to key business questions/requirements at the application. In IoT, hundreds of IoT queries vary in complexity and functionality using different expertise stacks as functional models. Today, various applications are built on top of IoT steps that can recommend application-related preferences with extraordinary tools for data mining, patterns, and additional analysis capabilities.

4.6 Business layer

IoT systems gather and preprocess data, which then aids in issue resolution and decision-making with superior outcomes. The business layer is distinguished as a distinct stage that is sophisticated and challenging to explain in a single application layer.

4.7 Layer of security

Any application about networks can appropriately use the phrase security. The security layer in the Internet of Things comprises all of the aforementioned services and layers. IoT security issues are difficult to discuss in a single paragraph or section. IoT models have different levels of security:

1. Device Security - IoT-related devices require low-resource authentication services, physical metal shields, and chips that can run unauthorized code. The primary method of connection security is the transmission of Internet of Things data via wireless networks, which facilitates data theft and modification by adversaries. As a result, data must be encrypted before being transferred across a network or device.

2. Cloud Security - detected data stored in the cloud must be encrypted with code to prevent sensitive data from being exposed to intruders. So, always pay attention to security protocols to ensure high security at all stages, from the smallest device to a versatile analysis system.

5. Technologies that facilitate

5.1 Radiofrequency identification (RFID)

RFID technology is specifically developed to track transportation using tags and scanners. RFID is thought to be an automatic recognition system that unintentionally detects a target identifying signal that contains useful data. As a result, it has been extensively utilized in a variety of passive and hazardous environments. As previously stated, readers and tags complete the RFID system. The address strips on the tag are affixed to various things, like a tiny microchip that is processed by the antenna. Data recordings are transmitted and collected using an electromagnetic pitch for each unit identification.

Professional equipment uses RFID tags because they can automatically manage warehouse operations, track products and their life cycles, track payments, track shipments or deliveries, and update data without the need for human or third-party interaction. RFID technologies can enhance the usability and efficiency of models and systems and integrate them into a variety of design fields. RFID deployment does have drawbacks, though, since the majority of IoT WSN devices are designed to withstand hostile settings where they may be subject to signal jamming, eavesdropping, or even collapse.

5.2 Power-line-communication (PLC)

When data are transmitted via a PLC, it is further stored via associated wires. This indicates that the data records are modulated by the transmitter and then sent to the transmission medium. Upon arrival, the data records are demodulated and read by the receivers. In this way, the data are transferred via the power connections, enabling both data reception and response time from its half-duplex manner, as well as turning on. Thus, intelligent meters (AMI), HEMS, BEMS, and solar-intensive maintenance systems that comprehend the intelligent society are drawn to the PLC communication paradigm. PLCs come in two varieties: slow and rapid, each with its communication protocol.

5.3 EPC- electronic product code

RFID tags are identified by their EPC, which is a type string 96-bit embedded in a tag or chip. Of these 96 bits, 8 bits are a header used to indicate the protocol version, 28 bits are the system’s unique address that controls the label data, and 24 bits are the product kind. The 36-bit identifier’s serial number is mentioned to be identified; the organization that developed the tag is contained in the final 2 bits.

5.4 Actuator

Actuators respond strongly to motion and are appropriate for use with specialized equipment. It generates a variety of movements, including linear, spherical, oscillatory, and rotating movements; it then uses these movements as kinetic energy to generate forces. Pneumatic actuators employ compressed air; hydraulic actuators use hydraulic fluids; and electric actuators are used in engines.

5.5 Machine-to-machine (M2M)

Similar to LANs and WANs, M2M communication involves devices gathering data from several sources and returning it to other devices connected to the network. Depending on the application, the M2M recordings are automatically monitored and carry out roughly predetermined activities. Furthermore, software-controlled communication between machines and devices is essential to the M2M performance.

5.6 Wireless fidelity (Wi-fi)

A key component of wireless networking, Wi-Fi is ideal for IoT-based applications that require a lot of data. With a sophisticated transport mechanism, it offers strong wireless connectivity within a compact space. There are multiple collective versions of it: IEEE 802.11a offers 54 Mbps of data speed and boosts that data rate to 2.4 GHz.

