Ad Hoc Networks 75–76 (2018) 52–79

Contents lists available at ScienceDirect

Ad Hoc Networks

journal homepage: www.elsevier.com/locate/adhoc

Survey paper

Heterogeneous vehicular communications: A comprehensive study

∗ Abdennour Zekri a, , Weijia Jia a,b a Department of Computer Science and Technology , Shanghai Jiao Tong University , Shanghai , China b Centre of Data Science, University of Macau, SAR Macau, China

a r t i c l e i n f o a b s t r a c t

Article history: Vehicular communications have developed rapidly contributing to the success of intelligent transportation Received 27 October 2017 systems. In VANET, continuous connectivity is a huge challenge caused by the extremely dynamic network

Revised 25 March 2018 topology and the highly variable number of mobile nodes. Moreover, message dissemination efficiency Accepted 26 March 2018 is a serious issue in traffic effectiveness and road safety. The heterogeneous vehicular network, which Available online 30 March 2018 integrates cellular networks with DSRC, has been suggested and attracted significant attention recently. Keywords: VANET-cellular integration offers many potential benefits, for instance, high data rates, low latency, and VANET extended communication range. Due to the heterogeneous wireless access, a seamless handover decision V2I is required to guarantee QoS of communications and to maintain continuous connectivity between the Heterogeneous wireless networks vehicles. On the other hand, VANET heterogeneous wireless networks integration will significantly help

LTE autonomous cars to be functional in reality. This paper surveys and reviews some related studies in the Vertical handover literature that deals with VANET heterogeneous wireless networks communications in term of vertical Data dissemination handover, data dissemination and collection, gateway selection and other issues. The comparison between Autonomous cars different works is based on parameters like bandwidth, delay, throughput, and packet loss. Finally, we outline open issues that help to identify the future research directions of VANET in the heterogeneous environment. © 2018 Elsevier B.V. All rights reserved.

1. Introduction possible [4] . Recent development in automobiles and wireless com- munication technologies have enabled the evolution of ITS which Millions of persons are killed every year around the world in addresses numerous vehicular traffic issues like information dis- the road accidents. According to the World Health Organization semination and traffic congestion. (WHO) reviews fact sheet on road traffic injuries (9 May 2016), Vehicular Ad-hoc Network (VANET) is an integral element of around 1.25 million people die as a result of road traffic crashes ITS in which moving vehicles are connected and communicate each year (3400 deaths per day) [1] . Furthermore, the forecasts are wirelessly. Wireless communication technologies play an essen- even worse; it is estimated that by 2020 road traffic crashes are tial role in assisting both Vehicle-to-Vehicle (V2V) and Vehicle-to- predicted to increase to become the 7th leading reason for death Infrastructure communication (V2I) in VANET. V2V communication [2] . This obviously demonstrates that it has been a challenge to allows vehicles to communicate with each other and to share in- stop these accidents, which mean urgent actions and intensive ef- formation regarding their state (e.g., position, velocity, acceleration, forts are required to prevent and reduce car accidents as well as etc.) or information about the traffic (e.g., state of traffic lights, ac- improving road safety. cidents, traffic jams, the line works, etc.). However, V2I commu- In order to save lives, money, time, and the environment, the nication allows the cooperation between road infrastructure and Intelligent Transportation System (ITS) has been recently attracted vehicles. both academia and industry attention. It is the hope of such tech- Moreover, VANETs are used to support safety-critical applica- nologies that countries such as Japan and Sweden have publicly tions and non-safety infotainment or entertainment based appli- announced an objective of reaching “zero traffic fatality” societies cations. Safety applications such as collision avoidance, pre-crash by 2020 [3] . Intelligent Transportation Society of America (ITSA) sense or lane changing are aimed to minimize road accidents by summarizes its mission declaration as "vision zero" meaning its using traffic monitoring and management applications. Non-safety objective is to decrease the fatal accidents and delays as much as applications enable passengers to access various services like in- teractive communication, internet access, payment services, online games and information updates when vehicles are on the move. ∗ Corresponding author. E-mail addresses: [email protected] (A. Zekri), [email protected] (W. Jia). https://doi.org/10.1016/j.adhoc.2018.03.010 1570-8705/© 2018 Elsevier B.V. All rights reserved. A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 53

Research on VANET led to the approval of the IEEE 802.11p capability of LTE supporting vehicular applications is briefly as- standard [5–7] as an amendment to the well-known IEEE 802.11. sessed in [8] . On the other hand, Worldwide Interoperability for The enhanced version IEEE 802.11p forms the standards for Wire- Microwave Access (WiMAX) network has been proposed to cope less Access in Vehicular Environments (WAVE) at a frequency of with the coverage problem in VANETs [38] . Studies in [8] and 5.9 GHz. WAVE protocols (IEEE 802.11p/1609) provide interoperabil- [39] focused on the use of LTE in vehicular networks over hetero- ity between wireless devices On-Board Unit (OBU) of vehicles and geneous wireless networks. Vinel et al. in [39] compared 3GPP LTE infrastructure situated near the roads, Road Side Unit (RSU). Thus, and IEEE 802.11p/WAVE technologies to find which technology can V2V and V2I communications can be established in the vehicu- support cooperative vehicular safety applications. Mane and Jun- lar network. Despite the significant research that has been done narkar in [52] surveyed the techniques and fundamentals of in- on IEEE 802.11p, it suffers from scalability issues, limited coverage ternet access in VANET-Internet integration scenarios. They ded- area, and unbounded delay. icated to improve the performance of mobile gateways and data On the other hand, cellular networks have been developed in collection for giving priority to emergency messages dissemination. recent decades. Some of the disadvantages related to IEEE 802.11p, Moreover, the recent study in [40] reviewed V2I communication and the potential for the use of existing networks, have motivated over heterogeneous multi-tier with diverse Radio Access Technol- researchers to investigate the possibility of using cellular networks ogy (RAT) network environments. Shahid et al. in [54] presented in vehicular applications instead of IEEE 802.11p. By reviewing the a survey about different VANET technologies; whereas, a compari- characteristics of different access technologies, it is concluded that son between UMTS and LTE for vehicular safety communication at the cellular network is the best option as an alternative to IEEE intersections shown in [55] . Besides, Mir et al. in [56] presented 802.11p for supporting vehicular applications. Some of the distin- a hybrid communication system between LTE and WAVE protocol. guishing characteristics associated with LTE are high data rate, high However, hybrid approaches suitable for heterogeneous vehicular spectral efficiency, and low latency in the control plane [8] . communication combining both LTE and 802.11p were proposed In order to ensure vehicles to access the network, even in places in [44] . Various LTE-VANET collaborations were offered in [45–47] . uncovered by RSUs, existing radio access networks such as cellular The study in [36] by Zheng et al. concluded that the heterogeneous networks (3 G/LTE) and Wi-Fi may be employed to improve vehic- vehicular networking with LTE for V2I communications and DSRC ular communications. The potential impact of heterogeneous wire- for V2V communications is one of the best solutions for supporting less networks has been confirmed by an ever-increasing amount of vehicular services. mobile internet traffic, which cannot solely be absorbed by cellu- In the current study, we present a comprehensive overview lar data communication networks. Then, it can form heterogeneous of vehicular communications in the heterogeneous environment. vehicular networks that can be a combination of VANETs and cel- Therefore, in contrast with existing surveys in the literature, this lular networks for vehicular communications. work focuses on comparing various solutions proposed by the Seamless handover is the first necessary step when internet VANET research community in term of handover, data dissemi- connection needs to migrate between heterogeneous networks. nation, gateway selection, QoS and other concerns. The compari- The necessity for vertical handover can be initiated for convenience son between different works is based on metrics like bandwidth, rather than connectivity reasons (e.g., according to user choice for throughput, delay, communication overhead, channel capacity and a specific service). Two of the major challenges in vertical han- packet loss. Besides, we also explore some open issues for future dover management are seamlessness and automation aspects in investigations in heterogeneous vehicular communications. Many network switching. challenges need to be addressed to make vehicular communica- Major car companies, governmental organizations, and the aca- tions vision a commercially viable reality in 5 G networks and au- demic community have recognized the increasing importance of tonomous cars. This work would motivate VANET researchers, au- interworking over VANETs [9] . Many government projects have tomakers and newcomers to develop VANET- heterogeneous wire- been implemented in USA, Japan, and the European Union. The less networks integration technology. federal communications commission has allocated spectrum for The remainder of this paper is structured as follows. Inter-Vehicle Communications (IVC) and similar applications [10] . Section 2 presents VANET architecture and background. The stan- Governments and prominent industrial companies, such as Toy- dards for wireless access in VANET are described in Section 3 . ota, BMW, and Daimler-Chrysler, have started important projects Section 4 delineates VANET characteristics briefly. Possible het- for IVC communications. Advanced Driver Assistance Systems erogeneous vehicular communications scenarios are discussed in (ADASE2) [11] , Crash Avoidance Metrics Partnership (CAMP) [12] , Section 5 . Section 6 surveys VANET heterogeneous wireless net- Chauffeur in EU [13] , CarTALK20 0 0 [14] , FleetNet [15] , California works in different classifications such as vertical handover, data Partners for Advanced Transit and Highways (California PATH) [16] , dissemination and collection, gateway selection and others. An au- and DEMO 20 0 0 by Japan Automobile Research Institute (JSK) are tonomous car review is given in Section 7 . Open issues, chal- few notable projects, which are a significant step for the realization lenges, and future research directions are discussed in Section 8 . of intelligent transport services. Section 9 concludes the paper. In the recent years, few studies have surveyed the heteroge- neous vehicular communications. Most of the existing survey re- searchers focus on either the overview of ITS or a single net- 2. VANET: architecture and background work [8, 38–42] . Wu et al. [179] addressed the challenges of using Dedicated Short-Range Communication (DSRC) for vehicular com- Vehicular Ad-hoc Networks (VANET) are ad-hoc networks munications and proposed solutions. A comprehensive survey of where the devices making up the network are vehicles. VANETs vehicular ad-hoc networks is presented in [20] . The authors in should not be confused with Intelligent Transportation Systems [180] discussed the challenges and the solutions of connected ve- (ITS). ITS cope with all kind of communications inside the vehi- hicles. Furthermore, the heterogeneity between different networks cle, between cars or with the roadside unit, but are not limited to is essential as each system offers their unique benefits [181] . Since road transport. It also includes rail, water, and air transport. Thus, the vehicular network environment is highly dynamic, Viriyasita- VANET is a component of ITS. vat et al. in [48] analyzed the appropriate channel and propa- VANET communication system architecture comprises of three gation models for this heterogeneous system. Some studies tack- types of domains: in-vehicle domain, ad-hoc domain, and infras- led the different vehicular applications [49–51] . Additionally, the tructure domain as shown in Fig. 1 . 54 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

802.11p radio frequency channel, and it is in charge for the com- munications with other OBUs or with RSUs. The primary roles of OBU are wireless radio access, network congestion control, ad-hoc and geographical routing, reliable message transfer, data security and IP mobility.

B. Application units (AUs)

An Application Unit (AU) is an in-vehicle entity; many AUs can be worked with a single OBU and share the OBU processing and wireless resources. An AU interacts exclusively via the OBU, which manages all networking utilities and mobility on the AU behalf [17] . The AU can be connected to the OBU via a wired or wireless connection and may exist with the OBU in a single physical unit. Additionally, the AU communicates with the network just via the OBU, which is responsible for all mobility and networking func- tions.

C. Road-side units (RSUs)

A Road-Side Unit (RSU) is a device that is situated at station- Fig. 1. VANET communication system architecture. ary positions along roads and highways, or at dedicated locations such as at junctions, near parking spaces, hospitals, shopping com- 2.1. In-Vehicle domain plexes, restaurants, etc. RSU is equipped with at least a network device based on IEEE 802.11p. Internet connectivity to the OBUs is This domain is composed of an On-Board Unit (OBU) and one the main function of RSUs. Each vehicle equipped with an OBU and or multiple Application Units (AUs). Several AUs can be integrated a set of sensors collects information and sends to other vehicles or with a single OBU; the connection between them could be wired RSUs through the wireless links. RSU can also connect to the inter- or wireless. An OBU provides the communication links between net as a bridge to allow AU’s from multiple vehicles to connect to V2I and V2V. An OBU is fitted with a single network device based the internet [18] . on IEEE 802.11p radio technology. According to Car 2 Car Communication Consortium [19] , the primary functions and procedures associated with RSU are [20] : 2.2. Ad-hoc domain 1. Extending the communication range of the ad-hoc network by re-distributing the information to other OBUs and by The ad-hoc domain comprises of vehicles equipped with OBUs sending the information to other RSUs in order to forward and stations beside the roadside, Road Side Units (RSUs). An OBU it to other OBUs. can be seen as a mobile node of an ad-hoc network and RSU is 2. Running safety applications such as a low bridge warning, ac- a static node similarly. An RSU can be connected to the internet cident warning or work zone, using V2I communication and via the gateway; RSUs can communicate with each other directly acting as an information source. or via multi-hop as well. The key purpose of RSU is to provide the 3. Providing internet connectivity to OBUs. internet connectivity to the OBUs. Besides, OBUs form a mobile ad- hoc network that allows communications between vehicles with The goal of a VANET architecture is to allow the communica- no need for a centralized coordination existence [17] . tion among nearby vehicles, between vehicles and fixed roadside equipment leading to the following three categories as shown in 2.3. Infrastructural domain Fig. 2 .

a. Vehicle-to-Vehicle (V2V) communication The RSU can connect to the infrastructural networks or the internet as shown in Fig. 1 , allowing the OBU to access the in- Also referred to as inter-vehicle communication, V2V commu- frastructure network. In this situation, it is possible that the AUs nication allows the direct vehicular communication without rely- are recorded with the OBU to connect to any internet based host. ing on fixed infrastructure support. V2V can be mainly employed OBU can also communicate with other hosts for non-safety appli- for safety, security, and dissemination applications. In V2V com- cations, using the cellular radio networks communication (GSM, munication, the vehicles exchange data which can describe the ve- GPRS, UMTS, HSDPA, WiMAX, and 4 G). hicles’ internal states and environment, to expand the communi- VANET has various specific components like application unit, cation partner’s perceptual horizon. Any valuable information col- on-board unit, and road-side unit. A brief description of this ar- lected from sensors on a vehicle may be sent to neighboring vehi- chitecture is given below. cles. Hence, V2V communication is necessary to prevent fatal road accidents and has always been present in the form of automotive A. On-board units (OBUs) lightings, such as turn signals and brake lights. It appears natural Alternatively, an OBU is a device usually attached on-board a for it to evolve into V2V communication in the digital age. vehicle used for exchanging information with RSUs or with other b. Vehicle-to-Infrastructure (V2I) communication OBUs. It involves a Resource Command Processor (RCP), a user in- terface, a specialized interface to connect to other OBUs and a V2I allows a vehicle to communicate with the roadside infras- network device for short-range wireless communication based on tructure mainly for information and data gathering applications. IEEE 802.11p radio technology. Furthermore, it may include an ad- The vehicles use cellular gateways and wireless space network. ditional network device for non-safety applications based on other United State Department of Transportation (US DoT) defines V2I radio technologies such as IEEE 802.11a/b/g/n. The OBU connects communication for safety purposes as: “Vehicle-to-infrastructure to the RSU or to other OBUs through a wireless link based on IEEE (V2I) Communications for safety is the wireless exchange of critical A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 55

Fig. 2. VANET communication categories. safety and operational data between vehicles and roadway infras- tructure, intended primarily to avoid motor vehicle crashes.” [21] This wireless exchange of data can also be used to provide in- formation and entertainment services to drivers and passengers. Automotive entertainment devices could, for example, download current weather forecasts, traffic information, or music and other media. Vehicles can interact with infrastructure, for instance, to al- ter the duration of traffic light phases or to report traffic situation and the internal state of the vehicle. Besides, roadway infrastruc- ture could also be used for routing purposes in a VANET. A VANET incorporating roadway infrastructure might be connected to the internet, thereby integrating with car devices into the internet of Fig. 3. DSRC spectrum allocated by the FCC. things.

c. Hybrid architecture (V2X)

Also referred to as inter-road communication, combines both V2V and V2I communications. In this scenario, a vehicle communi- cates with the roadside infrastructure and shares the data received from infrastructure with other vehicles. According to [22–24] , two types of communications are available in the ad-hoc domain. On the one hand, vehicle communicates directly with another vehicle if there exists a direct wireless connection available between them, forming a single hop V2V communication. However, when there is Fig. 4. Multichannel operation in vehicular networks according to IEEE 802.11p Eu- no direct connection between them, a dedicated routing protocol is ropean Standard. used to forward data from one vehicle to another until it reaches the destination point, forming multi-hop V2V communication. On the other hand, vehicle communicates with an RSU to increase the not charge for usage of DSRC spectrum. DSRC supports Orthogonal range of communication by receiving, sending and forwarding data Frequency Division Multiplexing (OFDM) technique for data mul- from one node to another or to benefit from the ability of the RSU tiplexing. OFDM technique divides input data stream into parallel to process particular applications forming V2I communication. bit streams and maps individual bit streams onto overlapping sub- carriers. 3. Standards for wireless access in VANET As shown in Figs. 3 and 4 , the DSRC spectrum is divided into seven channels starting from channel number 172 ending with There are many standards used in VANET such as Dedicated channel number 184. The spectrum comprises a 5 MHz guard band, Short-Range Communication (DSRC) and Wireless Access in Vehic- one 10 MHz Control Channel (CCH) and six 10 MHz Service Chan- ular Environment (WAVE). nels (SCHs). Channel 178 is the control channel (CCH), which is used exclusively for safety communications. Additionally, channels 3.1. Dedicated short-range communication (DSRC) 172 and 184 are reserved for safety applications, while the other service channels (SCH) have for both safety and non-safety uses. In 1999, the United States Federal Communication Commission According to the ETSI institute [29] , the whole spectrum in (FCC) allocated 75 MHz of the dedicated short-range communica- DSRC is divided into time slots of 50 ms and messages have two tion (DSRC) spectrum at 5.9 GHz between 5.850 - 5.925 GHz for different priorities: low for data dissemination messages transmit- ITS [25, 26] . Similarly, in 2008, the European Telecommunications ted in the SCH channels, or high for safety or control messages Standards Institute (ETSI) allocated 70 MHz of spectrum in the transmitted in the CCH channel. If the CCH channel is active, all 5.8 GHz band [27] for DSRC applications. DSRC was started with nodes are bound to stop their communication during the CCH time the ISO Communication Access for Land Mobiles (CALM) standard- frame to receive and transmit security messages in the CCH chan- ization in 2001 and then the work on IEEE 802.11p which has been nel. finalized in 2010 [28] . Today, DSRC is used exclusively for V2V and V2I communications. DSRC technology allows high-speed communication between 3.2. IEEE 1609-Standards for wireless access in vehicular vehicles and the roadside or between vehicles within the com- environments (WAVE) munication range from 300 m to 1 Km. Furthermore, DSRC sys- tem supports a vehicle speed up to 200 km/h. This standard offers Wireless Access for Vehicular Environments (WAVE) is also half duplex and 6–27 Mbps data transferring rate. Besides, FCC does known as IEEE 802.11p. It is an approved amendment to the IEEE 56 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

802.11 standard. WAVE is required to support ITS applications in at random, they move along lanes following routes. Vehicles the short-range communications. usually follow a specific mobility pattern constrained by roads, Vehicular scenarios demand high-speed data transfer and fast streets and highways, traffic lights, speed limit and traffic con- communication because of their high topological change and high ditions. Thus, given the mobility pattern, the future position of mobility. For this, ASTM 2313 working group renames DSRC to IEEE the vehicle is more feasible to be predicted. 802.11p WAVE [30] . This works on MAC layer and physical layer. - Large-scale network: The network scale could be enormous in WAVE consists of Road Side Unit (RSU) and On-Board Unit (OBU). dense urban areas such as the city center, the entrance of the In WAVE, V2V and V2I communications use 5.850–5.925 GHz big cities and highways [27, 32] . Therefore, it becomes a big frequency range. It gives real-time traffic statistics improving the issue to identify intended receiver before data dissemination performance of VANET. WAVE uses OFDM techniques to split the since VANET may span over multiple towns or cities. signals into various narrow band channels. - Real-time requirements and hard delay constraints: Proto- cols designed for VANET must fulfill application requirements.

