Internet of vehicle's resource management in 5G networks using AI technologies: Current status and trends
Abstract
The Internet of Vehicles (IoV) and Vehicle-to-Everything (V2X) concept have emerged from IoT technology, which refers to connecting many vehicles with various applications to the internet. The 5G new radio is based on a cloud-radio access network (CRAN), considered as the communication infrastructure for IoV. However, due to the significant challenges and issues, researchers have been working on IoV and V2X. One of the main challenges for V2X is resource allocation and management for a high-speed vehicular environment. This paper discusses and provides complete detail for resource allocation and management for IoV over 5G RAN networks focusing on artificial intelligence techniques. The paper also presented reviews on integrating the multi-layers of vehicular network architecture with AI strategy to identify advancement and future directions for resource allocation and management issues.
1 INTRODUCTION
The recent development in IoT large domain drives to change the conventional vehicle ad-hoc networks (VANET) to become what is known as the Internet of Vehicles (IoV) [1]. IoV is a specific IoT technology application for intelligent transportation systems. Recently, she has gained a lot of attention as a part of the Intelligent Transportation System (ITS). TheITS applications and other networks such as Vehicle to Infrastructure (V2I) and Vehicle-to-Vehicle (V2V) must provide very low latency [2].
The 5G network provides the basic infrastructure for smart IoV environment building. It drives the vehicle network capabilities to achieve extreme performance. 5G technology enables a new evolution of the Internet of Everything (IoE) [3] networks with capabilities of faster speeds, lower power, and massive connectivity, to achieve digitizing and contribute to all life aspects. In Industry, applications based 5G networks, the integration of artificial intelligence (AI) with other technologies such as visual and sensing technology, edge computing, in addition to augmented and virtual reality will lead to high technology evolution [4-5]. The 5G communications ensure the enhancement of applications that operate within different types of integrated technologies more efficiently and intelligently.
Resource allocation and management are considered complex and challenging goals to achieve the expected results in IoV networks. In general, wireless technologies provide an infrastructure for vehicle communications, as they play an important role in the efficient allocation and management of resources, in addition to providing adequate QoS and QoE [5].
The 5G networks enable more reliable communication at ultra-high speeds and low latency. Technologies such as software-defined networking (SDN) and network function virtualization (NFV) enable to development and leverage of the 5G networks. The use of SDN in 5G networks enables to provision of open flexible and programmable new services incorporation, which will ensure high-performance platform for autonomous vehicles based on 5G [6]. In SDN-based 5G networks, control and data are separated to achieve a centralized and efficient resources management with secure allocation. SDN provides flexible communication management and resource allocation for the 5G vehicle network infrastructure, enabling the safety and privacy of network environments, as well as improving performance [7].
The paper provides a deep concept about resource allocation and management mechanisms for 5G IoV, in addition to intelligent AI-based IoV applications. The rest of this paper is organized as follows; Section 2 provides a background to IoV communication and networking concepts. The resource allocation and management of IoV issues are presented in Section 3. A brief idea about IoV based 5G network architecture, network access computing, and function visualization related to 5th G based IoV communication system are given in Sections 4 and 5. The review of AI in IoV based 5G network and its application in resource allocation and management are provided in Sections 6 and 7 respectively. In Section 8, different AI technologies in IoV based 5G applications such as machine learning and swarm intelligence algorithms are briefly reviewed. In Section 9, future research directions are discussed. Finally, in Section 10 this paper is concluded.
2 IOV IN 5G NETWORK
IoV has several definitions and concepts that were discussed extensively in the literature. According to (APEC China group, 2014), IoV is considered one of the IoT technologies related to intelligent transportation technology (ITS), as it depends on the integration of both internal and external vehicle networks, in addition to the vehicle's mobile network [8]. According to this concept, the authors explained that IoV is a wireless communication system based on the exchange of data between a number of vehicles and V2X networks, where X is defined as everything on the road or in the path of vehicles. The authors also explained that in IoV networks, communication between vehicles, personal devices, and roadside units (RSU) takes place using various wireless communication protocols.
In [9] Ibanez et al. defined IoV as a technology that integrates the IoT and ITS, enabling the exchange of data between vehicles and the surrounding environment such as sensors, roadside units, and other portable devices [10]. The authors explained that the method that allows the exchange of data in IoV is the Internet by using different communication protocols, which provide several traffic services and entertainment applications on vehicles in addition to safety management capabilities.
In [11] Yang et al, explained that IoV is the merging of several entities such as people, vehicles, and the environment with a large smart network that provides various services as an application for roads and traffic management in big cities [12]. They also explained that IoV aims to improve and reduce transportation costs, driver safety, and traffic management with high efficiency, in addition to providing entertainment applications and services and safety information.
In [13] Cheng et al. presented the concept of IoV as an indication of the expansion of VANET networks and the diversity of its applications to enable communications between passengers and vehicles [14]. It also provides the possibility of exchanging information between vehicles and the surrounding environment to achieve traffic safety and driving efficiency.
[15] Hossan et al., in 2017 explain that IoV is an evolution of Vehicular Ad hoc Networks (VANETs). It enables the development of intelligent vehicle applications [5]. They present that IoV offers many kinds of applications and services by using different communication technologies such as Wi-Fi, WI-MAX, and Bluetooth. The authors also define many types of IoV communication networks such as vehicle-to-vehicle (V2V), vehicle to infrastructure (V2I), vehicle to the cloud (V2C), and vehicle to everything (V2X).
In [16] Benalia present IoV is a combination between two folds, network, and intelligence. The vehicles can communicate with their environment using different communication technologies such as 4G / LTE and 5G [17]. In addition, by using smart technologies, several communication models such as V2V, V2I and vehicle to roadside (V2R) can be configured by taking advantage of cloud and fog computing capabilities, enhancement technologies, and processing and storage upgrades methods.
In [18] Nassar and Yilmaz discussed the resources allocation in the network using slicing technique by DL and RL. DRL is used for optimal and adaptive vehicular network slicing approaches. Extensive simulation analysis is conducted for smart cities and intelligent vehicles network. The results showed that using DL enhanced the dynamic resource allocations performance by about 15%
In [19] Hlophe and Maharaj a systemic review for cognitive radio (CR) networks resources allocation using AI techniques is presented. The emphasis of the review was on the integration between deep architecture and cognitive radio networks for spectrum allocation and sharing [20]. The paper addressed the problem by describing how sophisticated and deep architecture can enhance the CRNs in terms of resources management and allocation issues.