5.7 IEEE 802.15.4

IEEE 802.15.4, also known as Low-Rate Wireless Personal Area Networks, or LRWPAN, functions as the MAC layer’s sublayer. Simultaneously supporting sensor nodes, it offers high security, cheap cost, and efficient communication for weak communication. The IEEE 802.15.4 standard is thought to provide the foundation for several communication technologies, including Bluetooth, Z-Wave, ZigBee, and others, based on these requirements. It is also an interesting topic to investigate; however it does not provide QoS.

5.8 Z-wave

Originally, door switches and other smart home automation devices were connected to a central controller via Z-wave communication technology. Z-wave and ZigBee operate similarly in that they both leverage mesh topology and low wireless standards to enhance low-resource applications. However, ZigBee uses 2.4 GHz, and Z-Wave uses the.868 MHz band. Additionally, ZigBee uses hardware-side 128-bit AES encryption, whereas Z-Wave abandoned software-side encryption.

5.9 Bluetooth LE

The term Bluetooth, also known as IEEE 802.15.1, describes the use of industrial, scientific, and medical (ISM) frequencies to transfer data over short distances between stationary and mobile devices. Numerous industries, including healthcare, security, sports, military gear, smart cities, and homes, have adopted it. The most recent iterations of Bluetooth SIG, Bluetooth BLE, Bluetooth 4.0, and Bluetooth 5.0 gather and aggregate data observed by the Internet of Things (IoT)-based sensor nodes. Bluetooth technology proved to be highly appropriate for proximity surveillance gadgets.

6. Artificial intelligence (AI) of the internet of things (IoT)

The Internet of Computers (IoC) and the Internet of Things (IoT) continuously provide the operational functions of the Internet. The importance of artificial intelligence technologies must be considered when enabling intelligent online communication. Wireless sensor networks are becoming a hot research topic today due to their real-world applications and incredible remote monitoring of events such as healthcare, weather forecasting, sea level, and event forecasting. Furthermore, smart sensors are increasingly utilized in mobile devices, smart cities, and electronics-based household goods [22, 23, 24, 25, 29, 30, 31, 32, 33, 34, 35, 36].

The ubiquitous existence of numerous items or things around us, including mobile phones, RFID tags, sensors, actuators, and other devices, is the concept behind the Internet of Things (IoT). Unique address systems allow them to communicate with each other and cooperate with their neighbors [36]. Therefore, the contours of the IoT are constantly changing the lives of people around the world.

The Internet of Intelligent Things is one example of such a future system (IIoT). These networking advancements are what led to the creation of sophisticated, pervasive, instantaneous, and intelligent Internet connections. It appears that known items should be endowed with the capacity to comprehend their context and make independent inferences [36, 37, 38, 39, 40, 41, 42, 43, 44, 45].

Currently, decisions or conclusions do not need to be forwarded to central decision nodes. Thanks to the high intelligence and skill turn of the sensors, thanks to the stimulus recognized by the sensors, allows the IIoT to be more responsive to time-critical conditions because the inferences are made in a decentralized way.

6.1 Intelligent sensing through artificial intelligence

Smart/clever sensor ML models that find useful patterns or predictions based on data gathered by smart sensors are the foundation of artificial intelligence (AI). For instance, the number of classes the model can identify is constantly increased via active sensor learning. Since data are gathered in real time, rigorous questioning or episodic retraining must be the standard method of data gathering [32, 33, 34, 35, 36]. Active serving systems that aim to classify events without prior knowledge of a wide variety of environmental noises are best suited for unsupervised models.

6.2 IoT decision tree

Using sorting strategies based on related features, DT resolves classification issues. Access to information and the Gini index are two of the numerous techniques used in DT to choose the most alluring characteristic that most closely resembles the training samples. The following describes the entire DT process:

1. 1st – before and after pruning used to reduce the size of the tree.

2. 2nd – requested status from among regulated goods.

3. 3rd – the optimized search model used for deletion of additional functions.

4. 4th – the resulting tree structure, which is transformed into an appropriate data structure, such as a rule.

IoT-based real-time applications like pattern recognition, decision-making, environmental monitoring, security parameter detection, healthcare management, etc. have been successfully deployed by DT.