4. VANET characteristics Safety messages are the primary goal of VANETs. Hence, safety messages should be given high priority and must be delivered

Vehicular ad-hoc networks (VANETs) can be considered as a on time. Safety applications have stringent delay constraints as subset of mobile ad-hoc networks (MANETs). There are similar messages received, only within a specified period we can justify characteristics between them such as low bandwidth, short radio the concept of wireless communication based proactive safety transmission range, self-organization, self-management and omni- measures. Thus, it becomes extremely crucial to meet the delay directional broadcast. However, VANETs have some particular fea- requirements. tures that distinguish them from typical mobile ad-hoc networks - Geographical communication: Vehicles to be reached typically and that make their design challenging. depend on their geographic location. This issue differs from

The unique characteristics of VANETs include: other networks where an ID or a group ID define the target vehicle or a group of target vehicles. - Highly dynamic mobility and topology: A vehicular network is - Urban versus Rural: Vehicle density is another challenge, as it considered as highly dynamic due to the speed of the vehicles differs with time and location. The other obstacles to reliable and radio propagation characteristics. Vehicles have high rela- V2V communication are network fragmentation, especially in tive velocities in the order of 50 km/h in urban environments rural areas, and higher interference contributed by shadowing to more than 100 km/h in highways [178] . They may also move and fading in urban scenarios. in different directions. Thus, vehicles can quickly join or leave - Propagation model: VANETs operate in three environments: the network in a very short period of time, leading to frequent highway, rural, and city. In a highway, the propagation model and fast topology changes. is usually assumed to be free space, but the signal can suf- - Rapid changes in network topology: VANET nodes are charac- fer interference by the reflection with the wall panels around terized by their high speed. High speeds typify moving vehicles, the roads. In a city, its surroundings make the communication especially on the highway leading to frequent and rapid net- complex due to the variable vehicle density and the presence work topology changes. Moreover, vehicles on the road change of buildings, trees, and other objects, acting as obstacles to the their position and speed very frequently and thus network signal propagation. Such obstacles cause shadowing, multipath, topology changes very dynamically. The lifetime of the link be- and fading effects. Usually, the propagation model is assumed tween vehicles is affected by the radio communication range to not be free-space due to those characteristics of the commu- and the direction of the vehicles; thus expanding the radio nication environment. In rural environments, due to the com- communication range leads to an increase in the lifetime of the plex topographic forms (fields, hills, climbs, dense forests, etc.), link. The lifetime of the link between vehicles moving in op- it is essential to consider the signal reflection and the attenu- posite directions is very short-lived compared with the case in ation of the signal propagation. Therefore, in this scenario, the which vehicles move in the same direction, which may result free-space model is not appropriate. As in any other network, in frequent link ruptures and higher packet loss with reduced the propagation model in VANET must consider the effects of throughput. The rapid changes in link connectivity cause the potential interference of wireless communication from other effective network diameter to be small, while many paths are vehicles and the existence of largely deployed access points. disconnected before they can be utilized [31] . - High computational ability: Because the nodes in VANET are ve- - Link availability is low (frequently disconnected): Due to highly hicles, they can be equipped with a sufficient number of sen- dynamic topology, high speed, and the limited transmission sors and computational resources, such as processors, a large range, the link between two vehicles can quickly disappear memory capacity, advanced antenna technology and Global Po- while they are transmitting the information. The link availabil- sition System (GPS). These resources increase the computa- ity is low (less than 1 minute), not only for vehicles moving tional capacity of the nodes, which help to obtain reliable wire- in opposite directions but also for vehicles driving in the same less communication. Additionally, these resources help to ac- directions. quire accurate information regarding the current node position, - Variable network density: The network density in VANET varies speed, and direction [34] . depending on the traffic density which means it depends on - Network administration: VANET operates in a distributed fash- the time and the area. The network density can be very high ion without any centralized administrative entity. Lack of cen- in the case of a traffic jam, or very low, as in suburban traffic. tral administrator places extra responsibilities on nodes. Fur- At rush hours the traffic is high, and it is usually low in rural thermore, in VANET, nodes will be equipped with systems with areas [27,32] . adequate storage capacity. - No power constraints: Unlike mobile ad-hoc networks, the power in VANET is not a critical challenge because vehicles can 5. Heterogeneous vehicular communication scenarios provide continuous power to the OBU via the extended life bat- tery [32–34] . Many wireless communication systems have been considered to - Constrained and predictable mobility: VANET differs from other support ITS services. It is difficult to provide satisfactory ITS ser- types of mobile ad-hoc networks in which nodes do not move vices only through a single wireless network, due to the dynamic A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 57 topology changes of VANETs and high mobility. Furthermore, the 5.2.2. Cellular networks authors in [182] showed that DSRC offers poor performance with a There are two transmission modes in cellular networks which large number of vehicles. However, LTE networks are easily over- can be utilized for V2I communications, namely unicast and mul- loaded with an increasing number of vehicles. Hence, heteroge- ticast/broadcast. Unicast is point-to-point communication between neous vehicular communications are expected to meet several re- a vehicle and the base station; it can be used for both uplink and quirements of ITS services, by integrating different wireless access downlink message dissemination. However, multicast/broadcast is networks such as DSRC and LTE. a point-to-multipoint transmission, which is specifically used for Various wireless access technologies can be used to support V2I the dissemination of downlink messages. On the one hand, Wide- and V2V communications. Thus, it is a challenge to select the suit- band Code Division Multiple Access (WCDMA) system can be used able and the efficient technology that meets the QoS metrics for for V2I communication. However, this system cannot well support a Vehicular User (VU). Zheng et al. in [36] presented a new layer, safety services in vehicular communications, since the delivery la- namely the Heterogeneous Link Layer (HLL). HLL operates on the tency is larger than the requirement. On the other hand, LTE is top of the MAC layer in each radio access network and offers a an excellent platform to support V2I communications. The flat ar- unified interface to the higher layers. This new layer can meet the chitecture of the LTE system is designed to the low transmission QoS requirements by facilitating coordination between several ra- latency. Besides, the evolved Multimedia Broadcast and Multicast dio networks. Service (eMBMS) is an efficient way to support multicast or broad- cast services in high-density vehicular environments. LTE networks 5.1. V2V communication can provide high capacity with wide coverage. However, several problems have to be solved. As we can see in Section 2 , V2V communication allows the di- The advantages and the challenges of using DSRC and LTE for rect vehicular communication without relying on fixed infrastruc- V2V and V2I communications are summarized in Table [1] . Since ture support. In this subsection, two candidate techniques for V2V the collaborations between heterogeneous networks are essential, communications are discussed. the study in [36] proposed the new layer HLL. Different candidate technologies can work together under this 5.1.1. DSRC new architecture. Finally, from Table 1 , it is clear that DSRC is more DSRC is a useful technology to support both safety and non- suitable for V2V communications than LTE D2D. On the contrary, safety services in V2V communications. V2V communications LTE cellular network is the best choice for V2I communications. based on DSRC do not interfere with cellular networks due to the use of different frequency bands. Nevertheless, there are still many 6. VANET integration with various heterogeneous wireless challenges to using DSRC in V2V communications [42, 183] . For in- networks stance, collisions occur so frequently in a densely populated vehic- ular environment, due to the limitation of the CSMA mechanism. Heterogeneous networks use different radio access technolo- gies (RATs) such as Wi-Fi, WiMAX, 3 G and 4 G networks and var- 5.1.2. LTE D2D ious cell formats. In the literature, many studies have surveyed In Device-to-Device (D2D) communication, the User Equipment the vehicular networks domain [20] . However, in the recent years, (UE) can directly communicate with each other [184] . Various chal- a few publications have focused on the heterogeneous environ- lenges are facing D2D communications in LTE as a candidate tech- ment of the vehicular communications [36–42] . Studies in [8] and nology supporting V2V [185, 186] . Firstly, interference is a major [39] focused on the use of LTE in vehicular networks over het- concern when employing D2D in the heterogeneous vehicular en- erogeneous wireless networks. Vinel at al. in [39] compared 3GPP vironment. Secondly, in contrast with medium or high speeds of LTE and IEEE 802.11p/WAVE technologies to find which technology vehicles, most D2D devices in LTE systems are usually static or can support cooperative vehicular safety applications. Superior net- with low-speed mobility. This issue may decrease the performance work capacity is offered by LTE with greater mobility support and of D2D communications. Furthermore, existing peer and service higher coverage, but LTE suffers from higher latency with an in- discovery of D2D communications are not well suited for vehicular crease in network load in comparison to 802.11p [43] . However, environments. Finally, D2D discovery is a time-consuming proce- hybrid approaches suitable for heterogeneous vehicular communi- dure. Hence, in many cases, the D2D discovery time is larger than cation combining both LTE and 802.11p were proposed in [44] . Var- the message transmission time, which is not acceptable for deliv- ious LTE-VANET collaborations were offered in [45–47] . The study ering safety messages with severe QoS metrics. in [36] by Zheng et al. concluded that the heterogeneous vehicu- lar networking with LTE for V2I communications and DSRC for V2V 5.2. V2I communication communications is one of the best solutions for supporting vehic- ular services. V2I communication allows a vehicle to communicate with the Since the vehicular network environment is highly dynamic, roadside infrastructure mainly for information and data gathering Viriyasitavat et al. in [48] surveyed the appropriate channel and applications. The cellular networks are widely proposed to support propagation models for this heterogeneous network. Some stud- V2I communications. However, using DSRC in V2I is another solu- ies investigated the different vehicular applications [49, 50] . Simi- tion. larly, the study in [51] classifies the existing VANET context-aware applications into three dimensions: the environment, system-and- 5.2.1. DSRC application, and context-awareness. In order to meet the requirements of vehicular communications, Various wireless technologies are used with VANET for effi- a group of standards is defined by the IEEE 1609 working group cient, robust and flexible internet access solution from a VANET. for DSRC networks. Additionally, the Enhanced Distributed Chan- Mane and Junnarkar in [52] surveyed the techniques and funda- nel Access (EDCA) mechanism in IEEE 802.11e is adapted to satisfy mentals of internet access in VANET-Internet integration scenarios. the QoS requirements of the MAC layer [187] . Many broadcast pro- This work dedicated to improve the performance of mobile gate- tocols proposed in the literature for DSRC safety broadcast services ways and data collection for giving priority to emergency messages [188–193] . Hence, there are several challenges of DSRC networks dissemination. However, Gerla et al. in [53] identified the urban in- need to be addressed when used for V2I communications. ternet infrastructure role in support of emerging vehicular applica- 58 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

is

in

in

high

D2D

in

users

discovery scheduling

by

disseminating

services

users degradation

between

in

ITS

vehicles service

overloaded

caused efficient

for other

state

to environments

delays of and

speed and

idle

Prone Interference High Lack performance Peer

density pair - - time-consuming high - the - schemes messages - Challenges -

D2D

on for

downlink flat

eMBMS

efficiency

efficiency and

and

scheduling

mechanisms management

.

uplink spectral energy coverage

[36]

Effective High Robust High Centralized High-efficiency High

-Large - resources - capacity - - - architecture mobility - - Advantages from

adapted

in

as with with

broadcast broadcast service

vehicles vehicles leakage

and and operations and

design design of of

link

band

congestion congestion

node node pilot pilot

D2D

problems problems

number number communications

Sparse Hidden Adjacent Channel Channel Sparse Hidden

- storm multi-channel - - - large - large -Prioritization -Unbalanced selection storm - Challenges LTE/LTE - vehicular

etc.

low low

Short

signal, and and message parking low

is heterogeneous

i.e.,

Wave

local

traffic the

of

for

mode in (WSM)

deployment deployment

Easy Suitable Easy Overhead Ad-hoc

costs - costs dissemination, - information, Advantages - Message - - DSRC candidates

main

of

mode

challenges

and

Communication 1 Communication

V2I Communication V2V Table Advantages A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 59

Table 2 Main wireless technologies for heterogeneous vehicular communications as adapted from [8] and [40] .

Feature Wi-Fi 802.1p UMTS LTE LTE-A

Frequency band(s) 2.4 GHz, 5.2 GHz 5.86–5.92 GHz 70 0–260 0 MHz 700–2690 MHz 450 MHz-4.99 GHz Channel Width 20 MHz 10 MHz 5 MHz 1.4, 3, 5, 10, 15, 20 MHz Up to 100 MHz Range Up to 100 m Up to 1 Km Up to 10 Km Up to 30 Km Up to 30 Km Bit rate 6–54 Mb/s 3–27 Mb/s 2 Mb/s Up to 300 Mb/s Up to 1 Gb/s Coverage Intermittent Intermittent Ubiquitous Ubiquitous Ubiquitous Mobility support Low Medium High Very high (up to 350 Km/h) Very high (up to 350 Km/h) V2I support Yes Yes Yes Yes Yes V2V support Ad-hoc Ad-hoc No No D2D

tions and discussed the future of the mobile internet in vehicular handoff. However, the various handover requirements and the im- networks. portant factors that affect the handover process were discussed Moreover, the recent study in [40] surveyed V2I communication in [67] . Multiple parameters involved in VHO process were also over heterogeneous multi-tier with diverse radio access technol- presented. Advanced methods of VHO decisions were investigated. ogy (RAT) network environments. Shahid et al. in [54] presented Various metrics can be adopted to trigger handover decisions in- a survey about different VANET technologies, whereas, a compari- cluding RSS measurements, QoS parameters, and mobile terminal son between UMTS and LTE for vehicular safety communication at location information. Most of the research is based on single crite- intersections given in [55] . Table 2 summarizes the main wireless ria handover decision-making algorithm. Similarly, Midya et al. in technologies for heterogeneous vehicular communications. [68] made a detailed comparison between different VHO mecha- Furthermore, Mir et al. in [56] presented a hybrid communi- nisms based on parameters like handoff delay, addressing, technol- cation system between LTE and WAVE protocol. Moreover, some ogy and handoff decision. studies [57, 58] suggested the use of satellite communication in vehicular network communications. Li et al. in [59] proposed the 6.1.1. Vertical handover algorithms use of satellite communication in the downlink only while using A VHO decision algorithm that maximizes the collective battery cellular systems in the uplink. lifetime of Mobile Nodes (MNs) was developed in [69] . This algo- Currently, many studies are addressed the seamless connectivity rithm allows the proxy nodes, which provide connectivity to the in heterogeneous multi-RAT environments. Most of these studies nearest AP or BS for the ad-hoc-mode MNs, to share transit loads emphasized on performing seamless vertical handover, data dis- with the goal of balancing their consumption of battery power. semination, and mobility, while others focused on security issues. The proposed algorithms perform much better than the conven- tional optimization based on the SSF method, which is based on 6.1. Vertical handover RSS alone. Deployment of a System Selection and Mobility Man- agement Agent (SSMMA) on both vehicles and networks is pro- Since the vehicles are always in a move, a Vertical Handover posed as a solution to handover issues in [41] . (VHO) between the vehicle and the infrastructure in V2I commu- For the switching problem in a vehicular heterogeneous net- nication is a necessity. However, increasing the number of han- work interface, Kang et al. in [70] presented an optimized vertical dovers introduces ping-pong effects. Some papers in the literature handoff algorithm. Multiple metrics are integrated with the pol- [60, 61] surveyed the VHO in VANET heterogeneous wireless net- icy including monetary cost, available bandwidth, and data transfer works. delay. Besides, it can accurately predict the distribution of RSUs. Furthermore, a new handover mechanism relies on the VANET The results show that the optimal stopping-based policy signif- multi-hop communication model proposed in [62] . This mecha- icantly outperforms several existing systems. Similarly, advanced nism minimizes layer-3 latency. Chen et al. gave this approach methods of VHO decision algorithm to select the best network by where a vehicle leaving the network passes its IP address informa- comparing multiple parameter values (such as network traffic and tion towards the back cars entering the cell coverage. A fast han- vehicle speed) were used in [71] . The number of handover and la- dover may also be achieved by predicting the movement of the tency can be minimized, whereas throughput can be maximized, vehicle [63] . For interference management, further improvements considering vehicle speed and network traffic parameters. have been proposed in [59] . Li et al. investigated the uplink in- Bi et al. [72] provided a performance guaranteed optimized terference by applying a Coordinated Multi-Point (CoMP) strategy, handover decision algorithm to solve the handover problem which where the user equipment in multi-coverage region chooses a cell is ubiquitous in heterogeneous networks. With this algorithm, ve- with a better link quality. hicles can handover through the heterogeneous environment, not In multi-tier RAT heterogeneous environments, it is a priority only to reach overall load balance among all access points but also for a moving vehicle to choose the appropriate target roadside and to maximize the data rate as well as the vehicles’ fairness. Addi- not the wrong one. Situations like packet losses, ping-pong and tionally, in the process of decision making, the data rates of han- even call drops can happen; a seamless and reliable handover is dover vehicles are estimated. necessary for these environments [64] . On the other hand, WiMAX technology was proposed to sup- port high-speed V2I communication since the performance of 6.1.2. Vertical handover approaches 802.11p may be impeded by the non-line-of-sight conditions. IEEE A cross-layer fast handover scheme was presented by Chiu et al. 802.16 m WiMAX version can support mobility up to 350 km/h in [73] , called Vehicular Fast Handover Scheme (VFHS). In this sys- with an expected handover latency of less than 30 ms [65] . tem, the physical layer information is shared with the MAC layer The essential phases presented in a VANET handoff process to reduce the handover delay. As a consequence of this scheme, were examined in [66] and some similar studies in the literature a rapid handover and a significant decrease in handover latency that decreases the handoff time for VANETs were analyzed. This and packet loss are shown. Ahmed et al. in [74] proposed an effi- work also reviewed the fast handoff schemes proposed in previ- cient network assisted handover scheme which decreases the scan- ous works to improve the different phases presented during the ning time faced by vehicles during the handover procedure. The 60 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 selected RSU for a vehicle is based on its velocity during the han- is a need of evaluating these techniques in a more realistic scenario dover. However, the existing solution needs a vehicle to perform and applying them to an actual wireless situation. the scanning until it finds all the neighboring RSUs, which causes Furthermore, Gramaglia et al. [80] proposed Seamless Internet an additional scanning delay. Through the results, it is observed 3 G and Opportunistic WLAN Vehicular Internet Connectivity (SIL- that the proposed scheme efficiently decreases the overall scanning VIO). It is a solution for providing internet connectivity in multi- delay and iterations usually faced by vehicles. hop vehicular ad-hoc networks. By using SILVIO, the users can ben- A new architecture, which supports seamless mobility of mo- efit from a higher bandwidth, while the operators can alleviate bile network across a heterogeneous network, was offered in [75] . their overloaded cellular networks. SILVIO provides seamless con- By using this system, the overlapped reception of the packet from nectivity without signaling overhead. Moreover, the obtained re- diverse Access Router’s (AR’s) significantly reduces packet losses sults of real traffic traces from Madrid City show that the cellu- during handover even without decreasing handover latency. Sim- lar network can be offloaded by a factor up to 80% using SILVIO. ilarly, Meneguette et al. [76] developed a Seamless Flow Mobil- Additionally, a new hybrid interworking scheme was proposed in ity Management Architecture (SFMMA) to provide a better qual- [81] , which enables access to mobile internet and general IP ser- ity of service for vehicular applications. This architecture is based vices over a mobility management mechanism. Furthermore, the on vehicular network application classes with network-based mo- solution focuses on urban vehicular scenarios and enables seam- bility management. The proposed SFMMA deals with several net- less communications regardless of roaming agreements between work interfaces at the same time, seeking to decrease the han- network operators. This novel hybrid scheme allows a seamless dover time, to maximize network throughput, and to satisfy mini- transferring of IP sessions, despite different patterns of mobility mum requirements of latency and packet loss for each class of ve- and the heterogeneity of the supporting radio access technolo- hicular network application. As a result of this work, the new ar- gies. The performance analysis has shown that the offered solu- chitecture presented a low handover time, with lower packet loss tion outperforms other protocols, such as the optimized version of and lower delay. Furthermore, Chung et al. [77] proposed WAVE MIPv6, NEMO BS, the standard HIP, and Novaczki’s micro-mobility Point Coordination Function (WPCF) protocol which is a time co- scheme for HIP. On the other hand, an extensible simulation envi- ordinated Medium Access Control (MAC) protocol for improved ronment was developed in [82] combining a simulation tool NS2 V2I communications and handover. The WPCF protocol can sig- and a user mobility model based on VANET-Mobisim. Media Inde- nificantly enhance the utilization efficiency and can support more pendent Handover (MIH)-based VHO scheme and multi-mode node users when compared to Point Coordination Function (PCF), En- model are established in NS2. The communication performances hanced Distributed Channel Access (EDCA), or Hybrid Coordination of multi-mode VUs in the heterogeneously integrated network of Function (HCF) controlled channel access (HCCA) for V2I communi- UMTS and WLAN are evaluated. The velocity of VUs plays a lit- cations. Hence, the proposed scheme can significantly improve the tle effect on the throughput of VUs. However, the frequent handoff handover latency compared to the performance of existing proto- resulted from the high speed of users may degrade user QoS in cols. Furthermore, a contention-free WPCF handover channel ac- terms of packet loss and handoff latency. Furthermore, while the cess method to support high priority real-time data access has light traffic load of other users may not affect user performance been proposed. The WPCF mechanism also eliminates any wasted severely, the heavy traffic load will degrade user performance, es- time slots that a Time-Division Multiplexed (TDM) transmission pecially in WLAN. Consequently, the handoff latency resulted from mechanism would have. Overall, the preregistration mechanism upward handoff is much larger than that of downward handoff. and the efficient stable scheduling of the proposed WPCF scheme On the other hand, Ghosh et al. [83] presented a new VANET during handover results is much smaller service discontinuity time testbed which is being deployed at Middlesex University, London. compared to conducting handover using EDCA. This work shows a new performance model for proactive handover, Dias et al. [78] proposed a mobility approach that integrates ex- which is then compared with traditional approaches. To accurately tended mobility protocols based on MIPv6 and PMIPv6, with a mo- calculate useful values of TBVH and NDT, a probabilistic approach bility manager that provides seamless communication between ve- based on accurate propagation models from a real testbed is re- hicles and the infrastructure. This method can select the best tech- quired. Furthermore, based on realistic TBVH and NDT values, this nology to maintain the vehicle connected without breaking any ac- work shows also how these can be used for proactive channel al- tive sessions. The proposed architecture deals with both Layer 2 location. and Layer 3 handovers. For Layer 2 handover, a mobility manager Further handover research works, Joseph and Rajagopal in that scans the available networks and triggers the handover was [84] proposed a new system provides efficient handoff between designed; whereas, for Layer 3 handover control, the mobility pro- WLAN and WiMAX and makes a large coverage and multimedia tocols were enhanced and modified to be coupled with the mo- application to the vehicle. By using Elliptic Curve Diffie-Hellman bility manager. Experiments were done in real vehicular environ- (ECDH) cryptography with a public key, location preservation and ments combining three technologies: IEEE 802.11p, IEEE 802.11 g, high security of vehicle are done. Table 3 compares different verti- and 3 G. The results demonstrate the advantages of a simplified cal handover works in the literature. communication standard for VANETs, since the traditional Wi-Fi 6.2. Data dissemination and collection standards introduce high handover latency and packet loss, being it critical to deploy IEEE 802.11p-enabled RSUs in high demand- OBUs in vehicles communicate with RSUs through beacons ing scenarios. The results also show that, if IEEE 802.11p is used in which are short message exchanges. Diverse information are ex- both vehicles and RSUs, the proposed approach can perform seam- changed between vehicles, such as velocity, acceleration, move- less handover with very low delay and no packet loss. Moreover, ment directions, geographical location, etc. RSUs periodically enhancement of FMIPv6 was presented in [79] . This improvement broadcast this information to the other vehicles. There are numer- is based on a handover management technique using the concept ous challenges related to data dissemination in V2I communica- of tunneling in a VANET scenario. Numerous parameters like Han- tion. Some studies in the literature surveyed data dissemination dover latency, signaling overhead, performance comparison using and collection in VANET [85–89] . tunneling, packet loss, service disruption time and network life- time were taken into consideration. There is a reduction in the 6.2.1. Data dissemination handover latency, packet loss, signaling overhead, number of pack- Joshi et al. in [90] analyzed the heterogeneous configuration ets required for handover and service disruption time. Hence, there which has better performance than the homogeneous configura- A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 61

