A survey on data dissemination in vehicular ad hoc networks

JID:VEHCOM AID:19 /REV [m5G; v 1.137; Prn:24/09/2014; 11:59] P.1 (1-12)

Vehicular Communications ••• (••••) •••–•••

Contents lists available at ScienceDirect

Vehicular Communications

www.elsevier.com/locate/vehcom

A survey on data dissemination in vehicular ad hoc networks

Moumena Chaqfeh ∗, Abderrahmane Lakas, Imad Jawhar

United Arab Emirates University, Al-Ain 15551, United Arab Emirates

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

Article history:Received 21 May 2014Received in revised form 17 August 2014Accepted 4 September 2014Available online xxxx

Keywords:VANETsData disseminationPerformance modelingBroadcast optimizationData cachingChallenges

The rapid evolution of wireless communication capabilities and vehicular technology would allow traffic data to be disseminated by traveling vehicles in the near future. Vehicular Ad hoc Networks (VANETs) are self-organizing networks that can significantly improve traffic safety and travel comfort, without requiring fixed infrastructure or centralized administration. However, data dissemination in VANET environment is a challenging task, mainly due to rapid changes in network topology and frequent fragmentation. In this paper, we survey existing data dissemination techniques and their performance modeling approaches in VANETs, along with optimization strategies under two basic models: the push model, and the pull model. In addition, we present major research challenges.

© 2014 Elsevier Inc. All rights reserved.

1. Introduction

Intelligent transportation systems (ITS) [1] can provide effi-cient solutions for road traffic problems by combining technology and improvements in communication, information systems and advanced mathematical models with the existing world of sur-face transportation infrastructure [2]. Vehicular Ad hoc Networks (VANETs) represent a major component of ITS, because of their po-tential to enable various applications to improve road safety and travel comfort.

VANET enables participating vehicles to communicate coopera-tively for exchanging information about road conditions and travel situations, by providing a self-organizing network environment, without requiring a fixed infrastructure or centralized administra-tion. With the increasing number of vehicles being equipped with communication capabilities, large scale VANETs are expected to be available in the near future [3]. The novelty of VANETs with re-spect to other ad hoc networks has been recently highlighted, and detailed VANET requirements are specified in [52] and [53]. In [4], it is shown that rapid changes in the network topology are dif-ficult to predict and manage. Moreover, the network is prone to frequent fragmentation leading to high variability of the connectiv-ity. In addition, redundancy should be limited. For these reasons,

* Corresponding author.E-mail addresses: [email protected] (M. Chaqfeh), [email protected]

(A. Lakas), [email protected] (I. Jawhar).

http://dx.doi.org/10.1016/j.vehcom.2014.09.0012214-2096/© 2014 Elsevier Inc. All rights reserved.

providing scalable data dissemination in VANETs is a challenging task [5].

Existing VANET dissemination techniques can be classified into three models: push, pull and hybrid. In the push model, data is dis-seminated proactively using periodic broadcast, while in the pull model, data is disseminated on-demand [6], where a routing pro-tocol carries data to relatively faraway distances. Such protocols also rely on broadcasting for data dissemination at each hop. There are few schemes that combine both dissemination models together so as to support different types of applications. The push-based model is often preferred for safety applications, when an imme-diate response is required, while the pull-based model is used for delay-tolerant applications, such as seeking a free parking slot or detecting congestion on far-away road. Compared to push-based model, pull model often requires less overhead, with less latency constraints. In delay-tolerant applications, the requester usually sends a query to the broadcast site, and gets a reply message from there. In such applications, users can tolerate more delays as long as a response eventually returns.

It is a major concern to optimize dissemination caused by blind data flooding in VANETs – which is the simplest dissemination style. Blind flooding suffers from a high percentage of redundant data that wastes the limited radio channel bandwidth. Moreover, packet collisions may lead to broadcast storm problem since a large number of vehicles in the same vicinity may rebroadcast the same data nearly simultaneously. In pull data dissemination, fur-ther optimization is required to cache data locally by participating vehicles, in order to increase dissemination efficiency by improv-ing the probability of finding a query answer with reduced delay.

http://dx.doi.org/10.1016/j.vehcom.2014.09.001

http://www.ScienceDirect.com/

http://www.elsevier.com/locate/vehcom

mailto:[email protected]



http://dx.doi.org/10.1016/j.vehcom.2014.09.001


2 M. Chaqfeh et al. / Vehicular Communications ••• (••••) •••–•••

Despite the proven performance of caching in VANETs, research ef-fort in this direction is still limited.

There is no standard approach for performance evaluation of data dissemination in VANETs. Most of the designed dissemination techniques are verified via simulation results, and few are analyzed using a mathematical model. To model dissemination in a VANET for performance analysis, the major challenge is to provide suffi-cient level of details to ensure realistic traffic scenarios and driving behavior. Three different models are to be considered: the road layout model, the mobility model, and data dissemination model.

The focus of this survey is three fold: First, existing data dis-semination techniques proposed for VANETs are presented under the three dissemination models. For each model, a set of represen-tative examples are provided. Second, existing performance mod-eling approaches of data dissemination in VANETs are reviewed. Third, dissemination optimization strategies are highlighted. Ac-cording to the author’s best knowledge, there is no other VANET study from these perspectives.

The remainder of this paper is organized as follows: in the sec-ond section, existing techniques for data dissemination in VANETs are presented. In the third section, existing performance modeling approaches are reviewed. The forth section suggests optimization strategies, while the fifth section highlights some issues and chal-lenges. The last section provides concluding comments and sugges-tions for future work.

2. Data dissemination techniques

Different data dissemination techniques for VANETs are pro-posed to fit different applications. Basically, two major applica-tions are heavily researched in this area: traffic safety, and travel comfort. Traffic safety applications are low data-rate, confined to limited number of neighborhood with strict latency constraints. While travel comfort applications are known as delay-tolerant ap-plications with more relaxed time constraints, but are expected to require data transmission spanning relatively faraway distances.

With the assistance of vehicular communication systems, traf-fic safety applications can be employed to considerably lower the accident rates [6] and decrease the loss of life on the road, mainly by providing warning systems such as intersection collision, lane change assistance, overtaking vehicle, emergency vehicle, wrong way driver, collision risk, hazardous location, and signal violation [53]. On the other hand, driver comfort applications can signifi-cantly improve our everyday lives, by enabling the delivery of traf-fic data in addition to announcements and advertisements, such as: the congestion status of a faraway road, the available parking slots at a parking place, the estimated bus arrival time at a bus stop, sale information at a department store, and the meeting schedule at a conference room. Only clients around the access point can directly receive the information, since the broadcast range is limited. How-ever, these data may be received by drivers and passengers who are far away. For example, a driver may want to query some de-partment stores to decide where to go. A passenger on a bus may query several bus stops to choose the best next stop for bus trans-fer. Such queries may be issued tens of kilometers away from the broadcast site. Within a VANET, the requester can send the query to the broadcast site and may tolerate more delay as long as the reply eventually returns [7].

There are three basic models for data dissemination in VANETs:Push, Pull and Hybrid. In the push model, data is disseminated proactively using periodic broadcast, while in the pull model data is disseminated on-demand [6]. There are few hybrid schemes that combine both models together to support different applications. The push-based model is often preferred for safety applications, when an immediate response is required, while the pull-based

Fig. 1. Dissemination models in traffic view. (A) The same direction. (B) The opposite direction.

model is used for delay-tolerant applications with the aim of im-proving traffic efficiency and travel comfort.

