Vehicular Ad Hoc Networks: Architectures, Research Issues, Methodologies, Challenges, and Trends

Vehicular ad hoc networks (VANETs) have been quite a hot research area in the last few years. Due to their unique characteristics such as high dynamic topology and predictable mobility, VANETs attract so much attention of both academia and industry. In this paper, we provide an overview of the main aspects of VANETs from a research perspective. This paper starts with the basic architecture of networks, then discusses three popular research issues and general research methods, and ends up with the analysis on challenges and future trends of VANETs.


Introduction
Recently, with the development of vehicle industry and wireless communication technology, vehicular ad hoc networks are becoming one of the most promising research fields.
VANETs which use vehicles as mobile nodes are a subclass of mobile ad hoc networks (MANETs) to provide communications among nearby vehicles and between vehicles and nearby roadside equipment [1] but apparently differ from other networks by their own characteristics. Specifically, the nodes (vehicles) in VANETs are limited to road topology while moving, so if the road information is available, we are able to predict the future position of a vehicle; what is more, vehicles can afford significant computing, communication, and sensing capabilities as well as providing continuous transmission power themselves to support these functions [2].
However, VANETs also come with several challenging characteristics, such as potentially large scale and high mobility. Nodes in the vehicular environment are much more dynamic because most cars usually are at a very high speed and change their position constantly. The high mobility also leads to a dynamic network topology, while the links between nodes connect and disconnect very often. Besides, VANETs have a potentially large scale which can include many participants and extend over the entire road network [2].
It is precisely because of both of these unique attractive features and challenging characteristics that VANETs could draw the attention from both industry and academia.
Therefore, several articles have tried to summarize the issues about vehicular networks. For example, in [3,4], the authors discuss the research challenges of routing in VANETs and then summarize and compare the performance of routing protocols; Hartenstein and Laberteaux present an overview on the communication and networking aspects of VANETs and summarize the current state of the art at that time [5]; Raya and Hubaux address the security of VANETs comprehensively and provide a set of security protocols as well [6]; in [7], the authors propose a taxonomy of a large range of mobility models available for vehicular ad hoc networks. These articles all reviewed specific research areas in VANETs. In addition, others papers like [8] provide comprehensive overview of applications, architectures, protocols, and challenges in VANETs and especially introduce VANETs projects and standardization efforts in different regions (i.e., USA, Japan, and Europe); Al-Sultan et al. provide detailed information for readers to understand the main aspects and challenges related to VANETs, including network 2 International Journal of Distributed Sensor Networks architecture, wireless access technologies, characteristics, applications, and simulation tools [9].
Compared with these current articles, this paper adds the introduction of layered architecture for VANETs so that the summary of network architecture is more complete. Also, we organize the overview of the vehicular ad hoc networks in a novel way. That is, we introduce the VANETs from the research perspective in the paper, including some current hot research issues and general methods, which do good to the progress of VANETs. Moreover, we provide a more comprehensive analysis on VANETs research challenges and future trends, beneficial for further systematic research on VANETs. In summary, this paper covers basic architecture, some research issues, general research methods of VANETs, and some key challenges and trends as well as providing an overall reference on VANETs.
The rest of this paper is organized as follows. Section 2 first introduces the vehicular ad hoc networks architecture, including network components, communication types, and layered network architecture. Then in Section 3, we discuss three aspects of VANETs research issues: routing, security and privacy, and applications. Section 4 focuses on VANETs research methodologies and some VANETs models and simulator tools are also given. Section 5 provides an analysis on VANETs research challenges and future directions. Finally, the paper is concluded in Section 6.

Architecture
This part describes the system architecture of vehicular ad hoc networks. We first introduce the main components of VANETs architecture from a domain view. Then, we explain their interaction and introduce the communication architecture. Besides, we provide a presentation of the layered architecture for VANETs.