rate

security data

data rate

Maximize High NA NA NA NA NA NA NA Data NA NA

throughput throughput

NA NA NA Maximize NA NA NA NA Maximize NA Throughput NA

data

real-time

delay

delay delay

priority low delay

Reduce NA NA NA Very High NA NA Low Reduce Delay NA

bandwidth

higher

NA NA NA NA NA NA A NA NA Bandwidth NA NA

losses loss

loss

loss

packet

loss packet

packet loss

packet

packet

NA NA NA No NA NA NA Decreases Reduce Low Minimizes Packet

latency

handover

latency latency latency

latency

reducing

latency

Available). NA NA Minimize NA Without NA Reduce Minimum Decrease Handover NA NA

Not

2012 2014 2014 2016 2013 2014 2014 2013 2009 Year 2011 2011 (NA:

comparison (SFMMA)

[84] (SILVIO)

[76] (WPCF)

[80] works’

(VFHS)

[71] al.

[74] al. [79] [75]

[77]

et

al.

Rajagopal

[73]

et [78] al. al.

al.

al.

et

[72] al. et al.

name et et and

et

handover

al. et et

3

et

Kumaran Dias Joseph Shukla Bi Ahmed Meneguette Author Chiu Gramaglia Chung Dahiya Vertical Table 62 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 tion for VANET dissemination scenario. The packet loss in hetero- lower negative impact on the conventional cellular traffic and H2H geneous configuration is lesser than homogeneous configuration; traffic. Therefore, H2H traffic and FCD traffic can be served by the other parameters like back off time and busy time are also small. LTE network simultaneously and have a relatively smaller impact Moreover, Zhang et al. in [91] proposed the mathematical formula- on each other. However, the effect of FCD on the LTE H2H traffic is tion of heterogeneous vehicular wireless networking and analyzed reduced at the expense of increasing the overhead on the VANET. the effect of bandwidth aggregation on the problem formulation as Similarly, Alotaibi and Mouftah in [98] proposed the Area Defer well as the complexity of computation. This work has suggested ef- Transmission (ADT) dissemination algorithm for the transmission ficient approaches to solve the problem optimally when bandwidth range in vehicular networks. ADT allows each vehicle to freely aggregation is employed. Additionally, the authors suggested both decide either to transmit or to stop transmission taking into ac- optimal and approximate strategies to solve the problem when count heterogeneous transmission ranges and the amount of cov- bandwidth aggregation is not applied. ered transmission area. The algorithm does not depend on network For efficient data dissemination in heterogeneous vehicular net- topology information; instead, every node makes independent de- works, some approaches were proposed in the literature including cisions by drawing on information in the message and local data the Multimedia Broadcast Multicast Services (MBMS and evolved at the receiving vehicle. Moreover, ADT orders nodes transmission MBMS), push and pull-based protocols, the hybrid network infras- according to the amount of area that would be covered by poten- tructure and fog computing [40] . Generally speaking, both eMBMS tial new transmission. The results prove that more vehicles receive and fog computing are still in their early stages and look promising a message transmitted by a single relay vehicle with ADT than for multitier 5 G networks in the future [92] . with other distance versions. Thus, relay nodes selected by ADT are Baiocchi and Cuomo in [93] reviewed an architectural model more effective than others, owing to their higher delivery ratios to provide location-based on push-mode dissemination services and propagation speeds and fewer hops that reach long distances. through VANET and 3 G/4 G systems. By using this model, location- As an improvement for ADT method, the same authors devel- based services can be provided to mobile users both when they oped their works in EADT scheme [99] . The enhancement gives are in the RSU coverage area and when they move out of this area. the priority of retransmission according to the amount of area Moreover, a definition of dissemination algorithms and a perfor- that would be covered by potential new transmission, while in mance comparison of those algorithms on an urban map modeling ADT priority was given according to the overlapped area between an area in Roma were provided. Therefore, it is possible to extend transmitter and receiver transmission ranges. The simulation re- the RSU area and to reach a significant percentage of vehicles with sults prove that the dissemination speed in EADT approach im- a quite reduced delay, by adding some geometric rules for the dis- proves dramatically, with better saved-retransmission and delivery semination of messages. ratio, compared to ADT. An effective heterogeneous solution was proposed in [94] to To achieve seamless data connectivity for vehicular multime- solve the congestion control problem in IEEE 802.11p-VANET, with dia communication services, an efficient multi-metric Cluster Head the support of eMBMS in LTE cellular systems, especially in (CH) election mechanism was proposed in [100] to manage VANET- crowded traffic situations. An exponential algorithm used to make LTE integrated sub-clusters. The proposed architecture allows LTE surrounding vehicles’ beacon rates decrease to save more CCH to effectively schedule multimedia sessions based on the service wireless resource for event-driven messages when an emergency requirements of the VANET gateways, thus satisfying QoS. The inte- event occurs. Even cars without IEEE 802.11p devices can also grated system shows acceptable values concerning LTE throughput be informed of emergency information with the support of eM- and end-to-end delay. Furthermore, Zhou et al. in [101] demon- BMS. Thus, timely dissemination of event-driven messages can be strated an effective solution aimed to enable multicast/broadcast guaranteed, and the average packet delivery ratio (PDR) in IEEE service provision in integrated VANET-cellular networks based on 802.11p-VANET can be improved. a two-phase cooperative transmission strategy. The proposal com- Rubin et al. in [95] studied the reliable delivery of the safety bines advantages of these two heterogeneous networks and the na- message flows across a hybrid of mBS based cellular access net- ture of multicast/broadcast technology so that the traditional prob- work and VANET. This work proposed a VANET scheme working lems in VANET or cellular networks might be better solved. under a Vehicular Backbone Network (VBN) through which vehi- On the other hand, network coding has a crucial role in the cles that are positioned close to adequately selected nominal loca- area of VANETs, to satisfy the challenges like high mobility, quickly tions along a linear highway segment are elected to serve as relay changing topology and intermittent connectivity. As a definition, it nodes. The system can employ vehicular CSMA/CA access schemes is a data processing technique in which the flow of digital data to emulate the operations of the system well when managed by is optimized in a network, by transmitting a composite of two or the use of spatial reuse TDMA schemes. The system parameters more messages to make the network more robust. The authors equalize the throughput rates incurred across the cellular wireless in [194] surveyed the network coding schemes in VANETs. They access and VANET components of the hybrid network system. classified different applications in which these schemes are imple- On the other hand, He et al. addressed the data transmission mented. Furthermore, network coding is also defined as a method problem in heterogeneous vehicular networks [96] and proposed of transmitting messages across the network, which allows the re- a software-defined network (SDN) based heterogeneous commu- covery of lost data [195] . nication coordination approach. With SDN, the network resources Network coding is essential for data dissemination in VANETs. can be well managed. Moreover, the SDN-based wireless commu- Transmitted data is encoded and decoded in order to increase net- nication solution can schedule different network resources to re- work throughput, reduce delay, and make the network more ro- duce communication cost. The results validate the efficiency of this bust [196] . Moreover, network coding has other advantages such novel scheme. as robustness against link failures, reducing packet loss, decreasing Furthermore, many transmission approaches were proposed. A data transmission time, and enhancing security by sending scram- novel Floating Car Data (FCD) transmission scheme for LTE VANET bled data. A comparison between various data dissemination and heterogeneous network was designed by Jia et al. [97] . The authors collection works in the literature is presented in Table 4 . evaluated the effect of FCD transmission in LTE network and LTE- VANET heterogeneous network on human-to-human (H2H) traf- 6.2.2. Data collection fic by using a Markovian model and compared this scheme with For data collection, Drira et al. focused on vehicles data col- previous approaches. As results, the novel scheme has a relatively lection using 3 G/LTE [102] . For evaluating the system in term of A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 63

network

emission vehicle

G

on and

overhead LTE/3

overhead

communication

on

impact

the

the

QoS

impact consumption

overhead NA estimates NA fuel Low NA NA NA NA Increasing Reducing Satisfy NA NA Significant Communication NA

values

decision

rate

rate

rate

node data ratio information

throughput throughput

data /

beacon

throughput the

problem loss

traffic rate

Good transfer NA Improve Decrease Equalize Acceptable NA NA Solve NA NA Minimum Independent Data Highest

priority

delivery priority

area values

packet delay

delay

in

delay

Improve NA NA NA Retransmission Acceptable Reduce NA NA Good Overlapped NA NA Low Delay

the

in

maps

data is

city of

delivery ratios

in of

time time

dissemination

performance time

error

Packet types

traffic the delivery

equilibrium delivery

in

good low

Guaranteed Dissemination different Better Minimum packets dissemination NA speed NA NA Higher Efficient Perfect Improve Very Very Dissemination Improve NA

and

control

speed

area usage

Available).

RSU

Not

congestion

the

(NA: the reachability propagation

bandwidth

Extend problem NA NA NA NA reachability NA NA Solve NA NA High NA Low High Bandwidth

comparison

2013 2016 2013 2011 2016 2016 2014 2014 2015 2016 2016 2014 2014 2015 Year

works’

(ADT) (EADT)

(RBRA)

collection [99] [98]

(THOR)

[93]

(DISCOVER)

.

and [106]

(eMBMS) [103] [111]

[105] [107]

Cuomo

[100]

Mouftah Mouftah

al. [95]

Tyagi

[102] al.

al Wu [94]

[108] al. et

al.

and [97] et et al.

and and al.

name and

et al. and

et

al. et

et

et 4

Felice

dissemination et

Shukla Baiocchi Mir Turcanu Sivaraj De Yang Agrawal Alotaibi Author Rubin Alotaibi Zhang Drira Jia Data Table 64 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 efficient use of bandwidth, an adaptive data collection scheme is over, a novel incentive approach to solve the problem of traffic in- developed and described using only a 3 G/LTE communication net- formation transfer in VANETs was designed by Zhang and Wu in work. This adaptive data collection scheme is based on a proactive [111] . Under this scheme, all the vehicle users will be willing to approach using variable polling periods depending on the vehicle contribute to the traffic information trading game and to behave positions in the network and travel time. Thus, the system will in harmony with a unique subgame perfect equilibrium. provide accurate traffic and travel time information to the Traf- fic Management Center (TMC). Using taxi traces in Qatar, the re- 6.3. Gateway selection sults show that proactive schemes produce the lowest delay and bandwidth usage but the highest loss ratio. However, it has a sig- Benslimane et al. in [9] introduced a novel architecture that in- nificant impact on vehicle fuel consumption and emission esti- tegrates 3 G/UMTS networks with VANET. By using this design, a mates. Similarly, a new algorithm depends on LTE and DSRC called minimum number of optimal gateways at an instance is selected to THOR (Traffic monitoring Hybrid ORiented service) was proposed connect ordinary vehicles with the UMTS network. Besides, bottle- in [103] . This algorithm can combine two different wireless tech- necks and congestion across the path towards a single gateway can nologies and deliver real-time information about vehicular traffic be eliminated. The signal strength of vehicles, route stability, and monitoring and incident detection. Results show that THOR col- mobility features are all taken into consideration when clustering lects the needed information with a very low error and a rel- vehicles and selecting vehicle gateways. Similarly, another hybrid atively small impact on the existing LTE/3 G network. The study architecture VMaSC-LTE that integrates 3GPP/LTE networks with in [104] presented new Big Data Processing and Mining (BDPM) IEEE 802.11p was introduced in [112] . The new algorithm is called methods for next-generation intelligent transportation systems in Vehicular Multi-hop algorithm for Stable Clustering (VMaSC). Fur- real-world scenarios. The effectiveness of decentralized cooperative thermore, this novel scheme has the objective of achieving low de- BDPM methods is evaluated using real-world data models from the lay and high data packet delivery ratio while keeping the cellular city of Hannover, Germany. architecture usage at a minimum level. Moreover, this approach re- On the other hand, a new protocol ‘DISCOVER’ can be used both duces overhead and decreases the number of cluster heads while for data dissemination and collection in a given city area was pro- increasing their stability. As a result, the proposed architecture can posed recently in [105] . DISCOVER is disseminated and adaptive achieve higher required reliability of the application calculated by to the diverse levels of vehicular traffic density. This new proto- the data packet delivery ratio at the cost of higher LTE usage. col achieves very high performance in various types of city maps Furthermore, Eltahir et al. in [113] investigated VANET-cellular (New York, Paris, Madrid, and Rome) when compared with other networks integration and classified the various approaches. More- approaches. over, a comparative analysis of the integration techniques was pre- sented. The results show that 3GPP LTE networks were found to 6.2.3. Dissemination and routing algorithms be suitable for integrating VANET with cellular networks due to its Many dissemination and routing algorithms were offered in the acceptable results in terms of throughput and end-end delay. Addi- literature. Shukla and Tyagi proposed RBRA (Road-Based Routing tionally, some methods used to support the communications trans- Algorithm) for the Heterogeneous Wireless Network (HWN) [106] . port of Floating Car Data (FCD) message flows in LTE-VANET hybrid This new algorithm comprises of two processes: road-based rout- architecture were presented in [114] . This new approach leads to a ing discovery process and load distribution process. The results significant reduction in the LTE channel capacity required for such show that HWN with RBRA algorithm has minimum request block data transport when compared with the capacity required by using rate and transmission time. For the rural highway, Agrawal et al. the previous techniques. The collection of FCD is accomplished by have combined three wireless technologies: WLAN, cellular net- having each vehicle connect through its own established channel works, and WiMAX to offer seamless connectivity in the HWN ar- across the LTE Radio Access Network (RAN). However, the estab- chitecture [107] . They also proposed a new position based rout- lished channel can be used during the full process or only during a ing algorithm. This algorithm helps in the efficient delivery of data SETUP period. The results confirm the significant performance gain, packets from source to destination nodes on the road and presents as expressed in terms of the saved number of LTE Resource Blocks a significant performance enhancement in comparison to AODV (RBs). and GPSR routing protocols. However, Mir et al. in [108] gave us On the other hand, a novel gateway selection algorithm was a location-based routing scheme for integrated VANET-LTE hybrid addressed in [115] . This algorithm is based on QoS and a multi- vehicular networks. The hybrid architecture enables both routing criteria related attributes approach. Moreover, this system is uti- decision and forwarding strategy to be applied and optimized in- lized to decide the suitable gateway that might be used to con- dependently of each other. Hence, the proposed routing scheme nect a source vehicle to the LTE-A infrastructure. Many criteria are performed significantly better in terms of packet delivery ratio and taken into consideration for making the decision including load, re- delay while considerably reducing the communication overhead. ceived signal strength and link connectivity duration of the cluster Similarly, data dissemination in the vehicular network could be head and other candidate gateways. The results demonstrate that improved by deploying a hybrid network adding a hybrid overlay this algorithm makes efficient decisions for selecting the appro- protocol layer located between the application and transport layers priate gateway with the best VANET and infrastructure features. [109] . Alawi et al. wanted to extend the coverage and enhance the fre- To solve the interference problem, Yang et al. in [110] investi- quent handoff process. They proposed a new end to end multi-hop gated the joint resource assignment and power allocation problem relay scheme called (SGS) for vehicular communication [116] . This in a Full-Duplex (FD) cellular-VANET heterogeneous network. Since new simplified scheme proposed for the integrated VANET-UMTS the cellular network works in FD mode, the interference relation- network to reduce dead spot and increase the network coverage ships among different communication links become much more area. Furthermore, AODV and DSDV were used as primary proto- complicated. This also leads to a more complicated interference col to evaluate this scheme. The proposed scheme can be applied scenario. To solve the problem, they constructed a graph to model on top of any VANET routing protocols. An integrated simulation the system and further proposed a graph coloring based resource environment combined of Vanet-MobiSim and NS2 is used to sim- sharing scheme. Additionally, the proposed graph coloring based ulate the proposed system. As a result, the simplified gateway se- resource sharing algorithm can effectively achieve a sub-optimal lection shows the best performance on top of AODV compares to solution for the network throughput with low complexity. More- DSDV. Moreover, a new hybrid, scalable, and adaptive gateway dis- A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 65 covery algorithm for VANETs (LAPGD) was offered in [117] . The al- Some problems may exist in V2I communication. Among these gorithm aims to provide an efficient and adaptive location-aided problems, the limited RSU bandwidth resources. This issue leads and prompt gateway discovery mechanism. Furthermore, the pur- to the competition of vehicles in the single community and among pose of this algorithm is to ensure each vehicle capable of find- multiple communities to share the limited bandwidth in the avail- ing its optimal gateway, to reduce the total number of gateways able RSUs. As a potential solution to problems like RSU access selected in VANETs, and to guarantee the average delay of pack- problem, evolutionary game theory for RSU selection is proposed ets within the available range. This system considers many metrics, in [123] to ensure that all vehicles receive same resources from like position, velocity, delay, etc. The results show the performance the available bandwidth. Another model apart from bandwidth al- of the algorithm is affected by such parameters as the maximum location, game theory has been used to analyze various conflicts coverage of a gateway, the number of gateways, as well as the re- which may rise heterogeneous vehicular networks [124] . However, questing delay. Hence, it is clear that the algorithm is significant in a new algorithm (CoR-VANET) for rate adaptation and joint chan- gateway selection. nel of vehicular users in heterogeneous wireless networks was pre- Zhioua et al. proposed a cooperative traffic transmission algo- sented in [125] . This algorithm uses the game-theoretic approach rithm for VANET- LTE-A network gateway selection [118] . This new for selecting the best channel and data rate from available wire- algorithm is based on a fuzzy logic modelization. Moreover, this al- less options of vehicular users. After the selection, this algorithm gorithm aims to select the appropriate gateway for a given source forwards upcoming traffic information with high throughput. This vehicle when this vehicle communicates with the infrastructure. In novel approach aims to achieve higher throughput for vehicular the simulation results, the packet loss is always lower than 1%, users by changing wireless networks channels and data rate in the and the average delay never exceeds the delay constraint of the heterogeneous environment. The algorithm can also be used to find QoS class. These results are entirely different from the determinis- the optimal number of base stations for a given VANET scenario. tic scheme results where the CH is used as a default gateway. Thus, Moreover, Li et al. in [126] introduced a cooperative protocol for this new approach ensures good performances and performs better effective data dissemination in cellular-VANET networks based on in both delay and packet loss than the deterministic approach. For coalition game theory. In the heterogeneous networks, a part of ve- intersection collision avoidance, a cluster-based architecture using hicles is selected as a mobile gateway to connect the two types of both Wi-Fi and LTE channels was proposed by Tung et al. in [119] . networks. The protocol is successfully able to overcome the hetero- In this architecture, Wi-Fi channels are used for in cluster commu- geneity of two systems and to stimulate cooperation among vehi- nication while LTE channels are used for exchanging Cooperative cles. Moreover, as a benefit, this scheme can maximize the effec- Awareness Messages (CAMs) between clusters. This algorithm only tive data rate. The results verify that as well as vehicles follow the requires signaling between the cluster head and the last cluster protocol, the maximum effective data rate can be obtained. member for cluster creation and maintenance. Moreover, a channel A Public Goods Game (PGG) group interaction model was pro- allocation algorithm is applied to reduce the interference of Wi- posed in [127] to investigate the effects of the topological features Fi channels between different clusters. As a result, the proposed on the cooperation evolution in vehicular networks. The authors heterogeneous architecture shows much better performance than analyzed how networking properties can impact the diffusion of other schemes regarding the delivery rate. The average delay of cooperation. Results show that cooperation diffusion in these net- this design meets the requirement of the safety application, but works differs with diverse networking conditions. Higher clustering the delay bound has not been verified. Table 5 provides the gate- favors collaboration due to the minimum difference in payoff and way selection works’ comparison in the literature. the reduced probability for nodes to change their strategies. How- ever, increased connectivity can still improve cooperation caused 6.4. Network selection with effective quality of service (QoS) by more hubs presence. Therefore, cooperation in such networks is highly influenced by their networking properties, and optimization The process of selecting the most suitable available network is hard to achieve. Nevertheless, it can be enhanced under appro- technology is essential for maintaining a satisfactory QoS for an priate network conditions by providing higher benefits. ongoing communication in the heterogeneous environment. This On the other hand, MCDM algorithms are gaining importance selection relies on several parameters, such as the Received Sig- for seamless connectivity in the case of fast-moving vehicles. These nal Strength Indicator (RSSI), the movement direction and speed of algorithms consist of numerous techniques, such as AHP (Analyti- the vehicle, network bandwidth, traffic load, delay, usage cost, etc. cal Hierarchy Process) which has received considerable attention in In V2I heterogeneous communication, services can have diverse the literature. AHP method was used in [128] for network selection requirements in terms of latency, bandwidth, delay variation, er- in heterogeneous environments for moving vehicles. ror rate, etc. depending on the type of application running [40] . Besides, a new heterogeneous architecture-based QoS design could 6.5. Vehicular cloud computing (VCC) Vs. Vehicular fog computing be used to maintain acceptable service quality for heterogeneous (VFC) communications in a multi-RAT environment [120] . Moreover, Haz- iza et al. proposed a software defined radio (SDR) platform [121] to Some studies surveyed vehicular cloud computing in the lit- coordinate between different wireless accesses technologies in ve- erature [139–141] . Feteiha et al. [142] focused on the subject of hicular networks. heterogeneous wireless relaying vehicular clouds. They investi- Other different solutions and algorithms for selecting the ap- gated the performance of a transmission scheme over LTE-A net- propriate underlying network for the heterogeneous vehicular en- work, where vehicles act as relaying cooperating terminals for a vironment were proposed in the literature. Among these solutions downlink session between a base station and an end-user. They game theory and Multi-Criteria Decision Making (MCDM) for net- proposed an enabling cooperative vehicular relaying transmission work selection. scheme to provide better networking capabilities for densely pop- Furthermore, game theory is a set of tools developed to model ulated urban areas. As a result, this new approach not only reduces interactions between multiple agents with conflicting interests the power consumption but also provides higher throughput, in- [122] and is well-suited to address many problems in communica- creases coverage, allows for faster roll out, and promises more flex- tions systems. In VANET heterogeneous communication, a vehicle ibility. attempts to select a suitable network offering acceptable QoS and A new Cloud-assisted Message Downlink dissemination Scheme the lowest usage cost [40] . (CMDS) was proposed in [143] , which is a safety message dissemi- 66 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