2.1. Push model

Push model is generally preferred for safety messaging sys-tems [6], such as collision warning systems, emergency message dissemination systems [8] and information systems specified for hazardous road conditions like ice, water or snow [9]. Nevertheless, other approaches also exist to support other types of applications such as arrival time estimation [10], speed expectation [11] and congestion detection [12]. In this section, a representative tech-nique is provided for each of those applications.

2.1.1. Safety messaging systemIn [6], the data push model is studied in the context of the

“Traffic View” vehicular information dissemination system. The study differentiates between the vehicle’s own data and the stored data about other vehicles. Three propagation models were com-pared: same-direction, opposite-direction and bi-direction. In the same-direction model (Fig. 1-A), a vehicle periodically broadcasts both its own data in addition to its stored data in a single packet, which is propagated “backward” by vehicles moving in the same direction. While in the opposite-direction model (Fig. 1-B), ve-hicles in the same direction only broadcast their own generated data, which are aggregated and propagated backwards by vehicles moving in the opposite direction. These two models are combined together in the bi-direction model, in which the vehicles generated and stored data are propagated backwards by vehicles moving in the same direction, and only stored data is propagated by vehicles moving in the opposite direction.


M. Chaqfeh et al. / Vehicular Communications ••• (••••) •••–••• 3

Fig. 2. (A) Zone-flooding protocol. (B) Zone diffusion protocol.

Simulation results indicate that traffic densities in both direc-tions of the road significantly affect the performance of data dis-semination model. Such a simple strategy poses a great overhead since all the vehicles in the desired direction participate in broad-casting, while it could be sufficient to broadcast by only a subset of vehicles.

2.1.2. Hazardous road condition warningIn [9], two dissemination protocols for VANETs are proposed

in the context of the Life WArning System (LIWAS) [13] research project, which is a traffic warning system that aims at provid-ing drivers with information about hazardous road conditions such as ice, water and snow. The first protocol is called the “Zone Flooding” protocol (Fig. 2-A). This protocol proposes three differ-ent optimization techniques to limit the forwarding of packets. The first technique is “hop-count”, which aims to ensure the maxi-mum number of hops before discarding a packet. The second tech-nique is “sequence-list”, which ensures that a certain packet can be forwarded only once. The “zone flooding” concept is also in-troduced to further limit the dissemination of packets. The second protocol is called the “Zone Diffusion” protocol (Fig. 2-B), which is a data-centric protocol that is based on data aggregation. This protocol assumes that every node maintains an environment rep-resentation (ER) for the surrounding environment. ER is updated whenever data is received from sensors, and data are periodically broadcasted to neighbors. Simulation results proved that these two protocols are robust to changes in network mobility and density. However, the three simple techniques utilized for dissemination optimization are not sufficient to mitigate broadcasting overhead under high densities.

2.1.3. Arrival time estimationAn example of estimating arrival time to vehicles’ destinations

is proposed in [10], where the road map is divided into areas in which vehicles can measure the time required to pass through each area. A sufficient number of vehicles is required in each area in order to keep accurate traffic information statistics continuously.

Fig. 3. RMDP communication.

Each vehicle periodically broadcasts area passage time to share with neighboring vehicles. The time required to pass through a certain area can be estimated by each vehicle locally, by averag-ing the area passage times collected from neighbors. The proposed approach is evaluated with realistic traffic flows on realistic road system and proved to achieve the traffic information sharing at a practical level. This approach was proposed to be improved based on a message ferrying technique in [14]. Broadcasting is not opti-mized in this approach, and the network could be easily flooded with data under high densities.

2.1.4. Speed expectationsSpeed expectations example approach is proposed in [11]

where vehicles are enabled to build their own local traffic maps of speeds experienced on visited roads, and share them with other vehicles. This allows a vehicle to build a map of expected speeds even on non-visited roads, which indicates traffic conges-tion through the network. This approach was applied on a simple Manhattan grid network map, and data is exchanged only on areas of unexpected traffic, by using a distributed clustering algorithm that does not require constant network connectivity. This approach performs well in sparsely connected dynamic network, but it is not evaluated in large-scale scenarios.

2.1.5. Congestion detectionIn [12], an example of congestion detection is presented based

on disseminating and propagating traffic data using Received Mes-sage Dependent Protocol (RMDP). Most of the vehicles can acquire the head of traffic jam in a short time using RMDP. A simple com-munication strategy is considered, in which a vehicle broadcasts its own information to its surrounding vehicles that are traveling on the opposite lane. As shown in Fig. 3, the proposed approach can be presented as follows: Assume that a moving vehicle A dis-seminates its locally stored information to vehicle B moving on the opposite direction lane. B moves forward and re-disseminates A’s information to a vehicle C moving in the same direction as A. In this case, C can know the head of traffic jam and may decide to change its route. From this simple illustration, it can be concluded that RMDP has a limited scope since it focuses on the congestion of the road directly ahead.

2.2. Pull model

The pull model techniques often follow the request-response paradigm for data dissemination. Compared to the push-based model, pull model often requires less overhead, with less latency constraints. In pull-based approach, the requester usually sends a query to the broadcast site, and gets a reply message from there. In such applications, users can tolerate more delays as long as a re-sponse eventually returns. Pull-based techniques often target travel comfort applications such as service discovery and delay-tolerant systems.



Fig. 4. ABSRP information dissemination model.

2.2.1. Service discoveryAddress Based Service Resolution Protocol (ABSRP) [15] in-

tegrates a pull-based technique to discover services in VANETs. When a vehicle needs a service, it creates a service request with the specification of the type of service and the desired service area, and then transmits it to the nearest roadside unit, as shown in Fig. 4. The receiving roadside unit checks if it has proactively learned about the service provider. If it is aware of the service provider’s IP address, it forwards the service request to the tar-get service provider. Otherwise, it broadcasts the service request destined to the target service provider over the backbone network.

In the case of Fig. 4, the request received by roadside unit 1 is broadcasted to the two nearby units. After receiving the request, the target service provider creates a service response and transmits it to the originating vehicle. A roadside unit can transmit a service request to the target service provider over the vehicular network or backbone network. In the former case, broadcasting can rapidly flood a congested vehicular network with data packets, since no optimization is proposed in ABSRP.

2.2.2. Delay-tolerant systemsVehicle-Assisted Data Delivery (VADD) is another pull-based ap-

proach for data dissemination in VANETs [16]. When a vehicle issues a request to a certain fixed site, VADD proposes techniques to efficiently route the packet to that site and receive the reply within a reasonable delay. Involved nodes carry the packet when routes do not exist and forward it to the new receiver that moves into its vicinity.

As shown in Fig. 5, vehicle A has a packet to forward to a cer-tain destination. Optimal direction for this packet is assumed to be north. Two contacts are available for the packet carrier: B moving south and C moving north. Thus, A has two choices for selecting the next hop for the packet. Both choices aim at forwarding the packet north. B may be selected because it is geographically closer towards the north and provides better possibility to exploit the wireless communication (e.g. B can immediately pass the packet to D, but C cannot). C may also be selected because it is moving in the packet forwarding direction. These two choices lead to two different forwarding protocols: Location First Probe (L-VADD) and Direction First Probe (D-VADD).

VADD makes use of the predictable mobility in a VANET, which is limited by traffic pattern and road layout. Extensive experi-ments were designed for performance evaluation. Results show that the VADD outperforms existing solutions in terms of data packet delay, packet-delivery ratio, and protocol overhead. Never-theless, VADD is designed specifically for applications in sparsely connected networks, and did not resolve communication issues un-der high-densities. Dense networks may require the selection of the next forwarding vehicle from sets of vehicles based on some criteria.