In [21] Zhang, proposed a resource allocation reservation technique for IoV based on services arrival rate. A simulation analysis is conducted and the results achieved showed that the introduced technique gets a fewer rate of rejections for the request of service and the resources allocation performance is enhanced by 7%.
In [22] Zhang utilized a fog scheme for optimization of VoI resources management usage and reduce access delay for resources. For formulization of the problem, integer-mix non-linear problem and then converted to convex form by Frobenius–Perron theorem and MMSE. The simulation results showed a dramatical enhancement in resources management and reduction in delay by using aggressive and high services requests.
In summary, IoV was defined as merging both the Internet of Things and Intelligent Transportation Systems into a network that enable several services related to traffic safety, driving, and backward entertainment applications [23]. The new concept known as IoV supports communication with anything in the surrounding vehicle environment to manage traffic safety operations and reduce accidents, crimes, and environmental disasters such as bad weather and other natural accidents.
Due to IoV network structure based on 5G communication, resource management is considered one of the most important issues, which can affect QoS and QoE. With cloud and network function visualization (NFV) transformation of communication networks, integration of 5G and IoV, and development of diverse industrial applications, the operations of IoV networks will face unprecedented challenges in the 5G era. These challenges involve complex networking, diverse services, and personalized experience.
2.1 Complex networking
The process of interaction between different networks such as 4G and 5G encounters difficulties in interoperability, and defecting faults in the hierarchical separation structure. In addition, the unified resource scheduling problems due to the cloud and virtual networks dynamic change will increase the networking complexity [24, 25].
2.2 Diverse service
The single man-to-man communication mode has gradually evolved into a full-scenario communication mode that involves man-to-man, man-to-machine, and machine-to-machine communications. The business scenarios will be more complicated and thus bring about differentiated SLA requirements such as high bandwidth, massive connections, ultra-high reliability and low latency, and associated complex network management.
2.3 Personalized experience
Drawing on 5G network capabilities and abundant business patterns, 5G service experiences will tend to be diverse and personalized, like immersive experience, real-time interaction, and nuanced awareness of emotions and intentions. The network support for the experiment will destroy the traditional model and enter into new challenges. Accordingly, the challenges that arise with the 5G will be significant and lead to a gap between advanced operations and traditional operations based on expert experience. Therefore, automatic and smart network operations will become available only in the 5G generation networks.
Artificial Intelligence (AI) technology has new features and effective solutions for analysing big data and provides smart methods for creating dynamic strategies and managing resource operations in the 5G network [26]. In the future, based on cloud infrastructure, a network that combines 5G, AI, and IoV will gradually become the intelligent centre of digital society and promote the intelligent interconnection of all things. Based on cloud and service-based architecture, IoV based 5G network has distinct differences at different network levels [26, 27]. The upper layer is more centralized and has higher requirements for cross-domain analysis and scheduling capabilities such as E2E slice orchestration and management and global cloud resource coordination that rely on a centralized smart engine (SE) for centralized global strategy training and reasoning. The lower layer closer to the end side will focus on intelligence enhancement of professional subnets or on single network elements. Access network, bearer network, and core network introduce lite SE (LSE) to enhance the intelligence of subnets or sub-slice domains such as management strategies and smart operations [28].
Edge devices such as MEC and 5G eNB introduce real-time SE (RSE) to achieve real-time or semi-real-time intelligence at the edge. The artificial intelligence algorithms and smart engines can be deployed in 5G networks based on hardware computing environments at various levels. The combination of application algorithms with engines and model components in different network functional entities will enable the 5G network intelligence.
3 IOV RESOURCE ALLOCATION AND MANAGEMENT
Due to the multiplicity of different vehicles in IoV network, the nature of their connection to other networks, and the availability of different sources, IoV faces one of the most important challenges, which are how to manage existing resources and arithmetic restrictions for them. These challenges require advanced calculation mechanisms to ensure quality satisfactory experience (QoE) [29]. IoV requires to develop an efficiently alleviate resource management scheme and ease the heavy execution burden of vehicles.
Network function virtualization (NFV) and SDN are the most recent key technologies in 5G, which promises to achieve efficient network management. These techniques enable the organization of resource management process, as the level of control is separated from the level of data, which allows entering a logical central control of the network. SDN controllers are used in IoV networks at servers and cloud computing levels [30-32]. RSUs send local traffic information to vehicles and among them, enable to collects of macroscopic visualizations in the network to facilitate decision-making at different levels of SDN controllers. SDN has been envisioned as an emerging approach providing programmability, adaptiveness, and flexibility [33].
In IoV architecture, SDN enters into new layers that serve the application, control, and data levels, responsible for resource management. The development of new technologies, that is, network function virtualization (NFV) and software-defined networking (SDN) are important to enable 5G technology in IoV. The use of SDN in the 5G network provides an excellent platform for self-driving vehicles [34]. SDN also optimizes flexible communications and resource management for vehicle networks. Where these management processes are of utmost importance when considering the special nature of networks IoV in terms of safety and hydration [35]. Figure 1 shows the categorization of resources management in 5G.
Recently, several studies have been developed on resource management in IoV networks taking into account reliability and energy control constraints. Some of them have been investigated in the wider social IoV systems, and studied in improvement methods safety, traffic efficiency, and decision-making mechanisms for resource management in V2I and V2V applications. Other studies propose a resource allocation scheme to maximize the sum capacity of pedestrian users and V2V links [36].
3.1 AI for 5G resources management
The use of artificial intelligence has become one of the most important tools required to work efficiently to manage the network resource among smart vehicles with high-level intelligence in IoV networks. Machine learning methods achieve a means of intelligent policy control and do decision-making for IoV resource management. ML enables to manage IoV traffic and balance traffic load in clustering-based IoV networks [37]. ML with SDN achieves a cognitive capability for IoV networks, and optimal routing policy under dynamic environments.
Artificial intelligence allows strong learning and thinking ability, in addition to intelligent recognition ability of 5G network architecture to learn and adapt and support various services accordingly without human intervention [38]. Figure 2 shows the IoV resource management-based AI
AI technologies can be applied to achieve network intelligence, improve network management, and intelligent wireless connectivity for 5G networks. ML enables to achieve intelligent resource management, automatic network modification, and smart service provision with a high level of intelligence, as the architecture consists of four layers: sensor layer, data extraction and analytics layer, control layer, and application layer [39, 40]. ML is able to intelligently extract valuable information from big data, learn and support various functions of self-configuration, self-improvement in IoV-based 5G networks, in order to address optimal physical layer design, complex decision-making, network management, and resource optimization tasks.