capacity

channel

LTE

the

in

reduction

capacity

NA NA NA NA Significant NA NA NA Channel

delivery

1% supported

packet

loss than

data traffic

CAM Lower High NA NA NA NA Packet NA

results

NA NA NA Acceptable NA NA NA throughput NA

gateway

congestion

appropriate

selection and

the

gateway

electing

performance in

in bottlenecks

selection

NA Efficiency NA NA Eliminate Significant Significant NA Gateway

QoS

Available). account

results

Not

into

respect

it

(NA:

Good Delay NA NA NA NA Acceptable Take Delay

2013 2014 2016 2012 2015 2011 2014 2014 Year comparison

[9]

works’

al.

[118] [115] [113]

et [114] [119]

[112]

al. al. al.

al. [117] al. al.

Name

et et et selection

et et

al. et

5

et

Tung Zhioua Zhioua Salvo Ucar Eltahir Author Benslimane Ju Gateway Table A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 67

Table 6 Comparison between vehicular fog computing (VFC) and vehicular cloud computing (VCC).

Features Vehicular fog computing (VFC) Vehicular cloud computing (VCC)

Communication type Real-time load-balancing Bandwidth constrained Decision making Local Remote Geo-distribution Yes No Computation Capacity Large Medium Deployment cost Low High Latency Low High Mobility support High Limited Storage Limited Highly scalable nation framework to disseminate traffic information efficiently. Ad- into the latest promising technology of vehicular fog computing ditionally, this scheme uses advantages of both wireless network- and discussed issues and challenges of its implementation as well. ing and cloud computing technologies. In this context, the cloud However, a vision and key features of VFC main applications were collects massive traffic flow information and chooses some gate- outlined in [204] . ways that are buses equipped with both VANET and cellular in- Chaudhary et al. [205] presented a relative comparison and terfaces. When a gateway receives the message from the cellular analysis of fog and cloud computing, with respect to network ser- network, it will further distribute it to nearby vehicles via V2V vice chaining in the 5 G environment. They discussed the major communication. The results prove the efficiency of CMDS in nu- issues of security, cost, latency, and data offloading at the core merous urban scenarios. These results show that the proposed Data Centers (DCs). In addition, a new hybrid computing model scheme not only can disseminate messages rapidly and efficiently for Vehicle-to-Grid (V2G) networks named “Foud” was proposed but also can significantly reduce the cost of cellular communi- in [206] . V2G technology is considered a robust approach, enabling cation as well. Moreover, Ashok et al. [197] designed a dynamic renewable energy sources to provide electrical services, managing system for offloading specific vehicular application modules. They and monitoring power usage in the smart grid. Fog computing and developed heuristic mechanisms for placement and scheduling of cloud computing assisted V2G systems in future 5 G mobile net- these modules in the on-board unit versus the cloud. The design works are expected to provide new realistic services for moving can efficiently reduce the response time to compute intensive ap- Electric Vehicles (EVs). Foud is made up of two sub-models, tem- plications across variable network conditions and different urban porary fog, and permanent cloud. However, Foud architecture is di- environments. vided into three layers, the user layer, the service layer and the On the other hand, Fog Computing (FC) has been introduced as network layer. a technology to bridge the gap between remote data centers and Zhang et al. [207] presented a regional Cooperative Fog Com- Internet of Things (IoT) devices. This new paradigm broadens cloud puting (CFC) architecture to support big data in IoV applications. computing and services to the edge of the network instead of en- Possible services for CFC in IoV applications were discussed, in- tirely in the cloud. VANETs have been depending on cloud com- cluding multi-source data acquisition, distributed computation and puting services for computing and data storage. Besides the tradi- storage, mobility control, and multi-path data transmission. En- tional cloud characteristics, vehicular fog computing distinguishes ergy efficiency was optimized in intra-fog resource management, itself from existing techniques with its dense geographical distri- and the packet dropping rates of Local Fog Servers (LFSs) were bution, proximity to end users, location awareness, low latency and minimized in inter-fog resource management. However, a new support for mobility [198] . Table 6 presents a comparison between real-time mechanism for ITSs named FOX (Fast Offset XPath) was Vehicular Fog Computing (VFC) and Vehicular Cloud Computing shown in [208] . This new mechanism aims to detect and man- (VCC). According to Bonomi et al. [198] , the key characteristics of age traffic congestion in vehicular ad hoc networks. FOX is imple- fog computing are: a) low latency and location awareness, b) wide- mented in a fog computing environment, taking advantage of this spread geographical distribution, c) mobility, d) very large number platform. The proposed mechanism can reduce the average trip of nodes, e) predominant role of wireless access, f) strong presence time, fuel consumption and CO2 emissions. of streaming and real-time applications, g) heterogeneity. In order to solve mobility challenges, Lopez et al. [209] dis- The idea of utilizing vehicles as the infrastructures for com- cussed three different scenarios for communication between vehi- munication and computation in VFC was conceived by Hou et al. cles in the city. Two of these scenarios were with a fog comput- [199] to enlarge the available resources and to enhance the achiev- ing approach and one with a cloud computing approach. In each able capacities. In particular, they analyzed four types of scenarios of them, it is assumed that all vehicles have an OBU. Furthermore, of moving and parked vehicles as infrastructures for communica- Giang et al. [210] investigated how smart transportation applica- tion and computation, respectively. As a result, better connectiv- tions were developed following fog computing approach and their ity with higher capacity and more reliable communication can be challenges. However, the work in [211] addressed resource man- achieved by involving VFC. Similarly, Xiao et al. [200] configured agement in fog-enhanced radio access networks (FeRANs). The au- each vehicular fog node as a wireless access point to avoid the de- thors focused on management strategies at each fog node to im- lay caused by multi-hop communication. This enables computing prove QoS, particularly for real-time vehicular services. They pro- resources to be delivered when they are needed upon request of posed two schemes, namely Fog Resource reallocation (FRL) and the nearby vehicles and passengers. Fog Resource Reservation (FRR). In both schemes, real-time vehic- Hu et al. [201] summarized fog computing model architecture, ular services are a priority so that their respective vehicular users key technologies, applications, challenges and open issues. They can access the services with only one hop. compared fog computing with cloud computing and edge comput- Huang et al. [212] proposed an adaptive content reservation ing to highlight the similarities and differences. A typical use case distributed strategy for reliable real-time streaming in vehicular in VFC was presented in [202] . This issue can result in significant cloud-fog networks. Moreover, the key aspect of this strategy is bandwidth saving for the heterogeneous communication network. that the streaming content is pre-allocated from computing service Similarly, Menon in [203] discussed the deployment of fog layer providers aimed to sustain the QoS of real-time streaming. This ap- between the vehicles and the cloud servers. He provided insights proach utilizes the resources without any parameters tuning. 68 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

Nowadays, it is difficult to find a parking space, especially in 6.7. Other VANET-Heterogeneous networks integration issues crowded cities. The study in [213] focused on solving parking prob- lem to relieve the traffic congestion, enhance driving efficiently 6.7.1. Routing protocols and mobility and reduce air pollution. From this perspective, fog computing and Some studies [144–146] surveyed the routing protocols for het- roadside cloud are utilized to find a vacant spot. By using this in- erogeneous vehicular communication. Bilal et al. in [144] analyzed frastructure, any available parking space at many places can be different position-based routing protocols for V2V and V2I and pre- shared. Additionally, the parking problem was solved using the sented a qualitative comparison between numerous routing pro- matching theory. Hence, this proposal not only helps drivers find- tocols in both city and open environment. However, the study ing an ideal available space but also brings profit to the parking’ in [145] provided five different taxonomies of routing protocols, owner. whereas the main research challenges of routing in VANETs were Since security and privacy issues are important in IoT environ- discussed in [146] . ments, Alrawais et al. in [214] offered a mechanism that employs Mir et al. proposed a new routing scheme for VANET-LTE hy- fog computing for security enhancement among IoT devices. There- brid vehicular networks based on location [108] . The hybrid archi- fore, fog computing could provide a high security level that will tecture allows a functional split between routing decision and for- help minimize attacks in heterogeneous environments. Ni et al. warding strategy so that both can be optimized independently of [215] examined the security, privacy, and fairness requirements in each other. In the novel routing scheme, maintaining local neigh- fog-based vehicular crowdsensing. The authors described the pos- bor information and forwarding state results in fewer route re- sible solutions to achieve security assurance and privacy preserva- quests to be sent towards the remote routing server. This issue is tion. Furthermore, Huang et al. [216] proposed Meet-Fog, a scheme beneficial in most of the cooperative awareness applications where based on Meet-table and cloud computing for accurate distribu- messages are destined within proximity of the source vehicles. For tion of negative messages such as CRL (Certificate Revocation List) this reason, links with minimum delays are selected to reach the in VANET. This scheme can significantly reduce the bandwidth and nearby neighboring cars. However, for scenarios where the route to the storage requirements of cloud, and completely move the com- the destination is not available locally, the remote routing server puting requirements from cloud to the edge. calculates the shortest path by taking into consideration current Basudan et al. in [217] presented a new efficient certificateless topology state. Finally, routing server transmits route updates to aggregate signcryption (CLASC) scheme. On the basis of the pre- all vehicles on the road between the source and the destination. sented system, a privacy-preserving protocol for monitoring road Hence, simulation results show that this new approach performs surface conditions was designed using fog computing. Besides, the significantly better regarding delay, packet delivery ratio, and con- proposed protocol meets the security requirements such as in- trol message overhead if compared with GPSR and AODV routing tegrity, data confidentiality, anonymity, and mutual authentication. protocols. Additionally, Soleymani et al. [218] proposed a fast, accurate and A new analytical mobility model was presented in [147] to com- trust model to ensure the correctness of the information received pute the link availability, the vehicles’ distribution on the road and from authorized vehicles. The authors applied fog nodes as a facil- the maximum number of vehicles within the range of the trans- ity to assess the level of accuracy of event’s location. Moreover, this mitter. This work also confirmed that Most Forward within Range solution does not serve only to detect faulty nodes and malicious (MFR) is not a valid scheme in VANETs. Moreover, due to highly attackers but also tackles the uncertainty data in the vehicular net- dynamic mobility in VANET, the analytic and simulation results work in both LOS and NLOS. show that not all vehicles within the transmitter range will receive the broadcasted packets successfully. Additionally, selecting the far- 6.6. Security of vehicular communication thest vehicle to rebroadcast the message is not valid in vehicular networks. Furthermore, Guillaume Rémy et al. in [148] proposed The security challenges issues in VANET were studied in [129] . LTE4V2X, a novel framework for a centralized vehicular network Similarly, Shrikant et al. in [130] surveyed the various possible at- organization. The LTE4V2X framework uses the eNodeBs of the LTE tacks in VANET and offered some trust management solutions. network to self-organize the vehicular network. This framework In multi-tier, multi-RAT V2I communication, there are two was compared to a decentralized DCP and All-LTE approaches for broad categories associated with possible security problems: at- different highway scenarios with the aim to evaluate the impact of tacks on the user and attacks on the system. Attacks on the user high mobility and vehicles’ density. Performance evaluation shows seek to cause vehicle crashes, congestion, making the driver take that LTE4V2X has better performances than DCP. More the vehi- a wrong direction or reducing the user’s faith in the system due cles number increases, more framework is efficient in the matter to unreliable messages. To mitigate these attacks, public key cryp- of overhead besides LTE bandwidth usage or packet loss. Moreover, tosystems are proposed to enhance vehicular network security the velocity does not impact the LTE4V2X efficiency, whereas DCP [131, 132] . Additionally, an attacks classification was presented in efficiency decreases when the vehicles’ velocity increases. The re- [133] according to their characteristics, the requirements involved, sults show also that LTE4V2X leads to performance improvement and the defenses that could be used. Moreover, Engoulou et al. in in term of decreasing the Floating Car Data (FCD) packet losses, this work also proposed a global security architecture in VANETs. and can efficiently work with high vehicles’ speeds. The main security requirements defined in [133, 134] are au- Based on the microscopic and macroscopic parameters, the traf- thentication, integrity, confidentiality, non-repudiation, availability, fic engineers have defined several traffic flow models to implement access control, real-time constraint and privacy protection. Most of the traffic codes. Such models describe the relationship between these requirements are related to general security issues, and oth- the characteristics of moving vehicles, such as density, speed, flow, ers are specific to VANETs. Manvi et al. in [135] focus on the au- safety distance and length [233] . The available models of traffic thentication schemes in VANET as they play an essential role in se- flow theory, such as the fluid dynamics model and the car follow- cured communications. The authentication schemes are classified ing model are used to represent traffic flow in various VANET stud- into three categories: cryptography techniques, digital signatures, ies. Ho et al. in [234] used a fluid model for the density estimation and message verification techniques. On the other hand, many so- and then applied the stochastic model for analyzing the connectiv- lutions have been proposed in the literature to address the secu- ity issues. rity problem of VANETs [136] . Furthermore, other vehicular secu- On the other hand, the study in [235] presented a new hetero- rity schemes were introduced in both studies [137, 138] . geneous traffic mobility model for the vehicular network. Micro- A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 69 scopic parameters of varying safety distance between the vehicles isting quorum-based RLSMP [220] and hierarchical hashing-based and vehicular length were considered in this scheme, to achieve HRHLS [221] protocols. The results show that the time to deliver more accurate and realistic data about road conditions. The den- update packets location, signaling overhead and the probability to sity dynamics under different road scenarios are calculated under deliver packets successfully is better than RLSMP and HRHLS pro- the influence of these constraints with the use of the fluid dy- tocols. The handovers produced by ZGLS are much lower than the namic model. Besides, this model can capture the impact of road other two protocols. Thus, ZGLS is the better choice for large-scale constraints such as traffic lights and road incidents on the traf- sparse and dense VANET environment. fic flow. This scheme aims to capture the impact of traffic flow Soleimani and Boukerche proposed a Speed and Location Aware theory constraints on the vehicular density under the heteroge- (SLA) scheduler for LTE which is suitable for vehicular applications neous traffic flow on the road. The vehicular density depends on [149] . This approach has an essential role in allocating resources the changing road conditions; it can significantly affect the net- for the vehicles which are more sensitive to delay. Moreover, the work connectivity. Furthermore, Dynamic Transmission Range As- location and speed of vehicles are the main parameters for SLA signment (DTRA) algorithm was suggested in the study [236] to scheduler to assign priorities for resource allocation. Additionally, improve dynamic transmission range capabilities of VANET nodes this work considered the speed as a factor in scheduling metric for better connectivity conditions. A safety distance characteristic to assign a higher priority to faster vehicles. The results prove of the traffic flow has been implemented in a heterogeneous traf- that SLA scheduler is better than the existing algorithms since it fic environment using the car following model. The implementa- reduces high delays and large position errors for faster vehicles. tion of these parameters provides more realistic traffic flow and Moreover, Prasad and Priyanka [150] tried to solve the uncertainty road conditions. The fluid dynamic model provides the traffic flow in time schedule when passengers use buses transport system. This and density as a function of time and space as presented in [237] . problem can be solved with the help of Real Time Passenger Infor- This model comes with more realistic data as compared to previ- mation System (RTPIS) which is a system provides real-time in- ous studies. Besides, the mobility of vehicles can be manipulated formation regarding the position of a bus. In this work, Wi-MAX with the optimal use of safety distance between the vehicles. has been compared with Wi-Fi and WLAN to check for better per- Umer et al. [238] proposed a heterogeneous traffic flow based formance ratio. Furthermore, an efficient scheme was proposed for dual-ring connectivity model to improve both message dissemina- enhancing the connectivity of each bus with the stations. This en- tion and network connectivity. The new model takes into account hancement is passed through selecting the best seamless and con- the distribution and the availability of different types of vehicles tinuous network for the public transport during the communica- on the road. This scheme is based on the dual-ring structure that tion. To achieve seamless and fast handoff, handoff detection and forms the primary and the secondary rings of vehicular commu- handoff execution used to avoid packet loss, the handoff metrics nication. Fast speed vehicles (cars) moving in the faster lane on such as the cost of service, available bandwidth, power require- the road, establishing a primary ring. However, slow speed vehicles ments, and QoS. The results show that such a system is ideal for (buses) moving in the slower lane form a secondary ring, providing real-time public transport. a backup path of communication for high-speed vehicles driving A new algorithm has a repetition mode decision based on the on the primary ring. The authors implemented this scheme for dif- distance of two nearby vehicles for vehicular communication was ferent road scenarios under a cluster-based routing protocol known presented in [151] . This vehicle packet repetition and position data as Warning Energy Aware Cluster-head (WEAC) protocol [239] . Fur- updating strategy based on Kalman filter predicting. Both simula- thermore, the vehicular density was calculated by using a fluid dy- tion results in highway and realistic urban road show that, by us- namics model under the effect of microscopic parameters of traffic ing this position updating strategy, the vehicle position data updat- flow theory as presented in [240] .This model considers the het- ing frequency could be decreased. Furthermore, the communica- erogeneity of vehicular types on the road and its impact on the tion reliability is greatly enhanced through packet repetition mech- safety distance constraint between the vehicles. The used model anism. shows an improvement in network coverage and connectivity even under the multi-hop communication system. Additionally, the av- erage packet delay is minimized under the heterogeneous traffic 6.7.3. Video on demand (VoD) flow. Xu et al. proposed an efficient user-centric mobile Video-on- Demand (VoD) solution called QUVoD in an urban vehicular net- 6.7.2. Positioning work environment [152] . The QUVoD solution offers high Qual- In VANET, the location management techniques have been ity of Experience (QoE) service level to vehicle passengers. It is categorized into flooding-based, flat hashing-based, hierarchical a big challenge to support high QoE for interactive mobile VoD hashing-based and hierarchical quorum-based techniques. The hi- services because of the high devices mobility and the users’ inter- erarchical quorum-based technique suffers from handover signal- active viewing behavior. Based on the proposed efficient designed ing between servers, high load on the top hierarchy and location multi-homed hierarchical P2P/VANET structure, four novel mech- query delay when source and destination are apart. To overcome anisms were introduced: 1) distributed grouping-based video seg- these problems, Rehan et al. [219] proposed ZoomOut Geographic ments storage scheme, 2) video segment seeking scheme, 3) multi- Location Service (ZGLS) protocol, which introduces flat quorum- path data delivery mechanism, and 4) speculation-based prefetch- based location management service model for cities and highways. ing strategy. The storage scheme distributes the segments along The proposed ZGLS consists of adaptive neighbor discovery, Loca- the chord network uniformly by groups. This distribution allows tion Update (LU) and Location Query (LQ) modules. For each ve- each video segment to be stored in multiple nodes, and each node hicle, the neighbor discovery module finds a maximum two front to store multiple video segments. However, this not only balances neighbors and two behind neighbors. These neighbors are called the nodes’ load but also reduces the resource-seeking times. Ad- relatives. Moreover, this new scheme brings new research contri- ditionally, the video segments are downloaded from the located butions. ZGLS has shifted the location server role from fixed geo- node by the multipath data delivery mechanism over multi-homed graphic regions to available vehicles; then it does not suffer from 4 G/VANET networks, with high data rate and reliability. Further- frequent handovers. Since this protocol is non-hierarchical, LU de- more, the presented prefetching strategy can sense the user’s livery and LQ response are speedy. Besides, ZGLS has higher packet viewing behavior and make intelligent decisions which greatly delivery. Furthermore, ZGLS protocol was compared with the ex- smooth the playback experience for interactive users. Hence, the 70 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 simulation-based results show how QUVoD is an efficient interac- tive mobile VoD solution in urban vehicular networks.