6.3 Random forest in IoT

Among the class of supervised learning models is Random Forest (RF). RF is made up of numerous randomly generated trees that can select superior classes through voting. The ultimate classification outcome is determined by choosing the category that received the most votes. The voting subset criteria are often constructed by RF using decision trees, and the result score is the mean of the DT results. Furthermore, the accuracy of the RF calculation wins for feature selection because it requires the fewest input parameters, but is not applicable for real-world applications. RF models have been very suitable for IoT devices in various fields. For example, RF models that recognize features derived from network traffic will correctly identify classes of IoT devices because RF has a precise real impact on the classification of unapproved IoT gadgets.

6.4 Clustering

K-means - K-means’ primary goal is to classify unlabeled data characteristics into sets or K-clusters; in this case, the data points that match the same cluster need to share certain similarities. K-means is often a quick and very scalable machine learning method. In some cases, MapReduce was used to analyze several smaller datasets and then provide a clustering method based on the K-means procedure for large small datasets [39, 40, 41]. The researchers drew K-Means clusters and then assessed the similarity of travel patterns.

A density-based approach to spatial clustering with noise (DBSCAN) clusters unlabeled datasets based on data point density (data point with nearest neighbors) values. It is a widely used clustering system with several real-world applications such as temperature data anomaly detection, traffic management, emotion detection, and crystallography via X-ray.

6.5 One class support vector machine (OCSVM)

OC-SVM is an extension of SVM and a member of the semi-supervised method family. If, following an operation, the new data is an outlier or deviates from the intended boundary line, and it establishes a border between the trained data. Owing to the nature of their function, OC SVMs are helpful for IoT-based machine performance evaluation, network intrusion detection, and anomaly identification in WSNs.

6.6 IoT ensemble learning models

Ensemble learning (EL) creates a collaborative and effective outcome by combining many fundamental categorization techniques. A study of EU exams shows that learning patterns vary depending on the exact application. So, a community of researchers begins to combine different classifications to increase accuracy. In addition, EL models use many learning techniques that reduce variance, which strongly resists overfitting. EL has been effectively used for Internet-based intrusion and anomaly detection in IoT-based environmental data and for evaluation of IoT precision devices and real-time data decision.

6.7 Neural networks

Neural networks (NNs) with compressed representations are the fastest models for processing new data instances. According to updates, NN has different NN with distinct structures and equipment. According to results, the most prevalent kind of neural network in functional devices is the forward neural network (FFNN), also known as a multilayer perceptron.

In FFNN, a non-linear or active function dictates how well each layer performs. When enough hidden units are present, an FFNN with at least two hidden layers can estimate a random mapping from a bounded input space to a bounded output space.

Finding the ideal weights for FFNNs that are subject to the NP-complete problem is problematic. The model incorporates several learning techniques, including RMSProb, adaptive delta, adaptive slope, adaptive moment estimation, Nesterov’s accelerated gradient, and stochastic gradient descent. In the Internet of Things, FFNN is a good option for energy management, feature selection, decision-making, computational complexity reduction, and energy efficiency.

6.8 In the UIoT, support vector machines (SVM)

SVMs use distance and data attribute computations to create a partitioning hyperplane between two distinct classes to do classification. SVMs are chosen for big datasets with lots of distinguishing features but few sample points. SVMs’ primary benefit is their ability to conduct real-time penetration exposure and then dynamically inform training patterns. Variants of SVM, like QS-SVM, CESVM, and SVDD, are less sophisticated and require less memory, making them popular in many security applications like anomaly and intrusion detection.

6.9 Social networks and the internet of intelligent things (IoIT)

In today’s digital world, where millions of individuals routinely interact, communicate, and express their ideas, opinions, and proposals, social media plays a crucial role. Many people can solve complex problems more efficiently than a single person when they connect and share ideas like this. These days, human activity in crowded regions is automatically classified in real time by smart sensors.

IoT has proven to be a model used for other forms of networking beyond computing, such as IoT and robotics as a service on the other side of the Internet world. These new models effectively add intelligence to online things or think about things like robots, cobots, services, and users. Applying social networking principles to the Internet of Things can bring enormous change and benefits.

