A new networking paradigm, Vehicular Edge Computing (VEC), has been introduced in recent years to the vehicular network to augment its computing capacity. The ultimate challenge to fulfill the requirements of both communication and computation is increasingly prominent, with the advent of ever-growing modern vehicular applications. With the breakthrough of VEC, service providers directly host services in close proximity to smart vehicles for reducing latency and improving quality of service (QoS). This paper illustrates the VEC architecture, coupled with the concept of the smart vehicle, its services, communication, and applications. Moreover, we categorized all the technical issues in the VEC architecture and reviewed all the relevant and latest solutions. We also shed some light and pinpoint future research challenges. This article not only enables naive readers to get a better understanding of this latest research field but also gives new directions in the field of VEC to the other researchers.
A significant component of future intelligent transportation systems is the vehicular networks. Various mobility-based services are offered by these networks, which vary from content-sharing services (marketing purposes and infotainment) to information spreading applications (emergency operations like information regarding natural calamities, etc.) [
Vehicular Ad hoc Networks (VANETs) depend predominantly on cloud computing [
Contrary to the Vehicular Cloud Computing services, which are centralized, VEC aims at applications with extensively distributed deployments. VEC furthers the benefits yielded by cloud computing services to the edge of the network [
Comparison between VEC and VCC.
Features | Vehicular Edge Computing | Vehicular Cloud Computing |
---|---|---|
Location | At user’s proximity | Remote location |
| ||
Latency | Low | High |
| ||
Mobility support | High | Limited |
| ||
Decision making | Local | Remote |
| ||
Communication | Real Time | Constraints in Bandwidth |
| ||
Storage Capacity | Limited | Highly Scalable |
| ||
Context awareness | Yes | No |
| ||
Device Heterogeneity | Highly Supported | Limited Supported |
| ||
Computing Capability | Medium | High |
| ||
Cost of Development | Low | High |
Software Defined Networking (SDN), characterized by the decoupled control plane and data plane, provides comprehensive network control. The communication between the data plane and the control interrupts the flow of the packet, resulting in an increase in both cost and delay in communication. Thus, SDN focuses on minimizing the communication between data plane and control plane [
VEC has received tremendous attention from researchers since its birth, and several surveys have studied various aspects of VEC. In [
The rest of the paper is organized as follows: in Section
Organization of the survey on vehicular edge computing.
This section explains the architecture, role, and mode of operation for each component. The VEC architecture lays its foundation on three layers: cloud layer, edge cloud layer, and smart vehicular layer as shown in Figure
Three-layer vehicular edge computing architecture.
Most important advantages of cloud layer are data aggregation, data mining, analysis optimization, storage, batch processing, and computation of complex data [
The edge cloud layer ensures the connection between the smart vehicular layer and the cloud layer. For achieving this, the vehicles contain devices that make use of wireless communication protocols, such as 802.11p, 3GPP, 3G, 4G, LTE, and 5G. The purpose is to provide low latency, location awareness, emergency management, caching, content discovery, and computation and improves quality of services since it is at the proximity to vehicles, and it is used for real-time interaction [ Infotainment as a service (IaaS): it provides useful information to the users, about the events like emergency situations, connected to the VEC. It also offers entertainment services like music, games, movies, and so forth. The aim of these services is to enhance the quality of user experience and safety of the passengers by offering both information and entertainment. Network as a service (NaaS): the users who have an Internet connection can facilitate other users by providing them connection through the vehicle, roadside units, or micro base stations. This facility is very valuable, particularly in emergencies [ Storage as a service (SaaS): vehicles need additional storage for running their applications, which require high storage resources, and for making a backup for a temporary purpose. This need is fulfilled by the edge server, which provides free storage for the particular client to run their applications [ Computation as a service (CaaS): those vehicles which are in a parking lot, centers, or a traffic congestion spent their several hours having unused computational resources. There is an opportunity for those vehicles/users who desire to augment the computational resources of their vehicles or mobile devices to achieve massive computational tasks [
This layer also provides a communication facility for vehicles to vehicles (V2V) and with an external infrastructure vehicle to Infrastructure (V2I). In V2V, vehicles interact with each other which are in the same range of communication, so information can be propagated through vehicles until it arrives on the edge. For instance, if any vehicle exhibits strange behavior, in case of direction change, violation of the speed limit, or mechanical failure, emergency messages will be sent to neighboring vehicles as well as to edge, which contain the position, speed, and moving direction of that specific vehicle. The V2I is liable for exchange of operational data among vehicles through infrastructures like roadside units, micro base stations, and edge servers over wireless networks. This layer comprises many components and all are being managed through SDN.