6.7.4. Other issues The vehicular Wi-Fi offloading performance was investigated by Cheng et al. in [153] . They presented an analytical framework for offloading cellular traffic by outdoor Wi-Fi network in the vehicu- lar environment. Under this scenario, the Wi-Fi offloading perfor- mance, characterized by offloading effectiveness, was analyzed in terms of desired average service delay. The authors established an explicit relation between offloading effectiveness and average ser- vice delay by an M/G/1/K queuing model. The analytical framework has validated through simulations based on a VANET simulation Fig. 5. Toyota Prius Google self-driving car [162] . tool VANET Mobisim and real map data sets. However, Park et al. in [154] presented a feasibility study along with guidelines and the VoCell application development framework for enabling vehicular • Driverless cars may need very high-quality specific maps networking applications with smartphones. The observations made [157] to function properly. in this work suggest that all the ingredients for realizing such sys- • The radio spectrum competition desired for the car’s communi- tems are ready. Nevertheless, the fact that over the next decade ve- cation. hicular networks based on the WAVE standards will start to appear • Existing road infrastructure may require changes for driverless in the cars that we drive every day. The VoCell framework and ap- vehicles to work optimally [158] . plications presented in this work can act as catalysts that bring im- portant vehicular applications to reality sooner. Therefore, studies In addition to the technological challenges, there are other ob- and commercial efforts suggested that such cellular-based vehicu- stacles such as the period required to change a current stock of lar networking systems can be well-accepted by a large population. vehicles from non-autonomous to autonomous, customer anxiety Moreover, the active participation of drivers can revolutionize road about the safety of autonomous cars and the implementation of le- safety effort by actively providing relevant suggestions to drivers. gal government regulations for self-driving cars [159] . On the other hand, self-driving cars could be loaded with explosives and used 7. Autonomous cars as bombs by hackers or terrorism, which means the risk of confi- dentiality loss and security concerns. However; there are a lot of An autonomous car also called a driverless car, robotic car or fears about the consequential loss of driving-related jobs in the self-driving car is an automobile that has an autopilot system al- road transport industry. lowing it to move from one place to another safely without help from a human driver. The only role of a human in such a vehicle would be indicating the destination [155] . There are various poten- 7.2. Google self-driving car tial benefits to using a driverless car instead of a traditional vehi- cle. Driverless cars could lead to several enhancements in trans- In June 2011, the state of Nevada voted for a law to authorize port systems, like a significant reduction of traffic accidents, a sub- the use of autonomous cars [160] , which went into effect on March stantial increase in road capacity and more well-organized trans- 1st, 2012 (A Toyota Prius modified with Google) [161] . Therefore, portation. Therefore, there would be a less need for traffic police Nevada became the first authority in the world where driverless or even road signage and also an enhanced ability to deal with cars might be officially operated on public roads. This legislation traffic movement. was supported by Google to conduct further testing of its Google The driverless cars can use different techniques can be used by autonomous vehicles legally. Furthermore, Nevada’s rules require a to detect surroundings such as radar, GPS, and computer vision. person behind the wheel and one in the passenger’s seat during Google is the most well-known company working on these types tests. Fig. 5. of vehicles. However, various research organizations and major cor- In August 2012, Google proclaimed that they had finished porations have developed working prototype autonomous vehicles, over 30 0,0 0 0 autonomous-driving miles (50 0,0 0 0 km) accident- including Mercedes-Benz, General Motors, Bosch, Nissan, Renault, free. In late May 2014, Google revealed a new prototype fully au- Toyota, Audi, Tesla Motors, Hyundai Motor Company, Volvo, Peu- tonomous of its autonomous vehicle, with no gas pedal, steering geot, Vislab from Parma University, Oxford University, and Google. wheel, or brake pedal. As of June 2016, Google had test driven Although autonomous cars have been tested on ordinary roads in their driverless cars in autonomous mode, a total of 1,725,911 mi regular weather, Ford has been testing its autonomous vehicles on (2,777,585 km) [163] . snow-covered roads [156] . The United Kingdom allowed the testing of autonomous vehi- In Europe, many countries like Belgium, France, Italy, the UK, cles on public roads in 2013. In 2015 the government of France Germany, the Netherlands, and Spain have allowed testing robotic opened 20 0 0 km of road in Bordeaux, in Isère, Île-de-France, and cars in traffic. Strasbourg for testing the autonomous cars. In 2016, seven states in the USA (Nevada, California, Florida, Michigan, Hawaii, Wash-

7.1. Main obstacles ington, and Tennessee), along with the District of Columbia, have legislated laws for driverless cars.

There are several obstacles and difficulties due to an extensive The first fatal accident concerning a driverless car took place in adoption of autonomous cars. The possible technological barriers Williston, Florida on May 7th, 2016, while a Tesla Model S electric are: car was engaged in autopilot mode. The occupant was killed in a crash with an 18-wheel tractor-trailer. As of July 2015, Google’s 23 • Software reliability. test cars have been involved in 14 collisions, according to Google’s • Artificial intelligence still is not able to work appropriately in accident reports, of which other drivers were at fault 13 times. It chaotic inner city environments. was not until 2016 that the car’s software caused a crash [164] . A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 71

Alheeti et al. in [173] developed their works and designed an intelligent hybrid security system BP-IDS to secure external com- munications for self-driving and semi self-driving cars. The pro- posed system based on the behavior of connected communicating vehicles in the heterogeneous communicating environment. Results show that BP-IDS have high detection rate and can guarantee low false alarm. However, some evaluation systems and methodologies to smoothly move from simulation into real world car driving were discussed in [174] . Rane et al. [175] proposed an algorithm for successful conflict resolution between autonomous cars using Game theory and im- plementing the same with Cellular Automata on MATLAB. Their objective is to come up with an optimum solution to avoid the collision so that the vehicles involved reach their destination. Fur- thermore, a vehicular cloud model was presented in detail by Gerla Fig. 6. Google self-driving car ‘Waymo’ [166] . et al. in [176] with a discussion of possible design perspective. The authors also discussed the evolution of intelligent vehicle grid However, on February 14th, 2016, a Google autonomous car failed to autonomous, internet-connected vehicles, and vehicular cloud. to avoid sandbags on its path; the car struck a bus [165] . They claimed that the autonomous driving would be the main ben- The Google self-driving car project had been expanded to be eficiary in the vehicular cloud architecture. Pozna et al. [177] stated an independent developing autonomous cars company under Al- cultural parts related to an autonomous car, since driving a vehi- phabet Inc., called ‘Waymo’ . The company began in 2009 as a cle describes our lifestyle. Therefore, the transformation is also a project under Google. Google transitioned the project into this new cultural issue. The proposed control architecture is composed of a company in December 2016. Furthermore, Google plans to make strategical level, a tactical level, and an operational level. The cur- these cars available to the public in 2020. The project team has rent matters of autonomous cars design consist on solving the un- equipped different types of cars, including the Toyota Prius, Audi expected driving problem; these issues refer to the second level of TT, and Lexus RX450h. Fig. 6 presents the self-driving car from the proposed architecture. google called ‘Waymo’. 8. Open issues, challenges and future research directions 7.3. Open opinion surveys This section presents several open issues, research challenges as In the Accenture 2011 online survey of 2,006 US and UK con- well as future research directions for heterogeneous vehicular net- sumers, 49% said they would be comfortable using a "driverless works. Addressing these issues is important for successful and effi- car" [167] . cient heterogeneous communications. Fig. 7 summarizes the main Another survey across ten countries of 1,500 consumers in 2017 issues, challenges and future research directions for the heteroge- made by Cisco systems found 57% "stated they would be likely to neous vehicular communications. ride in a car controlled entirely by technology that does not require a human driver", with Brazil, India, and China the most willing to 8.1. Big data management trust autonomous technology [168] . In 2016, a survey in Germany studied the opinion of 1,603 per- With millions of vehicles on the road, a huge volume of data is sons, who were representative in terms of age, gender, and ed- generated by different network participants. These data should be ucation, towards partially, highly, and fully automated cars. Re- managed and aggregated to reduce bandwidth consumption and sults presented that women and men differ in their readiness to memory. Due to the heterogeneous environment features, unique use them. Women felt more anxiety and less joy towards auto- challenges are posed by data management [222, 223] . mated cars, in comparison with men who showed the exact op- In general, big data are physically and logically decentralized posite [169] . but virtually centralized [224] . Management of this volume of dy- namic data includes several processes from gathering, aggrega- 7.4. Academia and research organizations tion, validation until data dissemination. Indeed, it is a funda- mental challenge that requires the design and the development The subject of autonomous cars is a very hot topic in academia of smart routing protocols, for establishing effective communica- and automobile companies. VANET-heterogeneous wireless net- tion and improving data dissemination between vehicles. Advanced works integration will significantly help autonomous vehicles to data processing and mining techniques are needed to collect, ag- be functional in reality. Most of the existing research papers tar- gregate, process and analyze data in heterogeneous vehicular net- geted the various malicious attacks of the driverless cars. However, works [36] . Moreover, vehicular clouds provide an efficient way to there are many types of attacks in VANETs that have a direct effect exploit the utilized computation and storage resources of vehicles. on the advance of self-driving vehicles, including Denial of Service The driver’s behavior can be considered as a factor to reduce the (DoS) attacks, grey hole, and black hole attacks. dissemination time. Alheeti et al. [170] proposed an intelligent Intrusion Detection System (IDS) which can detect grey hole and rushing attacks in 8.2. Security and privacy VANETs. These attacks try to drop some or all received messages in an attempt to stop transmission between vehicles and roadside Security and privacy are significant aspects in VANETs. With- units. IDS can provide safety to self-driving and semi self-driving out considering the security restriction, the vehicles can access or by CAMs and data control that were exchanged between the ve- modify the stored data on other vehicles. Hence, the sensitive data hicles in that zone. Similarly, Alheeti et al. [171, 172] designed an that are stored in each one has to be encrypted to protect it against enhancement of IDS for VANETs. This new IDS uses to detect black the unauthorized access. Privacy measures are required to ensure hole attacks. a reliable communication and computation environment, whereas 72 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

security procedures are needed to cope with network threats. Es- tablishing a trust relationship between several participants is nec- essary for safe communication and computation. To overcome the security challenges, the researchers should es- tablish novel protocol solutions, taking into consideration the fol- lowing characteristics [178, 225–227] :

(1) Minimum hops communication among nodes, (2) Low overhead due to time sensitivity, (3) Pre-stored information about the participating routing nodes, (4) Optimized data dissemination solutions.

However, new secure communication protocols must be investi- gated by involving the unique characteristics of heterogeneous ve- hicular networks.

8.3. Handover issues in V2I communications

There are various wireless technologies such as DSRC, LTE and LTE-A in heterogeneous VANET. Due to the fast movement, vehicu- Fig. 7. Main issues, challenges and future research directions. lar users may frequently switch among different networks. It is de- sired that a vehicle always stays connected with the most suitable system. Traditional handover mechanisms for cellular networks are under the constraints of high mobility and different networks re- not well suited for the hybrid-distributed vehicular architecture; quirements. they are mostly centralized. Therefore, the handover decision usu-

ally depends on a single threshold, which is affected by some fac- 8.6. Cross-layer design approaches tors such as channel condition, network load, and received signal

strength. Besides, the handover of vehicular users is more frequent The vehicular environment is highly dynamic causing severe

than cellular users, resulting in excessive signaling overhead. Since concerns, such as connectivity break problem and frequent topol-

the handover plays a vital role in vehicular communication, we ogy changes. Traditional layers designs are hard to meet QoS re-

motivate the researchers to find more seamless, fast, and reliable quirements. Hence, it has been increased attention to employ the

handover mechanisms for the heterogeneous environment. interaction among various protocol stack layers for better perfor- mance [228-231] . However, the cross-layer approaches offer better 8.4. Coexistence and cooperation between different networks performance, but the implementation complexity should be seri- ously addressed. The primary challenge is how to design the up- It is expected that the next generations of ITSs reflect a more per layer functions based on lower layers feedback [36] . Addition- holistic approach to network solutions. This issue would require ally, a mathematical framework which characterizes the interac- support to the coexistence of different co-located wireless net- tion among the layers is desirable. Reliable algorithms are urgently works, to provide ubiquitous access to broadband services [178] . needed for real-time vehicular applications. The cooperation between different networks will allocate advan- tageous services to the vehicular user, for instance, information 8.7. Vehicular cloud computing (VCC) and vehicular fog computing about traffic conditions, routes, and weather. Cooperative commu- (VFC) nication can reap the benefits of spatial diversity gains and in- crease the link capacity. Current studies have shown that cognitive The concepts of VCC and VFC have been rapidly developed. radio technology provides more opportunities for cooperative com- These computing systems have specific features, differently from munications. Moreover, cooperative Multiple Input Multiple Output the traditional ones. There are a lot of issues and problems need (MIMO) techniques offer attractive benefits for vehicular networks to be addressed. These computing systems should be designed [35, 37] . Finally, Schemes such as relay selection, link adaptation, to function with the heterogeneous environment. Therefore, new and radio resource management in cooperative communications computing architectures have to be investigated. are essential to improve system performance. On the other hand, resource management in multiple access networks should be ad- dressed to maximize the resource utilization, stability, and robust- 8.8. Vehicle-to-grid (V2G) ness of the network. As a smart grid application, Electric Vehicles (EVs) are gaining 8.5. QoS metrics significant attention in both academic and industrial areas. There are many benefits of using EVs, including lower cost, lower carbon Quality of service (QoS) is the ability to provide different pri- emission via making the ‘green’ choice, and increased energy ef- ority to different applications or users. It is a guarantee offered by ficiency. Hence, the Canadian government’s vision is to have one the network to satisfy a certain level of performance to a data flow, electric vehicle among every 20 vehicles by 2020 [232] . with regard to jitter, end-to-end delay, and available bandwidth. QoS is necessary for real-time streaming multimedia applications. 8.9. Emergency situations and priority Considering data congestion problem in VANET, several modifica- tions to the transport layer were proposed in the literature. How- Some vehicles such as ambulances, police cars, and fire service ever, these solutions need to be investigated in a realistic commu- vans need to be given a higher priority in VANET network archi- nication scenario. The resulting absence of network congestion re- tecture, as their interventions are crucial during an emergency sit- duces or eliminates the need of QoS mechanisms. In VANET het- uation. Researchers should address these aspects in heterogeneous erogeneous environment, satisfying the QoS metrics is necessary vehicular architectures. ( Fig. 7 ). A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 73

