Cloud Computing in VANETs: Architecture, Taxonomy, and Challenges

ABSTRACT Cloud computing in VANETs (CC-V) has been investigated into two major themes of research including vehicular cloud computing (VCC) and vehicle using cloud (VuC). VCC is the realization of autonomous cloud among vehicles to share their abundant resources. VuC is the efficient usage of conventional cloud by on-road vehicles via a reliable Internet connection. Recently, a number of advancements have been made to address the issues and challenges in VCC and VuC. This paper qualitatively reviews CC-V with the emphasis on layered architecture, network component, taxonomy, and future challenges. Specifically, a four-layered architecture for CC-V is proposed including perception, coordination, artificial intelligence and smart application layers. Three network components of CC-V, namely, vehicle, connection, and computation are explored with their cooperative roles. A taxonomy for CC-V is presented considering major themes of research in the area including design of architecture, data dissemination, security, and applications. Related literature on each theme is critically investigated with comparative assessment of recent advances. Finally, some open research challenges are identified as future issues. The challenges are the outcome of the critical and qualitative assessment of the literature on CC-V.


INTRODUCTION
Recently, cloud computing has witnessed significant attention in vehicular communication. It is because of the realization of smart intelligent transport system (ITS) applications, and architectural similarity between mobile cloud computing (MCC) and vehicular ad hoc networks (VANETs) [1][2][3]. Cloud computing in VANETs (CC-V) has been investigated in two major dimensions of research in the area. First, vehicular cloud computing (VCC) has been explored. In VCC, a group of connected vehicles forms a cloud with the aim of sharing their abundant resources.
VCC increases resource utilization in vehicular communication [4]. The abundant resources include storage, computing capability, sensing, and communication capability. The resources became available for sharing, when several vehicles assemble at the places including parking lots, garage, canteens, highway hold-up, and traffic lights. At these places, the idea of VCC has been realized. VCC dynamically allocates abundant resources to authorized users. Second, vehicle using cloud (VuC) has been explored in which the connected vehicles are considered. VuC enables the access to the services of conventional cloud at vehicles via Internet. VuC helps in realizing smart ITS applications including real-time traffic prediction and web-based services [5].
The investigations are in initial stages in both VCC and VuC. There are opportunities and challenges considering a new area for cloud computing. In fact, huge resources of vehicles are underutilized at various places (where a group of vehicles assemble), which is needed to be explored and tap into [6]. The framework of CC-V improves the usability, reliability, and efficiency of ITS applications [7]. Subsequently, it enhances safety in transportation, reduces traffic congestion, decreases air pollution, and augments comfort in driving [8].
Some of the cloud computing services realized in CC-V include network as a service (NaaS), storage as a service (STaaS), cooperation as a service (CaaS), computing as a service (COaaS), and sensing as a service (SEaaS) [9][10][11]. Since vehicles are considered to be able to connect to the Internet, NaaS can be used by passengers to connect to the Internet in VuC. Applications and data of vehicles, which require more memory for storage, can utilize STaaS of VCC [12]. CaaS can be utilized to share traffic information among vehicles in case of accident, and road maneuverer. Vehicles can realize smart ITS applications using COaaS to fulfill higher computation power requirement of smart applications. SEaaS can be utilized for monitoring real-time condition of car as well as driver's behaviour [13]. Whaiduzzaman [14] have explored VCC by presenting a taxonomy based on strategic management, security and privacy, cloud formations, inter-cloud communication, and applications issues. An architecture for VCC, comparison with cloud computing, and open research issues have been presented. However, the architecture has not been explicitly defined considering layer-wise function, representation, and protocol. The network components related to VCC are also not explored.
Although various contributions have been made for realizing these cloud services in CC-V, yet there are a number of issues that need to be addressed in near future. In this paper, a qualitative review of CC-Vs has been presented. The review focuses on layered architecture, network component of CC-V, taxonomy of recent advances, and open challenges and issues for future direction of investigation in the area. A four-layered architecture for CC-V is designed including perception, coordination, artificial intelligence, and smart application layers. The representative components, responsibilities, and protocols of each layer are described. The three network components of CC-V, namely, vehicle, communication, and computation have been explored with their cooperative roles. From the best of our knowledge, there is no layered architecture available for CC-V. The major network components with their cooperative roles have not been explored previously. The taxonomy of CC-V is presented considering four major issues including design of architecture, data dissemination, security, and application development. Each issue has been qualitatively and critically reviewed with comparative assessments of recent advancements. Finally, some future challenges are identified according to the qualitative and critical review of related literature.
The rest of this paper is organized as follows. Section 2 presents four-layered architecture of CC-V. Section 3 describes three network components of CC-V. Section 4 presents a taxonomy of CC-V with qualitative and critical review of related literature. Some future challenges in CC-V are identified in Section 5, followed by conclusion made in Section 6.

LAYERED ARCHITECTURE
In this section, a four-layered architecture for CC-V is designed in terms of representation, functionalities, and protocol perspective. In a broader operational and structural point of view, CC-V has four major components, namely, perception physical devices, coordination, artificial intelligence, and smart application services. Therefore, the architecture consists of four layers including perception, coordination, artificial intelligence, and smart-application (see Figure 1). The perception layer is correlated with physical devices. The coordination layer is linked with communication networks. The artificial intelligence layer is associated with computation, and the smart application layer is correlated with the services. The layers effectively divide the functionalities of CC-V into four groups. The layers are unambiguously differentiable in terms of both functions and representations. Each layer is described below considering representative components and functionalities.

Perception Layer
The perception layer of the architecture represents invehicle devices including sensors, actuators, display unit, smart mobile devices, global positioning system (GPS) receiver, actuators, etc. The two main functions of the layer include gathering traffic data through sensing and delivering information at end level. The layer defines vehicle-to-physical world interactions through smart devices. Apart from the two functionalities, a number of services are provided by the perception layer to the next layer as interface services. The interface services include error detection and correction on sensed data, result verification on inferred information, etc. In terms of protocol perspective, physical layer part of the protocols including IEEE 802.11p (PHY-802.11p) of WAVE [15], 802.11a/b/g of WLAN [16], Wi-Max [17], and 4G/LTE [18] are considered. This is due to the consideration of heterogeneous in-vehicle devices containing different types of sensors.

Coordination Layer
The coordination layer represents network devices and internetworking technologies. The network coordination is significant due to the consideration of different types of networks including VANETs, Wi-Fi, and 4G/LTE. The main functionalities of the layer include transfer of data packet and efficient handoff among these networks. The heterogeneous network architecture makes both the tasks quite challenging. The other functionalities of the layer include data dissemination in heterogeneous networks, location-based networking support, network level authentication, and authorization.
In terms of protocol perspective, two sub-layers are considered. In the first sub-layer, medium access control (MAC) layer part of the protocols including IEEE 1609.4 [19] in WAVE, 802.11p, and LLC are considered. In the second sub-layer, network and transport protocols including fast application and communication enabler (FAST) [20][21][22][23] in CALM, car-to-car (C2C-net) [24] and short message protocol (SMP) [19] in WAVE are considered apart from traditional IP and TCP/UDP combinations.