SC: SDN controller is the overall central intelligence point of the network. It governs the entire behaviors of the network. It also acts as edge orchestration and resources management for the edge [ SES: SDN-based edge server, a group of SRSUs is connected to an edge server over broadband connections. SES provides the vehicles storage and computational facility at the edge of the network. It also stores regional road system information and performs emergency services. SMBS: SDN micro based station is not only capable of carrying voice data, but it is also more sophisticated. SMBS works under the supervision of the SC, running OpenFlow, and capable of delivering edge services. SMBS is also local intelligence edge device [ SRSU: SDN-based roadside unit, running OpenFlow and managed by the SC. It is an edge device. It can exchange data with those vehicles that are in its communication range [ SSV: SDN-based smart vehicle acts as a forwarding device (end user), armed with On-Board Unit (OBU) and operating OpenFlow. It can communicate with other vehicles, SRSU, and SMBS. It is also considered as Data Plane element.
For control channel, every SSV is equipped with WiMax/(3G, 4G LTE, 5G) and Dedicated Short-Range Communication (DSRC)/LTE-V interface for Data Channel. Under any circumstances, if SDN controller’s connection is lost to deal with this situation, the SSVs support emergency backup mechanism plan and go back to conventional operations like AODV, DSDV, and OLSR routing protocols [
Vehicles are likely to perform more communication, exchange onboard services, and offer storage [
Therefore, this layer can promote sensing of not only the environment but also the behavior of passenger and drivers in a vehicle. The main component of this layer is a vehicle; vehicles in this paradigm are considered as a smart vehicle and are fully armed with many of the latest sensors and communication equipment. We will discuss the term smart vehicle, its basic components, communication, services, and application in detail in Section
The vehicle that is capable of computing and has storage as well as communication facilities and can learn from its environment by making decisions accordingly is known as a smart vehicle. Smart vehicles are equipped with different types of sensors and multi-interface cards, internally (onboard) and externally. With the rising number of smart vehicles equipped with onboard wireless devices (3GPP, IEEE 802.11p, Bluetooth, etc.) and sensors (radar, lidar, etc.), efficient management and transport applications are aiming at enhancing flows of vehicles by minimizing the travel period of time and averting any traffic congestion.
Smart vehicles, with a group of innovative functionalities (i.e., information exchanging, positioning information, etc.), can support specific applications (i.e., safety messages and warnings, gossip-based applications, etc.). Mostly the vehicles within a VEC are equipped with an onboard wireless device, particularly OBU; for example, the onboard radar of smart vehicle is able to slow down the vehicle automatically after detecting traffic congestion. In accident warning systems, sensors can be used to verify if the airbags were deployed when the accident took place. This information is then spread through V2V/V2I within the network [
Various intelligent systems are deployed within vehicles to offer real-time measurement and safety services [
The architecture of the system should be designed to help drivers to avoid accidents. To attain reliable and ample environment information, the smart vehicle is commonly equipped with multibeam lidars, microwave radars, high-resolution cameras, and so forth. As shown in Figure CPU: it computes the instructions by performing arithmetic, logical, and I/O operations in an efficient way. Moreover, the communication and applications protocols are also implemented. Wireless transceiver: it transmits data and information among vehicle to vehicle and vehicles to infrastructure. GPS receiver: it receives the Global Positioning System’s information and gives support to provide navigation services. By integration with a communication means, a GPS could automatically report the vehicle’s precise position even with an accuracy of less than 1m. This gathered information facilitates organizing the formation of platoons and could offer relevant and prompt information about accidents, road condition, and so forth. Sensors: various sensors are located internally and externally in a vehicle to measure different factors like speed, distance from surrounding vehicles, and so forth, for instance, an ultrasonic sensor based on sound waves, reflected by the objects. The reflected sound waves can be used to identify the distance and relative velocity of nearby objects. I/O interface: it provides user’s friendly human-vehicle interaction. RADAR: Radio Detection and Ranging monitors the position of neighboring vehicles. Automotive radars are classified as long-range radars and short-range radars. Such sensors are already used in adaptive cruise control systems. LIDAR: Light Detection and Ranging, where the sensor has a one-dimensional scanning capability that can precisely measure the comparative distance of the vehicle by scanning the horizontal surface with laser beams. The sensor uses high-energy rays of the laser to transfer infrared light pulses with a wavelength, ranging from 850 to 950 nm [ OBU: Onboard Unit, where each vehicle in this system is armed with an OBU, which controls communication of vehicle with SRSUs, SMBSs, and other vehicles through DSRC/LTE-V [ LCS: Local Camera Sensor is the sensor which observes the behavior of the driver and also provides accurate and reliable object detection.