9. Conclusion CMDS Cloud-Assisted Message Downlink Dissemination Scheme

In this paper, we surveyed and compared the integration of CoMP Coordinated multi-point VANET heterogeneous wireless networks in the literature. The sur- CoR-VANET Cognitive Radio-VANET veyed papers were chosen based on different classifications. Fur- CRL Certificate Revocation List thermore, vertical handover, data dissemination and collection, CSMA/CA Carrier sense Multiple Access with Collision Avoidance gateway selection and others are all taken into consideration. Ad- D2D Device-to-Device DCs Data Centers ditionally, our work highlights the recent development in the au- DCP Decentralized congestion Control Protocol tonomous cars. The integration of VANET heterogeneous wireless DoS Denial of Service networks has great importance in the success of driverless vehicles DSDV Destination-Sequenced Distance-Vector Routing projects. On the other hand, this work will help future researchers DSRC Dedicated Short-Range Communications DTRA Dynamic Transmission Range Assignment to obtain ideas about vehicular heterogeneous networks collab- EADT Enhanced Area Defer Transmission oration in developing VANETs and enhancing the drivers’ safety. ECDH Elliptic Curve Diffie-Hellman Hence, the researchers are suggested to validate their studies not EDCA Enhanced Distributed Channel Access only in small-scale homogeneous networks but also in such large- eMBMS evolved MBMS scale heterogeneous ones. ETSI European Telecommunications Standards Institute EVs Electric vehicles

In the future, the integration of VANET-5 G network is a promis- FC Fog computing ing area in such VANET heterogeneous networks. Using 5 G net- FCC Federal communication commission work with VANET will lead to a significant enhancement and more FCD Floating car data efficiency in vehicular communications. We believe that in the next FD Full-duplex FDD Frequency division duplex years addressing such topics will be an influential driving factor for FeRANs Fog-enhanced Radio Access Networks vehicular communication research in both academic and industrial FMIPv6 Fast Mobile IPv6 areas. FOX Fast Offset XPath FRL Fog Resource reallocation FRR Fog Resource Reservation GPRS General Packet Radio Service GPS Global Positioning System

Acknowledgments GPSR Greedy Perimeter Stateless Routing GSM Global System for Mobile communication This work is supported by DCT-MoST Joint-project No. H2H Human-to-Human (025/2015 / AMJ); University of Macau funds Nos: CPG2018- HCCA Hybrid Coordination Function Controlled Channel Access 0 0 032-FST & SRG2018-0 0111-FST ; Chinese National Research Fund HCF Hybrid Coordination Function

(NSFC) Key Project No. 61532013; National China 973 Project No. HLL Heterogeneous Link Layer 2015CB352401 ; Shanghai Scientific Innovation Act of STCSM No. HIP Host Identity Protocol 15JC1402400 and 985 Project of Shanghai Jiao Tong University : HSDPA High Speed Downlink Packet Access

WF220103001 . HWN Heterogeneous Wireless Network IDS Intrusion Detection System IEEE Institute of Electrical and Electronics Engineers IoT Internet of Things IoV Internet of Vehicles Appendix A IP Internet Protocol IPv6 Internet Protocol version 6

List of acronyms used in this paper ISO International Organization for Standardization ITS Intelligent Transportation Systems ITSA Intelligent Transportation Society of America IVC Inter-Vehicle Communications LAPGD Location-Aided and Prompt Gateway Discovery LFSs Local Fog Servers 3G Third Generation LQ Location Query 3GPP Third Generation Partnership Project LTE Long Term Evolution 4G Fourth Generation LTE-A LTE-Advanced 5G Fifth Generation LU Location Update ADT Area Defer Transmission MAC Medium Access Control AHP Analytical Hierarchy Process MANET Mobile Ad-hoc Network AODV Ad-Hoc on-Demand Distance Vector mBS micro Base Station AP Access Point MBMS Multimedia Broadcast Multicast Services AR Access Router MCDM Multi-Criteria Decision Making ASTM American Society for Testing and Materials MFR Most Forward within Range AUs Application Units MIH Media Independent Handover BDPM Big data processing and mining MIMO Multiple Input Multiple Output BP-IDS Back Propagation Neural Network- Intrusion Detection MIPv6 Mobile IPv6 System MNs Mobile Nodes BS Base Station Mobisim Mobility Simulator CALM Communication Access for Land Mobiles NDT Network Dwell Time CAMP Crash Avoidance Metrics Partnership CAMs Cooperative Awareness Messages ( continued on next page ) CCH Control Channel CFC Cooperative Fog Computing CH Cluster Head CLASC CertificateLess Aggregate SignCryption ( continued on next page ) 74 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

[7] IEEE standard for information technology – telecommunications and infor- NEMO BS Network Mobility Basic Support mation exchange between systems - local and metropolitan area networks - NS2 Network Simulator-2 Specific requirements –“part 11: wireless LAN medium access control (MAC) OBU On-Board Unit and physical layer (PHY) specifications, Wireless Access Veh. Environ. (July) OFDM Orthogonal Frequency Division Multiplexing (2010) 1–51 IEEE Std 802.11p . P2P Peer-to-Peer [8] G. Araniti , C. Campolo , M. Condoluci , A. Iera , A. Molinaro , LTE for vehicular

PCF Point Coordination Function networking: a survey, IEEE Commun. Mag. 51 (May(5)) (2013) 148–157.

PDR Packet Delivery Ratio [9] A. Benslimane, T. Taleb, R. Sivaraj, Dynamic clustering-based adaptive mobile gateway management in integrated VANET - 3G heterogeneous wireless net- PGG Public Goods Game works, IEEE J. Sel. Areas Commun. 29 (March(3)) (2011) 559–570 . PMIPv6 Proxy Mobile IPv6 [10] T. Taleb , E. Sakhaee , A. Jamalipour , K. Hashimoto , N. Kato , Y. Nemoto , A stable QoE Quality of Experience routing protocol to support ITS services in VANET networks, IEEE Trans. Veh.

QoS Quality of Service Technol. 56 (November(6)) (2007) 3337–3347 .

QUVoD Quality oriented User-centric mobile VoD [11] M. Heddebaut , J. Rioult , J.P. Ghys , C. Gransart , S. Ambellouis , Broadband vehi- RAN Radio Access Network cle-to-vehicle communication using an extended autonomous cruise control RATs Radio Access Technologies sensor, Meas. Sci. Technol. 16 (June(6)) (2005) 1363–1373 . RBRA Road-Based Routing Algorithm [12] M. Shulman , R. Deering , Third Annual Report of the Crash Avoidance Metrics RBs Resource Blocks Partnership April 20 03–March 20 04, National Highway Traffic Safety Admin- RCP Resource Command Processor istration (NHTSA), Washington, DC, 2005 February . RSS Received Signal Strength [13] C. Bergese , A. Braun , E. Porta , Inside CHAUFFEUR, in: Proceedings of 6th ITS

RSSI Received Signal Strength Indicator World Congr, Toronto, ON, Canada, 1999 Novemeber.

RSU Road Side Unit [14] L. Andreone, C. Ricerche, Activities and applications of the vehicle to vehicle

RTPIS Real Time Passenger Information System and vehicle to infrastructure communication to enhance road safety, in: Pro- ceedings of 5th Euro Congress Exhibiton ITS, Hannover, Germany, 2005 June . SCHs Service Channels [15] R. Kruger , H. Fubler , M. Torrent-Moreno , M. Transier , H. Hartenstein , W. Ef- SDN Software-Defined Network felsberg , Statistical Analysis of the FleetNet Highway Movement Patterns, Uni-

SDR Software Defined Radio versity of Mannheim, 2005 Tech. Rep. TR-2005-004, July .

SFMMA Seamless Flow Mobility Management Architecture [16] W.-B. Zhang , H. Tan , A. Steinfield , B. Bougler , D. Empey , K. Zhou , M. Tomizuka ,

SGS Simplified Gateway Selection Implementing advanced vehicle control and safety systems (AVCSS) for high- SLA Speed and Location Aware way maintenance operations, in: Proc. 6th Annu. World Congr. ITS, Toronto, SSF Strongest-Signal First ON, Canada, 1999 November . SSMMA System Selection and Mobility Management Agent [17] N.Sawal Mahendri , A survey on vehicular Ad-hoc networks (VANETs), Int. J. TBVH Time Before Vertical Handover Adv. Res. Comput. Eng.Technol. (IJARCET) 4 (May(5)) (2015) . TDD Time Division Duplex [18] IEEE Trial-use standard for wireless access in vehicular environments (WAVE), TDM Time-division multiplexed 2006. Resource Manager. IEEE Std 1609. 1-2006c1-63 http://dx.doi.org/10.

TDMA Time Division Multiple Access 1109/IEEESTD.2006.246485.

THOR Traffic monitoring Hybrid ORiented service [19] C.C. Communication Consortium, Car2car communication consortium mani- festo.  Http://car-to-car.org/index.php?id=31  . TMC Traffic Management Center [20] S. Al-Sultan , M. Al-Doori , A. Al-Bayatti , H. Zedan , A comprehensive survey UE User Equipment on vehicular Ad Hoc network, J. Netw. Comput. Appl. 37 (January) (2014) UMTS Universal Mobile Telecommunications System 380–392 .

US DoT United State Department of Transportation [21] United States Department of Transportation, Office of the assistant secre-

V2I Vehicle-to-Infrastructure tary for research and technology: vehicle-to-infrastructure (V2I) communica- V2G Vehicle-to-Grid tions for safety. Accessed 2015-10-14, URL: http://www.its.dot.gov/factsheets/ V2V Vehicle-to-Vehicle v2isafety _factsheet.htm . VANET Vehicular Ad-hoc Network [22] S. Olariu , M.C. Weigle , Vehicular networks: from Theory to Practice, 1st ed., VBN Vehicular Backbone Network Chapman and Hall/CRC, 2009 March 17ISBN 9781420085884 - CAT# C5883 . VCC Vehicular Cloud Computing [23] A. Festag , et al. , Now-network on wheels: project objectives, technology and VFC Vehicular Fog Computing achievements, in: Proceedings of the 5th International Workshop on Intelli-

VFHS Vehicular Fast Handover Scheme gent Transportation (WIT), Hamburg, Germany, 2008, pp. 211–216. March.

VHO Vertical Handover [24] M.L. Sichitiu, M. Kihl, Inter-vehicle communication systems: a survey, IEEE Commun. Surveys Tutorials 10 (August(2)) (2008) 88–105 . VMaSC Vehicular Multi-hop algorithm for Stable Clustering [25] D. Jiang , L. Delgrossi , IEEE 802.11p: Towards an International Standard for VoD Video on Demand Wireless Access in Vehicular Environments, Mercedes-Benz Research & De- VU Vehicular User velopment North America, Inc., 2008 .

WAVE Wireless Access in Vehicular Environments [26] FCC allocates spectrum in 5.9 GHz range for intelligent transportation sys-

WCDMA Wideband Code Division Multiple Access tems uses: Action will improve the efficiency of the nation’s transportation WEAC Warning Energy Aware Cluster-head infrastructure, Washington, D.C. 20554. Oct. 21, 1999. Report No. ET 99-5. WHO World Health Organization http://www.fcc.gov . Wi-Fi Wireless Fidelity [27] Y. Toor , P. Muhlethaler , A. Laouiti , Vehicle ad hoc networks: applications and WiMAX Worldwide Interoperability for Microwave Access related technical issues, IEEE Commun. Surveys Tutorials 10 (3rd quarter(3)) WLAN Wireless Local Area Network (2008) 74–88 . WPCF WAVE Point Coordination Function [28] A. Islam , N.H.M. Bhuyan , The Effect of Radio Channel Modelling on the Net-

WSM Wave Short Message work Performance in VANET Master’s Thesis, Department of Electrical and

ZGLS ZoomOut Geographic Location Service Information Technology, Faculty of Engineering, LTH, Lund University, March 2015 . [29] European Telecommunications Standards Institute, Intelligent Transport Sys- tems (ITS); Radio Communications Equipment Operating in the 5 855MHz to References 5 925MHz Frequency Band, ETSI, France, May 2013 Technical Report ETSI EN 302 571 V1.2.0 .

[1] W. H. Organization, WHO global status report on road safety (9 May [30] A. Sharma, B. Raj, K. Kapoor, D. Jyoti, A novel approach for detection of traffic

2016), Reviews Fact Sheet on Road Traffic Injuries, World Health Orga- congestion in NS2, Int. J. Adv. Res. Ideas Innov. Technol. (IJARIIT) 2 (4) (2016).

nization, 2016 https://communitymedicine4asses.wordpress.com/2016/05/10/ [31] J.J. Blum, A. Eskandarian, L.J. Hoffman, Challenges of inter-vehicle ad hoc net-

who-reviews-fact-sheet-on-road-traffic-injuries-9-may-2016/ . works, IEEE Trans. Intell. Transp. Syst. 5 (December(4)) (2004) 47–51.

[2] W. H. Organization, WHO fact sheet on road traffic injuries (November 2016), [32] S. Yousefi, M.S. Mousavi, M. Fathy, Vehicular ad hoc networks (vanets): chal-

Road Traffic Injuries, World Health Organization, 2016 . lenges and perspectives, in: Proceedings of the 6th International Conference

[3] J. Roth, Harmonizing cars and humans in Japan’s era of mass automobility, on ITS Telecommunications, 2006, pp. 761–766. July.

Asia Pacific J. 9 (November(3)) (2011). No. http://japanfocus.org/- Joshua- Roth/ [33] J. Jakubiak, Y. Koucheryavy, State of the art and research challenges for

3643/article.pdf . VANETs, in: IEEE Consumer Communications and Networking Conference

[4] Website: http://www.itsa.org/ . CCNC, 2008, pp. 912–916.

[5] D. Jiang , L. Delgrossi , IEEE 802.11p: Towards an international standard for [34] M. Nekovee, Sensor networks on the road: the promises and challenges of

wireless access in vehicular environments, in: IEEE Vehicular Technology vehicular ad hoc networks and grids, Workshop on Ubiquitous Computing

Conference, May, 2008, pp. 2036–2040 . and e-Research, 2005.

[6] N. Choi , S. Choi , Y. Seokt , T. Kwon , Y. Choi , A solicitation-based IEEE 802.11p [35] T.D. Nguyen, O. Berder, O. Sentieys, Energy-efficient cooperative techniques

MAC protocol for roadside to vehicular networks, in: Mobile Networking for for infrastructure-to-vehicle communications, IEEE Trans. Intell. Transp. Syst.

Vehicular Environments, May, 2007, pp. 91–96 . 12 (September(3)) (2011) 659–668. A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 75