Fig. 5. VADD protocols used in the intersection mode: select the next vehicle to forward the packet.

2.3. Hybrid model

Along with the push and pull models that were presented, there are few schemes that combine both models in order to support different types of applications within a VANET environ-ment. Information transfer protocol for vehicular computing “VITP” [17] supports the establishment of distributed service infrastruc-ture over VANETs, by specifying the syntax and the semantic of messages between vehicles. VITP uses both of the information dis-semination models. For safety messages such as alerts about emer-gencies or hazardous traffic conditions, a push-based technique is used, while a pull-based technique is proposed to retrieve informa-tion by location-sensitive queries issued by vehicles on demand.

The push-based technique proposed by VITP disseminates alert messages among vehicles moving into the affected area. Whenever a vehicle detects such a condition, it generates an alert message and transmits it via the underlying VANET. The generated push message is transported to its target location area using geographic routing. Once arrived, the push message is broadcasted to all vehi-cles within the target location area. On the other hand, the usage of pull-based technique to disseminate messages is issued on de-mand in a context of service provision scenario, such as estimating the traffic-flow condition in a target location. When a vehicle ini-tiates such a request, it submits that request to the target area, assuming that there is a connection from the requesting vehicle to that area through the VANET (as shown in Fig. 6). The propaga-tion of such a request is done through intermediate nodes using geographic routing, in a way that is similar to transporting a push message. The semantic of the query determines the way to treat it once it arrives to the target location area, where vehicles construct a virtual ad hoc server (VAHS) to provide a reply message. As Fig. 6shows, the request message propagates through the virtual server until a certain return condition is satisfied. The vehicle that detects such a return condition immediately creates a reply message, and posts it towards the source area, where the requesting vehicle is located.

Simulation results have proven the feasibility of VITP in the VAENT environment. However, there is a high drop rate for queries, which grows substantially with increasing query distances, and with decreasing vehicle densities. That’s why optimization tech-niques are required to enhance the performance of VITP.

3. Performance modeling

There is no standard methodology for performance evaluation of VANET information dissemination algorithms. Such techniques are commonly verified via simulations, and are rarely analyzed



Fig. 6. VITP communication.

Fig. 7. Model assumptions of time probabilistic analysis.

using a mathematical model. This is mainly due to the major chal-lenge of providing sufficient level of details to ensure realistic traf-fic scenarios and driving behavior. In fact, three different models are to be considered: the road layout model, the mobility of vehi-cles [11], and message dissemination model. In this section, recent studies on performance modeling of data dissemination in VANETs are reviewed.

3.1. Modeling push-based data dissemination

3.1.1. Time-probabilistic analysisIn [18], the authors consider two algorithms to transfer warn-

ing messages in VANET. They have developed analytical models to obtain time-probabilistic characteristics of these algorithms. A lin-ear network topology is assumed as shown in Fig. 7. At a broadcast transmission from a node i of the network, its message is received by all the vehicles within its transmission range r with probabil-ity p. If two nodes transmit simultaneously in the vicinity of a recipient node j, the transmissions interfere at node j.

For each algorithm A, G A(t, d) is the probability of the event in which a node located at distance d from the message origina-tor receives the warning message at the t-th step of the system operation. The primary performance metric of A is the mean dis-semination delay, which is given by

D A(d) =∞∑

t=0

tG A(t,d) (1)

Considering such a simple model significantly simplifies the task of comparing different algorithms, which is an advantage. However, only limited set of algorithms can be modeled with such simplicity.

Fig. 8. Transmission interference in modeling with priority.

3.1.2. Modeling with priorityIn [19], the analysis uses two priority classes of traffic, assum-

ing that safety messages have higher priority compared to the other network traffic. A highway of length R meters is assumed, and the transmission range is set to cover a constant distance dfor each node. Nodes are dispersed in the highway according to a Poisson process with rate ϕ . Low-priority messages are assumed to arrive according to a Poisson process with rate λ0. Message transmission time is exponentially distributed with rate μ. Two concurrent transmissions interfere with each other whenever the distance between the transmitting nodes is less than 2d. If the distance is less than d, the interference is referred to as internal interference. Otherwise, it is referred to as external interference. Fig. 8 illustrates the two types of interference.

First, the probability of interference between two nodes is de-rived. Then, a birth–death process analysis is used to derive the probability distribution of lower priority messages, which are con-currently transmitted at the steady-state, and also to derive the percentage of destination node population which is affected by the interference, and thus cannot receive the message correctly. Finally, the performance of high-priority traffic is studied in the presence of low-priority traffic. Three performance metrics were considered. Firstly, the average message forwarding distance in a hop d is de-rived, and expressed as:

d = ω

1 − P s(0)

∞∑k=1

kP s(k)

1 + k(2)

where ω is the distance of the border point between the non-interference and interference regions from the sending node. P s(k)

is the probability that k nodes receive the high-priority message successfully. Secondly, the average number of nodes N that would receive the safety message successfully is expressed as:

N = (1 − Pe)

Pe(3)

where Pe is the probability that the forwarding of a high-priority message stops. Thirdly, the average number of communication hops that a message travels in the network nn is given by:

nh = Z

d(4)

where Z denotes the average distance that a safety message covers until its propagation terminates.

Numerical and simulation results show that the probability of a receiving node being exposed to interference increases as a func-tion of the transmission range, and thus, increasing the transmis-sion range does not necessarily improve the forwarding distance of safety messages, since more nodes may be exposed to inter-ference, especially under high density scenarios. The importance of the result provided in [19] is that it can be used to study the performance of different message dissemination algorithms and to determine the optimum range assignment in VANETs.

3.1.3. Warning delivery modelingThe work in [20] also analyzes the problem of dissemination

of safety messages. A safety area around the point where a haz-



Fig. 9. Model assumptions.

ard happens is introduced and the goal is to optimize the message dissemination strategy such that all vehicles within this area can receive the message. Multiple broadcast cycles are assumed, so that within a certain time all the designated vehicles are guar-anteed to be informed. Three performance measures are derived. The average delay, the probability that a vehicle is informed, and the average number of duplicate messages received by a vehicle within the defined safety area.

A linear network topology of a highway with at least two lanes is assumed. Every vehicle has a transmission range that is equal to R . Wireless channel is error-free which means that all vehicles within the transmission range of a source node can forward a re-ceived message correctly. In addition, no initial contention phase is considered. Whenever a vehicle traveling on the highway has detected some safety condition at any point of the highway, it trig-gers the dissemination of a warning message by exploiting multi-hop ad hoc communications (Fig. 9). The objective is to advertise all traveling vehicles within a certain dissemination area of exten-sion d. For simplicity, we only consider the analysis of a single broadcast cycle. The number of vehicles within the dissemination area n∗ is assumed to be constant. Every time a vehicle receives a new warning message, it decides, with probability α to act as re-lay to forward the message further. Three performance measures are derived: The average number of informed nodes (vehicles), the average delay, and the average number of duplicate messages re-ceived by a vehicle.