Artificial Intelligence Layer
The artificial intelligence layer represents cloud-based computing infrastructure. It includes data centres, and servers of conventional cloud in case of VuC. The computing resources of vehicles are also included in case of VCC. The functionalities of the layer include big data computation, analysis, and inferring intelligent decisions for real-time applications. Apart from cloud computing and decision-making, the other operations of the layer include mining traffic data for new business models, ensuring computing efficiency, scheduling and paralleling computation, etc. The operations of the layer are very much similar to the operations of cloud-based services. The layer significantly enhances the computing capability of vehicles through VuC architecture. It also improves the utilization of resources of vehicles through VCC architecture.
In terms of protocol perspective, the protocols related to service, big data analysis (BDA) and VCC are considered. The protocols include IEEE 1609.6 of WAVE, COaaS, STaaS, picture as a service (PICaaS), infrastructure as a service (INaaS), CaaS, NaaS, and gateway as a service (GaaS). The design of protocols for BDA and VCC is an open research theme in CC-V due to the growing volume of traffic data.

Smart Application Layer
The smart application layer represents smart cloudbased applications of three categories including safety, efficiency, and infotainment. The layer is responsible for delivering end-user smart services. The layer operations include application management, service management, application-and service-based data management, application-based authentication, and authorization. Although the applications for safety and efficiency are also implemented in VANETs, yet the cloud-based operation significantly enhances the intelligence, and usefulness of these applications. This is due to the advancement of computing power in applications operating through cloud support, for computation and traffic data. Some of the smart applications include real-time traffic forecasting, smart toll collection through car payment, smart challan for traffic rule violation, ad hoc multimedia sharing, smart black-box, and smart emergency call.
In terms of protocols perspective, resource handler protocol IEEE 1609.1 of WAVE is considered for efficient resource management among smart applications. Apart from this, the business models related to advertisement, sale, service, and insurance are considered. The protocol development to support traffic data-based business models is an open research theme in CC-V.

ELEMENT
In this section, three major elements which formed CC-V are explored focusing on networking aspect of the proposed layered architecture (see Figure 2). In a broader physical structural point of view, the components including vehicle, connection, and computation represent the networking aspects of CC-V. The vehicles are end-users, and, thus, the services are delivered at vehicles. This component is closely related to the perception layer of the proposed architecture. This is due to different types of sensors for monitoring speed, direction, and position which are considered to be attached to the vehicles. The second component, namely, connection includes network or communication devices of the heterogeneous networks. A heterogeneous network including VANETs, Wi-Fi, and 4G/LTE is primarily considered. This component is closely related with the coordination layer of the proposed architecture. The last component named computation refers to the computing and storage devices, responsible for the efficient processing of big traffic data, and inferring intelligent decisions. This last component is closely related to the artificial intelligence layer of the proposed layered architecture. This is due to the related functionalities of the layer. The relationship among these network components is potential due to the collective operation requirement for the services offered by the architecture. The vehicle-to-connection relationship determines the efficiency of delivery and acceptance of the services provided. The connection and computation determine the quality of information of the services. Each element and their role in CC-V are defined below.

Vehicle
The vehicles have several in-built technologies and resources, which are underutilized in traditional vehicular communication architecture. The technologies and resources include telematics wire devices (TWDs), GPS, sensors, dedicated short-range communication (DSRC), actuators, computer, and smartphones. The utilization of these resources can be improved through a cooperative collaborative network architecture. The underutilization of the resources motivates the concept of CC-V. This is due to the fact that the vehicles do not have some of the challenges of mobile devices in MCC, such as lower computing capability, smaller memory size, and battery life. The resources of vehicles can be shared under the framework of VCC at many places while travelling. It includes parking lot, garage, and traffic light. In case of extensive computing need for longer duration, vehicle's in-built technologies can be utilized for establishing durable connection to a conventional cloud. The communication is known as VuC. For both the cases, VCC and VuC, vehicles equipped with modern communication technologies are one of the most important constituents.

Connection
The connection refers to the network or communication devices of the heterogeneous networks architecture considered in CC-V. The network devices are utilized to establish reliable communication between vehicles and cloud. In the case of VuC, the connection is either a direct communication between vehicles and cloud infrastructure, or a multi-hop communication using road side units (RSUs) with vehicles. In the case of VCC, connection is mostly direct communication between a vehicle and vehicular cloud infrastructure. The connection also defines services level agreement (SLA) between a vehicular client and cloud infrastructure. It is due to the fact that SLA represents the level of Quality of Service (QoS) of an application based on the architecture of CC-V. Therefore, connection determines delivery and acceptances of the services.

Computation
The computation refers to the computing and storage devices where big traffic data are processed and intelligent decisions are inferred. The computation is broadly divided into three models including computation provider, computation consumer, and hybrid. In computation provider model, the vehicles are required to register with service provider for availing their vehicular resources to cloud computing architecture using SLA. In computation consumer model, vehicles need to register for utilizing the cloud computing services as their own resources. In the case of hybrid model, vehicles need to register for both as a provider and as a consumer. The provider model would generate revenue by improving resources utilization. The consumer model would enhance computing capability of vehicles by utilizing cloud infrastructure. The hybrid model would be more complex due to the need for maintaining provider or consumer state for each vehicle.

CLOUD COMPUTING IN VANETs
In this section, CC-V is qualitatively reviewed on the basis of a taxonomy depicted in Figure 3. Cloud computing is an emerging research theme in VANETs. It is evolving due to the growing need of higher computing capability at vehicles to expand the various commercial services currently available in the Internet to VANETs. CC-V has four major issues, namely, design of architecture, data dissemination, security, and applications. Each issue has been investigated in several directions which are qualitatively and critically reviewed in following sections.

Design of Architecture
In this section, related literature on designing architecture to access cloud computing services in VANETs is reviewed. The challenges of lower utilization of VANETs resources and unstandardized architecture of CC-V are the main focus in the literature. The design of architecture has been divided into three categories including services, computation, and communication. Architectures focusing on how to utilize vehicular resources as cloud resources have been categorized in services. Architectures focusing on how to utilize traditional cloud computing services in vehicles, and how to improve communication between traditional cloud and vehicular cloud, are categorized in computation and communication, respectively.