Smart vehicle.
Technically the smart vehicle must be able to make accurate and reliable environment perception based on incomplete and uncertain information from heterogeneous and multimodal onboard sensors. Moreover, by integration of learning-based into rule-based decision making methods to address the challenges in imperfect environment detection and low predictable traffic participants, a smart vehicle can be able to make a decision and react accordingly.
The vehicular networks pose many challenges that were unknown in the recent past using traditional wireless communication systems. This fact is a result of the highly dynamic vehicular environment and the diverse QoS requirements. Various communication mechanisms have been introduced to tackle these challenges, for example, DSRC, the Federal Communications Commission (FCC) alloted 75 MHz in 5.9 GHz band for DSRC for V2V and V2I communications in USA, and the ITS-G5 in Europe, both based on the IEEE 802.11p standard. The 3rd Generation Partnership Project (3GPP) has initiated projects lately which support vehicle to everything (V2X) services in Long Term Evolution networks and the 5G cellular networks [
Generally, there are two types of vehicular communication: V2V and V2I. In addition, other entities such as vehicles, pedestrians, and roadside infrastructure can gather information of their surrounding environment (e.g., receive information from other vehicles or other sensor equipment in range) to process and share it so that more intelligent services could be provided. These services include cooperative collision warning and autonomous driving. A unique virtual vehicle coordination framework [
Communication of smart vehicle.
This term is coined to express communications within the vehicle. The OBUs, which are installed in the vehicles, are able to communicate within the vehicle (i.e., various sensors provide information like traffic congestion, brakes, accelerator, or other nearby objects).
It signifies communications between different vehicles or among vehicles and sensors that are installed in or on different locations, like roadways, parking lots, and so on. Intervehicle communication implies greater technical challenges as the vehicle communication needs to be once supported when the vehicles are stationary and while they are moving. There can be direct communication between vehicles. The communication of this nature permits the sharing of information between vehicles irrespective of the infrastructure. Nevertheless, V2V communication applies to a limited range. An extensive study has been carried out in the V2V communication environment. Alghamdi et al. [
Vehicle to vehicle and vehicle to infrastructure communication.
The applications like Road-Accident and Street Parking support V2V communication to enhance the communication range of the vehicular network. These applications share information between vehicles and roadside units when they are not in the range of each other. During this whole process, other vehicles act as intermediaries; they receive information and forward it to enter in the range of the SRSU.
The term extravehicular communication represents the communication between vehicle and the outside world, that is, V2I (vehicle to edge nodes, SRSU, SMBS, etc.).