Main Components.
According to the IEEE 1471-2000 [10,11] and ISO/IEC 42010 [12] architecture standard guidelines, we are able to achieve the VANETs system by entities which can be divided into three domains: the mobile domain, the infrastructure domain, and the generic domain [13].
As is shown in Figure 1, the mobile domain consists of two parts: the vehicle domain and the mobile device domain. The vehicle domain comprises all kinds of vehicles such as cars and buses. The mobile device domain comprises all kinds of portable devices like personal navigation devices and smartphones.
Within the infrastructure domain, there are two domains: the roadside infrastructure domain and the central infrastructure domain. The roadside infrastructure domain contains roadside unit entities like traffic lights. The central infrastructure domain contains infrastructure management centers such as traffic management centers (TMCs) and vehicle management centers [13].
However, the development of VANETs architecture varies from region to region. In the CAR-2-X communication system which is pursued by the CAR-2-CAR communication consortium, the reference architecture is a little different.
CAR-2-CAR communication consortium (C2C-CC) is the major driving force for vehicular communication in Europe and published its "manifesto" in 2007. This system architecture comprises three domains: in-vehicle, ad hoc, and infrastructure domain.
As shown in Figure 2, the in-vehicle domain is composed of an on-board unit (OBU) and one or multiple application units (AUs). The connections between them are usually wired and sometimes wireless. However, the ad hoc domain is composed of vehicles equipped with OBUs and roadside units (RSUs). An OBU can be seen as a mobile node of an ad hoc network and RSU is a static node likewise. An RSU can be connected to the Internet via the gateway; RSUs can communicate with each other directly or via multihop as well. There are two types of infrastructure domain access, RSUs and hot spots (HSs). OBUs may communicate with Internet via RSUs or HSs. In the absence of RSUs and HSs, OBUs can also communicate with each other by using cellular radio networks (GSM, GPRS, UMTS, WiMAX, and 4G) [2].

Communication Architecture. Communication types in
VANETs can be categorized into four types. The category is closely related to VANETs components as described above. Figure 3 describes the key functions of each communication type [15].
In-vehicle communication, which is more and more necessary and important in VANETs research, refers to the invehicle domain. In-vehicle communication system can detect a vehicle's performance and especially driver's fatigue and drowsiness, which is critical for driver and public safety.
Vehicle-to-vehicle (V2V) communication can provide a data exchange platform for the drivers to share information and warning messages, so as to expand driver assistance.
Vehicle-to-road infrastructure (V2I) communication is another useful research field in VANETs. V2I communication enables real-time traffic/weather updates for drivers and provides environmental sensing and monitoring.
Vehicle-to-broadband cloud (V2B) communication means that vehicles may communicate via wireless broadband mechanisms such as 3G/4G. As the broadband cloud may include more traffic information and monitoring data as well as infotainment, this type of communication will be useful for active driver assistance and vehicle tracking.

Layered Architecture for
VANETs. The open systems interconnection (OSI) model is well known to most readers, which groups similar communication functions into one of seven logical layers [16]. The session layer and presentation layer are omitted here, and a given layer can be further partitioned into sublayers in this architecture, as illustrated in Table 1 [17].
Generally, the architecture of VANETs may differ from region to region, and thus the protocols and interfaces are also different among them. For instance, Table 2 illustrates the protocol stack for dedicated short-range communication (DSRC) in the US. DSRC is specifically designed for automotive use and a corresponding set of protocols and standards [17]