Service
In [25], a concept that merges VANETs with cloud computing (MVCC) has been suggested to address underutilization of vehicles' on-board devices and lack of standard architecture for VCC. The on-board devices have the capabilities including computation, communication, and storage of information. An architecture for cloud computing in vehicular communication has been defined and further divided into three architectural frameworks including VCC, VuC, and hybrid vehicular cloud (see Figure 4). In addition, security and privacy issues related to VCC are outlined. However, elements of the cloud computing in vehicular communication are not explicitly defined. The protocols and functions of the architecture are not considered in terms of applications, network coordination, and intelligence layer of the proposed CC-V-layered architecture.
The paradigm shift from VANETs to VANET-based clouds (VNVC) [26] is an extension of the work in MVCC. The VNCN proposed an architecture including communication paradigm, cloud services, computation, and traffic information dissemination through cloud architecture. The major contribution of VNCN is traffic information as a service. It handles complex traffic information computations using cloud. It also provides services including big traffic data analysis, remote configuration, car performance checking, smart  location-based advertisements, and vehicle witnesses; its application can be related to smart application and artificial intelligence layer of the proposed CC-V layered architecture. Figure 5 shows a network model for the traffic information as a service. In traffic information dissemination through cloud, moving vehicles serve as cooperative forwarder to send coarse-grained information to the cloud, and to receive fined-grained information from the cloud. The functional modules including cloud processing module (CPM), cloud knowledge base (CKB), and cloud decision module (CDM) and authenticators have been considered at the cloud layer, which are also applicable to artificial intelligence layer of the CC-V-layered architecture. However, some performance metrics have been measured, yet the distance between communicating vehicles and infrastructure, and the delay in connectivity, are not considered.
Another idea of taking VANETs to the clouds (TVC) has been suggested by Olariu [27] to address the problem of standardization of VANETs to cloud integration architecture. Furthermore, privacy and security issues related to integration of VANETs to the cloud have been outlined. The architecture includes two different services, namely, NaaS and STaaS. The cloud computing vehicular communication scenarios for the TVC have been depicted in Figure 6. It shows the NaaS and STaaS concepts in vehicular environment. From the authors' deduction, NaaS is more suitable for cloud computing in vehicular communication due to the consideration of 3G-or Wi-Fi-based Internet connection at vehicles. Therefore, Internet connections can be used by passengers on-board to surf Internet. SaaS is also applicable in cloud computing over vehicular communication, since vehicles are considered interconnected to form local clouds. These vehicles can share their memory, and processor for storage, and computation of large data. The implementation of services of traditional cloud computing has been theoretically presented for the vehicular cloud. Thus, the investigation can be considered very close to the functions considered under artificial intelligence and smart application layers. However, the technical details of the implementations have not been presented.

Computation
In [28], a generic cloud computing model (GCCM) for VANETs has been suggested to enhance the availability of on-board system for other client vehicles. VANETs cloud model has been unveiled which consists of two concepts including permanent and temporary cloud layer model. The permanent cloud layer serves as conventional cloud, and the temporary cloud serves as VCC. The permanent cloud layer is the same as the tradition cloud which handles related functions of smart applications and artificial intelligence layer of the CC-Vlayered architecture, whereas the temporary cloud can be related to the perception layer in the proposed CC-Vlayered architecture. The infrastructure components are grouped into three layers including client, communication, and cloud. The novel intelligent transportation system applications supported by VANETs cloud include business and research applications, vehicular software, web services and processing cloud backup, and safety applications. However, protocols and functions of the architecture are not defined.
Yan [29] proposed a vehicular cyber-physical systems (VCPS) based on mobile integrated architecture (VMIA). The system addresses the increasing demand from MCC users to access VCC services. VMIA consists of the conceptual architecture for VCPS with MCC  capabilities (see Figure 7). VMIA is the integration of VCPS and MCC to provide mobility support to users. Cloud-supported components are grouped into trafficaware mobile geographic information system and dynamic vehicle routing algorithm. In order to provide driving assistance, a paradigm called traffic-aware mobile geographical information system with traffic cloud support system has been discussed. The functions of mobile geographical information system can be used for traffic cloud support by incorporating traffic dynamics with base map management. A decentralized and proactive dynamic vehicle routing algorithm has been developed to enable drivers to self-organize the traffic and shift the system state from either dynamic all-ornothing or dynamic user equilibrium to dynamic system optimal. VCPS based on MCC support architecture have been explained. The CC-V elements are related to VMIA in terms of mobile phone devices which are situated in the vehicle, the communication between mobile device and the cloud is aided by connection, and the computation is carried out at the MCC layer. The VMIA findings can be considered very close to the functions of smart applications and artificial intelligence layer of the CC-V-layered architecture. Nevertheless, frequent intermittent connection might arise due to high mobility of vehicles. Security-related issues and implementation of the architecture have not been dealt with.

Communication
In [30], an idea of whether to migrate or not has been investigated, by exploring virtual machine migration in roadside cloudlet-based vehicular cloud (M-EVMVC).
The concepts are to address the challenges of sharing resources with high-mobility vehicles. The roadside cloudlet-based vehicular cloud architecture and twophase polynomial heuristic algorithm have been presented. The problem has been formulated as a mixed-integer quadratic programming problem. The programming problem is based on four constrains including path selection, virtual machine placement, resource capacity, and link capacity. The virtual machine placement issue has been addressed using static offline placement approach. Virtual machine migration and system model of cloud computing in vehicular communication has been explored. The performance of the two algorithms has been evaluated to test the effect of cost on the density and resource requirement of virtual machines, and roadside cloudlets, and traffic rate requirement of virtual machine. Although implementation has been carried out, yet well-known network simulator has not been used for implementing the architecture in order to evaluate its effectiveness. The virtual machine migration is closely related to the mentioned CC-V-layered architecture in terms of artificial intelligence and perception layer, since it deals with migration technology of traditional cloud.
Vehicular cloud networking (VCN) and design principles have been suggested to address the need for intelligent computation for safety and comfort applications in vehicular environment [31]. The VCN architectural formation can be seen in relation to smart applications and artificial intelligence layer of the CC-V-layered architecture. VCN architecture and design principles have been discussed. The VCN architecture is constituted based on cloud computing for vehicular communication and information-centric networking for handling safety, comfort, and privacy. In VCN routing, it does not need to know who sent the information. Furthermore, new model for application and networking has been discussed. However, the elements are not clearly defined in terms of the CC-V. Also, functions and protocols are not considered.

Comparative Discussion on Design of Architecture
The aforementioned literature review on design of architecture is summarized in Table 1. The summary is based on the parameters including contribution, type of architecture, technique, implementation, and remarks. The contribution points out the progressive impact of the articles on the research theme of architecture design for CC-V. The architecture determines the category from the three types including VuC, VCC, and hybrid vehicular cloud (HVC). The techniques identify the approach followed for addressing the raised issue. The implementation shows implementation tools and performance metric. The critical remarks have been also made. A comparison is also presented in Table 2. The comparison is based on three parameters including technique, architecture, and implementation. The technique is defined using conceptual model, complete framework, and algorithm. The architecture is defined using VuC, VCC, and HVC. The summary and comparative assessment attests that M-EVMVC [30] has more viable architecture than other proposed architectures when related to the proposed CC-V-layered architecture, CC-V elements, better algorithm with complexity analysis, and implementation proof with a wide range of performance metrics

Data Dissemination
In CC-V, clustering is the preferred option for data dissemination among vehicles. This is due to the higher possibility of cloud-based resource sharing while disseminating data in case of clustering. In this section, related literature on designing clustering schemes for transmitting data in CC-V is critically explored and comparatively assessed.