Deploying edge computing environments along with the road can ensure communication between the moving vehicles. A vehicle can interact with the other approaching vehicles and notify them in case of any risk or traffic bottleneck and the number of pedestrians on the road. Moreover, edge computing facilitates accessible, trustworthy, and distributed environments that are in synchronization with the local sensors [
V2I allows the exchange of wireless information between vehicles and infrastructure (e.g., SRSU, SMBS etc.). As the OBU of the vehicles has limited processing and storage capacity, some applications rely on the edge servers as platform or middleware. In some cases, V2I communication is expected to access global information, for example, in gas-station systems. Likewise, some applications can extract weather information and traffic congestion via V2I communication. Figure
The smart vehicles provide numerous types of services. Some of these services, which include assistant driving, autonomous vehicle, platoon, and parking lots, are being described in the following sections.
These days, there are vehicles like cars, buses, and trains which are designed in such a way that they provide useful information, that is, accidents, closure of roads, and traffic jams with the help of sensors, actuators, and processors. This provides greater safety and better navigation of the vehicles. This information about the traffic patterns can be beneficial for various profit and nonprofit organizations [
Five developing stages of smart vehicles: primary automation, assisted driving, partially self-driving, high-level self-driving, and fully self-driving (autonomous driving) [
With the evolution of the smart vehicles towards autonomous driving, there is a need to have maximum connectivity among these vehicles. Therefore, the rising vehicular networks are considered as the most significant component in the development of intelligent transportation system and smart cities. The vehicular networks are expected to provide various high-end applications, which range from road safety to improvement in the traffic efficiency and from automatic driving to uninterrupted access to the Internet services [
The vehicle is designed in such a way that it can anticipate the safety-critical issues and perform functions accordingly while monitoring the road conditions at the same time during the entire journey. This type of design assumes that the driver will provide the destination or navigation but she/he is not likely to be there for control during the entire trip. Therefore, safe operation is the function of the automated vehicular systems [
Though the automated vehicular system is quite different from the connected vehicular technology, it is related to the latter. The connected vehicular networks enable the vehicles to share information through the transportation infrastructure. Vehicles might also share sensor information with the neighboring vehicles, which can in return provide an AV with additional information on which the decision making can be based. Experts from the international forums like Business Insider and World Economic Forum foresee the impact of the automated vehicles on the overall economical avenue and the transition of the mode of moving from one place to another. In the future, we might see a switchover from the personal vehicles towards the ride of the automated vehicles like Uber or Lyft [
A platoon is a group of smart vehicles, which have driving support systems where one vehicle follows the other one. The platoon is formed with a number of vehicles driven by technology and connected with each other through shared communication. Platooning is possible due to the development of various technologies named Advanced Driver Assistance Systems, Adaptive Cruise Control, Lane Departure Warning System, Blind Spot Information System, and Drowsiness Detection System which are fully equipped with sensors and actuators. Modern vehicles control processes; the human-machine interface and the amalgamation of these technologies help the vehicles to participate in cooperative platooning.
The fuel efficiency of heavy trucks can be enhanced using cooperative platooning as the speed changes can be anticipated, thereby causing the vehicle to maintain a steady speed. As the emission of carbon dioxide is directly proportional to fuel consumption, the cooperative platooning indirectly may reduce environmental pollution. Another benefit of cooperative platooning is the enhancement of road safety. In case of emergency, the message is delivered to all the vehicles in a platoon, and then a suitable mechanism is devised by the automated system accordingly; for example, in case of collisions among vehicles, an automated braking system will be enabled [
In the urban areas, the number of parked vehicles in the parking lots is huge. Moreover, the parked vehicles are extensively distributed geographically in street parking, outer parking, and so on. The parking vehicles do not change their locations, unlike moving vehicles, for a certain long period. Therefore, these parked vehicles are not expected to carry information from place to place. Nonetheless, the parked vehicles act as communication infrastructures and have their own specified features. As the SSV is complemented with wireless communication devices and the rechargeable battery, it is easier for the parked SSVs to communicate with each other and these parked vehicles can even connect to the nearby moving SSVs. The parked SSVs, therefore, serve as static backbones for the improvement of connectivity among the vehicles. These SSVs can act as communication infrastructures by carrying and transferring data packets on to other vehicles. The parked SSVs become credible and easy communication nodes due to their abundance, long time of staying, and wide distribution within the urban area. The number of parked vehicles in a parking lot and the parking time are the key factors that influence the usage of parked vehicles as communication infrastructures [
Heavy computation tasks can be carried out through the collaboration of SSVs when they join under suitable communication conditions, that is, parking lots. According to [
Parking lots.