Research Issues
This part is a brief introduction to three aspects of VANETs research issues: routing, security, and privacy, as well as applications. Firstly, we discuss the classification of routing protocols and some algorithms. Then state-of-the-art security and privacy researches are discussed. Finally, we introduce two types of applications, namely, safety applications and nonsafety applications.     and conducts the geographic forwarding), SADV (which finds the best path to forward the packet), an interference aware routing scheme (which equips the node with a multichannel radio interface and switches the channels based on the SIR evaluation), FROV (which selects the retransmission spans further node to rebroadcast a message), and a multihop broadcast protocol (which divides the road into segments and chooses the vehicle in the farthest nonempty segment). Other researchers propose some algorithms such as V-TRADE, UMB, AMB, MHVB, and MDDV [20].
(ii) Multicast. Multicast is necessary to communications among a group of vehicles in some vehicular situations, such as intersections, roadblocks, high traffic density, accidents, and dangerous road surface conditions. In [19], the authors categorize the multicast protocols into two main types. One is topologybased approaches, such as ODMRP (which generates a source-based multicast mesh and forwards based on the group address), MAODV (which generates a group-based multicast tree), and GHM (which generates group-based multicast meshes). The other one is location-based approaches, such as PBM (which is based on positions of all one-hop neighbors and positions of all individual destinations), SPBM (which introduces hierarchical group membership management), LBM (which uses a multicast region as destination information for multicast packets), and RBM and IVG (which define a multicast scope for safety warning messages).
(iii) Unicast. Researchers investigate the unicast communication protocols for VANETs in three ways: (1) greedy: nodes forward the packets to their farthest neighbors towards the destination, like improved greedy traffic-aware routing (GyTAR); (2) opportunistic: nodes employ the carry-toward technique in order to opportunistically deliver the data to the destination, like topology-assist geo-opportunistic routing; and (3) trajectory based: nodes calculate possible paths to the destination and deliver the data through nodes along one or more of those paths, like trajectory-based data forwarding (TBD) [21]. Some of the routing protocols and algorithms are summarized in Table 3.
The above-mentioned protocols are all used to overcome the existing problems of routing in VANETs. Nevertheless, some researchers focus on presenting some novel protocols for V2V and V2I communications of data dissemination. The main contributions are various communication protocols from physical to application layers, even cross-layers. Kim et al. survey ten papers about data dissemination including clusters and channel schemes, mechanisms, protocols, and security in vehicular environments [22]. The results are depicted in Table 4.

Security and Privacy.
Nowadays more and more intelligent on-board applications may store lots of personal information and vehicular trajectory data, which can disclose individuals' activities, habits, and traces. These threats have to be overcome before communication architecture in VANETs is deployed. Otherwise, the reliability, dependability, and individuals' acceptance of the VANETs system are likely to be low, because attackers may manipulate messages or track the trajectory of vehicles [23,24].
To address the security and privacy issues, many approaches have been proposed in the literatures over the past few years. Most of them pay more attention to two main aspects: communication and architecture of VANETs [25]. Security architecture is an important part of VANETs to satisfy the requirements of users' security and privacy. In [26], the authors describe the security architecture from several different viewpoints, such as the functional layer view, the organizational/component view, the reference model view, and the information centric view. In [27], the authors present a novel security architecture focusing primarily on securing the operation of the wireless part of the vehicular communication system and on enhancing the privacy of its users. Different from architecture, the vehicular communication system focuses primarily on secure communication schemes and algorithms [24,28]. Raya and Hubaux present a communication scheme, in which entities would like to establish a share session key if they need to securely communicate for a long time. This scheme pays much attention to safety-related applications, while the nonsafety-related applications are neglected [6,29]. In [28], the authors present an advanced secure communication scheme based on Raya and Hubaux's scheme, which extends its session key to be used in nonsafety-related applications and considers two session keys: pairwise and group keys. In [30], the authors discuss many security solutions that have been proposed in detail, such as VPKI (vehicular public key infrastructure), CA (certificate authority), and the group signature.

Application.
Applications in vehicular environment usually can increase the road safety, improve traffic efficiency, and provide entertainment to passengers. In most cases, VANETs applications can be roughly organized into two major classes: safety applications and nonsafety applications.

Safety Applications.
Traditionally the intention of safety applications is accident prevention, and thus this kind of applications is also the main motivation for developing vehicular ad hoc networks. Such applications like crash avoidance have a great requirement for the communication between vehicles or between vehicles and infrastructure [9]. Vehicles equipped with various sensors collect traffic data and monitor the environment continuously, and then cooperative vehicular safety applications can change real-time traffic information and send/receive warning messages through V2I or V2V communication to improve road safety and avoid accidents.
Several transportation departments in the US have identified eight safety applications in 2006, which are considered to provide the greatest benefits, that is, traffic signal violation, curve speed warning, emergency brake lights, precrash sensing, collision warning, left turn assist, lane change warning, and stop sign assist [17].

Nonsafety Applications.
With respect to their specific intended purpose, nonsafety applications can be classified into several subclasses, such as traffic convenience and efficiency applications, infotainment applications, and comfort/entertainment applications.
Since convenience and efficiency applications can be offered on an individual basis, they do not require standardization and cooperation among vehicles. The growth of such applications and services in the market can be seen in the recent years, including some mobile service offerings on smartphones [13]. Convenience and efficiency applications usually provide drivers or passengers with some useful information, such as weather or traffic information and the location of restaurants or hotels nearby [9]. Entertainment applications may provide services like media downloading and online games.