4 THE 5G BASED IOV COMMUNICATION SYSTEM
IoV concept is related to the communications of vehicles and things within the transportation infrastructure, which depends mainly on the Internet protocol, as it enables the exchange of information in an appropriate manner that achieves efficiency and safety in transport. The deployment of IoV environment depends on a certain number of aspects [41]. These aspects are considered as a scenario approach as a part of internet-connected vehicle applications, such as intelligent traffic management, safety, and emergency management, and environmental driving for energy efficiency.
Self-management algorithms provide a model capable of improving the operating performance of connected vehicles to the internet, enhancing safety standards, traffic management, and providing various services for internet-based vehicles. The use of AI mechanisms improves data collection processes for various and diverse vehicles connected to the internet and enables modelling techniques for both vehicles and the environment, which helps in deducing the variables that occur on the IoV network [42-44]. Recently, AI technologies have provided effective solutions to visualize large-scale IoV network that use a very large number of vehicles, sensors, and computing with higher efficient communication capabilities.
Previously, cellular communications have always been used in-vehicle networks such as GSM and UMTS, even the most recent networks like LTE. Recently, the emergence of fifth-generation (5G) wireless networks has enabled different devices to communicate and exchange data more efficiently with very high data capacities compared to 4G networks [45, 46]. 5G technology is characterized by high data exchange speed with low latency while preserving low power and provides very high bandwidth width, which led to the rapid development of IoV applications. 5G communication network technology can replace or complement some of the current car network communication systems. The new 5G based on D2D terminal direct communication technology is one of the most important features that would help the growth of IoV services and applications.
The benefits of 5G communication are related to the concept of vehicle-to-vehicle (V2V) communication which mostly relies on device-to-device (D2D) communication in a 5G cellular environment. D2D communication increases spectral efficiency, improves user experience, and extends communication applications [47].
The solution to this problem depends on the use of Long Term Evolution (LTE) technology, but it cannot maintain V2V connections normally. However, D2D, as a key technology in the 5G network, can send information directly between devices and devices, so the infrastructure-assisted D2D communication technology can be used as a natural way to achieve reliable and efficient V2V communication. However, the infrastructure-assisted D2D communication technology can be used as a natural way to achieve reliable and efficient V2V communication in 5G networks [48]
4.1 Transmission timeliness
In IEEE802.11p, the V2V communication suffers the transmission delay, about 10 ms. This transmission delay is reduced to 1ms using a 5G coverage network for V2V communication. But 5G car network V2V communication transmission delay up to 1 ms, can effectively improve the IEEE802.11p communication delay problem, to ensure timely reception of information.
4.2 Transmission rate
5G vehicles network compared to IEEE802.11p data transfer rate will increase about ten times to achieve the car and terminal equipment to achieve high-quality audio and video communications.
4.3 Communication distance
5G vehicles network communication distance up to 1000 m, can effectively solve the IEEE802.11p V2V communication in the connection time is short.
4.4 High speed mobility
Compared to IEEE802.11p standard communication, D2D supports 350 km/h vehicle speed, can meet the faster vehicle communication.
4.5 Increase spectral efficiency
Data is transmitted directly between vehicles without routing through a cell network and thus results in a hop gain. Gain can be increased by reusing resources between vehicle users and between vehicle and cellular networks. Wireless spectrum efficiency and network productivity can be increased by obtaining hop gain and resource reuse gain.
4.6 Enhance user experience
The development in mobile phone technology and the multiplicity of services are important for the business growth on the wireless platform. In addition, sharing of data between nearby vehicles and users' location-based social and business activities determines how well the quality of the user experience is improved. IoV based on 5G network will enhance the user experience in these service modes.
4.7 Expanding the communication applications
In conventional wireless networks, the communication system may collapse if the basic network facilities or access network devices are damaged. However, IoV based 5G communication makes it possible for cellular vehicles to create ad hoc networks. A multi-hop communication can be used between vehicles or access to the cellular network in the event of any malfunction or damage to the wireless infrastructure.
5 THE 5G BASED IOV NETWORK ARCHITECTURE
Various IoV based 5G architecture has been proposed by different researchers in last few years. (Benalia et al.) proposed an architecture called IoV based on 5G communications as shown in Figure 3 [49]. In this architecture, the authors used four main technologies which include cloud computing, fog computing, SDN, and 5G technology. (Kaiwartya et al.), proposed a five-layered architecture including perception, coordination, Artificial Intelligence (AI), application and business layers [50].
The functionalities of each layer are described below in detail and a summarized view is shown in Figure 4. (Contreras-Castillo et al.) proposed a seven layers-based architecture as appeared in Figure 5. They designed seven layers’ architecture by reducing the complex layer's functionalities by grouping the very similar functions in a same and appropriate layer, thus, making its implementation easy [51].
Other IoV network architectures also have been developed pending on layering design. The layers designed according to the IoV network applications purposes. Most of these architectures support many communication models, that is, V2V, V2I, V2X etc. Table 1 reviews a survey on different proposed IoV architecture schemes between years 2011 and 2020.
Scheme | Year | Proposed layers | Conceptapplied | Models supported | Security |
---|---|---|---|---|---|
Liu Nanjie[12] | 2011 | 3 layers |
|
V2V, V&R V&P, V&I |
Security as a service |
Flavio Bonomi [16] | 2013 | 4 layers |
|
V2V V&I |
Cross-layered |
Wan et al. [13] | 2014 | 3 layers |
|
V2V V&R |
Cross-layered |
Kang et al. [14] | 2015 | 3 layers |
|
OVS V2V, V2I, V2P |
Not specified |
Kaiwartya et al. [10] | 2016 | 5 layers |
|
V&I, V2V, V&S, V&P, V&R |
Security plane |
Gandotra et al. [15] | 2017 | 3 layers |
|
D2D-B, D2D-C, D2D-D, M2M-D and D2D-N |
Not specified |
Castillo et al. [ 11] | 2017 | 7 layers |
|
V&I, V2V, V&S, V&P, V&R, R&P, R2R, S&A |
Cross-layered |
Alouache et al. [17] | 2018 | 6 layers |
|
V2V, V&I, V2X | Security plan |
Darwish et al. [18] | 2018 |
Multi- Dimensional (13 layers) |
|
V&I, V2V, V&S, V&P, V&R, R&P, R2R, S&A, V2X |
Cross-layered |
Benalia et al. [6] | 2020 | 3 layers |
|
V2V, V21, V2S, V2R, V2P | Not specified |
Nassar and Yilmaz [22] | 2021 | 3 layers |
|
V2X, V2I, V2R | Not specified |
Zhang et al. [23] | 2021 | 5 Layers |
|
V&R, R&P, R2R, S&A | 5G privacy protocol |
5.1 Multi-access edge computing (MEC)
Edge computing is defined as a method that improves data processing in cloud computing systems that is implemented on the edge of the network to be closer to the data source which will reduce the delay dramatically [52]. Edge computing technologies offer many benefits associated with latency challenges, as they provide a dense bandwidth close to the user. In addition, advanced computing enables the future grid infrastructure to process power and analyse data. As shown in the Figure 6, the edge helps with real-time data processing closer to the source or even on the-premises [53]. Multiple access edge computing or mobile edge computing (MEC) is defined as a network architecture that allows computing and storage resources to be placed within a radio access network (RAN). MEC is improving network efficiency and delivering content to end users. Network efficiency can be improved and the need for long-distance transmission reduced by adapting to the load on the radio link by MEC unit [54].