Distributed Clustering
A distributed multi-hop clustering algorithm based on neighborhood follow (DMCNF) has been presented to enhance robustness in clustering algorithms [32]. The DMCNF basically focuses on how the vehicular cloud nodes communicate among each other to improve efficiency. Hence, its functional suitability is in coordination and artificial intelligence layer of the proposed CC-V-layered architecture. It is a routing scheme that uses proactive and shortest path methods. An algorithm based on one-hop neighborhood follow strategy has been suggested. The algorithm considers three factors for choosing a follow vehicle including relative mobility, current number of follows, and history of cluster. In proactive clustering scheme, high signalling load overhead might occur due to the frequent update of its follow information, and dynamic change of node's state. In this work, quality of network (QoN) at each node has not been considered during the selection of follow node. The aforementioned elements of CC-V is the constituent of the DMCNF, but has not been defined in this study.
A cluster-based vehicular cloud system with learningbased resource management (COHORT) has been presented. COHORT deals the challenges faced during deployment of new applications and advancement of ITS services [33]. An example of cluster-based VCC system and case scenario for resource management issue is depicted in Figure 8. A VCC system, q-learning technique, and queuing strategies have been discussed. VCC system has been designed to conform to clustering procedures. It then further presents the resource limitation difficulties, by grouping vehicles and cooperatively @ @ @ @ VNVC [26] @ @ @ @ TVC [27] @ @ GCCM [28] @ @ VMIA [29] @ @ M-EVMVC [30] @ @ @ @ VCN [31] @ @ CM, conceptual model; CF, complete framework; AL, algorithm; IM, implementation. @ = Yes.  Kumar [34] has suggested an optimized clustering for data forwarding using stochastic coalition game in VCPS (OC-DSG) to address the issue of lesser contact time for vehicles with access points. A stochastic coalition game for an optimized clustering and an algorithm for data dissemination in VCPS environment have been developed. Stochastic coalition game is employed as a selection strategy in VCPS. The vehicles are represented as players in the coalition game. A vehicle accesses a fixed number of resources from the cloud. Learning automata techniques are utilized in vehicles to gather and process information from the surrounding based on pre-stated policies. The optimization clustering scheme is typically related to coordination layer of the CC-Vlayered architecture, since it deals with communication between clustered node and RSUs. However, the authors have claimed, through the evaluation of some performance metrics, that the scheme outperformed the existing state-of-the-art schemes. Implementation of the CC-V-layered architecture will enhance performance of the clustering techniques employed.
A Bayesian coalition game for contention-aware reliable data (CARD) forwarding in vehicular mobile cloud addresses the issue of performance degradation due to the unicast sender-based data forwarding [35]. The problem of reliable data forwarding is formulated as Bayesian coalition game using an adaptive learning automata concept. An adaptive learning automata-based contention aware data forwarding algorithms for critical applications in the vehicular mobile cloud has been developed. The approach is based on coordination and artificial intelligence layer of the proposed CC-V-layered architecture. The vehicles represent the players in the game. These players are used for taking adaptive decisions with regard to effective and reliable data forwarding. Each player monitors the moves of the other players in the game for reliable data forwarding. Additionally, security issues have not been considered. The implementation environment has not been discussed.
In [36], a replication-aware data dissemination (RADD) is presented for VANETs using location prediction to address the challenges of disconnection due to high node mobility. The new replication-aware scheme has been suggested to estimate the location of nodes. An algorithm for position estimation, accessing, and routing messages from remote vehicles to the destination has been developed. The bloom filters are used for searching suitable vehicles for replica assignment. It makes searching faster and improves the total performance of RADD. Additionally, radio frequency identification (RFID) tags are employed on the vehicles and RSUs serve as RFID reader to gather data from these tags. The tags data serve as location information for short-range communication in case of global positioning system failure. The RADD is closely related to the coordination and perception layer of the CC-V-layered architecture because it deals with connectivity of vehicular nodes. The security issues of RADD have not been explored.
Enabling cooperative relaying in VANETs cloud over LTE-A networks has addressed connectivity and device heterogeneity issues in highly populated urban area [37]. A cooperative vehicular relaying transmission scheme has been designed. The scheme contributes towards the formation of an advanced heterogeneous telecommunication network. It provides increased networking capabilities for heavily populated urban areas. This scheme made use of vehicles equipped with low-elevation antennas and short-and medium-range wireless communication technologies. The authors claimed that reasonable diversity gains and minimized error rate were achievable. Furthermore, there is a significant reduction in the required transmitting energy when compared to the existing transmission scheme, and also improvement in distance area coverage. Despite enhancing connectivity and heterogeneity, low density and rural environment for VANETs setting have not been considered. The investigations can be considered very close to the functions considered under coordination and perception layer of the CC-V-layered architecture for cooperative relaying transmission in VANETs cloud.