Research Methodologies
In order to evaluate the performance of different architecture approaches, protocols, algorithms, and applications, an effective research methodology is required in VANETs. Such methods enable researchers and developers to check the drawbacks as well as ensure the availability of new proposed approaches to the above-mentioned aspects. Since VANETs have a potentially large scale, the introduction of a new technology into VANETs requires long development and the experimental implement is very expensive. In general, there are two important and necessary steps before the market introduction: (1) analysis and evaluation by simulations and (2) analysis and verification by field operational testing [13]. In this section, we first introduce the different models which are the essential basis for setting up respective methodologies, and then the simulations and field operational testing are discussed in the following contents.

VANETs Models.
VANETs are a large and complex overall system model, which consists of four submodels for the different aspects: driver and vehicle model, traffic flow model, communication model, and application model [13].

Simulation Methods.
Simulation is no doubt an essential step before the implement of new technologies in VANETs. The simulation of VANETs requires two different components: a traffic simulator and a network simulator.
(i) Traffic Simulators. In order to analyze vehicular ad hoc network characteristics and protocol performances, traffic simulators are needed to generate position and movement information of a single vehicle in VANETs environment. In [13], the authors list some existing traffic simulators in detail, like SUMO (simulation of urban mobility) and VISSIM (simulation of the position and movement for vehicles as well as city and highway traffic).

Field Operational Testing.
Although the simulation method makes great contributions to the investigation of the VANETs, it does not reflect the real vehicular world. In order to overcome these issues, field operational testing (FOT) has attracted the attention of researchers, which aims to test and evaluate these applications at scale and covers a much wider range of real-world scenarios. Such testing can make the VANETs system closer to the market and generate economic value. Due to the high financial costs and the number of partners, FOT still depends on the reliable results of simulations. On the contrary, the data from the FOT can make the network models more reasonable and improve the performance of protocols. Finally, FOT has four important characteristics: (1) real system components, (2) real vehicles and traffic, (3) including all stakeholders, and (4) large and heterogeneous fleet [13].

Challenges and Future Trends
Based on the previous discussion of VANETs, we can see that VANETs are a fantastic self-organizing network for the future intelligent transportation system (ITS). Although researchers have achieved much great progress on VANETs study, there are still some challenges that need to be overcome and some issues that need to be further investigated (e.g., communication, security, applications, stimulation, verification, services, etc.) [15,33].

Top Challenges.
Compared with MANETs, the specific features of VANETs require different communication paradigms, approaches to security and privacy, and wireless communication systems [34]. For example, network connections may not be stable for a long time period. In order to improve the performance of communication, researchers have investigated the efficient use of available infrastructure, such as roadside units and cellular networks. Although some specific challenges of VANETs have been overcome, many key research challenges have only partially been solved [34]. Thus, researchers need to do deeper work to solve these challenges. In the following discussion, we will summarize the key challenges.
(i) Fundament Limits and Opportunities. Surprisingly little is known about the fundamental limitations and opportunities of VANETs communication from a more theoretical perspective [35]. We believe that avoiding accidents and minimizing resource usage are both important theoretical research challenges.
(ii) Standards. The original IEEE 802.11 standard cannot well meet the requirement of robust network connectivity, and the current MAC parameters of the IEEE 802.11p protocol are not efficiently configured for a potential large number of vehicles [15]. Thus, researchers must do more work about standards.
(iii) Routing Protocols. Although researchers have been presenting many effective routing protocols and algorithms such as CMV (cognitive MAC for VANET) and GyTAR (greedy traffic-aware routing), the critical challenge is to design good routing protocols for VANETs communication with high mobility of vehicles and high dynamic topology [33].
(iv) Connectivity. The management and control of network connections among vehicles and between vehicles and network infrastructures is the most important issue of VANETs communication [36]. Primary challenge in designing vehicular communication is to provide good delay performance under the constraints of vehicular speeds, high dynamic topology, and channel bandwidths [37].
(v) Cross-Layer. In order to support real-time and multimedia applications, an available solution is to design cross-layer among original layers [37]. In general, cross-layer protocols that operate in multiple layers are used to provide priorities among different flows and applications. In [34,38], the authors address the importance of cross-layer design in VANETs after analyzing the performance metrics.
(vi) Cooperative Communication. In [36], the authors consider the VANETs as a type of cloud called mobile computing cloud (MCC), and in [15] the authors present a broadband cloud in vehicular communication. Thus, the cooperation between vehicular clouds and the Internet clouds in the context of vehicular management applications has become a critical challenge to researchers.
(vii) Mobility. Mobility that is the norm for vehicular networks makes the topology change quickly. Besides, the mobility patterns of vehicles on the same road will exhibit strong correlations [38]. In [29], the authors address the idea that mobility plays a key role in vehicular protocol design and modeling.
(viii) Security and Privacy. Reference [39] presents many solutions that come at significant drawbacks and the mainstream solution still relies on "key pair/certificate/signature. " For example, key distribution is a key solution for security protocols, but key distribution poses several challenges, such as different manufacturing companies and violating driver privacy [38]. Besides, tradeoff of the security and privacy is the biggest challenge under the requirement of efficiency.
(ix) Validation. It is necessary not only to assess the performance of VANETs in a real scenario but also to discover previously unknown and critical system properties. Besides, validation has become more and more difficult under the wider range of scenarios, and Altintas et al. present can use field operational tests (FOTs) to solve this problem, but conducting meaning FOTs is a challenge like a large and complex system with technology components [36].
Thus, considering the characteristics of high mobility and high dynamic topology, researchers still need to study further and find solutions to the challenges we discussed above.