As IoV network requirements increase dramatically, the technologies and devices that support the IoT and 5G are evolving. Mobile edge computing allows operators to deal with excessive traffic and more resource intelligence. MEC is helping to define the foundations for smart future IoV networks [55, 56]. It also provides enhanced location, augmented reality and IoT services support, in addition to giving intelligent vehicles industries a head starts and time to adapt to new 5G technologies. Figure 6 illustrates the MEC deployed to provide storage, computation, and connectivity capabilities in addition to resource allocation and management within a radio access network (RAN) at a radio network controller (RNC), LTE base station (eNB), and multi-technology cells aggregation sites [57]. The MEC server is managed by an operator and consists of hardware resources, a virtualization layer, and an application hosting platform. RAN infrastructure provides real-time network information, proximity, and location awareness.
Mobile edge computing (MEC) enables the extension of the centralized cloud computing capabilities to the edge of the terminal devices, which meets the requirements of critical vehicular networks applications. The use of MEC provides many benefits including an increase in mobile networks efficiency. However, the continuous demand to enable communication at any location, in addition, to enabling energy-saving calculations and obtaining low latency are considered the most important challenges facing the application of MEC to IoV networks [58]. MEC is considered as fog computing that provides a potential environment for information service and cloud computing capabilities at the edge of the network, meaning that RSUs or adjacent vehicles in IoV environment help to meet the productivity performance and response time requirements [59]. Vehicular edge computing (VEC) frameworks can make tasks for efficient offloading. Vehicle edge computing depends on the availability of adjacent vehicles as well as RSUs and a cloud server. It requires task scheduling for an available vehicle, RSUs, or the cloud, and specifying the offloading and downloading methods [60-63].
5.2 Network function virtualization (NFV)
Software-defined networking (SDN) provides a core network model with flexible programming and management capabilities and easy to instantly modify the behaviour of IoV network elements. The degree of flexibility is shown in the possibility of making many networks functions virtually so that they are deployed as software packages by using network function virtualization (NFV) [64]. NFV is a new concept of network architecture, in which the network functionality must move from dedicated hardware to virtual software systems. The defined VNF functions can be created or destroyed dynamically, and they can be grouped and restricted to implement traditional or novel services. The cooperation between the NFV and SDN achieves new flexible and robust IoV architectures [65].
The network operators can use the NFV to operate and extend their network capabilities on demand by virtual software applications. This enables load balancing, scaling and miniaturization, and transfer of functions across distributed hardware resources where physical hardware once stood in network architecture. Operators can keep things running on the latest software without interrupting their customers through constant updating. NFV has the advantage of improving network flexibility whilst reducing overall overhead. The use of NFV helps to improve network resilience while reducing overheads, as well as achieving a significant reduction in energy consumption and simplifying network functionality [66]. Telecommunication network operators have proposed the use of network function virtualization (NFV) to address the lack of business flexibility and meet the ongoing requirements of reliable infrastructures. ESTI developed reference architecture as shown in Figure 7 for NFV depending on network service, management, and orchestration domains.
In the network service domain, three layers presented which are responsible for NFV infrastructure, virtual network function, and operational support infrastructure is serving as a central data cloud to enable hosting and virtual functions connectivity. It provides a virtualization concept to the physical resources, in addition to the bearing of all the elements required for the virtual network function layer [67].The management and orchestration domains are responsible to operate three layers, two managers are virtualized infrastructure and VNF, in addition to the NFV coordinator. These layers help to schedule the strategy management and network service lifecycle. IoV physical network infrastructure is virtualized by the virtual network function (VNF) layer based on the NFV infrastructure (NFVI), that is, vehicle gateway, local server, and security models [68].
6 AI ENABLED 5G IOV NETWORK ARCHITECTURE
Artificial intelligence (AI) promises great solutions to issues related to complex perception and optimization, in addition to related to predication and adaptation to the surrounding environment [69]. Researchers validated that 5G IoV networks operation may need AI to work more efficiently and effectively [70]. AI enables archives of several solutions to perform different 5G applications at many levels in IoV network architecture. Machine and deep learning algorithms help to create methods for essential predictive and proactive processes in 5G IoV applications. Table 2 reviews different kinds of AI algorithms for 5G wireless communications technology.