In general, online communities are formed by humans and robots, or a mixture of them, but such groups are formed by intelligent avatars in the virtual world of the Internet of Things. Ongoing research links social networks for the intelligent processing of other biological beings automated qualified. Co-location object relationship, social object relationship, and property object relationship are SIoT examples.

6.10 Analyzing the principal components

Principal component analysis (PCA) plots the data orthogonally in an L-dimensional linear subspace called the principal subspace. PCA applications include data visualization; PCA deals with high-dimensional data sets using data compression and an iterative expectation expansion convention. As a result, PCA is regarded as the most crucial ML preprocessing step. Two or more variables are dealt with in the PCA version known as Canonical Correlation Analysis (CCA). Finding a consistent pair of highly cross-correlated linear subspaces is the major objective here. As a result, each factor in one subspace has a relationship with a distinct element in the other subspaces.

6.11 Bagging

The goal of bagging is to decrease overfitting by increasing the precision and stability of machine learning systems. By choosing data points at random from a single training set with replacements, training data are produced using this method. Thus, ML exercises are taught for each training set that was first constructed.

There are many approaches in ML such as DT, RF, and neural networks where the bagging method helps to achieve results.

6.12 Artificial intelligence in analytical skills (IoT)

Various business organizations have employed analytical skills for some decades; today, many organizations attract their AI talent to shape. To support decision-making, organizations and businesses have integrated their experience with efficient data utilization, statistical analysis, and quantitative techniques in recent decades. For now, though, the main focus of these businesses is on identifying artificial intelligence and applying it to improve one another. Unlike ML and DL, which quickly boost dominance along with demand, AI is not statistical [29, 30, 31, 32, 33, 34].

Analytic-oriented clusters in management may want to focus primarily on these machines or acquire new skills in non-statistical parts. Innovation analytics has evolved into different versions, some of which are mentioned below.

Artificial intelligence is constantly growing and plays an important role in analytics 4.0 because it can change business models. Therefore, the impact of Analytics 4.0 is potentially greater and more confusing than previous automation developments. Additionally, organizations transitioning to Analytics 4.0 are times faster than those not using any AI model. The procedure to understand the achievements of artificial intelligence begins with a primitive consideration of artificial intelligence, how artificial intelligence affects creativity and new skills, and what work practices should be implemented. Companies that manage their current analytics capabilities can get started with AI much faster and more proactively.

6.13 Analytics using deep learning (IoT)

IoT-based devices gather a ton of sensed data from their deployed settings, thanks to the development of different networks and miniaturized technologies.

Depending on the applications, these IoT objects/devices also result in quick and real-time data streams. In this case, it is critical to apply analytical models to these significant data streams to find novel information, forecast future structures, and control outcomes. As a result, IoT applications are now a legitimate business standard and a feature that raises the standard of living.

It was important to describe the features of IoT data and how they differed from those of ordinary big data to better understand the requirements for IoT-based data analysis [40, 41, 42, 43, 44]. A selection of them is listed below:

1. Massive Streaming Data: IoT implemented with a large number of devices distributed collect massive amounts of data from IoT applications, resulting in a large amount of continuous streaming data.

2. Heterogeneity: The Internet of Things is a heterogeneous connected network, so the numerous IoT data collection devices account for different results for data heterogeneity.

Time-space relationship: Currently, among most IoT devices in the real world, here are sensor devices involved in a specific location, then location and time stamp for each data.

1. High Noisy Data: Dynamic environmental changes, small error bits, and noisy data produced in IoT queries must be removed before implementation in any decision system. Otherwise, it will affect the output results.

Although extracting confidential information from big data is a talented technique to improve our lives and avant-garde, this is not an easy and simple job. It is necessary to go beyond outdated models of inferential learning and skills, innovative capabilities, practices, and infrastructure to tackle such complex and thought-provoking tasks.

6.14 Edge computing in IoT

IoT-connected devices typically produce enormous volumes of data, gathering, and processing as much information as they can in a single appropriate object to turn the data into meaningful information. Therefore, all IoT configurations today use important data functions; big data supports IoT applications as colossal identification law courts have stimulated the data stored in the Internet of Things. Furthermore, because of its varied connections, IoT gathers unstructured multivariate data that requires additional analysis to extract meaningful information [38, 39, 40, 41, 42, 43, 44]. Given how quickly a variety of technologies are developing, the Internet of Things will become the next technological revolution; However, it confuses a lot of data, processing, and system analysis capabilities. IoT uses real-time applications to work with continuous streaming, which disrupts the data storage of many sites. Therefore, more data centers are needed to process data collected by IoT devices. One possible answer is to move data to the cloud using an application platform as a service. Cloud computing is one of the established technologies today, and it provides data processing capabilities or data storage over the Internet.