Let’s firstly show how the average number of informed nodes is computed. Since the probability that a node forwards the warn-ing message is Bernoulli with parameter α < 1, and the distance between nodes is assumed to be exponentially distributed with parameter γ , the distance between two consecutive relay nodes is exponentially distributed with αγ . Let Pr(n) indicate the proba-bility that the number of connected relay nodes is equal to n:

Pr(n) ={

(1 − e−αγ R)ne−αγ R n < n∗(1 − e−αγ R)n∗

n = n∗0 n > n∗

(5)

where n∗ = �αγ D�, and the average number of relay nodes:

r = (1 − e−αγ R)n∗(

1 − eαγ R) + e−αγ R − 1 (6)

Given Pr(n), the distribution of the number of informed nodes, S(n) can be estimated by considering the average number of nodes covered by message propagation. The average distance covered by the warning message when there are n connected relay nodes is:

d(n) = Pr(n)

α(7)

where 1/αγ is the average distance between two consecutive relay nodes. The number of informed nodes can be obtained by:

S(n) = d(n)γ = Pr(n)

α(8)

and the average is:

S = (1 − eαγ R)n∗(1 − eαγ R)

(9)
α
Fig. 10. Upper bound and lower bound of delay-tolerant message propagation.

Now we show how the delay D is computed. Assume single broadcast cycle, D can be obtained from:

D = Ttx + d

c(10)

where Ttx is transmission time of the warning message, d is the covered distance, and c is propagation speed. Lastly here, the av-erage number of messages M received by each informed node is computed as:

M = [2αgR − (αgR − 1)(1 − P s)

]P I (11)

where P I is the probability to inform a node. Despite its simplified assumptions, this model can accurately and effectively compute the three derived performance metrics [20].

One drawback of the proposed analysis is the error-free wire-less channel assumed, which means that all vehicles within the transmission range of a source node can forward the safety mes-sage correctly and collisions are not taken in consideration. It is also assumed that topology does not change during each broadcast cycle, and topology modifications are considered only at the be-ginning of new cycles. However, the impact of this assumption is limited, since transmission, propagation, and back-off time scales are much smaller than those of vehicles’ movements.

3.2. Modeling pull-based data dissemination

3.2.1. Delay-tolerant message propagationIn [21], an analytical model is presented for delay tolerant mes-

sage dissemination in VANETs. A bidirectional highway is assumed, in which vehicles and messages travel either upstream or down-stream as shown in Fig. 10. Nodes traveling in one direction are separated by distances X that are exponentially distributed. For transmission range R , two vehicles are connected if Xi ≤ R . Con-nectivity is modeled as the probability P (Xi ≤ R). The roadway is divided into cells of size l as shown in Fig. 10, and two bounds are defined for the cell size: an upper bound of size R , and a lower bound of size R/2. For the lower bound, vehicles located in adjacent cells are surely connected. While for the upper bound, ve-hicles traveling in two adjacent cells are not necessarily connected. Distances remain unchanged since vehicles are assumed to travel at a fixed speed.

Upper and lower bounds are derived for message propagation as a function of traffic density, vehicle speed and transmission range. Message propagation rate is the performance metric used for the evaluation of the analyzed model. Simulation results imply that the message propagation rate experiences a phase transition behavior as a function of traffic density. Extended analysis is pro-vided in [22] for characterizing such a behavior.

Another lower bound is also presented in [23], for the proba-bility that a vehicle receives a safety message through multi-hop



Fig. 11. VADD model example.

communication from a source at a distance d away and within tseconds. This probability p is derived as a function of single-hop communication reliability. The analysis studies the tradeoff be-tween the parameters t , p and the inter-vehicle distance, d. Again, it is assumed that inter-vehicles distances are fixed, which is unre-alistic.

3.2.2. Vehicle-Assisted Data Delivery (VADD)Another approach for delay-tolerant VANETs is called Vehicle-

Assisted Data Delivery (VADD) [16]. When a vehicle issues a delay tolerant data query to a certain fixed site, techniques are proposed to efficiently route the packet to that site, and receive a reply within reasonable delay, by using the predicable vehicle mobil-ity, that is limited by traffic pattern and road layout. Based on the existing traffic pattern, a vehicle can find the next road to forward the packet with the aim of reducing the delay. Selecting the next hop that is closer to the destination is usually efficient in geographic routing, but VADD assumes sparsely connected net-work, where such a selection is not always possible. VADD always tries to transmit through wireless channels as much as possible. If the packet has to be carried through certain roads, the road with higher speed should be chosen. Dynamic path selection is contin-uously executed throughout the packet forwarding process.

VADD is analyzed in three packet modes: Intersection, Straight-way and Destination based on the location of packet carrier. A stochastic model is used to estimate the data delivery delay, which is used to select the next road (intersection). Fig. 11 shows an example of VADD delay model. For a packet at Im , the expected delay of delivering the packet through road rmn is given by:

Dmn = dmn +∑

j∈N(n)

(Pnj × Dnj) (12)

where Dij is the expected packet delivery delay from intersection Ii to the destination if the packet carrier at Ii chooses to deliver the packet following road ri j . Pij is the probability that the packet is forwarded through road ri j at Ii . Finally, N( j) is the set of neigh-boring intersections of I j . The latter equation can be applied on a bounded area that includes the source and the destination of a certain vehicle in a connected graph, in order to find and the in-tersection with minimum expected delay, and select it for packet delivery. In addition to data delivery delay, two other performance metrics are considered: data delivery ratio and data traffic over-head.

3.3. Discussion

This section provides a comparison among the reviewed analyt-ical models from different perspectives. In the following, two main comparison criteria are discussed. Comparison is summarized in Table 1. Most of the reviewed modeling techniques rely on IEEE 802.11 communication standard, which is a set of standards for implementing wireless local area network (WLAN) computer com-munication in certain frequency bands. IEEE 802.11p [24] is an

approved amendment to the IEEE 802.11 standard that aims to add wireless access in vehicular environments. It defines enhancements to the original 802.11 to support Intelligent Transportation Systems (ITS) applications.

3.3.1. Performance metricsPure network performance metrics can be used for the evalua-

tion of VANET data dissemination algorithms, such as packet loss, packet error, packet delivery ratios, end-to-end delay, normalized network load, and packet duplication. However, some techniques propose other corresponding metrics that can better evaluate spe-cific application scenarios, such as in [25].

Models for traffic safety and delay-tolerant systems rely on two basic performance metrics: Mean dissemination delay and prob-ability of successful message reception. Delivery ratio is an alter-native measure for the latter that can be gathered via simulation results. In safety applications, it is required to minimize the de-lay and maximize data delivery. If an accident occurs on a road for example, it is required to disseminate a warning message in or-der to inform vehicles that are planning to visit the same road, so that they may search for an alternative route. Under high density circumstances, the arrival of more uniformed vehicles can shortly block the road. In this case, the data traffic overhead becomes sig-nificant metric, which can be measured by the average number of duplicate messages received per vehicle.

Despite that delay-tolerant techniques accept relatively longer delay compared to traffic safety models, it is still required to keep the delay in a reasonable range. For example, if a vehicle sends a delay-tolerant query to indicate congestion status in a certain far-away road, it can wait for some time before receiving a reply. Since data is expected to travel far away from the requesting vehicle, message propagation rate measures the speed of message delivery process in delay-tolerant models.

3.3.2. AssumptionsAnalytical models should address sufficient level of complexity

to simulate realistic traffic scenarios and realistic driving behav-ior, which seems a real challenge in VANETs. Since there is no standard modeling approach, each of the reviewed models uses a different way for analysis. Specifically, to model a VANET, it is re-quired to provide assumptions about three basic components: the road layout, vehicular mobility, and networking (including density criteria and communication techniques). In the following, assump-tions made by the reviewed models on each of those components are discussed.