Non-Distributed Clustering
A data dissemination model for cloud-enabled VANETs using in-vehicular resource system based on road side et al.
access point (RSAP) has been suggested to handle connectivity and resources availability issues [38]. The technology of RSAP is closely applicable to the coordination and perception layer of the CC-V-layered architecture since it functions as networking support layer. Different services and applications of VANETs including connectivity have been pointed through RSAP. The deployment of cloud, need of VCC service provider, and classification of VCC have been discussed. The classification of VCC includes private or public, dynamic or static, invehicular or out-of-vehicle, NaaS or communication as a service. It has been further divided as data centric or address centric, distributed or hybrid. The classification is based on the factors including participation, mobility, integration point, content management criteria, and integration with backbone networks. However, security issues with regard to the outlined model are not discussed.
Cloud computing-based message dissemination protocol for VANETs (ClouDiv) has been presented to deal with the issue of intermittent connectivity due to the higher speed of vehicles and their restricted capacity in terms of bandwidth [39]. Figure 9 shows data centre and VANETs node's routing table in ClouDiV. ClouDiv has provided an adaptive dissemination of safety and nonsafety messages through cloud computing architecture. In ClouDiV dissemination, a proactive routing approach for data centres and reactive approach for vehicle have been adopted. Stochastic routing method has been used during dissemination. Hence, since both proactive and reactive methods are employed, high signalling load might occur due to the adoption of proactive routing table discovery. The data centre has to update the table frequently. ClouDiV is closely related to functions of coordination layer of the proposed CC-V-layered architecture. QoN at each vehicular node has not been considered.
Cloud-supported seamless Internet access in intelligent transportation system has been recommended for accessing high-quality ITS services [40]. Cloud-supported gateway model, GaaS, and a link lifetime prediction scheme have been developed. GaaS model is depicted in Figure 10. GaaS has been divided into two gateways including mobile gateway and static gateway. In mobile gateway, access point is mounted on highmobility vehicles such as public bus in the city. It serves as a gateway for other vehicles to connect to the cloud. In static gateway, the conventional RSUs serve as access point, and RSUs are used as gateway for vehicles to connect to the Internet. Link life time prediction scheme considers time of entering or exiting from the gateway coverage. In this scheme, qualitative packet delivery has been achieved since link lifetime is considered. However, fault detection has become complex due to large numbers of gateways in the network setup. The GaaS model and link lifetime prediction scheme are nearly related to the functions considered under coordination and perception layer of the CC-V-layered architecture due to the handoff operation and location-based networking support.
Ikeda [41] have suggested a Performance of Optimized link state Routing Protocol (PORP) for video streaming application in VCC. PORP has addressed the challenges of advancement in communication and the need of efficient connection. NaaS architecture for VANET cloud computing has been considered to investigate the performance of Optimized Link State Routing (OLSR) protocol for video streaming application. OLSR has used proactive and shortest path approaches which might cause high signalling load. It is required to measure signal strength at each node to choose the best path for routing. The PORP protocol can be considered under the functions of coordination and artificial intelligence layer in terms of the CC-Vlayered architecture.   IPv6-based VCN (IP6-VCN) has been suggested in [42]. The flooding techniques adopted in most of the recent studies tend to increase the cost of content acquisition due to the content-centric approach for dissemination in VCC. The vehicular cloud domain system, VCN, and performance evaluation have been discussed. VCN includes addressing structure, vehicular cloud construction, vehicular cloud management, and content acquisition. The addressing structure creates the relationship between IP and content data for effective routing. The vehicular cloud domain system effectively minimizes the cost of content acquisition. However, in this scheme, communication between vehicle-to-infrastructure (V2I) has not been considered for assisting the content sharing and acquisition. The IP-VCN can be related to the aforementioned functions of coordination layer of the CC-Vlayered architecture.
In [43], a Reliable Adaptive Resource Management for cognitive Cloud Vehicular networks (RAR-MCV) has been suggested to address limited computing capability and energy of smartphones in car in order to utilize the available V2I Wi-Fi connections for traffic data offloading [43]. An optimal joint controller and related supporting access protocol have been discussed. The protocol has been claimed to be adaptive, scalable, and distributive. The developed optimal controller dynamically manages the access time windows at the serving RSUs. It also manages the access rates and traffic flows at the served VCC system in a distributed and scalable way. Its implementation complexity is fully independent from the number of serving RSUs and served VCC system. Nevertheless, optimized routing management and implementation of the concept have not been considered. The findings can be closely related under the functions of coordination and artificial intelligence layer of the proposed CC-V-layered architecture. The basic elements of the RAR-MCV have not been adequately explored. However, our proposed CC-V elements can be applicable in this study.
Zheng [44] proposed a semi-Markov decision process (SMDP)-based resource allocation in VCC system in order to address underutilization of vehicular resource. The SMDP is closely suitable in relation to the aforementioned CC-V-layered architecture in terms of artificial intelligence and perception layer. A computation resource allocation scheme has been presented. Furthermore, the resource allocation problem has been formulated as an infinite horizon problem for SMDP. SMDP defines state space, action space, reward model, and transition probability distribution of the VCC system. In order to develop optimal scheme, iteration algorithm has been used to define the action taken under a specific state. In addition, resource allocation and decision-making schemes, and a reward system were developed. Authors claimed that reasonable performance gain has been achieved by the SMDP-based scheme within the permissible complexity. However, effects of tolerance parameter to the optimal scheme have not been investigated.

Comparative Discussion on Data Dissemination
The above reviewed literature on data dissemination in CC-V focusing on clustering approach is summarized in Table 3. The summary considers the parameters including contribution, type of architecture, technique, implementation, and remarks. The contribution represents the progressive impact of the articles on the design of data dissemination technique for CC-V.
The architecture determines the category from the three types including VuC, VCC, and HVC. The technique defines the approach followed for addressing the raised issue. The implementation shows simulation tools and performance metric. The critical remarks have also been made. A comparative study is also presented in Table 4. The comparative study is based on three parameters including technique, architecture, and implementation. The technique is defined using conceptual model, complete framework, and algorithm. The architecture is defined using VuC, VCC, and HVC. The summary and comparative study of data dissemination techniques suggest that RAR-MVC [43] is a more feasible and better data dissemination technique. RAR-MCV considered Markovian random walk as its mathematical modelling technique which effectively optimizes routing objectives. The performance evaluation is widely explored with a range of metrics and environments. Both the distributive and non-distributive clustering can be efficiently enhanced based on coordination, artificial intelligence, and perception layer of our proposed CC-V-layered architecture. The coordination layer will handle clustering and connectivity issues. The artificial intelligence layer will handle vehicle clustering in order to perform cloud computing operations. The perception layer is linked with the in-vehicular devices and vehicular node itself.

Security
The major challenges of security in CC-V include privacy, intrusion detection, and authentication. The communication with neighbour vehicles using location, speed, and direction information, without unfolding the identity of each other, is privacy in CC-V. The identification of uncooperative neighbour vehicle is intrusion detections. The verification of neighbour vehicle base on some reputation is authentication in CC-V. All these major CC-V security issues can be improved on, if the CC-V-layered architecture is critically analyzed and incorporated with the security model. Security challenges of smart applications are basically authentication issues including automatic safety, information management, and services authentication. For artificial intelligence, the major issues with security is in intrusion detection of servers and data centres or vehicles which form cloud when a computing request from external client is being sent to the machines. Coordination layer can be used as a channel for intrusion and bridging of authentication process. Hence, intrusion detection and authentication issues can be tackled effectively at this layer. The perception layer constitutes in-vehicle devices and the vehicle nodes, and, thus, the major challenge in this layer is privacy and intrusion detection. With implementation of the CC-V-layered architecture, a more secured CC-V can be achieved.

Privacy
Hussain [45] suggested a secure and privacy-aware traffic information as a service (TIaaS) for VNVC to address the vehicle user privacy issue. The privacy issue is one of the major concerns why many vehicle users do not want to take part in the information sharing among vehicles. Hence, a privacy-aware TIaaS has been discussed. The network model of TIaaS is shown in Figure 11. The privacy issues cause lower usage of VANETs infrastructure including communication, computation, and on-board storage. A revocation mechanism, TIaaS thin-client concept for vehicles, and efficient mobility vectors framework have been designed.  [32] @ @ @ COHORT [33] @ @ @ @ OC-DSG [34] @ @ @ @ CARD [35] @ @ @ @ RADD [36] @ @ @ LTE-A [37] @ @ @ @ RSAP [38] @ @ @ @ ClouDiV [39] @ @ @ GaaS [40] @ @ @ @ PORP [41] @ @ @ IP6-VCN [42] @ @ @ RAR-MCV [43] @ @ @ @ @ SMDP [44] @ @ @ CM, conceptual model; CF, complete framework; AL, algorithm; IM, implementation. @ = Yes. TIaaS model has used VuC framework. It provided finegrained traffic information to vehicles from the cloud due to the subscriber cooperation with the cloud in secure and privacy-preserving way. Despite its strength in concealing location information, it might take longer processing time due to the encryption scheme adopted. The TIaaS is a security scheme that is best applicable under the functions of perception and coordination layer of the CC-V-layered architecture. Implementation is required to test the feasibility of the scheme.
In [46], a secure and privacy-preserving protocol for cloud-based vehicular delay tolerance networks (DTNs) has been presented to address the issue of privacy in incentive system and packet forwarding protocol. Network model for vehicular DTNs is depicted in Figure 12.
A threshold credit-based incentive (TCBI) mechanism has been designed for privacy preserving packet forwarding. TCBI encourages vehicles to cooperate with each other by calculating security and privacy and sharing resources with a certain rule. The privacy-preserving packet forwarding protocol has been used to address the challenges of layer attack by contracting out privacy-preserving transmission proof generation for resource-constrained vehicles. In TCBI, the vehicular privacy is well secured from both the cloud and transportation manager for performing any one-way trap-door function. The failure at either cloud or transportation manager side might hinder communication of the vehicles. As discussed in the proposed CC-V-layered architecture, the layer functions closely related to the TCBI are perception and artificial intelligence layer.