[36] K. Zheng , Q. Zheng , P. Chatzimisios , W. Xiang , Y. Zhou , Heterogeneous ve- [65] S.K. Ray , K. Pawlikowski , H. Sirisena , Handover in mobile WiMAX networks: hicular networking: a survey on architecture, challenges, and solutions, IEEE the state of art and research issues, IEEE Commun. Surv. Tutorials 12 (3) Commun. Surveys Tutorials 17 (2015) 2377–2396 . (2010) 376–399 . [37] M. Pan , P. Li , Y. Fang , Cooperative communication aware link scheduling for [66] Y.T. Chang , J.W. Ding , C.H. Ke , I.Y. Chen , A survey of handoff schemes for ve- cognitive vehicular networks, IEEE J. Sel. Areas Commun. 30 (May(4)) (2012) hicular ad-hoc networks, in: IWCMC ’10 Proceedings of the 6th International 760–768 No. . Wireless Communications and Mobile Computing Conference, Caen, France, [38] F. Liu , L. Shan , Heterogeneous vehicular communication architecture and key 2010, pp. 1228–1231. June 28–July 02 . technologies, ZTE Commun. 8 (August(4)) (2010) 39–44 . [67] U. Kumaran , Vertical handover in vehicular ad-hoc networks - a survey, Int. [39] A. Vinel , 3GPP LTE versus IEEE 802.11p/WAVE: which technology is able to J. Latest Trends Eng. Technol. (IJLTET) 3 (March(4)) (2014) 132–138 . support cooperative vehicular safety applications? IEEE Wireless Commun. [68] S. Midya , K. Majumder , A. Roy , D. De , Vertical handoff mechanisms in VANET: Lett. 1 (February(2)) (2012) 125–128 . a survey, ACM ICTCS ’16, 2016 March 04–05 . [40] E. Ndashimye , S.K. Ray , N.I Sarkar , J.A. Gutiérrez , Vehicle-to-infrastructure [69] S. Lee , K. Sriram , K. Kim , Y.H. Kim , N. Golmie , Vertical handoff decision algo- communication over multi-tier heterogeneous networks: a survey, Comput. rithms for providing optimized performance in heterogeneous wireless net- Netw. 112 (2017) 144–166 . works, IEEE Trans. Veh. Technol. 58 (February(2)) (2009) . [41] E. Hossain , G. Chow , V.C.M. Leung , R.D. McLeod , J. Mišiè, V.W.S. Wong , [70] J. Kang , Y. Chen , R. Yu , X. Zhang , H. Chen , L. Zhang , Vertical handoff in ve- O. Yang , Vehicular telematics over heterogeneous wireless networks: a sur- hicular heterogeneous networks using optimal stopping approach, IEEE 8th vey, Comput. Commun. 33 (2010) 775–793 . International Conference on Communications and Networking in China (CHI- [42] G. Karagiannis , O. Altintas , E. Ekici , G. Heijenk , B. Jarupan , K. Lin , T. Weil , NACOM), 2013 . Vehicular networking: a survey and tutorial on requirements, architectures, [71] U. Kumaran , R.S. Shaji , Vertical handover in vehicular ad-hoc network using challenges, standards and solutions, IEEE Commun. Surveys Tutorials 13 (4) multiple parameters, IEEE International Conference on Control, Instrumenta- (2011) 584–616 . tion, Communication and Computational Technologies (ICCICCT), 2014 . [43] G. Remy , S.-M. Senouci , F. Jan , Y. Gourhant , LTE4V2X: LTE for a centralized [72] S. Bi , C. Chen , R. Du , X. Guan , Proper handover between VANET and cellu- VANET organization, in: IEEE Global Telecommunications Conference (GLOBE- lar network improves internet access, IEEE 80th Vehicular Technology Con- COM 2011), Houston, TX, USA, 2011, pp. 1–6. 5–9 December . ference (VTC, Fall), 2014 . [44] Z.H. Mir , F. Filali , LTE and IEEE 802.11 p for vehicular networking: a perfor- [73] K.L. Chiu , R.H. Hwang , Y.S. Chen , Cross-layer design vehicle-aided handover mance evaluation, EURASIP J. Wireless Commun. Netw. (2014) 1–15 . scheme in VANETs, Wireless Commun. Mobile Comput. 11 (July(7)) (2011) [45] M. Kihl , K. Bür , P. Mahanta , E. Coelingh , 3GPP LTE downlink scheduling strate- 916–928 . gies in vehicle-to-infrastructure communications for traffic safety applica- [74] S.H. Ahmed , S.H. Bouk , D. Kim , Reducing scanning latency in WiMAX enabled tions, in: IEEE Symposium on Computers and Communications (ISCC), Cap- VANETs, Research in Adaptive and Convergent Systems Conference RACS’14, padocia, Turkey, 2012, pp. 448–453. July . ACM, USA, 2014 October . [46] G. Remy , S.-M. Senouci , F. Jan , Y. Gourhant , LTE4V2X—impact of high mobility [75] R.S. Shukla , N. Tyagi , Global mobility and handover management for het- in highway scenarios, in: Global Information Infrastructure Symposium (GIIS), erogeneous network in VANET, Int. J. Next Gener. Netw. (IJNGN) 5 (Septem- Da Nang, Vietnam, 2011, pp. 1–7. August . ber(3)) (2013) 23–34 . [47] L.-C. Tung , M. Gerla , LTE resource scheduling for vehicular safety applications, [76] R.I. Meneguette , L.F. Bittencourt , E. Madeira , A seamless flow mobility man- in: the 10th Annual Conference on Wireless On-demand Network Systems agement architecture for vehicular communication networks, J. Commun. and Services (WONS), Canada, 2013, pp. 116–118. March . Netw. 15 (April(2)) (2013) 207–216 . [48] W. Viriyasitavat , M. Boban , H.-M. Tsai , A. Vasilakos , Vehicular communica- [77] J.M. Chung , M. Kim , Y.S. Park , M. Choi , S. Lee , H. Seo Oh , Time coordinated V2I tions: survey and challenges of channel and propagation models, IEEE Veh. communications and handover for WAVE networks, IEEE J. Sel. Areas Com- Technol. Mag. 10 (June(2)) (2015) 55–66 . mun. 29 (March(3)) (2011) . [49] M.S Sahasrabudhe , M. Chawla , Survey of applications based on vehicular [78] J. Dias , A. Cardote , F. Neves , S. Sargento , A. Oliveira , Seamless horizontal and ad-hoc network (VANET) framework, Int. J. Comput. Sci. Inf. Technol. (IJCSIT) vertical mobility in VANET, in: IEEE Vehicular Networking Conference (VNC), 5 (3) (2014) 3937–3942 . 2012, pp. 226–233. November . [50] S. Kato , M. Hiltunen , K. Joshi , R. Schlichting , Enabling vehicular safety applica- [79] P. Dahiya , S. Deswal , Performance analysis of advanced FMIPv6 for han- tions over LTE networks, in: International Conference on Connected Vehicles dover management in VANET, Int. J. Innov. Res. Sci. Eng. Technol. (IJIRSET) and Expo (ICCVE), Las Vegas, NV, USA, 2013, pp. 747–752. 2–6 December . 3 (June(6)) (2014) 13572–13578 . [51] H.V. Nejad , A. Ramazani , T. Mohammadi , W. Mansoor , A survey on contex- [80] M. Gramaglia , C.J. Bernardos , M. Calderon , Seamless internet 3G and oppor- t-aware vehicular network applications, Veh. Commun. 3 (2016) 43–57 . tunistic WLAN vehicular connectivity, EURASIP J. Wireless Commun. Netw., [52] U.L. Mane , A .A . Junnarkar , A survey on techniques and fundamentals of inter- 2011 Springer . net access in VANET-INTERNET integration scenarios, Int. J. Comput. Sci. Inf. [81] S. Céspedes , X.S. Shen , On achieving seamless IP communications in hetero- Technol. (IJCSIT) 6 (3) (2015) 2035–2039 . geneous vehicular networks, IEEE Trans. Intell. Transp. Syst. 16 (December(6)) [53] M. Gerla , L. Kleinrock , Vehicular networks and the future of the mobile inter- (2015) 3223–3237 . net, Comput. Netw. 55 (February(2)) (2011) 457–469 . [82] D. Mu , X. Ge , R. Chai , Vertical handoff modeling and simulation in VANET [54] M.S. Anwer , C. Guy , A survey of VANET technologies, J. Emerging Trends Com- scenarios, IEEE International Conference on Wireless Communications and put. Inf. Sci. 5 (September(9)) (2014) 661–671 . Signal Processing (WCSP), 2013 . [55] T. Mangel , T. Kosch , H. Hartenstein , A comparison of UMTS and LTE for ve- [83] A. Ghosh , V.V. Paranthaman , G. Mapp , O. Gemikonakli , J. Loo , Enabling seam- hicular safety communication at intersections, in: IEEE Vehicular Networking less V2I communications: toward developing cooperative automotive applica- Conference (VNC), Jersey City, NJ, USA, 2010, pp. 293–300. December . tions in VANET systems, IEEE Commun. Mag. (December) (2015) 80–86 . [56] Z.H. Mir , J. Toutouh , F. Filali , E. Alba , QoS-Aware radio access technology [84] S. Joseph , R. Rajagopal , Efficient handoff based privacy preservation for (RAT) selection in hybrid vehicular networks, Springer Commun. Technol. VANET, Int. J. Sci. Res. (IJSR) 5 (March(3)) (2016) 1703–1708 . Veh. (2015) 117–128 . [85] W. Chen , R.K. Guha , T.J. Kwon , J. Lee , I.Y. Hsu , A survey and challenges in rout- [57] G. Giambene , M. Marandola , Internet access in hybrid terrestrial and satel- ing and data dissemination in vehicular ad-hoc networks, IEEE International lite mobile communication systems, in: IEEE Vehicular Technology Confer- Conference on Vehicular Electronics and Safety, 2008 September . ence (VTC), 2004, pp. 2811–2815 . [86] B. Dhawan , T.P. Singh , Efficient data dissemination techniques in VANETs: a [58] B. Aslam , P. Wang , C. Zou , Pervasive Internet access by vehicles through satel- review, Int. J. Comput. Appl. 116 (April(7)) (2015) 22–32 . lite receive-only terminals, in: Proceedings of 18th International Conference [87] D. Sutariya , S. Pradhan , Data dissemination techniques in vehicular ad hoc on Computer Communications and Networks, San Francisco, CA, USA, 2009, network, Int. J. Comput. Appl. 8 (October(10)) (2010) 35–39 . pp. 1–6. 3–6 August . [88] S. Ilarri , T. Delot , R.T. Lado , A data management perspective on vehicular net- [59] Q. Li , Y. Wu , S. Feng , P. Zhang , Y. Zhou , Cooperative uplink link adaptation in works, IEEE Commun. Surv. Tutorials 17 (4) (2015) 2420–2460 . 3GPP LTE heterogeneous networks, in: IEEE 77th Vehicular Technology Con- [89] X. Shen , X. Cheng , L. Yang , R. Zhang , B. Jiao , Data dissemination in VANETs: ference (VTC), 2013, pp. 1–5 . a scheduling approach, IEEE Trans. Intell. Transp. Syst. 15 (April(5)) (2014) [60] J. M-Barja , C.T. Calafate , J.-C. Cano , P. Manzoni , An overview of vertical han- 2213–2223 . dover techniques: algorithms, protocols and tools, Comput. Commun. 34 [90] P. Joshi , J. Kaur , I.S. Gill , Heterogeneous configuration: a better approach to (June(8)) (2011) 985–997 . analyze VANET, Int. J. Comput. Appl. 118 (May(19)) (2015) 26–30 . [61] K. Zhu , D. Niyato , P. Wang , E. Hossain , D. In Kim , Mobility and handoff man- [91] H. Zhang , F. Bai , X. Ju , Heterogeneous vehicular wireless networking: a the- agement in vehicular networks: a survey, Wireless Commun. Mobile Comput. oretical perspective, IEEE Wireless Communications and Networking Confer- 11 (April(4)) (2011) 459–476 . ence (WCNC), 2015 . [62] Y.S. Chen , C.S. Hsu , C.H. Cheng ,Network mobility protocol for vehicular ad [92] I. Stojmenovic , Fog computing: a cloud to the ground support for smart hoc networks, Int. J. Commun. Syst. 27 (11) (2014) 3042–3063 November . things and machine-to-machine networks, in: Australasian Telecommunica- [63] A. Magnano , X. Fei , A. Boukerche , Predictive mobile IP handover for vehicu- tion Networks and Applications Conference (ATNAC), Southbank, VIC, Aus- lar networks, in: 40th IEEE Conference on Local Computer Networks (LCN), tralia, 2014, pp. 117–122. November . Clearwater Beach, Florida, USA, 2015, pp. 338–346. 26–29 October . [93] A. Baiocchi , F. Cuomo , Infotainment services based on push-mode dissem- [64] S.K. Ray , H. Sirisena , D. Deka , LTE-advanced handover: an orientation match- ination in an integrated VANET and 3G architecture, J. Commun. Netw. 15 ing-based fast and reliable approach, in: 38th Annual IEEE Conference on Lo- (April(2)) (2013) 179–190 . cal Computer Networks, Sydney, Australia, 2013, pp. 280–283. 21–24 October . [94] Y. Yang , P. Wang , C. Wang , F. Liu , An eMBMS based congestion control scheme in cellular-VANET heterogeneous networks, IEEE 17th International Confer- ence on Intelligent Transportation Systems (ITSC), 2014 October . 76 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

[95] I. Rubin , Y.Y. Lin , A. Baiocchi , F. Cuomo , P. Salvo , Micro base station aided less networks, in: IEEE 7th International Conference on Wireless and Mobile vehicular ad hoc networking, IEEE International Conference on Computing, Computing, Networking and Communications (WiMob), Wuhan, china, 2011, Networking and Communications (ICNC), 2014 . pp. 17–24. October . [96] Z. He , D. Zhang , J. Liang , Cost-efficient heterogeneous data transmission [121] N. Haziza , et al. , Multi-technology vehicular cooperative system based on in software defined vehicular networks, IEEE 17th International Conference software defined radio (SDR), in: Communication Technologies for Vehicles, on High Performance Computing and Communications (HPCC), 2015 IEEE Springer, 2013, pp. 84–95 . 7th International Symposium on Cyberspace Safety and Security (CSS), and [122] A.B. MacKenzie , S.B. Wicker , Game theory in communications: motivation, ex- 2015 IEEE 12th International Conference on Embedded Software and Systems planation, and application to power control, in: IEEE Global Telecommunica- (ICESS), 2015 . tions Conference (GLOBECOM’01), San Antonio, TX, USA, 2001, pp. 821–826 . [97] S. Jia , S. Hao , X. Gu , L. Zhang , Analyzing and relieving the impact of FCD traf- [123] D. Wu , Y. Ling , H. Zhu , J. Liang , The RSU access problem based on evolution- fic in LTE-VANET heterogeneous network, 21st IEEE International Conference ary game theory for VANET, Int. J. Distrib. Sens. Netw. (2013) Hindawi . on Telecommunications (ICT), 2014 . [124] D. Niyato , E. Hossain , M. Hassan , Game-theoretic models for vehicular net- [98] M.M. Alotaibi , H. Mouftah , High speed multi-hop data dissemination for het- works, in: Game Theory for Wireless Communications and Networking, first erogeneous transmission ranges in VANETs, IEEE International Conference on ed., CRC Press, 2012, pp. 61–98 . Ubiquitous Wireless Broadband (ICUWB), 2015 . [125] D.B. Rawat , B.B. Bista , G. Yan , CoR-VANETs: game theoretic approach for chan- [99] M.M. Alotaibi , H.T. Mouftah , Probabilistic area-based dissemination for het- nel and rate selection in cognitive radio VANETs, 7th IEEE International Con- erogeneous transmission ranges in vehicular ad-hoc networks, IEEE Inter- ference on Broadband, Wireless Computing, Communication and Applica- national Conference on Computer and Information Technology; Ubiquitous tions, 2012 . Computing and Communications; Dependable, Autonomic and Secure Com- [126] Y. Li , K. Ying , P. Cheng , H. Yu , H. Luo , Cooperative data dissemination in cel- puting; Pervasive Intelligence and Computing, 2015 . lular-VANET heterogeneous wireless networks, 4th IEEE International High [100] R. Sivaraj , A.K. Gopalakrishna , M.G. Chandra , P. Balamuralidhar , QoS-enabled Speed Intelligent Communication Forum, 2012 . group communication in integrated VANET-LTE heterogeneous wireless net- [127] S. Shivshankar , A. Jamalipour , An evolutionary game theory-based approach works, IEEE 7th International Conference on Wireless and Mobile Computing, to cooperation in VANETs under different network conditions, IEEE Trans. Networking and Communications (WiMob), 2011 . Veh. Technol. 64 (May(5)) (2015) 2015–2022 . [101] W. Zhou , Ch. Ren , Ch. Ma , H. Yu , Multicast/broadcast service in integrated [128] R.K. Goyal , S. Kaushal , Network selection using AHP for fast moving vehicles VANET-cellular heterogeneous wireless networks, International Conference on in heterogeneous networks, in: Advanced Computing and Systems for Secu- Wireless Communications and Signal Processing (WCSP), 2013 . rity, Springer, 2016, pp. 235–243 . [102] W. Drira , K. Ahn , H. Rakha , F. Filali , Development and testing of a 3G/LTE [129] J.M. de Fuentes , L.G. -Manzano , A.I.G. -Tablas , J. Blasco , Security models in adaptive data collection system in vehicular networks, IEEE Trans. Intell. vehicular ad-hoc networks: a survey, IETE Tech. Rev. 31 (2014) 47–64 . Transp. Syst. 17 (January(1)) (2016) 240–249 . [130] S.S. Tangade , S.S. Manvi , A survey on attacks, security and trust management [103] M. De Felice , F. Cuomo , A. Baiocchi , I. Turcanu , S. Zennaro , Traffic monitor- solutions in VANETs, IEEE 4th International Conference on Computing, Com- ing and incident detection using cellular and early stage VANET technology munications and Networking Technologies (ICCCNT), 2013 July . deployment, ACM IoV-VoI’16, 2016 July . [131] B. Mishra , P. Nayak , S. Behera , D. Jena , Security in vehicular ad hoc networks: [104] J. Fiosinaa , M. Fiosinsa , J.P. Müllera , Big data processing and mining for next a survey, in: International Conference on Communication, Computing & Se- generation intelligent transportation systems, Jurnal Teknologi 63 (3) (2013) curity, New York, NY, USA, 2011, pp. 590–595 . 23–38 . [132] Y. Kim , I. Kim , Security issues in vehicular networks, in: International Con- [105] I. Turcanu , P. Salvo , A. Baiocchi , F. Cuomo , in: An Integrated VANET-based ference on Information Networking 2013 (ICOIN), Bangkok, Thailand, 2013, Data Dissemination and Collection Protocol for Complex Urban Scenarios, 52, pp. 468–472. January . Elsevier Ad Hoc Networks, 2016, pp. 28–38. December . [133] R.G. Engoulou , M. Bellaïche , S. Pierre , A. Quintero , VANET security surveys, [106] R.S. Shukla , N. Tyagi , Routing algorithm for heterogeneous wireless network Comput. Commun. J. 44 (May) (2014) 1–13 . in VANET, IOSR J. Comput. Eng. (IOSR-JCE) 13 (July–August(6)) (2013) 68–77 . [134] V.S. Yadav , S. Misra , M. Afaque , in: Security of Wireless and Self-Organiz- [107] S. Agrawal , N. Tyagi , A.K. Misra , Seamless VANET connectivity through hetero- ing Networks: Security in Vehicular Ad Hoc Networks, CRC Press, 2010, geneous wireless network on rural highways, ICTCS ’16, 2016 March 04–05 . pp. 227–250 . [108] Z.H. Mir, J. Kim, Y.B. Ko, F. Filali, “Improved multi-hop routing in integrated [135] S.S. Manvi , S. Tangade , A survey on authentication schemes in VANETs for VANET-LTE hybrid vehicular networks,” in ACM IMCOM ’16, Danang, Viet secured communication, Veh. Commun. J. 9 (2017) 19–30 Vol. . Nam, January 2016. [136] R. Mishra , A. Singh , R. Kumar , VANET security: issues, challenges and solu- [109] S. Gopinath , L. Wischhof , C. Ponikwar , H.-J. Hof , Hybrid solutions for data tions, in: IEEE International Conference on Electrical, Electronics, and Opti- dissemination in vehicular networks, Wireless Days Conference (WD), 2016 mization Techniques (ICEEOT), 2016, pp. 1050–1055. March . March . [137] X. Zhu , Y. Lu , X. Zhu , S. Qiu , Lightweight and scalable secure communication [110] T. Yang , R. Zhang , X. Cheng , L. Yang , A graph coloring resource sharing in VANET, Int. J. Electron. 102 (5) (2015) 765–780 Vol. . scheme for full-duplex cellular-VANET heterogeneous networks, IEEE Inter- [138] Y.-H. Park , S.-W. Seo , Fast and secure group key dissemination scheme for national Conference on Computing, Networking and Communications (ICNC), out-of-range V2I Communication, IEEE Trans. Veh. Technol. 64 (2) (2015) Workshop on Computing, Networking and Communications (CNC), 2016 . 5642–5652 . [111] T. Zhang , F. Wu , Stimulating traffic information transfer in non-cooperative [139] M. Whaiduzzaman , M. Sookhak , A. Gani , R. Buyya , A survey on vehicular vehicular ad hoc networks, International Wireless Communications and Mo- cloud computing, J. Netw. Comput. Appl. 40 (April) (2014) 325–344 . bile Computing Conference (IWCMC), 2014 . [140] L. Gu , D. Zeng , S. Guo , Vehicular cloud computing: a survey, in: IEEE Globe- [112] S. Ucar , S.C. Ergen , O. Ozkasap , Multi-hop cluster based IEEE 802.11p and LTE com Workshops (GC Wkshps), 2013, pp. 403–407. Dec. . hybrid architecture for VANET safety message dissemination, IEEE Trans. Veh. [141] E. Lee , M. Gerla , S. Oh , Vehicular cloud networking: architecture and design Technol. (2015) . principles, IEEE Commun. Mag. 52 (February(2)) (2014) 148–155 . [113] A .A . Eltahir , R.A . Saeed , R.A . Mokhtar , Vehicular communication and cellu- [142] M.F. Feteiha , H.S. Hassanein ,Enabling cooperative relaying VANET clouds over lar network integration: gateway selection perspective, 5th IEEE International LTE-A networks, IEEE Trans. Veh. Technol. 64 (April(4)) (2015) 1468–1479 . Conference on Computer & Communication Engineering, 2014 . [143] B. Liu , D. Jia , J. Wang , K. Lu , L. Wu , Cloud-assisted safety message dissemina- [114] P. Salvo , I. Turcanu , F. Cuomo , A. Baiocchi , I. Rubin , LTE floating car data ap- tion in VANET–cellular heterogeneous wireless network, IEEE Syst. J. (2015) . plication off-loading via VANET driven clustering formation, 12th IEEE Annual [144] S.M. Bilal , C.J. Bernardos , C. Guerrero , Position-based routing in vehicular net- Conference on Wireless On-demand Network Systems and Services (WONS), works: a survey, J. Netw. Comput. Appl. 36 (2013) 685–697 . 2016 . [145] J.J. Cheng , J.L. Cheng , M.C Zhou , F.Q. Liu , S.C. Gao , C. Liu , Routing in internet [115] G. Zhioua , H. Labiod , N. Tabbane , S. Tabbane , An efficient QoS based gate- of vehicles: a review, IEEE Trans. Intell. Transp. Syst. 16 (October(5)) (2015) . way selection algorithm for VANET to LTE advanced hybrid cellular network, [146] B.T. Sharef , R.A. Alsaqour , M. Ismail , Vehicular communication ad hoc routing MSWiM’12, ACM, Paphos, Cyprus, 2012 October 21–25 . protocols: a survey, J. Netw. Comput. Appl. 40 (2014) 363–396 . [116] M.A . Alawi , R.A . Saeed , A .A . Hassan , Cluster- based multi-hop vehicular com- [147] K.A. Hafeez , L. Zhao , Z. Liao , B.N. Ma , Impact of mobility on VANETs’ safety munication with multi-metric optimization, IEEE International Conference on applications, IEEE Global Telecommunications Conference GLOBECOM, 2010 . Computer and Communication Engineering (ICCCE 2012), 2012 3–5 July . [148] G. Rémy , S.M. Senouci , F. Jan , Y. Gourhant , LTE4V2X - impact of high mobility [117] K. Ju , L. Chen , H. Wei , K. Chen , An efficient gateway discovery algorithm in highway scenarios, IEEE Global Information Infrastructure Symposium – with delay guarantee for VANET-3G heterogeneous networks, in: Wireless GIIS, 2011 . Personal Communications, Springer science Business Media, New York, 2014, [149] H. Soleimani , A. Boukerche , SLA: speed and location aware LTE scheduler for pp. 2019–2036 . vehicular safety applications, MobiWac’15, ACM, Nov. 2015 . [118] G. Zhioua , H. Labiod , N. Tabbane , S. Tabbane , FQGwS: a gateway selection al- [150] N. Prasad , S.P. Priyanka , Performance of tracking public transport in heteroge- gorithm in a hybrid clustered VANET LTE-advanced network: complexity and neous networks, IEEE International Conference on Research in Computational performances, IEEE International Conference on Computing, Networking and Intelligence and Communication Networks (ICRCICN), 2015 . Communications, Mobile Computing & Vehicle Communications Symposium, [151] Y. Mo , D. Yu , J. Song , K. Zheng , Y. Guo , Vehicle Position Updating Strategy 2014 . Based on Kalman Filter Prediction in VANET Environment, Hindawi Publish- [119] L.C. Tung , J. Mena , M. Gerla , C. Sommer , A cluster based architecture for inter- ing Corporation Discrete Dynamics in Nature and Society, 2016 . section collision avoidance using heterogeneous networks, 12th IEEE Annual [152] C. Xu , F. Zhao , J. Guan , H. Zhang , G.M. Muntean , QoE-driven user-centric VoD Mediterranean Ad Hoc Networking Workshop (MED-HOC-NET), 2013 . services in urban multi-homed P2P-based vehicular networks, IEEE Trans. [120] R. Sivaraj , A.K. Gopalakrishna , M.G. Chandra , P. Balamuralidhar , QoS-en- Veh. Technol. 62 (June(5)) (2013) . abled group communication in integrated VANET-LTE heterogeneous wire- A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 77