AI schemes | Study ref. | Learning models | 5G applications |
---|---|---|---|
Reinforcement learning | Khan et al. |
Reinforcement learning (RL) Based on long short-term memory cells. With network-assisted feedback. |
Proactive resource allocation in LTE-U Networks [19]. Non-cooperative game to learn for LTE-U traffic loads [19]. Heterogeneous radio access technologies (RATs) selection [19]. |
Nassar and Yilmaz |
Deep reinforcement learning for adaptive network slicing in 5G for intelligent vehicular systems infinite-horizon Markov decision process (MDP) |
Network slicing [23] resource block scheduling cloud-RAN (C-RAN) C2X protocol |
|
Supervised learning | Wang et al. |
Machine learning (ML) Statistical logistic regression |
Dynamic frequency and bandwidth allocation in self-organized LTE networks [20]. |
Sadreddini et al. | Support vector machines (SVM) | Path loss prediction model for urban environments [21]. | |
Yang, et al. | Neural network-based approximation | Channel learning for channel state information (CSI) [22]. | |
Zhao J et al.(2016) | Supervised machine learning frameworks |
LTE network traffic conditions prediction. Adjust LTE TDD uplink-downlink configuration [23]. |
|
Zhang et al. |
Supervised fog-based Vehicular networks Mixed-integer nonlinear Convex by perron-frobenius A weighted MMSE [85] |
HetNet radio resources management (RRM) massive machine-type communications (mMTC) [24] cloud radio access network (C-RAN) CRAN and FogRAN architecture heterogeneous cloud RANs (HCRANs) |
|
Liangmin et al. | Artificial neural networks (ANN) | Objective functions modelling and approximations for link budget and propagation loss [24, 25]. | |
Feng et al. | Multi-layer perceptron's (MLPs) | ||
Unsupervised learning | Pham et al. |
K-means clustering Gaussian mixture model (GMM) |
Cooperative spectrum sensing [26]. Relay node selection in vehicular networks [27]. |
Abbas et al. | Expectation maximization (EM) | ||
Kreutz et al. | Hierarchical clustering | Anomaly, fault, and intrusion detection in mobile wireless networks [28]. | |
Borcoci> |
Unsupervised soft-clustering Machine learning framework |
Latency reduction in heterogeneous cellular networks by clustering fog nodes to predict nodes' power [29]. | |
Nguyen et al. | Affinity propagation clustering | Data-driven resource management [30]. |
In 5G IoV networks, energy efficiency is a required aspect, especially with spectrum distortion effects due to phase and square imbalance and non-linear distortion. Artificial intelligence helps boost energy efficiency by facilitating effective resource allocation and management. In 5G networks, PHY and medium access control (MAC) layers are considered foundational layers [71]. AI helps to improve performances within these layers. In the 5G networks, MIMO operation depends on the channel's state information. In IoV applications, various services need a large number of MIMO systems. Based on 5G network standards limitations, the number of signals is much smaller than the number of MIMO antennas [72]. Learning methods like deep learning enable to assist in the estimation of channels for MIMO systems, in addition to learning the MMSE channel estimation with low complexity.
Learning-based mechanisms deliver high performance without regard to detailed channel models. A study presented by Samuel et al. and Mosleh et al., reviews models demonstrating the potential of AI-like deep learning algorithms to detect traditional MIMO code using incomplete CSI receivers [73]. In these models, deep learning helps to eliminate interference and improve the performance of the receiver, which will turn to improve the efficiency of 5G IoV communication systems [74]. For the channel decoding process, deep learning neural networks are utilized for obtaining performance gains [75].
7 AI FOR 5G BASED IOV RESOURCE ALLOCATION MANAGEMENT
Although AI technologies offer unique solutions in computing and communication networks managements, they face some factors that affect their application to 5G networks [76]. These factors are related to configuring a new platform, data-based products, and the sense of machine learning supported by distributed computing and resources [77]. Artificial intelligence algorithms are designing new models to improve mobile networks with a high level of intelligence. In addition, AI provides unique solutions for the Internet of Vehicles (IoV) for operations related to transportation safety and collision avoidance, in addition to information transmission services in ITS [78]. AI enables to intelligently drive mobile communication and coordination among several kinds of vehicles networking, that is, vehicle to vehicle (V2V), vehicle to infrastructure (V2I), and vehicle to everything (V2X).
In 5G IoV, AI promises to reduce accidents and road traffic congestion besides ensuring traffic safety improvement. Moreover, AI helps to optimize the standards of the end users’ quality-of-service (QoS) and quality of experience (QoE). However, these standards are dependant strongly on how to efficiently do resource allocation and management [79]. Various resource allocation approaches have been proposed in IoV communication networks to ensure effective and improved QoS and QoE standards. Distributed and cloud-based computing resources, in addition to big data resource allocation and management processes, are the most important trends in resource allocation and management, whose efficiency can be increased by using AI.
7.1 Intelligent distributed computing resource
The process of enhancing resource sharing is due to a group of autonomous computers known as a distributed system. In distributed systems, resource discovery and resource scheduling are related to the concept of resource allocation and management. Resource discovery processes can be carried out in a centralized or decentralized manner [80]. The major important factors of distributed computing resources are considered as resource sharing, concurrency, scalability, transparency, and fault tolerance.
Recently, modern distributed systems have been developed, including independent processes operated on the same physical device and interacting with each other by exchanging messages. Such systems are distributed real-time systems, parallel processing, distributed database systems, and distributed artificial intelligence. Many advanced and smart industries use several kinds of real-time systems such as flight control systems, automatic control systems in factories, as well as smart tracking systems [81]. All these real-time systems require an intelligent mechanism to optimize the processes related to distributed resources and systems [82].
Large-scale computing and the parallel processing power of large datasets are stimulated by AI learning methods. Moreover, multiple data models and different database management systems are enabled by heterogeneous distributed databases [83]. Recently, distributed systems architectures are based on IoT which developed with web applications into three kinds of layers as follows.
Multi-layers: In enterprise web services, a multi-layered architecture has been defined that works to popularize application servers, as they interact with data and display layers.
Three layers: This three-layer architecture depends on the clients' processing and decision-making depending on the middle layer, which is called the agent or recipient of customer requests, as is the case in most first web applications that process data and return it to the servers.
Peer to peer: In this architecture, no central or private machine is used to lift heavy loads and smart work, as all decision-making processes and responsibilities are divided among the concerned agencies that do client or server tasks.
Advances in distributed operating systems have positively influenced the field of AI. The performance of AI in IoV applications has largely been improved by using distributed systems and the ability to manage resources in smart ways.
7.2 Intelligent cloud-based computing resource
Cloud computing, enables data access for multiple users over the Internet and acts as a virtual centre for data distribution and sharing. While edge computing is related to the connection for users relatively closed to the cloud [84]. Cloud computing provides higher storage and remotely enables computing resources to the end-users [85]. In IoV, cloud computing enables to provision of services to end-users due to the different architectures of cloud computing, that is, SaaS, PaaS, and IaaS according to the application. Many IoV applications such as web services, multimedia can be provided by the SaaS cloud computing architecture. PaaS and IaaS provide the ability of software frameworks and hardware facilities respectively.
Artificial intelligence works side by side with cloud computing to provide services in an intelligent way that enables network management in a fast and highly efficient manner, thus achieving user satisfaction [86]. Cloud computing with AI allows leveraging cloud services with network function virtualization (NFV), collecting all networking events, and initiating machine learning. Machine learning can be done by integrating with APIs, SDN, multiple database sources, in addition to virtual networks. The use of the machine learning method enables to improvement of the experience of end-users on 5G IoV networks and to maintain the network at an optimal level through the adoption of self-management. The AI-powered NFV network automatically optimizes the network by making use of data analysis. It uses multi-resource AI engines with database validation and ML algorithms [87].