Future Trends.
In the future intelligent society, the potential value of VANETs is unpredictable with safety and entertainment applications. New vehicle applications have recently emerged in several areas ranging from navigation safety to location aware content distribution, commerce, and games [36]. Thus, the VANETs are worthy of further exploration and research, and we believe that there must be more applications and research results in the future. For instance, DSRC technology is proposed to provide a communication link between vehicles and roadside beacons [40] so as to support many applications in vehicular environment. Wu et al. discuss the evolution paths of DSRC and propose several possible enhancements in future versions of DSRC: better channel interleaving and channel coding, better MAC congestion control protocols, more flexibility in channelization, and so forth, [41]. The advanced DSRC can offer the necessary guarantee for the deployment of DSRCbased communication in VANETs. We now discuss some possible future trends of VANETs in three aspects including architecture, algorithm, and application, which we called 3A.
Architecture. In the future, a main research issue of vehicular ad hoc networks focuses on designing an integrated system architecture that can make use of multiple different technologies (e.g., IEEE 802.11p DSRC, WAVE, ITS G5, Wi-Fi, or 3G/4G) and heterogeneous vehicular networks [36]. Besides, in order to deploy the FOT mentioned above, researchers need to design a large scale and complex system architecture which should cooperate with different partners and manufacturers [13]. Thus, developing a reliable and flexible system architecture is one of the main research trends.
Algorithm. Although the existing algorithms have provided some solutions to some data dissemination problems in VANETs, it is still difficult to examine their performance and security because of the unique features of VANETs. For example, due to the nonpersistent network connections, the end-to-end communication path may not exist. In [42], the authors present the idea that the opportunistic routing algorithm can solve this problem with the carry-forwarding pattern. So the advanced algorithms should be designed with the low communication delay, the low communication overhead, and the low time complexity.
Application. Due to the requirement of continuous awareness of the road ahead, safety applications are still the key research trend in the mobile vehicular environment. However, the authors find no applications following the VANETs application guidelines after studying the most popular vehicular applications in the Android marketplace [43]. So researchers should do more work on standards and security of VANETs applications and investigate the question of "how to use model checking to automatically explore whether these applications meet the standards. "

Conclusion
In this paper, we first introduce the VANETs architecture, including network components, communication types, and layered network architecture. Then we discuss three aspects of VANETs research issues: routing, security, and privacy, as well as applications. We also focus on VANETs research methodologies and some mobility models and simulator tools are also given. Finally, we provide an analysis on VANETs research challenges and future trends. This paper introduces the vehicular ad hoc networks from the research perspective, covers basic architecture, critical research issues, and general research methods of VANETs, and provides a comprehensive reference on vehicular ad hoc networks.