• Road layoutTo model the road layout, analysis approaches usually rely on linear network topology [18–21,25]. However, linear layout cannot be generalized, since it represents only one possible road representation, which is the simplest. For safety appli-cations, it could be sufficient to analyze dissemination in a simple linear layout, which is not the case in delay-tolerant models. In [16], a graph is considered to represent three dif-ferent road layouts (intersection, linear road and destination) which make the model more realistic.

• MobilityFor mobility modeling, [18], [19] and [21] assume a fixed ve-hicular speed, since distances between vehicles were set fixed for simplicity. In [19] and [20], vehicle overtaking is assumed to provide more realistic mobility. In contrast, no constrains were assumed on the mobility of vehicles in [16].

• NetworkingTo model the networking characteristics, [18] and [25] assume that all network nodes are restricted to start their message transmissions synchronously at the time moments. At the zero



Table 1Performance modeling approaches for VANETs (comparison summary).

Modeling techniques Standard Metrics Road layout assumptions

Mobility assumptions

Network assumptions

Addressed challenges

Traffi

csa

fety

Time-probabilistic analysis [18]

Not mentioned

(1) Mean dissemination delay

– Linear network topology

– Fixed speed – Synchronous transmission

– Interference– Fixed distances

between vehicles

Modeling with priority [19]

IEEE 802.11

(1) Average number of nodes that receive the high-priority message

– One-direction highway (linear topology)

– Vehicles may take over each other

– Two priority classes for messages (low, high)

– External and internal interference

(2) The per-hop message forwarding distance.

– Fixed-speed time intervals chosen from Gaussian distribution

Warning delivery service [20]

Not mentioned

(1) Average delay – One-direction multi-lane highway (linear topology)

– Lane change (overtaking)

– Error-free wireless channel

–(2) Probability of

successful reception(3) Average number of

duplicate messages– Vehicles do not

move during broadcast

– Worst-case (high density)

Del

ay-t

oler

ant

Delay-tolerant message propagation [21]

IEEE 802.11p

(1) Message propagation rate

– Bidirectional highway

– Fixed speed – Delay tolerance network

–– Fixed distances

VADD [16] IEEE 802.11

(1) Data delivery delay – Graph with different traffic modes: intersection, straightway and destination

– – Sparse network

– Dynamic path finding(2) Data delivery ratio

(3) Data traffic overhead.

time moment of [18], a message is always transmitted by its originator. In the realistic vehicular network, it could be always assumed that each vehicle has access to the global positioning system and clocks, which ensure the practical implementation for such synchronization. Consideration of asynchronous trans-missions under the deterministic packet arrival process would increase the complexity of the model. Existing approaches that are dealing with asynchronous transmissions assume Poisson packet arrival process and take a benefit of memory-less na-ture of exponential distribution. In [18] and [20], an error-free wireless channel is assumed, and interference phenomenon is neglected.

In traffic safety models, it is required to analyze the network in the worst case, by assuming a high-density network, such as in [9]. On the other hand, delay-tolerant models may assume a sparsely connected network [21,16]. Nevertheless, it is sometimes required to analyze delay-tolerant models under high-density scenarios so as to fit different applications. For example, congestion detection is considered as delay-tolerant application, but is required to be analyzed in dense networks.

To conclude the discussion, it can be noted that modeling is generally not sufficient for evaluating dissemination in VANET, and it should be verified using simulation results, in which more so-phisticated details can be considered. It is required to provide suf-ficient level of details for the simulation, and then evaluate the impact of assumptions simplified in model analysis, such as ne-glected collisions, neglected transmission delays, the constant ve-hicles speeds assumed while the message is propagating. In many cases, despite the simplified assumptions, the model estimations are closed to simulation results [20]. This is due to the fact that connectivity, or node density, is the factor that mainly determines the network performance in VANETs. In [26], statistical properties of the connectivity with user mobility at the steady state are stud-ied.

4. Data dissemination optimization

Despite that pull-based data dissemination in VANETs can use similar optimization strategies like push-based methods, further optimization can be achieved via data caching. In this section, we firstly focus on broadcasting schemes as optimization strategies for push model, and secondly on data caching for further optimization in pull model.

4.1. Optimization in push-based dissemination

Data dissemination in VANET systems often requires the notion of broadcast, which forms the basis of all types of communica-tion in ad hoc networks [27]. Plain flooding is the simplest style of broadcasting, where the originating vehicle broadcasts a data packet to all its one-hop neighbors. In multi-hop dissemination, all receiving neighbors would rebroadcast the packet to their one-hop neighbors, and so on. In dense networks, simple broadcast may easily lead to broadcast storm problem, when many vehicles in the same vicinity broadcast simultaneously and too many packets col-lide. Moreover, plain flooding is not scalable, since the same data packet may be excessively disseminated, and the limited available bandwidth of the radio channel is wasted.

More specifically, plain flooding is extremely costly because it may result in the following [27]:

– Redundant rebroadcasts, that occur when a node decides to rebroadcast data to its neighbors; however, all neighbors have already received the same data before.

– Medium contention, that occurs when neighboring nodes re-ceive a broadcast data and decide to rebroadcast the message. These nodes must contend with each other for the broadcast medium.

– Packet collisions, which result in packet loss or corrupted mes-sages.

Since VANET is fully connected under high density, a data packet that is disseminated by plain flooding would be received by



all the nodes, and every node will rebroadcast a copy of the same data. For a number of packets P , assuming N connected nodes, the total number of messages M sent through the network grows ex-ponentially with the network density:

M ≈ P × (N − 1) × N = O(N2) (13)

The basic approach that is commonly used for data dissemina-tion optimization is to decrease the percentage of data redundancy so that the delivery overhead is reduced, by selecting only a subset of vehicles to rebroadcast [28], basically by exploiting position-ing information. Several protocols are specifically designed for data dissemination in VANETs, and they present lightweight solutions in terms of data redundancy overhead. These protocols can be cate-gorized into two basic schemes: The probabilistic scheme and the delay-based scheme.

4.1.1. Probabilistic broadcastIn the probabilistic scheme, a different rebroadcast probability

is assigned to each vehicle. Since only some vehicles will partic-ipate in rebroadcasting the packet data redundancy overhead as well as the number of collisions is reduced. The main challenge in the probabilistic broadcasting scheme is determining an opti-mal probability assignment function that decreases data redun-dancy and keeps a high delivery ratio. Simple probabilistic-based broadcasting protocols assign constant probability to participating vehicles, while more sophisticated protocols let each vehicle dy-namically determine its forwarding probability, basically based on its location or the network density that is locally detected. Like plain flooding, in probabilistic broadcast, the total number of mes-sages M sent through a connected network of size N (when a packet is to be disseminated) is still exponential, but grows more slowly according to probability Φ:

M ≈ ΦN(N − 1) (14)

Weighted p-persistence [29] is a well-known probabilistic broadcasting method that uses the distance as a parameter to determine the forwarding probability, where the farthest vehi-cles always rebroadcast with the highest probability. Whenever a packet is received from vehicle i, the receiving vehicle j checks the packet ID and rebroadcasts with a certain probability if it receives the packet for the first time. Otherwise, it discards the packet. The forwarding probability Pij is calculated by:

Pij = dij

R(15)

where dij is the distance between i and j, and R is the transmis-sion radio range. The major drawback of weighted p-persistence is that it does not consider the traffic density and therefore it is not efficient under high-density scenarios.