7.3 Intelligent big data resource allocation and process
Recently, the concept of intelligent big data analysis has emerged as a model for using AI in a number of applications. The massive deployment of IIoT and IoV at the 5G platform will enforce the industry to reconsider the big data analytics for this sector. 5G already supports Cloud-radio access network C-RANS, Fog-RANs, and mobile edge computing (MEC); which are the main drive for big data treatment at the 5G network. Also, the existence of technologies like Artificial intelligence (AI), Deep Learning (DL), Machine Learning (ML), and Context-Aware Engine (CAE), will accelerate big data to be one of the facilities that easily can be accommodated within the 5G network especially in SDN/NFV services which involve massive of data storage and exchange. In many IoV applications, the network collects data on a very large scale to enable the provision of vital social and information services. Due to the big data being exchanged across the network, data analysis processes face challenges in several aspects associated with modern data processing and management systems [88]. These challenges are due to the arbitrary complexity of large data structures that cannot be stored efficiently in a relational database. The ability to access widely distributed data in a duplicate and complexity limits the possibility of efficient storage, in addition to the complexity of recovering mechanism to handle the extensive data failure processes.
- Level 1, for command-line tools, API and management console
- Level 2 performs validation, monitoring, deployment, and automation.
- Level 3, as a PaaS to enable messaging, content delivery, and parallel processing services
- Level 4, as an IaaS to enable cloud storage and computing services, in addition to network and database services.
Future implementation of big data in IoV applications will need access to diverse data sources. To achieve this, it can be enabled by creating graphical user interfaces (GUI) to allow users to access data-intensive frameworks [91]. However, due to complex issues related to heterogeneous devices, new APIs should be developed [92]. The ability of artificial intelligence to work well with data analytics is one of the main cause's development of resource management mechanisms for big data. Both machine learning and deep learning analyse all data inputs and use that input to create new bases for future business analytics [93].
8 AI APPLICATIONS IN 5G IOV
There is great competition among vehicle industries companies in the use of different communication technologies and the addition of applications based on artificial intelligence. Recently, a great momentum has appeared by vehicle manufacturers to develop cars that are ready to connect to the surrounding environment, depending on self-control and connection ability to various mobile communication infrastructures [94]. The mobility applications of the 5G network differ in terms of services as it can achieve road-planning processes, in addition to providing emerging autonomous driving services, connected vehicles, and the expanded participation economies of smart transportation.
Intelligent vehicles mobility enables to archive many benefits to IoV applications, that is, efficient traffic balancing, accident prevention, emission reduction, and energy saving. The use of 5G networks in IoV applications leads to support low latency and energy consumption, in addition, to ensuring widespread secure connectivity with high speed [95]. In IoV applications, 5G develops and improves services related to the automotive industry on top of release 13. The improvement falls on several related IoV processes such as autonomous driving and infotainment inside the vehicle. The use of 5G technology helps to make IoV systems more connected and collect more data from different resources, in addition to the possibility of remote diagnosis and insurance on the basis of payment when driving and assisting drivers [96].
8.1 Machine learning for resource management
The concept of connected vehicle networks has become a reality due to the great momentum of using vehicle communication mechanisms, in addition to computing and sensing technologies that play an important role in enabling intelligent transportation systems (ITS) and smart city applications [97]. AI-based IoV enables to achieve easy and fast mobility, safe and comfortable driving, which deeply impacts the society. AI technologies are helping to cobble the use of autonomous driving with the fifth-generation (5G) cellular systems. Moreover, AI helps to enhance the operations related to resource allocation and management in the 5G IoV network to ensure efficient communications with reliable speed.
AI technologies such as machine learning (ML) and reinforcement learning (RL) offer unique solutions to address reliability problems in low-latency vehicle communication networks when applied to 5G and beyond. The lack of resources causes a particularly important problem concerning the radio spectrum, as it affects traditional scheduling decisions for real-time resources with serious delays. Several researchers presented studies that investigate the problem of delay reduction with spectrum and energy constraints in the network of vehicle networks based on the fifth generation with the aim of reducing the delay caused by high load and lack of radio resources. In the study presented by Huang et al. it was found that the use of reinforcement learning (RL) technology helps in developing a scheme for allocating spectrum and managing energy resources [98]. According to the presented study, it was found that this technique gives an appropriate approximation to the learning process in the allocation of energy resources compared to other methods as noted in Figure 8, so that the RL-based vehicle internet system ensures an appropriate delay constraint at high loads.
Machine learning emphasizes the ability to learn and adapt to the changes occurring in the IoV network environment. It provides unique solutions in solving problems related to the multiple sources and the data that is generated and stored in IoV networks, especially those related to network topology, vehicle behaviour patterns, and energy calculations. Moreover, machine learning helps to recognize the dynamics in the environment and extract appropriate features for use in a variety of tasks for communication purposes, as is the case in resource allocation and management processes in the IoV [95].
Different ML techniques can provide a great solution for IoV over 5G networks. In supervised learning techniques, the estimation and prediction of unknown parameters are performed by prior models. The most common supervised learning algorithms are decision tree (DT), support vector machine (SVM), and random forest. These algorithms provide solutions for spectrum sensing, channel estimation, and localization problems. The challenges related to IoV network load balancing, clustering issues, and user association can be enhanced by the unsupervised learning algorithms [99].
The dynamism of the IoV environment is one of the issues in the stability of the IoV network and its impact on various QoS requirements. The use of reinforcement learning algorithms interacting with the dynamic environment leads to satisfactory policies for both QoS and QoE. Reinforcement learning (RL) techniques enable interaction with the environment without the need for training data sets and agent learning as in issues related to resource allocation. It is possible to determine the optimal policies first and accordingly, vehicle dealerships allocate channels depending on the variables that affect the IoV environment. The changing environments are characterized by several factors including link conditions, locally perceived interference, and kinetics that are all mathematically modelled to track these dynamic changes [100].
Q-Learning is a widely RL technology used. IoV sensors can use Q-learning for decision-making and deduction in circumstances of unknown network conditions, and that dynamically changes. Q-learning-based IoV enables to share spectrum in cognitive radio networks with the existence of unknown channel and unknown resource conditions for small-cell networks [101]. Moreover, it can improve BS performance, especially with unknown power and load condition. However, it suffers from a low convergence rate in terms of optimum policy learning in the state of the system, which creates challenges for its use in dynamic IoV networks.