VEC and the concept of utilizing smart vehicles as infrastructures have opened an arena of various associated vehicular applications, for example, driving safety, AR, infotainment services, and video streaming. For the applications where high computational processing is the demand, the VEC networks play an important role in rushing up computing, thereby minimizing the delay like if an accident takes place, we need to formulate a solution to reschedule traffic lights and to dissipate large traffic backlog in a suitable way. This has an exceptional demand in the computation resources [
Applications with their requirements.
Application | Requirements | |||
---|---|---|---|---|
Bandwidth | Delay | Data | ||
Source | Time | |||
Health Monitoring System [ | High | Real Time | On-board Sensors | Up to seconds |
| ||||
Infotainment [ | High | Real Time | On-board Sensors | Up to seconds |
| ||||
Multi User Gaming | High | Real Time | On-board Sensors | Up to seconds |
| ||||
Nearby Driver Collaboration | Low | Low | Nearby Vehicles | Up to minutes |
| ||||
Platooning [ | Low | Real Time | Nearby Vehicles | Up to minutes |
| ||||
Parking Lot’s Information [ | Low | Low | Nearby Vehicles | Up to minutes |
| ||||
Vehicle Tracking | Low | Low | Edge Coverage | Up to hours |
| ||||
Traffic Light Management [ | Low | Low | Edge Coverage | Up to hours |
| ||||
Emergency vehicle Warning [ | Low | Low | Edge Coverage | Up to minutes |
| ||||
File Sharing (Multimedia) | High | High | Entire network | Up to days |
| ||||
Driver Behavior [ | Low | High | Entire network | Up to days |
| ||||
Maps Update | High | High | Entire network | Up to days |
This category of applications focuses on improving safety by limiting the chances of accidents. These applications keep track of the driving environment and notify the drivers of expected hazardous points to prevent accidents.
Context Acquisition: gathers context-specific information from different sensors. Processing: it applies reasoning techniques to obtain the contextual information. Acting: contextually facilitating the users according to their current place and time.
By the knowledge of context, it can also help to produce a context-aware concise information, which will utilize less radio resource for transmission. For instance, if a BS would like to transmit text information to a user, the BS can transmit only its contextual coded data. The user would extract the desired content just from the context by utilizing an appropriate decoder and big-data analytics technique such as NLP [
VEC applications stress not only upon the safety services but also on developing nonsafety applications, for example, multimedia applications like video streaming, AR, and infotainment services. The number of streaming applications has increased significantly and these contribute a large proportion of the network traffic.
Now, it is quite possible to visualize innovative solutions to the parking lot monitoring problem. As, in [
The output of the device’s camera is analyzed by the edge computing application, thereby covering the objects viewed with AR content. AR involves complex storing operations and tedious data processing tasks; therefore, it needs an increased level of data storage, computation, and communication. The VEC is regarded as the best alternative to fulfill the demand of AR applications in a vehicular network with specific requirements of mobility, location awareness, and low latency.
In this section, some technical issues of VEC have been categorically highlighted. Table
VEC technical issues categorization I.