Intrusion Detection
An intelligent clustering scheme for distributed intrusion (ICDI) detection in VCC has been suggested to address the issue of security, alteration and misuse of information in VCC [47]. A network module of ICDI is depicted in Figure 13. A learning automata-assisted distributive intrusion detection system has been developed.
The system is based on clustering, standard cryptographic techniques, and reward penalty stochastic scheme. The system takes intelligent decision, and uses pseudo-dynamic clustering technique to select the CH, and then determine the cluster structure. The CH handles well-organized dissemination of information and storage through cloud-based infrastructure. However, to secure the learning automata system from malicious vehicles, a standard cryptographic technique has been employed. Nevertheless, a lower processing speed and complexity issue might occur due to the complex cryptographic approach and dynamic nature of the system. The intrusion detection approach in ICDI is relevant to the functions under artificial intelligence and coordination layer of the CC-V-layered architecture in terms of efficient intrusion detection implementation.
Kang [48] suggested a VCC service-oriented security framework (VCC-SSF) to solve the challenges of insufficient internal or external security in vehicles' infrastructure, and information leakage from sensors attached to the vehicles. Framework for VCC-SSF has been shown in Figure 14. The suggested framework has been used to handle user-oriented payment and accident avoidance management services. Furthermore, the framework has provided encryption, authentication, access control, confidentiality, integrity, and privacy protection of personal information related to users and vehicles. Accident avoidance management service uses the VCC model. The architecture consists of two models: one for before accident and another for after accident. The before accident model has utilized sensors attached  i.e.
with vehicles to monitor the status of driver's health and driving capabilities. However, this framework might not be economically viable due to the higher requirements and complexity. The findings can be considered very close to the functions employed under smart applications, artificial intelligence, and coordination layers of the CC-V-layered architecture. They can be applied in implementation of the VCC-SSF.

Authentication
Protecting vehicular cloud against malicious (PVCM) nodes using zone authorities has been suggested to address the issues of weak protection of VCC, and threats to data, resources, and services [49]. A zone authority framework has been depicted in Figure 15. A secured framework that uses key management and revocation technique to secure VCC from malicious nodes has been presented. The framework is a decentralized one. It has used multiple zone authorities. Each zone authority manages an area called zone which consist of RSUs, vehicles, and the clients at that zone. Every zone authority serves as a gateway which authenticates actions of that zone. It manages the service requests and the data flow and preserves the privacy of the cloud entities including vehicles and the client. In this kind of node protection, too many authentications occur while moving from one zone to another. This might degrade the communication between vehicles. Implementation is also required to examine suitability of the framework. The investigation can be considered to be close to the functions considered under coordination and smart application layer of the CC-V-layered architecture. The PVCM framework can be improved based on the CC-Vlayered architecture.

Privacy Preserving Authentication.
The privacy-preserving authentication is also one of the major issues in CC-V under privacy and security issues. Privacy-preserving authentication is broadly divided into three categories including pseudonym changing, silent period, and mix zone. Some of the recent advances in these categories are critically reviewed. In [50], a conditional privacy-preserving authentication scheme (CPAS) for vehicular sensor networks has been suggested to secure communication between vehicle and infrastructure in VANETs. CPAS uses pseudo-identity-based signature to secure V2I communication. It enables RSU to validate numerous received signatures in parallel. It significantly reduces the total verification duration. The performance of the CPAS has been investigated in terms of verification delay. It shows great potentials when compared with identity based batch verification (IBV). Privacy-preserving authentication of vehicle-to-vehicle (V2V) communication has not been considered in this work. The coordination and perception layer are best suitable for the CPAS scheme when related to functions and its protocol of the CC-V-layered architecture.
Lu [51] suggested a Dynamic Privacy-preserving Key management scheme for Location-based Services in VANETs (DPK-LSV) to preserve privacy of vehicles, while improving efficiency of location-based services in VANETs [51]. DPK-LSV provides anonymous authentication to vehicles and enables dual registration detection. The efficient location-based service sessions have been used. The service sessions are based on several time slots to hold the session key. An integration of dynamic threshold technique with V2V and V2I communication  has been performed to accomplish the session key's backward secrecy. The authors claimed about the effectiveness and efficiency of the scheme in relation to fast key update ratio and low key update delay. A pseudonymous authentication scheme with strong (PASS) privacy preservation for vehicular communications has been suggested to enhance privacy preservation in vehicle communication [52]. PASS has applied pseudonymous authentication in preserving vehicles' privacy. It also supports RSU-aided distributed certificate services that allow the vehicles to update the information on road. It has been claimed that PASS outperformed previous schemes in relation to certificate updating overhead and revocation cost. However, only highway scenario was considered during implementation. From the findings, it demonstrates that PASS is very close to the functions considered under coordination and perception layer of the CC-V-layered architecture. In DPK-LSV privacypreserving scheme, the functions of coordination and perception layer of the CC-V-layered architecture are suitable layers that can enhance authentication of the system.

Comparative Discussion on Security
The aforementioned literature review on security in CC-V is summarized in Table 5. The summary is based on the parameters including contribution, type of architecture, security technique, implementation, and remarks as security holes. The contribution points out the level of security enhancement provided by the articles in CC-V. The architecture determines the applicability of the security technique in the categories including VuC, VCC, and HVC. The technique tells the novel method used for providing security. The implementation shows experimental tools and metric for security attestation. The critical remarks in terms of security holes have also been identified. A comparative investigation is also presented in Table 6. The comparative investigation is based on three parameters including security technique, architecture, and implementation. The security technique is defined using the model for security, complete framework, and security algorithm. The architecture is defined using VuC, VCC, and HVC. The summary and comparative investigation of the security techniques affirms that ICDI [47] provides better security and privacy preservation in CC-V environments. The distributed security model is presented in ICDI. All the major CC-V security challenges including privacy, intrusion detection, authentication, and privacy-preserving authentication can be efficiently enhanced based on smart applications, coordination, artificial intelligence, and perception layer of our proposed CC-V-layered architecture. Authentication and intrusion detection will be best implemented on smart applications layer to attain reliability and efficiency. Intrusion detection, authentication, and privacypreserving authentication are more applicable to coordination layer. For the artificial intelligence layer, intrusion detection and privacy-preserving authentication are applicable. The perception layer is linked with in-vehicular devices and vehicular nodes, and is thus applicable to the privacy and privacy-preserving authentication in relation to CC-V-layered architecture.
can be divided into three categories: safety, efficiency, and infotainment. Safety applications are created to enhance vehicle's behaviour awareness, so as to eradicate or reduce vehicle crashes via V2V communication. Its applications include control loss warning, emergency electronic brake lights, blind spot/lane change warning, etc. V2I communication applications include oversize vehicle warning, railroad crossing warning, curve speed warning, etc. Vehicle-to-pedestrian (V2P) communication includes transit pedestrian indication. Some other applications in these categories are listed in Figure 16.
The major CC-V applications including safety, efficiency, and infotainment can be improved on, if the CC-V-layered architecture is critically analyzed and incorporated with the application models. Safety applications are most suitable at the smart applications, artificial intelligence, and perception layer in terms of functions and protocols. The efficiency applications are best linked to smart applications and perception layer of the layered CC-V. Infotainment can be represented at smart applications, artificial intelligence, and perception layer of the CC-V layered architecture.
In a generic term, safety application includes warning and support advisories, and infrastructure and vehicle controls [54].
The efficiency applications provide information on vehicles' and drivers' condition for passenger's comfort and health. The applications monitor both car and driver's performance during journey [55]. Safety applications can be best implemented by considering the functions and its protocols under smart applications, artificial intelligence, and perception layers of the CC-V-layered architecture.