Machine learning can be used on the increasing deployment of 5G networks and services dramatically in data traffic, storage, and processing, taking into account smartphones as gateways for remote access to resources through cloud computing. However, there are many challenges that should be addressed in cloud computing and integration to 5G protocols [102]. One of these challenges is related to cloud computing resource allocation and management, which is required to design ML-based management solutions.
The most major resource allocation and management problem has existed in cloud computing data centres. In cloud computing, virtual machine mode (VMP) enables the host of the selected virtual machines (VMs) in every available physical device (PM) of the cloud computing infrastructure. The challenge, therefore, lies within the problems facing VMP, which are considered in the specific context of 5G services, and ML techniques are mainly considered to address the relevant decision-making process regarding cloud computing infrastructure operations [103]. Moreover, the challenges of managing a network based on software defined networking (SDN) are also considered a perspective of VMP problems, as ML technologies may lead to a promising approach to support these types of operational decisions.
8.2 Swarm intelligent for resource management
Swarm intelligence (SI) is a mass behaviour of self-organizing decentralized systems, whether natural or artificial, which is used as an artificial intelligence methodology [104]. Swarm intelligence systems usually consist of a set of simple agents that interact locally with each other and with their environment. Agents follow very simple rules with no central control structure that determines how individuals behave [105]. Figure 9 illustrates the various benefits that can be obtained when applying swarm principles in the context of problem prediction.
The Swarm intelligence approach was developed to solve problems related to group behaviour by using algorithms based on the organizational behaviour of social insects. By studying the nature of insect colonies, it was found that they have a decentralized social system consisting of independent units in their environment, which depends on the concept of stimulus behaviour and the probability of response.
The rules governing insect interactions are applied on the basis of local information without the need for knowledge of global patterns. Organization at the colony level arises from the interactions that take place between individuals who exhibit these simple behaviours [106]. A set of dynamic mechanisms works to achieve self-organization, which shows the structures at the global level of the system from the interactions between its lower-level components without any coding at the individual level.
In the context of the V2X model, swarm intelligence is demonstrated by a set of vehicle factors, which are interacted locally with each other and with their environment. In IoV networks, vehicles follow simple rules that can be combined in one of the methods of intelligent colonial behaviour, such as colonies of ants, birds, and fishes, all of which generally depend on the concept of swarm intelligence [107]. The most widely use cases for swarm intelligence are, particle swarm optimization (PSO), ant colony improvement (ACO), and swarm casting. IoV based on 5G interfaces gains additional advantages over traditional IEEE802.11p V2V and V2X environments. 5G access interface consists of limited resources and is preferred for efficient use of resources.
In addition, it has sufficient bandwidth for any type of service, but it is preferred to use it as a backup infrastructure because it contains a limited number of channels [108]. In a heterogeneous network environment, communication between different devices takes place through several types of services, which are preferably through the application of clustering for strong communication [109]. In systems based on 5G innovation, a 5G client area server associated with the eNB get to the interface is utilized to perform portability administration operations. This organized foundation can be utilized as a framework that performs operations to find all vehicles associated with the 5G arrange. By implying the clustering technique, the problem can be resolved. The CHs forward the data to diminish the over-burden of the broadcast; this implies that CHs can trade data between the other bunches within the network [110]. Optimized swarm intelligent-based ML algorithms provide an efficient CHs selection mechanism.
8.3 Other AI technologies for intelligent resource management
Many IoV applications require flexible and intelligent mechanisms for resources management. For example, in smart transport systems, applications related to information services, collision avoidance, and road safety require sharing information through frequent access to Internet servers with a large data transfer rate [111]. These applications need to support the exchange of critical safety information via V2V links with the requirements of ultra-reliable low-latency (URLLC) communication. There are several approaches that have been studied recently to allocate and manage other resources in IoV telecommunications networks to archive the requirements of quality of service (QoS) and quality of experience (QoE), and to effectively improve network capacity [112].
Google designed cube meter technology as an open-source container manager. Kubemeter project as an open-source, enables to improve the community and sharing machines between applications. The sharing machines require ensuring that two applications do not try to use the same ports [113]. Dynamic port allocation brings a lot of complications to resource management systems. Kubemeter enables to provide container-centric infrastructure, in addition to scaling application across clusters of servers. Kubemeter approach consists of many components as shown in Figure 9. Containers perform the operations related to creating isolation boundaries at the application level for packaging portability. The dock units in Figures 10 and Figure 11 represent a runtime technique for managing containers to be run successfully. It enables the container to run a software application that is isolated from the host machine [114]. Pods are deployed in the machine known as Kubernetes node; where the nodes are enabled to operate the necessary services to run application containers. Nodes are managed by one or multiple kubemeter masters. The pod consists of the single container or a few numbers of containers, which are tightly coupled and share resources [115]. As shown in Figure 9, Kubemeter cluster many has one or more masters and nodes. It achieves high infrastructure availability when it has multiple masters. The kubemeter master provides a front end for the common cluster state through which all other components interact [116].
The architecture of the efficient transfer actor-critic learning (ETAC) approach is the approach proposed by Helin Yang et al. The authors show that ETAC enhances the learning efficiency, and improves the convergence speed of learning. This approach supports reliable sensitive services in IoV networks. ETAC is considered as an enhanced transfer actor-critic (AC) RL mechanism to provide intelligent resource management for IoV networks [117]. It enables to increase the network capacity and ensure URLLC requirements for V2V links. ETAC can effectively achieve a means of smart exploit the stochastic actions related to channel information, transmit power, and other IoV network stability considerations.
In V2X IoV communications, ETAC learning approaches enable to development of an intelligent resource management framework to intelligently make the decision. The process of ETAC is based on the Markov decision process (MDP) RL acts as a reward function and AC learning to describe the intelligent resource management issues [118]. Both AC learning and MDP work together to enable suitable solutions to optimize the policies for IoV networks resource management. AC enables to a selection of the commutation mode, assigned resource blocks (RBs) in a smart manner, in addition, to efficiently managing the power by using the gradient method to optimize the policy and the properties of computational cost and fast convergence [119]. In the ETAC approaches, the actor and critic parts being updated with the temporal difference (TD) error value causes the AC learning by the policy gradient method as shown in Figure 10. The TD blunder computations take put by the AC RL to discover the blunder between the assessed esteem and the genuine esteem [120]. In the case of dynamic IoVs, ETAC enhances the convergence speed and improve the learning efficiency toward the optimal policy in IoV networks.