Issues | Reference | Year | Contributions | |
---|---|---|---|---|
Latency | Routing | Lu. et. al. [ | 2018 | Proposed a routing scheme (IGR) for the vehicles moving in a city environment. |
Prasanth et. al. [ | 2009 | Presented EBGR, which augments the packet behavior for the network whose mobility is high and to enhance the reliability of the delivered messages. | ||
SDN | Truong et. al. [ | 2015 | Authors presented FSDN, which integrates SDN and Edge computing by considering various factors, e.g., physical medium, mobility, and capability. | |
Deng et. al. [ | 2017 | Authors revealed an entire series of latency control mechanisms, from radio access steering to processing the caches at base stations. | ||
Tomovic et. al. [ | 2017 | Presented a model for IoT, which backs real-time data mobility and scalability. | ||
5G | Ge et. al. [ | 2017 | A multihop approach is adopted for the vehicular communication within the fog cell. | |
Khan et. al. [ | 2018 | A 5G-VANET model with the integration of SDN, Cloud-RAN, and edge computing technologies is designed. This model provided better throughput and minimized delay. | ||
Tao et. al. [ | 2017 | 5G technologies are used to overcome the issue of extreme growth of the vehicular terminals and mobile data traffic. | ||
| ||||
Scheduling & Load balancing | Chen et. al. [ | 2017 | Proposed a scheduling scheme established on queue length and response time. They also formulated a design for a vehicular cloud, based on a compositional approach (PEPA). | |
Lai et. al. [ | 2018 | The PV system supports a heuristic insertion algorithm and a cooperative strategy between vehicle nodes, edge, and the cloud for sending requests as well as schedules routes for PVs. | ||
Park et. al. [ | 2017 | Proposed scheduling algorithm to recover the lost connection and continue the services in case of an edge server failure. | ||
He et. al. [ | 2016 | Presented an SDN based (MPSO-CO) centralized load balancing algorithm. This optimized the workload between the edge/fog networks so the latency can be efficiently minimized. | ||
| ||||
Offloading | Zhang et. al. [ | 2017 | Presented a computational offloading infrastructure, which stresses upon the computational effectiveness of the transfer frameworks of V2I and V2V modes of communication. | |
Saqib et. al. [ | 2016 | Introduced a model for computation offloading as FogR. It could respond to any emergency with greater reliability in case of fog node failure. | ||
Bi et. al. [ | 2017 | Presented CVFH scheme in which, before entering the coverage area of the target AP, a vehicle takes relevant information from the qualified vehicle through its neighboring vehicle. | ||
| ||||
Resource Management | Miao et. al. [ | 2016 | Proposed an FLRM scheme, which determines each resource’s survival time by using the collected information with the help of fuzzy logic, which is based on popularity evaluation algorithm. | |
Li et. al. [ | 2017 | They addressed the local resource management in the FeRAN. To support this, FRR and FRL schemes were introduced. The on-hop probability for real-time vehicular services is enhanced. | ||
Brennand et. al. [ | 2016 | Proposed FOX, which detects and manages traffic congestion in VANETS. Through this, the time of the trip, | ||
| ||||
Security & Privacy | Basudan et. al. [ | 2017 | Introduced CLASC, a privacy preserving protocol, which aims at improving security in a crowd-sensing based road condition monitoring system. | |
Fan et. al. [ | 2018 | Proposed a data-sharing scheme, which analyzes a multiauthority CP-ABE by effective decryption, while protecting a CP-ABE system safe against the collusion attack. | ||
Soleymani et. al. [ | 2017 | Proposed a fuzzy trust model, which detects the defective nodes and unauthorized attackers and handles the uncertainty of data in the vehicular networks. | ||
Huang et. al. [ | 2017 | Analyzed DREAMS where edge server executes local management tasks by ensuring a trusted reputation. It optimizes resource allocation and detects and enhances the recognition rate of misbehaving vehicles. | ||
Huang et. al. [ | 2017 | Proposed Meet-Fog to accurately distribute negative messages such as CRL in VANET. | ||
Alrawais et. al. [ | 2018 | Presented a revocation architecture, which increases the effectiveness of the certificate status by using Edge and Merkle hash tree. |
Various upcoming vehicular applications need real-time mobility support (e.g., positioning systems and smart traffic lights). This factor makes network latency an important area to ponder upon. The long distance between the cloud and the vehicles is not the only parameter that causes latency. In addition, it may also be caused by the inability of routing an optimal path, delays in queueing, or various other factors. However, the new vehicular applications need a lot of computational capacity to process the complex tasks. To achieve less delay in the transmission of data and maximum throughput, some approaches have been categorized.