Safety
Transportation and communication plays a critical role in disaster response and management in order to combat or reduce loss of life and property [56]. An emergency disaster response system (EDRS) model has been introduced. ERDS system consists of the three main layers as shown in Figure 17. The layers include infrastructure as a service, intelligence, and system interface layer. Infrastructure as a service layer consists of base platform and environment intelligent emergency response system. Intelligent layer provides computational model and algorithms. The system interface layer acquires data from gateways such as Internet, roadside masts, mobile smart phones, and social networks. Lighthill-Whitham-Richards (LWR) model has been adopted for modelling the disaster system. Its effectiveness has been demonstrated in terms of improved disaster evacuation characteristics. Security issues related to the architecture have not been discussed. The result of the findings demonstrates that the functions considered are closely related under smart applications and coordination layer of the CC-V-layered architecture. The CC-V elements are suitable for emergency disaster response system, since it requires computation, connection, and vehicular node for its communication.
A wireless access technology for vehicular network safety applications (WATVSA) has been suggested to address the issue of non-reliable broadcast of safety messages, in order to realize standard road safety applications [57]. The adopted wireless technologies include time division Segata and Renato [58] presented an automatic emergency braking with realistic analysis of car dynamics (AEB-ACD) and network performance as one of the important applications for VANETs safety. The simulation and analysis of driver behaviour awareness have been conducted. Emergency braking application has been simulated by embedding mobility, cars' dynamic, and driver behaviour models in to the network simulator. Furthermore, a simpler message aggregation mechanism has been presented to enhance message repropagation during peak load. The complete system permits capturing the interactions of the communications with vehicle's automated break mechanism and driver's behaviour.
The system yields detailed information on the communication level during experimentation as claimed by authors. However, there is a need for refinement of communication channel model and development of the vehicular dynamics models. The car dynamics and network analysis can be linked to the CC-V elements and archive better performance. It represents the basic components of AEB-ACD including road side infrastructure, vehicle node, and cloud computation.
In [59], an analysis of information dissemination in VANETs with application has been presented. The analysis considers cooperative vehicle safety systems (CVSSs) to demonstrate functionalities and viability of the systems. Thus, analysis of the effects of different communication ranges and rates has been conducted. The novel models that measure network performance in terms of their ability to broadcast tracking information are presented. The study demonstrates that hidden nodes affect VANETs communication. The channel occupancy or busy ratio can be used as feedback measure that quantifies the success of information broadcasted. Consequently, these outcomes are used to develop feedback control system for transmission range adaptation. The findings can be closely related to the functions and its protocols considered under smart application and coordination layer of the CC-V-layered architecture. The major constituent elements of CVSSs related to the CC-V elements are including DSRC network and vehicular node. However, effects of some parameters such as contention window size on information dissemination rate have not been derived.

Efficiency
A secured incentive-based architecture for vehicular (SIAV) cloud has been suggested to address the issues of underutilization of computational, communication, and storage capabilities of vehicles because of non-participation in vehicular cloud [60]. A summary of SIAV is shown in Figure 18. Two design architecture approaches including system model and secure token reward system have been developed for encouraging the vehicles to participate in computation and sharing of information. Major components of system model include service provider manager, reward token system, revocation authority, trusted authority, RSUs, and on-board units. Secure token reward system has three major phases, namely, searching resources, requesting reward tokens, and using the token for cloud services. The efficiency of the system model has not been evaluated. Efficiency applications can be optimally achieved by considering functions under smart applications and artificial intelligence of the CC-V-layered architecture. The proposed CC-V Singh [61] suggested a secure and reliable cloud networks for smart transportation (SCST) services for accident prevention, monitoring, and controlling system [58]. A smart transportation system and security issues related to the transportation system have been discussed. Smart transportation system has four functional layers including application layer, support layer, network layer, and perceptual layer (see Figure 19). Application layer represents various user applications. Support layer covers the cloud services, and network layer entails the Internet for connection. And perceptual layer is the client layer. An algorithm for vehicle detouring procedure in smart transportation system has been developed. The algorithm has been used to solve cloud computing down-time routing problem. The functions of smart transportation system include preventing accident, finding destination, and transfer of accident information to the vehicles using cloud. Implementation of vehicle detouring has not been conducted. The investigation and implementation can be considered to be close to the functions under smart application and perception layer of the CC-V-layered architecture. CC-V elements can be linked to the major constituent of the smart transportation services for accident and emergency prevention, controlling, and monitoring.
A Real Time services concept for future Cloud computing-enabled Vehicle (RTCV) networks has been suggested to ensure real-time performance as well as to improve accuracy and comfort degree for drivers [62]. A cloud computing system, real-time vehicular cloud services, and context classifications have been presented. Vehicular cloud system is partitioned into three tiers including device, communication, and service levels (see Figure 20). The real-time vehicular cloud services are introduced as road traffic and health care monitoring, and other customized services. The context information is classified into low-and high-level context. In another point of view, it can be classified into driver, car, and road traffic contexts.

Infotainment
In [63], cloud-based ITS (CITS) has been suggested to address the increasing transportation problem with the help of infotainment applications. A system for multilayered vehicular data cloud has been presented. The system employs cloud computing and Internet of things (IoT) technologies. The system has three modules including intelligent parking cloud service, communication from VANETs to cloud, and vehicular data mining cloud. Intelligent parking cloud module handles the decision process of selecting an available parking space for vehicles, and the mobile device with android application service for communication with the cloud. The system has higher interdependence between layers, which might degrade the performance of the system. The CITS is closely related to the functions under smart applications, artificial intelligence, and perception layer of the CC-V-layered architecture. The component which serves as elements of CITS are also applicable for the elements of CC-V and CC-V-layered architecture.
Multimedia services have become one of the major research areas of interest in both cloud computing and VANETs because of its relevance in both infotainment  Figure 19: The cloud computing module for STS system and safety. Thus, multimedia services in cloud-based vehicular networks (MSCVNs) have been employed to integrate cloud computing and storage with vehicles, in order to increase the accessibility to multimedia services [64]. Different systems including LTE system for network access, and multimedia cloud computing system, have been suggested. Three-layered cloud-based vehicular network model, which includes cluster layer, physical layer, and perception layer, has also been presented. A dynamic road monitoring system has been discussed. In video up-linking scenario, the MSCVN performs closer to the optimum when compared with two well-known schedulers including maximum largest weighted delay first, and exponential. However, delay in connectivity might arise due to the large audio and video files that need to be transmitted.
Infotainment applications are best represented by considering the CC-V-layered architectures' functions and protocols in relation to smart applications, artificial intelligence, and perception layers.