9 FUTURE DIRECTIONS
5G IoV communication frameworks showed up with the most recent era of the fifth wireless generation. The rapid development of various modern mobile phone applications, particularly those supported by AI technology, has called for building smart vehicle networks as it opens new horizons shortly [121]. The improvement in 5G communications has persuaded the industry and the scholarly world to begin conceptualizing the following era of remote communication frameworks (6G) pointed at giving communication administrations for end of the requests.
9.1 Tends on 6G IoV ML and AI
Machine learning (ML) with artificial intelligence and deep neural networks (DNN) is making a major impact on the IoV revolution including 6G communications [122]. 6G communication link and system solutions are constructed using artificial intelligence and machine learning. AI-powered 6th era innovation is anticipated to grow numerous highlights counting self-configuration, self-connectivity, smart systems and intelligent configuration [123]. The capabilities of remote signals and the greatest recognition of brilliant radio transmission can be completely realized through AI-powered 6G with the assistance of ML calculations. In addition, it is expected to build the hardware basics for wireless equipment through reconfigurable and smart materials [124]. Rundowns of ML device-to-device (D2D) communication, gigantic MIMP optimization, and heterogeneous organize plan are given in [125-135]. Counterfeit insights can move forward to organize execution based on its solid learning and capacity to think [133-136].
The complete information and the complex network models lead to challenges to the learning and preparing that underpins AI. In expansion, constrained computing assets may be deficient to handle huge, high-dimensional information to meet the preparing precision rate. In addition, deep learning is for the most part characterized by a tall degree of computational complexity, which is very costly [127, 137, 138]. Planning a successful AI learning plan to progress both computation proficiency and precision postures is a major investigation challenge. More as of late, remaining organizing, illustrations preparing, including interference, mobility, synchronization, coordinating, and offline preparing are promising strategies with high-performance computing offices to energize merging speed, diminish complex computations, and progress preparing exactness [136-142].
9.2 Challenges of IoUAV
The vehicle systems that back IoUAVs appear high flow in a few angles, such as joins with BS, remote channels, arrange topologies and versatility elements. In specific, gadgets or terminals that connect or take-off systems may have diverse necessities for QoS and QoE. All these vulnerabilities said in dynamic systems require consistent overhauls of the parameters of AI learning calculations. The extraordinary quality, versatility, and adaptability of learning systems are basic perspectives to back a possibly unlimited number of connection substances and to supply high-quality administrations in energetic real-world systems. Subsequently, how to plan vigorous, adaptable, and flexible instructive systems for 6G systems remains an open issue [128, 139, 140].
9.3 IoV on release 16 and beyond
The IoV is one of 3GPP focus areas which has newly addressed it in Rel. 16-3GPP standard which comprises for the first time the idea of V2X standard. The new standard mainly uses the NR of 5G air interfaces. Future IoV2X should introduce cloud-based functionalities to provision internet with automated (autonomous) and connected vehicles (CAV) scenarios with rigorous requirements.
One of the topics that need to be addressed is the side-link (SL) IoV2X to device-to-device technology, where a direct link to be initiated between nodes without the help of the BSs. The SL concept is also addressed in V2X in LTE standard in Rel. 13; however, NR-V2X-SL is more advanced and based on NR rather than LTE-MTC. LTE-MTC (machines type communications) a.k.a. LTE-M, is a V2X technology that introduces over evolved-UMTS terrestrial radio access network (EUTRAN).
For NR-IoV2XSL, a few things need to be revisited like spectrum allocations, especially that associated with cognitive radio network (CRN). Also, quality-of-service (QoS), the improvement of mobility managements (MMs), and the seamless existing techniques for various V2X, that is, NR-SL, EUTRAN-SL and 802.11p are known as WAVE—unlicensed wireless accesses for vehicles environments.
9.4 Connected and automated 5G V2X
In 5G, automated and connected vehicle would be able to reach the cloud through C-RAN which can provide cloud-based ML/AI to take self-drive decision. AI technology is a key drive to provide concrete model for the challenged vehicle automation issues.
Edge computing and 5G can play great role to fastening and smartening data decryption in high mobility scenarios. In these scenarios, analysis and processing the data using edge computing are automatically running in the closest servers to vehicles. This may dramatically reduce data traffic in V2X network, relaxing the network limitations for handling fasters data rates, and reduce the transmissions cost. In these regards, issues related to cloud and data centres need to be resolved. There are a few domains need the researchers to focus in, that is,, network slicing or NaaS, software and infrastructure managements issues, that is, DevOps, and SON-based AI, where self-organizing network (SON) is one of the key and driving factors for V2X automation technique.
10 CONCLUSION
Due to the rapid increase in the use of 5G technologies, an investigation of 5G resource allocation and management strategies are important to achieve a high level of network responsibility and efficiency. In-vehicle communication networks, resource allocation and management are some of the most important challenges facing the performance of the network. The challenges are represented by the big data and huge information that is to be exchanged across the network to and from a different number of sources. Recently, data-based enabling machine learning (ML) approaches were developed to obtain optimum performance with affordable computational complexity.
Enhanced learning (RL) for the future smart road is one of the most important AI tools used in managing vehicle networks based on 5G networks. This paper has reviewed different AI algorithms, that is, reinforcement and deep learning mechanisms within smart approaches such as swarm Intelligence, Kubemeter scheme, transfer actor-critic (AC), and efficient transfer actor-critic learning (ETAC), which are considered to allocate radio resources in a 5G IoV network.
This paper touched on studying most of the challenges facing IoV applications dependent on the fifth-generation networks, which are related to cloud computing, and resource allocation and management. During this paper, several previous studies related to the fifth-generation vehicle network architecture have been reviewed, in addition to machine learning applications to improve IoV resource allocation and management. The paper also reviewed the intelligence applications that could revolutionize the vehicle of the Internet using the sixth-generation networks soon.
ACKNOWLEDGEMENT
This work has been conducted and supported by Faculty of Information Science and Technology, Universiti Kebangsaan Malaysia (UKM) under GGPM2020-028, Sudan University of Science and Technology, Sudan, Red Sea University Sudan, and Taif University, Saudi Arabia.
CONFLICT OF INTERESTS
The authors declare that they have no competing interests.
Open Research
DATA AVAILABILITY STATEMENT
Data used to support the findings of this study are already available in the manuscript.