Edge computing, also known as fog computing, can spread cloud computing more quickly than it can address the issues raised. Fog/edge computing, to put it briefly, replaces all computing operations in the cloud’s core by processing data and then storing it on devices at the system’s edge. Any node or item having fog computing capabilities can compute, store data, and effectively provide heterogeneous network connectivity. These items/devices are utilized across the network to transport Internet of Things objects with linked applications [38, 39, 40, 41, 42, 43, 44, 45].

Generally speaking, IoT items gather different types of data and send them to the right place for additional analysis depending on the application’s requirements. Here, fog/edge computing nodes closer to the IoT Campaigns that yield significant evidence can handle high-priority data that needs to be delivered instantly. After that, low-priority data records might be sent to a group node or object for additional processing and inspection.

In addition to its advantages, edge computing comes with obligations and restrictions when used with IoT. The most crucial job is developing edge computing and providing enough resources for Internet of Things devices. IoT devices only ever need a limited set of services; therefore, the communication, data processing, and storage capacities of each edge service node are insufficient. Each edge computing node in this scenario needs to be set and implemented as optimally as possible for IoT devices to consistently deliver the necessary service. Another challenging issue is modifying the resources allotted to the edge node. This indicates that a popular area of study in edge computing in the Internet of Things is resource management between edge nodes. Therefore, it is important to verify several requirements, including power consumption, node layer, and service availability, before utilizing fog/edge computing nodes for a certain service. Concerns with edge computing architecture’s security and privacy are equally crucial.

6.15 Federated learning

A machine learning technique called Federated Learning (FL) [46, 47, 48] uses an algorithm to train several dispersed edge devices or servers that store local data samples while maintaining their original format.

Users train local models at the base station using local data to update global models.

Devices can learn from a common shared model, thanks to federated learning. The shared model is trained on the server using proxy data first. After that, every device loads and enhances the decentralized data and trains the model using data that are available locally. An update containing the model and modifications will be sent to the cloud. Individual updates and training data are stored on the device [48, 49].

7. IoT system trustworthiness based on AI

Including the Internet of Things, many of us think that the Internet of Things will make people’s lives more convenient and stress-free. However, because it contains spam, copyright, malware, and other unwanted content, some academics have claimed that the Internet of Things is the “Internet of Garbage.” On the other hand, it will be constructed with stronger community management, stringent control, and greater communication. The most crucial step after gathering data from the trash network is determining an appropriate value. It is no secret that the Internet of Things is expanding quickly and posing new challenges. As we covered before, big data analysis, real-time streaming data monitoring, and other crucial topics include efficient communication skills and guaranteeing security requirements in a network this size. Secure software should also be deployed with appropriate network connections [50, 51, 52].

Because their information is publicly accessible online, customers—the owners of the intelligent Internet of Things—are extremely sensitive. Three key concerns for IoT devices and services are privacy, trust, and data secrecy. Before sharing data and utilizing a service, an IoT object or device needs to be authorized by an organization or individual.

Cybersecurity is the security paradigm for Internet of Things systems and associated parts. For small devices, where IoT-based cybersecurity systems frequently stop hackers from stealing critical data, cybersecurity protocols are crucial. There are countless approaches to cyber security such as encryption protocols, firewalls, antivirus, intrusion detection systems and scanners, and security sockets. ML, DL, Blockchain, and quantum-proof cryptographic technologies have been fully adapted to IoT systems to improve security. Recently, some problems have also appeared, how small wearable IoT devices collect user data that has become developer when they connect to the respective databases [43, 44, 45]. These device vendors then sell the collected user data to other commercial companies without user consent. Based on information, companies constantly make notifications and advertisements through social networks to the relevant user. Besides security requirements, the most important challenge is how to avoid this type of data ethics in IoT-based systems.