Routing approach considers the geographic routing, that is, position-based routing, which makes decisions locally. A specific node, which transmits a data packet, is required to consider three positions, that is, its current position, destination position, and its one-hop neighbor. In [
In [
This type of networking focuses on a logically centralized network control plane. It helps to implement a robust mechanism for the management of resources and traffic control. An SDN-oriented network provides flexibility, programmability, and knowledge of the network. Tomoviv et al. [
A software-defined mobile-edge vehicular networking was proposed in [
With the advent of 5G mobile communication networks, the performance of the existing vehicular networks can not only be enhanced but the new applications of vehicular networks can also be supported. In [
A new architecture named Foud is examined in [
Vehicle networks ensure an effective communication to improve the dissemination of data among vehicles. A large number of vehicles conduct the data dissemination, which results in an increase in the load. The recent scheduling algorithms are developed in such a way to adjust to the varying challenges of the queue length. One such algorithm is the classic shortest queue policy. The shortest queue would not imply a minimum waiting time; thus time-based scheduling proves to be more efficient and reliable. Therefore, Chen et al. [
If an edge server fails, then recovering the lost connection and continuing the services require a demanding scheduling algorithm. In [
The fragmented network devices in the edge computing networks have weak computational power. Therefore, how to allocate the tasks is important in IoV, which require computation and balancing the load according to the capacity of the equipment of edge networks. Therefore, in [
Edge servers can curtail transmission cost and generate a fast response in the offloading services because of the closeness to the vehicular users. Despite the fast response rate, the edge servers usually face the limitation of the resources as compared to the conventional cloud servers, which have a large computational capacity. The edge servers take a certain time to perform the computation tasks. This is especially true for the edge servers located at the road segments, which have a high density of vehicles in comparison to others. In [
Saqib et al. [
Y. Bi [
Connectivity is a feature that needs to be embedded in the vehicles [
To transfer services from a source node to the target node, an edge server must have a sufficient number of resources. In reality, an edge node has an inadequate collection of resources and therefore these can become overloaded when many user’s requests arrive at once; for instance, at the time of peak traffic, it results in degraded performance. Reference [
Traffic congestion causes the loss of billions of dollars annually due to fuel consumption and lost time. In order to address these concerns, Brennand et al. [
The security and privacy issues are usually prevalent in vehicular crowd sensing, which requires the user identity and location privacy to be protected. In this regard, various solutions have been proposed in the domain of crowdsensing and vehicle-based sensing. Edge computing, being an emerging paradigm, addresses these issues. Basudan et al. [
The vehicular communication network, which is considered as an important element of intelligent transportation systems [
In VANETs, trust must be developed among vehicles to secure the integrity and ensure the reliability of applications. Reliable sources ensure the collection of credible information from the surrounding vehicles. Soleymani et al. [
All of the components in a large VANET are not trustworthy; therefore negative messages [
The proximity and decentralization of the service infrastructure to the edge provide different benefits for networks like low latency, efficient energy utilization, and greater throughput. The latest vehicles are being embedded with different sensors for processing and wireless communication capabilities. This has enabled many potential benefits to be exploited such as safety, efficiency, and comfort while they are on the road. However, different challenges to the VEC arise, which are illustrated in this section.
Traditional models of sensor networks consider a static environment. Similarly, ad hoc network also focuses on limited mobility based on laptops and handheld devices carried by the users. However, mobility is a norm for vehicular networks. The patterns of mobility for the vehicles have a robust correlation. Each vehicle on the road has a constantly changing set of neighbors, some of which it has never come across before and is quite unlikely to have an interaction with in the future. This ever-changing nature of the vehicular dynamics can hinder the utility of reputation-based schemes. To rate different vehicles based on the reliability of their reports is dubious to prove useful; that is, any specific vehicle may not receive sufficient information from the same vehicle to make its decision about that vehicle. In addition, as two vehicles are likely to be in the communication range for a few seconds, we could not consider protocols, which require interaction between sender and receiver. An upgraded mobility model is needed so it can provide data related to the accurate vehicular behaviors, like vehicular speed, prediction of the vehicular reputation, and distributions in both space and time. In particular, we need to develop a more elaborate mobility model, which studies the mobility patterns accurately and precisely for different environments that are useful for practical applications. Knowledge of prevalent vehicular behaviors and mobility patterns can help conduct better communication and computational resource utilization. Therefore, mobility between edge nodes and between edge and cloud can also be studied. Contrary to conventional data centers, edge devices are geographically deployed over heterogeneous platforms. The QoS across platforms must also be optimized.