4.1.2. Delay-based broadcastIn delay-based broadcast scheme, different waiting delays are

assigned to receiving vehicles. Vehicles with shorter delays would rebroadcast first, and vehicles assigned to later times would cancel their transmissions upon the receipt of data duplication, since this indicates that the data has already been disseminated, and there-fore redundant rebroadcasts can be avoided. This is what is known as “broadcast suppression”.

Slotted 1-persistence [29] is a delay-based broadcast suppres-sion mechanism where vehicles are assigned to different time slots depending on their distance to the sender, such that vehicles with highest priority are given the shortest delay before rebroadcast-ing. When a packet is received, a node checks the packet ID and rebroadcasts with probability 1 at the assigned time slot T Sij if it receives the packet for the first time and has not received any

duplicates before its assigned time slot. Otherwise, it discards the packet. Given the relative distance dij between nodes i and j, the average transmission range R , and the predetermined number of slots Ns , T Sij can be calculated as:

T Sij = Sij × τ (16)

where τ is the estimated one-hop delay, which includes the medium access delay and propagation delay, and Sij is the as-signed slot number, which can be expressed as:

Sij = Ns

(1 −

⌈min(dij,R)

R

⌉)(17)

The problem with this approach is that Ns is a design pa-rameter that should theoretically be a function of the traffic den-sity, however, vehicles cannot predict traffic density using this approach, and it requires network designers to fix this value or adaptively change it according to the road condition.

Slotted p-persistence is another method, which mixes both the probability and the delay-based schemes by giving the highest pri-ority vehicles the shortest delay and the highest probability to re-broadcast. Whenever a packet is received, a node checks the packet ID and rebroadcasts with the pre-determined probability p at the assigned time slot, if it receives the packet for the first time and has not received any duplicates before its assigned time slot. Oth-erwise, it discards the packet. Similar to slotted 1-persistence, this approach relies on a good choice of the forwarding probability p, and it doesn’t consider traffic density to adapt the probability value according to the road traffic condition.

The major drawback of delay-based approaches is the uneven distribution of vehicles among timeslots, since these timeslots are determined according to location information, without consider-ing traffic density. Distributed Optimized Time (DOT) [30] is a re-cent delay-based approach that provides time slot density control, where the number of vehicles assigned to each slot can be deter-mined. However, it doesn’t dynamically determine this number; in other words, DOT does not automatically detect traffic density.

In addition to the distance as a parameter to determine the re-broadcasting probability and/or delay, recent studies consider traf-fic direction to support variety of traffic scenarios [54–56].

4.2. Optimization in pull-based dissemination

There is limited coverage of data access issues in VANETs [31]. Caching is a commonly used technique for improving data access, in which the network performance is significantly increased, since the overhead caused by global network flooding is reduced. Gen-erally, VANET caching schemes rely on the cooperative approach, which allows for sharing of cached data among multiple vehicles, where the potential of the caching can be further explored. How-ever, there exist some techniques which are non-cooperative.

Caching schemes for ad hoc networks are proposed in [32,33]. Mobile ad hoc networks (MANETs) caching schemes also ex-ist [34–36]. Caching for Internet-based VANETs are also proposed in [37,38]. A large positive impact in utilizing caching techniques for vehicular networks has been proven in [39–43]. In this section, we suggest cache formalization approach for VANETs [44] and then we provide a list of existing cache invalidation strategies.

4.2.1. Caching formalizationAssume a bidirectional graph G of N nodes and d links. Nodes

are vehicles and the links are connections between them. Each ve-hicle is connected to a set of other vehicles d called neighbors. Each vehicle stores data in a local cache of size k. Stored data are either locally generated or gathered from other vehicles. Caches are assumed to be in the steady state (each node stores k data items).



The content of each cache is assumed to be completely random (resources are a uniformly random subset of the R available re-sources).

When an inquiring node searches for some data x, where x is available in nx number of nodes, x is not known to that node if it cannot be generated locally and is not already stored in the cache. The probability F1 that x is not known to a particular node is:

F1 = 1 − (P [x is not locally known]

· P [x is not already stored in the cache]) (18)

Since x is offered by nx nodes in total, we have:

P [x is locally known] = nx/N (19)

The number of ways to choose k elements out of a set of R ele-ments such that a particular element is not chosen is

(R−1k

). Since

k elements of the node’s cache are assumed completely random, we obtain:

P [x is not already stored in the cache] =(

R − 1

K

)/(R

K

)= R − k/R (20)

Thus:

F1 = 1 − (1 − nx/N)(R − k/R) = (N − nx)(R − k)/N R (21)

To compute the number of messages in plain flooding, it is assumed that nodes reply directly to the inquiring node. If the desired data is not found at the inquiring node, there will be dtransmissions to the d neighbors, in addition to the internal trans-missions by those neighbors:

M = (1 − F1)(d + dm(t − 1)

)(22)

where m(t − 1) are the transmissions generated at a particular neighbor that spread to t − 1 hops, which is shown to have an exponential behavior [44]. Probabilistic flooding also has an expo-nential number of messages, but this number grows more slowly according to Φ:

M = (1 − F1)(dΦ + dΦm(t − 1)

)(23)

4.2.2. Caching invalidation strategiesWhen data caching is involved in communication, it is required

to state when a piece of data is invalid, which means that it is no longer beneficial and should be removed. In the following, cache invalidation strategies for VANETs are presented.

• Time-To-Live (TTL)In [17], the caching support in VITP is provided by using cache-control headers that can be included in messages to act as a caching decision directive. The cache replacement pol-icy used in VITP is TTL, which defines the maximum time for which cached information is considered valid. The eval-uation of caching support for VITP is provided in [39], with the main objective of investigating if caching extension for a proactive, location-aware communication protocol can main-tain acceptable levels of information quality, while sustaining the performance of the vehicular network. Simulation results show an improvement of information accuracy of more than 65%, and a reduction in network overhead of only 12% under both normal traffic conditions and unscheduled traffic events.

• Location-based invalidationCache-based routing strategy in VANET is proposed in [40], by utilizing the locality of vehicles’ traces, without requiring global network flooding or location servers. Two basic schemes

are proposed: the update scheme and the query scheme. In the update scheme, each vehicle sends update messages at intersections to disseminate location information in the vehic-ular network, so that the location information is stored in the caches of its neighboring vehicles. While in the query scheme, when a vehicle v1 needs to get a route to another vehicle v2, it initiates a local flooding to search for a vehicle v3 that has the location information of v2, without considering how old the information is. When the message is received by v3, it re-sends the query message to the location stored in its cache that v2 had ever located in. If another vehicle that receives the query message has more recent location information of v2, it redirects the query message to the newer location stored in its cache. Then, a limited flooding is used again to find v1 so as to send a reply message with the newest location informa-tion found. Simulation results show that the proposed protocol works effectively in city environments. However, it considers updating location information only at intersections, which may not be effective in other traffic scenarios.

• Randomized invalidationInfoshare [41] is a pull-based information sharing application for vehicular networks that aims to achieve the maximum spreading of information messages, where the broadcast over-head is limited using a smart caching technique, so as to re-duce useless queries and duplicated replies. Vehicles form an ad hoc network cooperate in disseminating information mes-sages that are pulled from fixed gateways connected to the Internet and are broadcasted along the road. This cooperation is aided by on-board caches in which the information shared by nearby vehicles is preserved. Each vehicle originates an in-formation request at a random time. The request message is broadcasted in a multi-hop fashion until a vehicle carrying the desired information is found, then a reply is carried back to the originator following the same path in reverse order. When the originator vehicle receives the reply, it caches the required information, which is discarded after a random time, and can be requested again later.