[153] N. Cheng , N. Lu , N. Zhang , X.S. Shen , J.W. Mark , Opportunistic WiFi offloading [185] L. Lei , Z. Zhong , C. Lin , X. Shen , Operator controlled device-to-device commu- in vehicular environment: a queueing analysis, IEEE Global Communications nications in LTE-advanced networks, IEEE Wirel. Commun. 19 (June(3)) (2012) Conference Globecom - Ad Hoc and Sensor Networking Symposium, 2014 . 96–104 . [154] Y. Park , J. Ha , S. Kuk , H. Kim , C.M. Liang , J. Ko , A feasibility study and develop- [186] 3rd Generation Partnership Project, Technical Specification Group Radio Ac- ment framework design for realizing smartphone-based vehicular networking cess Network, Study on LTE Device to Device Proximity Services - Radio As- systems, IEEE Trans. Mob. Comput. 13 (November(11)) (2014) 2431–24 4 4 . pects (Release 12), 2014 3GPP TR 36.843, V12.0.1, March . [155] D. Liden , What Is a Driverless Car?, 11, 2013 WiseGeek, Retrieved Oct. . [187] , in: IEEE standard for information technology-local and metropolitan area [156] J. Bryner , Ford takes autonomous cars for snowy test drive, Live Science networks-specific requirements-part 11: wireless LAN medium access control (2016) January 12 . (MAC) and physical layer (PHY) specifications amendment 6, 2010, pp. 1–51. [157] J. Boyd, Mitsubishi Electric joins race to make maps for self-drive cars, 2016. IEEE Std. 802.11p-2010, Jul. . Decemebr 8 www.atimes.com . [188] G. Korkmaz , E. Ekici , F. Özgüner , Ü. Özgüner , Urban multi-hop broadcast pro- [158] E. Badger , 5 Confounding Questions that Hold the Key to the Future of Driver- tocol for inter-vehicle communication systems, in: Proceeding of 1st ACM In- less Cars, 2015 Wonk Blog. The Washington Post, Jan. 15 . ternational Workshop on Vehicle Ad Hoc Networks, NY, USA, 2004, pp. 76–85 . [159] Will regulators allow self-driving cars in a few years? Forbes (September) [189] G. Korkmaz , E. Ekici , F. Ozguner , Black-burst-based multi-hop broadcast pro- (2013) . tocols for vehicular networks, IEEE Trans. Veh. Technol. 56 (September(5)) [160] Nevada enacts law authorizing autonomous (driverless) vehicles, Green Car (2007) 3159–3167 . Congress, 2011 June 25. [190] Q. Yang , L. Shen , A multi-hop broadcast scheme for propagation of emer- [161] M. Slosson , Google gets first self-driven car license in Nevada, Reuters, 2012 gency messages in VANET, in: Proc. 12th IEEE ICCT, Nanjing, China, 2010, May 8 . pp. 1072–1075. November . [162] Website: http://bigthink.com/natalie- shoemaker/googles- self- driving- cars- [191] M. Slavik , I. Mahgoub , Spatial distribution and channel quality adaptive proto- are- built- to- make- a- lot- of- assumptions . col for multihop wireless broadcast routing in VANET, IEEE Trans. Mob. Com- [163] Google Self-Driving Car Project Monthly Report, Google, 2016 June 201Re- put. 12 (February(4)) (2013) 722–734 . trieved July 15. [192] M. Barradi , A. Hafid , S. Aljahdali , Highway multihop broadcast protocols [164] M.M. Farland , For the First Time, Google’s Self-Driving Car takes Some Blame for vehicular networks, in: Proceeding IEEE ICC, Ottawa, Canada, 2012, for a Crash, 2016 Washington Post. February 29 . pp. 5296–5300. June . [165] Passenger bus teaches Google robot car a lesson, Los Angeles Times (Febru- [193] M. Slavik , I. Mahgoub , Stochastic broadcast for VANET, in: Proceeding 7th ary) (2016) 29 . IEEE CCNC, Las Vegas, NV, USA, 2010, pp. 1–5. January . [166] Website: https://www.washingtonpost.com/news/innovations/wp/2016/ [194] F. Jamil , A. Javaid , T. Umer , M.H. Rehmani , A comprehensive survey of net- 12/13/google- is- one- step- closer- to- making- money- off- self- driving- cars/ work coding in vehicular ad-hoc networks, Wireless Netw. 23 (November(8)) ?utm _term=.d6d34c6b98d0 . (2017) 2395–2414 . [167] Consumers in US and UK Frustrated with Intelligent Devices That Frequently [195] T. Ho , D.S. Lun , Network Coding: An Introduction, Cambridge University Press, Crash or Freeze, New Accenture Survey Finds, in Accenture, 2011 February 10. Cambridge, 2008 . [168] Autonomous Cars Found Trustworthy in Global Study, in 2013 autosphere.ca, [196] T. Matsuda , T. Noguchi , T. Takine , Survey of network coding and its applica- May 22. tions, IEICE Trans. Commun. 94 (3) (2011) 698–717 . [169] C. Hohenberger , M. Spörrle , I.M. Welpe ,How and why do men and women [197] A. Ashok , P. Steenkiste , F. Bai , Vehicular cloud computing through dynamic differ in their willingness to use automated cars? The influence of emo- computation offloading, Comput. Commun. J. (2017) In press, Available online tions across different age groups, Transp Res. Part A 94 (December) (2016) 19 Dec . 374–385 . [198] F. Bonomi , R. Milito , J. Zhu , S. Addepalli , Fog computing and its role in the [170] K.M. Alheeti , A. Gruebler , K.D. McDonald-Maier , On the detection of grey hole internet of things, in: Proceedings of the 1st MCC Workshop on Mobile Cloud and rushing attacks in self-driving vehicular networks, 7th Computer Science Computing, August, 2012, pp. 13–16 . and Electronic Engineering Conference (CEEC), 2015 . [119] X. Hou , Y. Li , M. Chen , D. Wu , D. Jin , S. Chen , Vehicular fog computing: [171] K.A . Alheeti , A . Gruebler , K.D. McDonald-Maier , An intrusion detection system a viewpoint of vehicles as the infrastructures, IEEE Trans. Veh. Technol. 65 against black hole attacks on the communication network of self-driving cars, (June(6)) (2016) 3860–3873 . in: Sixth International Conference on Emerging Security Technologies (EST), [200] Y. Xiao , C. Zhu , Vehicular fog computing: vision and challenges, IEEE Interna- 2015, pp. 86–91. September 3 . tional Conference on Pervasive Computing and Communications Workshops [172] K.A . Alheeti , A . Gruebler , K.D. McDonald-Maier , An intrusion detection system (PerCom Workshops), 2017 . against malicious attacks on the communication network of driverless cars, [201] P. Hu , S. Dhelim , H. Ning , T. Qiu , Survey on fog computing: architecture, key in: IEEE 12th Consumer Communications and Networking Conference (CCNC), technologies, applications and open issues, J. Netw. Comput. Appl. 98 (2017) 2015, pp. 916–921 . 27–42 . [173] K.M. Alheeti , K. McDonald-Maier , Hybrid intrusion detection in connected [202] C. Huang , R. Lu , K.K.R. Choo , Vehicular fog computing: architecture, use case, self-driving vehicles, IEEE 22nd International Conference on Automation and and security and forensic challenges, IEEE Commun. Mag. 55 (November(11)) Computing (ICAC), 2016 . (2017) 105–111 . [174] A. Belbachir , R. Boutteau , P. Merriaux , J. Blosseville , X. Savatier , From au- [203] V.G. Menon , Moving from vehicular cloud computing to vehicular fog com- tonomous robotics toward autonomous cars, in: IEEE Intelligent Vehicles puting: issues and challenges, Int. J. Comput. Sci. Eng. (IJCSE) 9 (February(02)) Symposium (IV), Gold Coast, Australia, 2013, pp. 23–26. June . (2017) 14–18 . [175] A. Rane , S. Krishnan , S. Waman , Conflict resolution of autonomous cars using [204] K. Kai , W. Cong , L. Tao , Fog computing for vehicular Ad-hoc networks: game theory and cellular automata, International Conference on Reliability, paradigms, scenarios, and issues, J. China Univ. Posts Telecommun. 23 Optimization and Information Technology - ICROIT, 2014 February 6–8 . (April(2)) (2016) 56–65 . [176] M. Gerla , E. Lee , G. Pau , U. Lee , Internet of vehicles: from intelligent grid [205] R. Chaudhary , N. Kumar , Zeadally S , Network service chaining in fog and to autonomous cars and vehicular clouds, IEEE World Forum Internet Things cloud computing for the 5G environment: data management and security (WF-IoT) (2014) . challenges, IEEE Commun. Mag. 55 (November(11)) (2017) 114–122 . [177] C. Pozna , C. Antonya , Issues about autonomous cars, 11th IEEE International [206] M. Tao , K. Ota , M. Dong , Foud: integrating fog and cloud for 5G-enabled V2G Symposium on Applied Computational Intelligence and Informatics, 2016 May networks, IEEE Netw. 31 (02) (2017) 8–13 Vol.No. . 12–14 . [207] W. Zhang , Z. Zhang , H.C. Chao , Cooperative fog computing for dealing with [178] F.D. Cunha , A. Boukerche , L. Villas , A.C. Viana , A.F. Loureiro , Data Communi- big data in the internet of vehicles: architecture and hierarchical resource cation in VANETs: A Survey, Challenges and Applications, INRIA Saclay, 2014 management, IEEE Commun. Mag. 55 (December(12)) (2017) 60–67 . Research Report RR-8498March . [208] C.A.R.L. Brennand , F.D. da Cunha , G. Maia , E. Cerqueira , A .A .F. Loureiro , [179] X. Wu , S. Subramanian , R. Guha , R.G. White , J. Li , K.W. Lu , A Bucceri , T. Zhang , L.A. Villas , FOX: a traffic management system of computer-based vehicles Vehicular communications using DSRC: Challenges, enhancements, and evo- fog, in: IEEE Symposium on Computers and Communication (ISCC), 2016, lution, IEEE J. Sel. Areas Commun. 31 (July(9)) (2013) 399–408 . pp. 982–987 . [180] N. Lu , N. Cheng , N. Zhang , X. Shen , J. Mark , Connected vehicles: solutions and [209] H.J.D. Lopez , M. Siller , I. Huerta , Internet of vehicles: cloud and fog comput- challenges, IEEE Internet Things J. 1 (August(4)) (2014) 289–299 . ing approaches, in: IEEE International Conference on Service Operations and [181] F. Dressler , H. Hartenstein , O. Altintas , O. Tonguz , Inter-vehicle communica- Logistics, and Informatics (SOLI), 2017, pp. 211–216. September . tion: Quo vadis, IEEE Commun. Mag. 52 (June(6)) (2014) 170–177 . [210] N.K. Giang , V.C.M. Leung , R. Lea , On developing smart transportation applica- [182] C. Han , M. Dianati , R. Tafazolli , R. Kernchen , X. Shen , Analytical study of the tions in fog computing paradigm, ACM DIVANet’16, November, 2016 . IEEE 802.11p MAC sublayer in vehicular networks, IEEE Trans. Intell. Transp. [211] J. Li , C. Natalino , D.P. Van , L. Wosinska , J. Chen , Resource management in Syst. 13 (February(2)) (2012) 873–886 . fog-enhanced radio access network to support real-time vehicular services, [183] P. Papadimitratos , A. La Fortelle , K. Evenssen , R. Brignolo , S. Cosenza , Vehic- in: IEEE 1st International Conference on Fog and Edge Computing (ICFEC), ular communication systems: Enabling technologies, applications, and future 2017, pp. 68–74 . outlook on intelligent transportation, IEEE Commun. Mag. 47 (November(11)) [212] C.Y. Huang , K. Xu , Reliable realtime streaming in vehicular cloud-fog com- (2009) 84–95 . puting networks, IEEE International Conference on Communications in China [184] L. Lei , Y. Zhang , X. Shen , C. Lin , Z. Zhong , Performance analysis of device-to (ICCC), 2016 . device communications with dynamic interference using stochastic petri nets, [213] O.T.T. Kim , N.D. Tri , V.D. Nguyen , N.H. Tran , C.S. Hong ,A shared parking IEEE Trans. Wireless Commun. 12 (December(12)) (2013) 6121–6141 . model in vehicular network using fog and cloud environment, in: 17th Asi- a-Pacific Network Operations and Management Symposium (APNOMS), 2015, pp. 321–326 . 78 A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79

[214] A . Alrawais , A . Alhothaily , C. Hu , X. Cheng , Fog computing for the internet [228] K. Zheng , Y. Wang , L. Lei , W. Wang , Cross-layer queuing analysis on multi- of things: security and privacy issues, IEEE Internet Comput. 21 (21) (2017) hop relaying networks with adaptive modulation and coding, IET Commun. 4 34–42 . (February(3)) (2010) 295–302 . [215] J. Ni , A. Zhang , X. Lin , X.S. Shen , Security, privacy, and fairness in fog-based [229] J. Nzouonta , N. Rajgure , G. Wang , C. Borcea , VANET routing on city roads vehicular crowdsensing, IEEE Commun. Mag. 55 (June(6)) (2017) 146–152 . using real-time vehicular traffic information, IEEE Trans. Veh. Technol. 58 [216] B. Huang , X. Cheng , W. Cheng , Meet-fog for accurate distribution of negative (February(7)) (2009) 3609–3626 . messages in VANET, ACM SmartIoT’17, 2017 October . [230] K. Lee , S.H. Lee , R. Cheung , U. Lee , M. Gerla , First experience with CarTor- [217] S. Basudan , X. Lin , K. Sankaranarayanan , A privacy-preserving vehicular rent in a real vehicular Ad Hoc network testbed, in: Mobile Networking for crowdsensing-based road surface condition monitoring system using fog Vehicular Environments Conference, USA, May, 2007, pp. 109–114 . computing, IEEE Internet Things J. 4 (June(3)) (2017) 772–782 . [231] V. Cabrera , F. Ros , P. Ruiz , Simulation-based study of common issues in [218] S.A . Soleymani , A .H. Abdullah , M. Zareei , M.H. Anisi , C. V-Rosales , M.K. Khan , VANET routing protocols, in: Proc. IEEE VTC Spring, Barcelona, Spain, 2009 S. Goudarzi , A secure trust model based on fuzzy logic in vehicular ad hoc April . networks with fog computing, IEEE Access 5 (2017) 15619–15629 . [232] S. Taha , X. Shen , Secure IP Mobility Management for VANET, Secure IP Mobil- [219] M. Rehan , H. Hasbullah , I. Faye , W. Rehan , O. Chughtai , M.H. Rehmani , ZGLS: ity Management for VANET, 2013 . a novel flat quorum-based and reliable location management protocol for [233] L. Elefteriadou , An Introduction to Traffic Flow Theory, 84, Springer, New VANETs, Wireless Netw. (January) (2017) . York, NY, 2014 a Book . [220] H. Saleet , O. Basir , R. Langar , R. Boutaba , Regionbased location-service-man- [234] I.W.-H. Ho , K.K. Leung , J.W. Polak , Stochastic model and connectivity dynam- agement protocol for VANETs, IEEE Trans. Veh. Technol. 59 (February(2)) ics for VANETs in signalized road systems, IEEE/ACM Trans. Netw. 19 (Febru- (2010) 917–931 . ary(1)) (2011) 195–208 . [221] M. Ayaida , M. Barhoumi , H. Fouchal , Y.G. Doudane , L. Afilal , Joint routing [235] Chapter 10 T. Umer , M. Amjad , N. Shah , Z. Ding , Modeling vehicles mobility and location-based service in VANETs, J. Parallel Distrib. Comput. 74 (2014) for connectivity analysis in VANET, in: M. Alam, J. Ferreira, J. Fonseca (Eds.), 2077–2087 . Intelligent Transportation Systems,” Studies in Systems, Decision and Control, [222] W. Lindner , S. Madden , Data management issues in disconnected sensor net- 52 Springer, 2016 . works, GI Jahrestagung 2 (4) (2004) 349–354 . [236] M. Artimy , Local density estimation and dynamic transmission-range assign- [223] O. Wolfson , B. Xu , H.J. Cho , Multimedia traffic information in vehicular net- ment in vehicular Ad Hoc networks, IEEE Trans. Intell. Transp. Syst. 8 (3) works, in: ACM Proceedings of the 17th SIGSPATIAL International Conference (20 07) 40 0–412 . on Advances in Geographic Information Systems, GIS ’09, NY, USA, 2009, [237] T. Umer , Z. Ding , B. Honary , H. Ahmad , Implementation of microscopic pa- pp. 4 80–4 83 . rameters for density estimation of heterogeneous traffic flow for VANET, in: [224] J. Fiosina , J.P. Müller , M. Fiosins , Big data processing and mining for next IEEE 7th International Symposium on Communication Systems, Networks & generation intelligent transportation systems, Jurnal Teknologi 63 (3) (2013) Digital Signal Processing (CSNDSP), 2010, pp. 66–70 . 23–38 . [238] T. Umer , M.K. Afzal , E.U. Munir , M. Alam , A dual ring connectivity model for [225] M. Raya , P. Papadimitratos , J. Hubaux , Securing vehicular communications, VANET under heterogeneous traffic flow, Wireless Person. Commun. 98 (1) IEEE Wireless Commun. (2006) 8–15 Special Issue on Inter-Vehicular Com- (2018) 105–118 . munications . [239] T. Sheltami , H.T. Mouftah , An efficient energy aware clusterhead formation in- [226] A. Stampoulis, Z. Chai, A survey of security in vehicular networks, 2007. http: frastructure protocol for MANETs, in: Proceedings of the 8th IEEE Symposium //zoo.cs.yale.edu/ ∼ams257/projects/wirelesssurvey.pdf . on Computers and Communications (ISCC), 2003, pp. 203–208 . [227] S. Zeadally , Y.R. Hunt , A. Irwin , A. Hassan , Vehicular Ad Hoc Networks [240] B.B. Dubey , N. Chauhan , S. Pant , Effect of position of fixed infrastructure on (VANETs): Status, Results, and Challenges, Vehicular Ad Hoc Networks data dissemination in VANETs, Int. J. Res. Rev. Comput. Sci. 2 (May(2)) (2011) . (VANETs): Status, Results, and Challenges, 2010 . A. Zekri, W. Jia / Ad Hoc Networks 75–76 (2018) 52–79 79

Abdennour Zekri received the B.Sc. and the M.Sc. degrees in telecommunications from Abou Bekr Belkaid University, Tlemcen, Algeria, in 2009 and 2011, respectively. He is currently a Ph.D. candidate at the Cyber Space Intelligence Computing Laboratory, Department of Computer Science & Technology, School of Electronic Information and Electrical Engineering (SEIEE), Shanghai Jiao Tong University (SJTU), China. His research interests include Vehicular Ad-hoc Networks (VANETs), smart cities, heterogeneous wireless networks, big data management, cyberspace sensing, and next- generation mobile networks.

Weijia Jia is currently a Chair Professor at University of Macau while he is taking no-pay leave from the Department of Computer Science & Technology, Shanghai Jiao Tong University, China. He received the B.Sc. and the M.Sc. degrees in computer science from Central South University, Changsha, China, in 1982 and 1984, and Master of Applied Sci. and Ph.D. degrees from Polytechnic Faculty of Mons, Belgium, in 1992 and 1993, respectively. He joined the German National Research Center for Information Science (GMD) in Bonn (St. Augustine) from 1993 to 1995 as a research fellow. From 1995 to 2013, he has worked in Department of Computer Science, City University of Hong Kong as a full professor. His research interests include smart cities, wireless communication and networks, next-generation Internet of Things, cyberspace sensing, distributed systems, QoS and routing protocols for the Internet. He has published more than 400 papers in various Transactions and prestige international conference proceedings. He has served as an editor and guest editor for international journals and as a PC chair and PC member/keynote speaker for several international conferences.