8. Internet of things applications

Since their introduction, Internet of Things (IoT) applications have brought incredible value to our daily lives. Innovations in wireless networks, smart sensors, and revolutionary computing capabilities provide a new IoT-enabled product every day. IoT applications aim to teach billions of everyday things/objects with connectivity and intelligence. This section attempts to provide an overview and discussion of many areas such as smart homes, structural health monitoring, environment, logistics, agriculture, health, lifestyle, and industry with IoT applications.

8.1 Agriculture

There was a significant increase in demand for food sources as the planet and its people developed. Farmers are assisted by highly developed research centers and governments in utilizing state-of-the-art techniques to boost food output. Smart agriculture is one of the fastest-growing areas of the Internet of Things. Here, farmers use the insights emerging from data to get a healthier return on investment. Smart Watering - With IoT-based sensors that determine soil moisture, you release water through irrigation pipes to control water use and regulate traditional lawns. Greenhouse management – information related to the indoor climate of greenhouses can monitor and control the most sensitive plant growth situations. Sense facts are saved from various sensors to a centralized server where they analyze and then improve various control strategies.

8.2 Augmented reality

Augmented reality (AR) improves the way people request, apply, and display information without disturbing the real world. Mobile Augmented Reality (MAR), where virtual elements are placed on the base material on the screen, adds value and improves the user interface with reality. This can increase efficiency and provide services by allowing workers to see basic sensor information, such as view selection, on the control panel. US-based DAQRI has designed a helmet that protects workers from falling objects and helps them avoid mistakes. The DAQRI gadget is also skilled at diagnostics in addition to using thermal vision to identify risks. The heavy machinery manufacturer Caterpillar employs augmented reality (AR) technology to observe the machine and instantly displays a visual overlay that indicates when certain parts need to be replaced, how well the filter is working, and how much gasoline is required. Because technical literature and user instructions are usually uninteresting, Bosch used augmented reality (AR) to provide movies, enhanced 3D simulations, and overlays for the equipment.

8.3 Virtual reality

The number of devices linked to the Internet will rise dramatically as virtualization continues to proliferate. Virtual reality (VR) features high-level resolution and evident, dynamic changes that provide substantial hurdles to reaching this promise since VR can transform the industry in comparison to traditional video systems. Virtual reality technologies have a significant connection to smart cities; China has already developed VR-based smart cities that have both virtual and actual instructions for emergency fire control. In the meantime, Japan unveiled the Tokyo Virtual Lab, which combines street and traffic data to mimic traffic conditions and assist drivers in emergency scenarios.

8.4 Mixed reality

There will be a considerable rise in the number of devices linked to the Internet as virtualization keeps growing. Compared to standard television systems, virtual reality (VR) has the potential to completely transform the market. However, because of its high definition and evident dynamic changes, VR poses substantial obstacles to achieving this promise. Virtual reality technology and smart cities are closely associated. China has developed VR-based smart cities that incorporate both virtual and actual instructions for emergency fire control. Simulating traffic conditions by merging street and traffic data, Japan unveiled the Tokyo Virtual Lab, which also assists drivers in dire circumstances.

8.5 Smart locks

Operators are now able to remove traditional locks and increase interest in smart locks, thanks to IoT in smart home security. Smart locks eliminate the need for a physical key, allowing a different person to access doors via biometric techniques like fingerprint, face, and iris scanning.

8.6 Smart intelligent grid (an intelligent smart grid)

In power generation, IoT makes ingenious use of power generation monitoring from different power plants. Additionally, IoT-based systems have been effectively implemented to monitor substations, towers, power consumption, and transmission lines. IoT devices also help customers with smart meters by measuring various parameters and networks. High-processing-power clever intelligent IoT devices can enhance the administration, disaster recovery, reliability, and alerting capabilities of the smart grid.

8.7 Intelligent robotics

With a multitude of uses, the Internet of Robotic Things (IoT) can analyze and improve machine performance in real time using information gathered from intelligent sensors. Logistics, delivery, rescue, agriculture, security, healthcare, defense, and entertainment all use service and humanoid robots. But as the current pandemic crisis has demonstrated, much more advancement in IoT technologies is required.