• Probabilistic estimationHamlet [42] is a fully distributed scheme that aims to pro-vide effective information caching without swamping the stor-age capacity with needless information. What is important in Hamlet is that it does not consider a fixed scheme for cache replacement and thus it helps users to decide on the infor-mation to keep and for how long, based on a probabilistic estimate of what other neighbors are caching. The objective of such an approach is to avoid network flooding with query messages whenever possible, by creating a content diversity within the node neighborhood, so that a requesting vehicle likely finds the required information nearby. When consistency becomes an issue, Hamlet allows for a quick replacement of the outdated information with the most recent version. Ham-let assumes that a node can “overhear” query/response mes-sages, which may raise a security issue when non-cooperative driving is assumed. Hamlet considers gateway nodes as the original and permanent source of information such that each participating vehicle can locally have a temporal part of this information.

5. Issues and challenges

VANET applications impose diverse requirements on the sup-porting technologies. This diversity leads to a number of challenges [47]. This section addresses the main research challenges to be considered for data dissemination in VANETs. These include scala-bility, security and trust, quality of service, node cooperation and simulation issues.



5.1. Scalability

As we have previously mentioned, high VANET density scenar-ios can lead to broadcast storm problem. The impact of this prob-lem is quantified in [44]. VANET researchers continue to focus on scalable data dissemination, by trying to reduce data redundancy. Novel solutions should be developed to detect traffic density such that scalability issues can be addressed.

5.2. Security and trust

In travel comfort applications, security issues are not often con-sidered, because cooperative driving is typically assumed. While in safety data dissemination, integrating security mechanisms is highly necessary within VANETs [45]. Warning systems will not be accepted by customers if trust, security and reliability are not provided. The introduction of trust by providing trustworthy ap-plications is considered as the most crucial security issue within a VANET [46]. However, integrating security schemes into VANET applications will increase the delay of message arrival.

5.3. Quality of service and traffic characterization

Quality of service (QoS) and traffic characterization are addi-tional important issues to be addressed by dissemination methods in VANETs, since different types of applications are expected to have different QoS requirements. QoS metrics for VANET are not well defined yet [47].

5.4. Node cooperation

The majority of VANET dissemination protocols assume that nodes are willing to cooperate and allow the use of their resources to provide connectivity for other nodes. However, this assumption might not always be valid. In addition, some nodes might exhibit selfish behavior and make user of services available by other nodes while not allowing similar use of their resources by the same nodes. Efficient reward, punishment, debit and credit mechanisms would have to be enforced by the corresponding protocols and sys-tems.

5.5. Simulation

The cost and complexity of implementing VANET data dissem-ination schemes and applications in large test-bed systems forces such an implementation to be within a simulation environment [46]. Three major challenges can be addressed in the context of VANET simulation. First, the credibility and feasibility of simulation systems require reliable and standardized simulation parameters so that verification techniques can be applied. Second, mobility models should address sufficient levels of complexity to simulate realistic traffic scenarios and realistic driving behavior [48]. Third, the scalability of simulation represents a huge challenge in this context. Specifically, it is currently impossible to simulate the full-stack of very large networks [49].

6. Conclusion

This paper firstly presents existing data dissemination tech-niques for VANET applications, which are classified under three ba-sic models: push-based, pull-based and hybrid. Secondly, it reviews existing performance modeling approaches for analyzing dissemi-nation in VANETs. Thirdly, it surveys dissemination optimization strategies. Finally, main technical challenges are listed.

In a previous work [50], a pull-based technique is evaluated in the context of congestion detection application. This technique

is enhanced in [51] by providing a caching strategy that has sig-nificantly reduced the network overhead. One possible avenue for future work, which we are currently investigating, is to design a reliable scalable data dissemination protocol that follows the hy-brid model, in order to fit the two basic styles of dissemination in VANETs. To fit safety applications, a method for broadcast suppres-sion is to be developed such that network overhead is decreased to the minimum. The broadcast scheme will be incorporated in a request-reply data dissemination protocol that fits travel comfort applications such as congestion detection.

References

[1] http://sites.ieee.org/itss/ (accessed on January 20th, 2014).[2] Joseph M. Sussman, Perspectives on Intelligent Transportation Systems (ITS),

Springer, New York, 2005.[3] J. Zhao, G. Cao, VADD: vehicle-assisted data delivery in vehicular ad hoc net-

works, in: IEEE INFOCOM’06, 2006.[4] R. Fracchia, M. Meo, Analysis and design of warning delivery service in inter-

vehicular networks, IEEE Trans. Mob. Comput. 7 (2008) 832–845.[5] S. Eichler, C. Schroth, T. Kosch, M. Strassberger, Strategies for context-adaptive

message dissemination in vehicular ad hoc networks, in: Proceedings of the 2nd International Workshop on Vehicle-to-Vehicle Communications, V2VCOM, 2006.

[6] T. Nadeem, P. Shankar, L. Iftode, A comparative study of data dissemination models for VANETs, in: The 3rd Annual International Conference on Mobile and Ubiquitous Systems – Workshops, 2006, pp. 1–10.

[7] Bejan, R. Lawrence, Peer-to-Peer cooperative driving, in: International Sympo-sium on Computer and Information Sciences, 2002.

[8] Y. Bi, H. Zhao, X. Shen, A directional broadcast protocol for emergency message exchange in inter-vehicle communications, in: Proceedings of the 2009 IEEE International Conference on Communications, IEEE Press, Piscataway, NJ, USA, 2009, pp. 4805–4809.

[9] J. Bronsted, L.M. Kristensen, Specification and performance evaluation of two zone dissemination protocols for vehicular ad-hoc networks, in: Proceedings of the 39th Annual Symposium on Simulation, 2006, pp. 68–79.

[10] N. Shibata, T. Terauchi, T. Kitani, K. Yasumoto, M. Ito, T. Higashino, A method for sharing traffic jam information using inter-vehicle communication, in: The 3rd Annual International Conference on Mobile and Ubiquitous Systems – Work-shops, 2006, pp. 1–7.

[11] S. Dornbush, A. Joshi, StreetSmart traffic: discovering and disseminating auto-mobile congestion using VANET’s, in: IEEE Vehicular Technology Conference, Ireland, 2007.

[12] M. Saito, J. Tsukamoto, T. Umedu, T. Higashino, Evaluation of inter-vehicle ad hoc communication protocol, in: Proceedings of IEEE 19th International Conference on Advanced Information Networking and Applications, 2005, pp. 78–83.

[13] LIWAS: Life Warning Systems, http://www.liwas.dk (accessed on April 24, 2011).

[14] T. Shinkawa, A technique for information sharing using inter-vehicle com-munication with message ferrying, in: Proceedings of the 2006 International Workshop on Future Mobile and Ubiquitous Information Technologies, 2006, pp. 221–225.

[15] B.K. Mohandas, A. Nayak, K. Naik, N. Goel, ABSRP – a service discovery ap-proach for vehicular ad hoc networks, in: Asia-Pacific Services Computing Con-ference, 2008, pp. 1590–1594.

[16] J. Zhao, G. Cao, VADD: vehicle-assisted data delivery in vehicular ad hoc net-works, IEEE Trans. Veh. Technol. 57 (3) (2008) 1910–1922.