In the case of routing and forwarding, many questions arise like the switching of edge servers and their services from source to destination according to the movement of the vehicle. Edge server switching: vehicles usually make decisions of their next move in a short span of time as the vehicles are constantly in motion at a blazing speed. Therefore, it is difficult to predict which particular vehicle is going to take its services from which one of the base stations or edge servers, based on traffic and public transportation information making the transition patterns of vehicles migration for predicting vehicles next location. Despite the fact that many techniques have been applied to determine the issue, still this is an open research problem and much work is needed to be done. Service switching: as discussed above, when the vehicle changes its position from one edge server to another, then the services, which it was using from the previous edge server, would be transferred to a new edge server. In [
Content caching such as prefetching and cooperative caching can be implemented in VEC. The caching contents could also include those elements that the vehicles have not requested but they catch those contents over the wireless connection. It may be useful for the vehicles to save and forward those unrequested contents (e.g., alarms generated in case of trouble). In addition to this, there are still gaps in caching policies that create the most effective temporal and spatial scope of the vehicular contents. By caching-in contents that are out of their spatial scope (e.g., emergency signals on the far-off side yet in the relevance area) and also caching-out the old contents (e.g., an hour ago traffic congestion information on a highway), furthermore, some more burning technical implications are as follows: Although vehicle-to-vehicle communication can enhance the capacity of the network in terms of content caching, yet it is still not able to validate a reliable and high-rate data service for the vehicles due to the extremely dynamic and uncertain network topology and strict channel condition. Since the SRSUs are deployed in different locations and different network operators own them, the cooperation of the SRSUs to provide contents to the vehicles must be considered in terms of the pricing model. A caching scheme needs to be developed to enhance content hit rate with the minimum handover cost, by identifying cache size splitting, prevalent content updates, and ensuring mobility-aware caching for smooth handover even with high mobility of the vehicles.
Moreover, these vehicular caching systems require such strategies, which by taking into account the topographies and network configurations efficiently explore the advantages.
A sufficient number of network elements enhance the network’s performance on a large scale. Since the deployment of the network equipment incurs a high cost, it is essential to optimally install an appropriate number of the network elements. The major concern is to find a suitable location so the efficiency of vehicular networks can be maximized. In addition to this, the cost must be optimized and also the edge servers and SRSUs should be deployed at such points where the available resources can be managed optimally. Due to varied traffic distribution in the urban environment, more servers are deployed in the congested areas. As servers play a vital role in sending traffic packets, the SRSUs present next to the servers cause the traffic packets to enter the infrastructure without any need for multihop communications. Through the infrastructure, these packets would be transferred to other nodes in the network. By accessing, the infrastructure through fewer hops exceedingly cuts the receiving time of servers that are sending messages to the other nodes. Therefore, it is ultimately required to develop an optimal model, which measures the minimum need of edge servers as well as SRSUs to be deployed in order to minimize the deployment cost and maximize the QoS.
The dynamics of the vehicular networks, their flexibility, and nonrigidness have put the security and privacy of data into question with the prime challenge being the authentication security [
In this paper, we have discussed the architecture for VEC, which was developed to support a high level of scalability, real-time data delivery, and mobility. VEC is regarded as a suitable model for vehicles as it is able to reduce latency for services that require real-time decision making. Thus, VEC also greatly enhances the computational performance as compared to traditional systems by promoting smart vehicular computing and making the best use of currently underutilized computational resources of the individual vehicle. Further, a wide range overview has been conducted on the study of previous work of VEC. Moreover, we have discussed several future directions and open challenges for academics and researchers related to this field.
The authors declare that there are no conflicts of interest regarding the publication of this paper.
This work was supported by the National Natural Science Foundation of China (61602054 and 61571066).