8.8 Near field communication (NFC) payment

These days, NFC is used for all retail payments as it allows customers to utilize their NFC-capable smart devices to make swipe payments. This speeds up payment processing and improves payment security and refunds.

8.9 AI-enabled internet of underwater things

IoUT-connected autonomous submarines equipped with sophisticated sensors can identify hostile submarines and underwater riches. IoT is also useful for metal, mineral, and coral reef detection. Underwater resource discovery generally requires sensors equipped with IoT-compatible video capture devices.

8.10 IoT-based forensic software/applications

With the resources at hand, countless models for IoT security and privacy have been implemented; yet, this is still an unanswered research subject. The Internet of Things has received little attention in the field of digital forensics today. IoT security breaches are likely as IoT security is still in its infancy. To track down similar attacks and locate trustworthy digital evidence that will identify the offender, active digital forensic techniques must be put into place alongside security safeguards. Personal forensic materials like EKG trends, heart rate, activity tracking, port scans, timeline logs, etc., are found during VitalPatch checks.

8.11 IoT-based intelligent healthcare systems

Wearable Internet of Things (IoT) devices make it possible to continuously monitor physiological constraints, which aid in the continuous monitoring of health state (e.g., exercise monitoring).

Moodables is a mood-enhancing device that tracks and improves our mood throughout the day. In particular, it is a wearable, head-mounted gadget that can be adjusted to improve mood by applying low-intensity currents to the brain.

Ingestable Sensors: These tiny sensors keep an eye on the medications we take and alert us to any anomalies they find, allowing doctors to diagnose patients early on.

In addition, it is suitable for reducing the waiting times in emergency rooms, improving medication management, patient monitoring, and ensuring the availability of critical equipment personnel.

8.12 Intelligent disaster management

Intelligent disaster management helps minimize the potential damage of future disasters and also confirms immediate and proper referral to victims for speedy recovery. So far, IoT capabilities have reached a high level and are likely to be useful in disaster situations. Risks are reduced with an Internet of Things system that has satellite connectivity and location data management. It encourages prevention, offers early warning, raises awareness through social media, steers clear of emergency and reaction operations, and looks for people who may have gone missing.

9. Open research challenges/problem for AI-based/powered IoT systems

9.1 Challenge/problem 1: how is processing power managed in industry 5.0, which is AI-based?

Tech and industrial companies suffer from computing power challenges. IoT devices collect massive amounts of data to build AI models that can explore these massive data using ML and DL, which require consistent power consumption. This is a big problem for product manufacturing and startups. In most cases, the power requirement of the learning algorithm drives the programmer away. However, ML and DL are excellent AI components that are very accurate yet powerful but require more cores than the GPU. The researcher implemented many ideas and plans for the progress of ML and DL models in several types of equipment. Additionally, AI-powered cloud computing and AI-powered parallel system processing are utilized to reduce power usage.

9.2 Challenge/problem 2: how AI gets beyond IoT’s technological obstacles

Several links to converse and connect to other networks. There are currently well-known platforms for hiding online organizations, but they are complicated and require a lot of effort. Service-Oriented Design: The performance and cost limitations of the Internet of Things present significant hurdles in terms of service orientation. Applications determine how many devices are connected, how scalable the system is, and how complex data processing, networking, and service delivery are. Detected data transmission over heterogeneous networks, in particular, frequently results in delays and communication issues.

High automated standards are needed that allow data to be easily collected from various devices and skillfully transformed into IoT systems. More importantly, the collaboration of connected devices between different entities requires proper allocation; identification and optimization are still open research questions. Since recently planted objects have minimal effect on the Internet of Things, suitable integration mechanisms—like a consistent data structure—are required. While AI-based methods take advantage of useful IoT data functionalities, they are complicated and require a lot of power.

9.3 Challenge/problem 3: what is the plan for handling different implementation tactics in industry 5.0?

Recent advancements in a variety of network technologies suggest that artificial intelligence (AI) has the power to digitize any industry or corporation. However, a major issue with AI is that there are not enough workable concepts. Succeeding in an enterprise environment requires a strategic approach, such as identifying successes, hotspots, and persistent bugs. To gain insight into the more mentioned issues, business leaders, managers, and technical teams need to have a broad understanding of AI capabilities, advances, and limitations and observe current AI challenges. Through adherence to these updates on Data Workstyle connected to AI, firms can readily identify places where AI can be enhanced.