Comparative Discussion on Applications
The aforementioned literature review on applicationbased developments in CC-V is summarized in Table 7. The summary considers the parameters including contribution as service, type of systems suitable for the application, the technique followed in application development, implementation detail of the application, and remarks as strength and weaknesses of the application. The contribution highlights the signification of the service provided by the application. The architecture tells the suitability of the application in the cloud-based vehicular communication categories including VuC, VCC, and HVC. The techniques are the development methods followed in application. The implementation shows the process and metric for quality attestation regarding the services of the application. The critical remarks are also made in terms of limitations of the application. A comparative analysis is also presented in Table 8. The comparative analysis is based on three parameters including application design technique, suitable system, and implementation performed. The application design technique is defined using the basic concept of the application, overall framework, and algorithm of the application operations. The application system is defined using the suitability in the cloud environment for vehicular communications including VuC, VCC, and HVC. The summary and comparative analysis of the application-based developments in CC-V suggest that MSCVN [64] is a more practical application concept with greater user friendly services. The implementation plan of MSCVN is widely acceptable in CC-V environments. The application model is more scalable due to the plug-in-based service concepts. The CC-V applications can be enhanced in terms of the representations, functions and protocols of the four layers of the aforementioned CC-V-layered architecture. The proposed basic elements of CC-V are closely related to the MSCVN elements.
need to be addressed for realizing CC-V. These research issues are discussed below: (1) Architecture design. Due to the fact that CC-V is still a new area of research, there is no generalized standard architecture for this new idea. Although many initial architectures for CC-V have been suggested, yet standard architectures with implementation details are unavailable [25][26][27][28][29][30][31]. Therefore, the issue needs thorough exploration. (2) Data dissemination. Efficient data dissemination in CC-V is a challenging issue, due to the high dynamicity of vehicles in changing their positions. The design and how to transmit data in CC-V need to be critically addressed. Even though some work has been done in [32][33][34][35][36][37][38][39][40][41][42][43][44] by suggesting solutions to handle various type of data dissemination issues using clustering, data-centric, and other routing approaches, yet many issues have not covered, such as location verification, data dissemination, and video-centric routing. The conventional VANETs routing [65][66][67] may not be suitable in the case of CC-V due to the connectivity challenges in mobility. (3) Data offloading. The complex unrefined data need to be offloaded to conventional cloud or to vehicular cloud for processing. After processing, this refined data is accessed by vehicles and other related organizations. However, the issue of how to offload unrefined data and access the refined data in a high-mobility vehicular scenario is required to be looked into. Data aggregation and computation are also required since vehicles use sensors and other on-board devices for collecting information relating to vehicle, environment, and traffic. Hence, the issue of data aggregation and computation are attached clearly, both at vehicle and cloud level. The data need to be aggregated and refined for users. (4) Application design and deployment. Some potential VANETs applications are yet to be designed or deployed at vehicles for efficient usage, such as picture and video coverage as a service (PVCaaS) in CC-V. PVCaaS is a service whereby vehicles on the road take pictures and video covering of their surroundings. The picture and video are automatically offloaded to the cloud for storage and analysis. This service is useful for road safety and security organization. Some of the recent works by [56][57][58][59][60][61][62][63][64] tackle some application designs for incentive-based disaster management, highway traffic control [68,69], and multimedia applications.
(5) Standardization and interoperability. The vehicles have heterogeneous devices, such as sensors, GPS, and smartphones. How to make the varied devices to work together efficiently is highly required for data gathering. Standardization is also required for these on-board devices, by looking into compatibility, quality, and implementation of guidelines, interoperability, and repeatability of on-board vehicle equipment and software. (6) Security and privacy. Security and privacy are also the major challenges in CC-V, because the participating users are always mindful of their privacy and whereabouts at any given time. Also, since vehicles rely on information from the cloud or other vehicle for their navigations, this might cause serious havoc; if there is information distortion by an intruder, it can lead to waste of time, fuel, or even loss of life. Several authors [45][46][47][48][49][50][51][52] have made attempt to see how these issues including intrusion detection, authentication, and location privacy could be handled. But the reality is that, extensive security-related research work need to be done considering security, privacy, authentication, and their efficiency in CC-V. (7) Delay in cloud--client communication. It is one of the fundamental issues in any cloud-based service due to the dynamic network environment and the consideration of cloud infrastructure in high mobile vehicular environment. The sparse distribution of vehicles and the dynamic nature of density of vehicles in the network environment are unavoidable. CC-V requires real-time communication decision in safety applications, which is quite challenging considering the delay issue in network access [70][71][72][73]. (8) Autonomous driving. It is one aspect of ITS which requires artificial intelligence, learning capabilities and storage, for making computational decision on possible route for achieving fast and safe navigation [74,75]. Cloud computing needs to be applied for computation of intelligent data and, subsequently, for analysis, which should be accessed by autonomous vehicle. (9) Learning-based data storage. It is another aspect of ITS which is required in order to achieve distributed ITS. It needs some initial storage of information in VANETs without any provision for device setup [76,77]. The device setup could be RSU or external access point. Hence, there is need for a robust strengthening learning-based, dynamic, and adaptive data storage techniques for VANETs in order to achieve distributive ITS.

CONCLUSION
We have reviewed related work and put forth a framework in vehicular communication for CC-V. The framework suffices from the merging of cloud computing, pervasive sensing, improved network mobility, and inbuild vehicle resources. The large volume of underutilized on-board vehicle resources, such as computing power, Internet, and storage, could be utilized by various clients Internet, in relation to the conventional cloud resources. Numerous of these resources can provide support for handling traffic incident. CC-V can generate income, and enhance security and safety. CC-V can also help minimize the losses in different types of emergency incidences including fire outbreak, flood, or accidents.
In this work, layered architecture for the CC-V is presented. Three elements of CC-V including vehicle, computation, and connection are identified, and their relations are discussed in the aspect of CC-V. Several application scenarios, security and privacy, and formation of CC-V have been identified and discussed. We present a taxonomy for CC-V and explored related literature based on the taxonomy. We have also identified open research issues and challenges with regard to CC-V.
However, a number of issues are still not been thoroughly explored by researchers, which include data dissemination to cloud, data offloading, data-centric routing, intermittent connection, VCN, VANETs application and deployment, security, and privacy-aware data sharing. In conclusion, CC-V could not be fully implemented, except if governments and automobile industries would come together to utilize the benefit of the framework of CC-V. Thus, CC-V could be the next paradigm shift that offers feasible and technologically viable and smart solutions for traffic issues with economical gains.