9.4 Challenge/problem 4: what obstacles did IoT data analytics have to overcome to be implemented?

We talked about IoT analytics using ML and DL techniques in the section above. Time-series-based data architectures can present challenges for IoT analytics implementations in certain situations. The long-term, massive static data samples that IoT-based smart sensor nodes routinely collect make it challenging to identify justifications for extrapolation analysis. It is also challenging to choose the best storage and analyze data stored there quickly. Information is sensitive and, with proper handling and accuracy, may yield valuable insights into any organization and product. Data scientists are therefore necessary for programmers, database experts with high skill levels, data processing specialists, and data scientists.

9.5 Challenge/problem 5: how AI addresses the security and system complexity issues with wireless networks and the internet of things

Any network has to deal with a great deal of complexity since AI applications in communication design make systems more complicated overall. The DL approach utilized to accomplish one goal in ad hoc ML of industrial networks overlooks the other objective goals. Furthermore, IoT devices can transmit data to a higher level without the need for computing pre-processing if they have enough resources. Therefore, AI strategies implemented in WSNs and IoTs need to balance regulating other restrictions like processing, latency, storage, and connection with optimizing individual goals. Optimizing one layer with AI techniques may help or benefit another layer.

9.6 Challenge/problem 6: how will AI enable intelligent communication in the UIoT at high speeds?

Dynamic AUVs, underwater magnetic inductions, and acoustic networks are all part of the complex heterogeneous network known as the Internet of Things. An unforeseen issue for a UIoT system is the sea’s variable environment, which affects temperature, wind energy, tides, and positioning accuracy underwater. Every UIoT network part is coordinated and optimized for communication using AI-based techniques, allowing the network to be deployed to its full potential.

9.7 Challenge/problem 7: what is the impact of AI-driven industry 5.0 on product management?

Many sectors employ radio frequency identification (RFID) tags for supply chain management, product tracking, and supply chain optimization. Products being shipped have RFID tags implanted, and readers follow the products’ whole path. Increased reader and location flexibility can be obtained from IoT equipment, and ongoing interoperability between RFID-based devices employed by various performers is made possible. Retail establishments are the main users of these IoT devices for tracking precise inventory levels and displaying product availability. More crucially, the food business may enhance manufacturing procedures and end-product value by merging sensors and biosensors with RFID technology.

9.8 Challenge/problem 8: man-power

The effectiveness of a company’s usage of AI in areas that require improvement is important, regardless of its size. Artificial intelligence is creating new technologies; hence, there’s a demand for experts in and implementation of AI models. That’s why, companies make extra budgets to train and hire AI experts.

9.9 Challenge/problem 9: aI training

The deployed AI system is prepared for training after it has gathered sufficient data. Every AI-based technique varies in terms of model type, data structure, outcomes, and decision-making. There is no perfect model in AI that fits all cases, and we must test each model. Based on the results, we determine what works best. Therefore, implanting smart sensors with additional capabilities is the first step toward analyzing a larger portion of the natural environment to teach machines to learn. Also, repeatedly run iterations on the proposed models with each new choice, so that the AI models naturally learn to enhance performance.

9.10 Challenge/problem 10: connected devices

The connected devices have digitalized almost every aspect of the world. However, simple IoT devices such as smart sensors communicate with other sources using Bluetooth/Zigbee, and connecting them to the Internet remains a challenge. Because it takes specialized technology, reinstallation, and moving to new devices to connect every device to the Internet, it is not an easy task. Not a simple operation, but a simple example of how to connect devices to the Internet is MQTT powered by AI.

10. Conclusions

This chapter outlined the process of comprehending vision, an IoT capabilities model based on AI, and its numerous applications. It also assists researchers and specialists in identifying project structure and AI algorithms with IoT and the latest IoT countermeasures. It provides a complete discussion of the functional framework, then the IoT information hierarchy, object recognition, smart sensors, learning, and analytics in smart IoT-enabled systems. This chapter explores AI-powered IoT paradigms that can be used for the betterment of humanity in the coming era. Specifically, comprehensive references and implementations of many jobs are found important for engineers and management.