[17] M.D. Dikaiakos, S. Iqbal, T. Nadeem, L. Iftode, VITP: an information transfer protocol for vehicular computing, in: Proceedings of the 2nd ACM International Workshop on Vehicular Ad Hoc Networks, 2005, pp. 30–39.

[18] A. Vinel, A. Dudin, S. Andreev, F. Xia, Performance modeling methodology of emergency dissemination algorithms for vehicular ad-hoc networks, in: The 7th International Symposium on Communication Systems Networks and Dig-ital Signal Processing, CSNDSP, 2010, pp. 397–400.

[19] M. Khabazian, S. Aissa, M. Mehmet-Ali, Performance modeling of message dis-semination in vehicular ad hoc networks with priority, IEEE J. Sel. Areas Com-mun. 29 (1) (2011) 61–71.

[20] R. Fracchia, M. Meo, Analysis and design of warning delivery service in inter-vehicular networks, IEEE Trans. Mob. Comput. 7 (7) (2008) 832–845.

[21] A. Agarwal, D. Starobinski, T.D.C. Little, Analytical model for message propa-gation in delay tolerant vehicular ad hoc networks, in: Vehicular Technology Conference, VTC, 2008, pp. 3067–3071.

[22] A. Agarwal, D. Starobinski, T. Little, Phase transition behavior of message prop-agation in delay tolerant vehicular ad hoc networks, MCL Technical Report No. 12-12-2008, 2008.

[23] G. Resta, P. Santi, J. Simon, Analysis of multi-hop emergency message prop-agation in vehicular ad hoc networks, in: Proceedings of ACM International

http://sites.ieee.org/itss/

http://refhub.elsevier.com/S2214-2096(14)00044-8/bib32s1

































http://www.liwas.dk
































Symposium on Mobile Ad Hoc Networking and Computing, MobiHoc, Montreal, 2007, pp. 140–149.

[24] http://standards.ieee.org/findstds/standard/802.11p-2010.html (accessed on May 3, 2014).

[25] A. Vinel, V. Vishnevsky, Y. Koucheryavy, A simple analytical model for the periodic broadcasting in vehicular ad-hoc networks, in: IEEE GLOBECOM Work-shops, 2008, pp. 1–5.

[26] M. Khabazian, M. Mehmet Ali, Generalized performance modeling of vehicular Ad Hoc networks (VANETs), in: The 12th IEEE Symposium on Computers and Communications, ISCC, 2007, pp. 51–56.

[27] J. Lipman, H. Liu, I. Stojmenovic, Broadcasting in ad hoc networks, in: S. Misra, I. Woungang (Eds.), Guide to Wireless Ad Hoc Networks, Springer-Verlag, Lon-don, UK, 2009, pp. 121–150 (Chapter 6).

[28] S. Panichpapiboon, W. Pattara-atikom, A review of information dissemination protocols for vehicular ad hoc networks, IEEE Commun. Surv. Tutor. (2011) 1–15.

[29] N. Wisitpongphan, O.K. Tonguz, J.S. Parikh, P. Mudalige, F. Bai, V. Sadekar, Broadcast storm mitigation techniques in vehicular ad hoc networks, IEEE Wirel. Commun. 14 (6) (2007) 84–94.

[30] R.S. Schwartz, K. Das, H. Scholten, P. Havinga, Exploiting beacons for scalable broadcast data dissemination in VANETs, in: The 9th ACM International Work-shop on Vehicular Inter-Networking Systems and Applications, UK, 2012.

[31] M.L. Sichitiu, M. Kihl, Inter-vehicle communication systems: a survey, IEEE Commun. Surv. Tutor. 10 (2) (2008) 88–105.

[32] L. Yin, G. Cao, Supporting cooperative caching in ad hoc networks, IEEE Trans. Mob. Comput. (2006) 77–89.

[33] F.J. González-Cañete, E. Casilari, A. Triviño-Cabrera, Proposal and evaluation of an application level caching scheme for ad hoc networks, in: Proceedings of the International Conference on Wireless Communications and Mobile Computing: Connecting the World Wirelessly, 2009, pp. 952–957.

[34] C. Casetti, C.F. Chiasserini, M. Fiore, C.A. La, P. Michiardi, P2P cache-and-forward mechanisms for mobile ad hoc networks, in: IEEE Symposium on Computers and Communications, ISCC, 2009, pp. 386–392.

[35] S. Gupta, A. Joshi, J. Santiago, A. Patwardhan, Query distribution estimation and predictive caching in mobile ad hoc networks, in: Proceedings of the Seventh International ACM Workshop on Data Engineering for Wireless and Mobile Ac-cess, 2008.

[36] Y. Wang, M. Martonosi, L.S. Peh, Situation-aware caching strategies in highly varying mobile networks, in: The 14th IEEE International Symposium on Mod-eling, Analysis, and Simulation of Computer and Telecommunication Systems, 2006.

[37] S. Lim, C. Yu, C.R. Das, Cooperative cache invalidation strategies for internet-based vehicular ad hoc networks, in: Proceedings of 18th International Confer-ence on Computer Communications and Networks, ICCCN, 2009, pp. 1–6.

[38] S. Lim, S.H. Chae, C. Yu, C.R. Das, On cache invalidation for internet-based ve-hicular ad hoc networks, in: The 5th IEEE International Conference on Mobile Ad Hoc and Sensor Systems, MASS, 2008, pp. 712–717.

[39] N. Loulloudes, G. Pallis, M.D. Dikaiakos, On the evaluation of caching in vehicu-lar information systems, in: Eleventh International Conference on Mobile Data Management, 2010, pp. 218–220.

[40] G.Y. Chang, J.P. Sheu, T.Y. Lin, K.Y. Hsieh, Cache-based routing for vehicular ad hoc networks in city environments, in: IEEE Wireless Communications and Networking Conference, WCNC, 2010, pp. 1–6.

[41] M. Fiore, C. Casetti, C.F. Chiasserini, On-demand content delivery in vehicular wireless networks, in: Proceedings of the 8th ACM International Symposium on Modeling, Analysis and Simulation of Wireless and Mobile Systems, 2005, pp. 87–94.

[42] M. Fiore, F. Mininni, C. Casetti, C.F. Chiasserini, To cache or not to cache? in: IEEE INFOCOM, 2009, pp. 235–243.

[43] C. Maihofer, R. Eberhardt, Geocast in vehicular environments: caching and transmission range control for improved efficiency, in: IEEE Intelligent Vehi-cles Symposium, 2004, pp. 951–956.

[44] V.V. Dimakopoulos, E. Pitoura, On the performance of flooding-based resource discovery, IEEE Trans. Parallel Distrib. Syst. 17 (11) (2006) 1242–1252.

[45] J.P. Hubaux, S. Capkun, J. Luo, The security and privacy of smart vehicles, IEEE Secur. Priv. 2 (3) (2004) 49–55.

[46] S. Eichler, C. Schroth, J. Eberspächer, Car-to-car communication, in: Proceedings of the VDE-Kongress – Innovations for Europe, 2006.

[47] S. Zeadally, R. Hunt, Y. Chen, A. Irwin, A. Hassan, Vehicular ad hoc networks (VANETS): status, results, and challenges, Telecommun. Syst. 50 (4) (2012).

[48] T. Camp, J. Boleng, V. Davies, A survey of mobility models for ad hoc network research, Wirel. Commun. Mob. Comput. 2 (5) (2002) 483–502.

[49] R.M. Fujimoto, et al., Large-scale network simulation: how big? how fast? in: Proceedings of the 11th Annual Meeting of the IEEE International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems, MASCOTS, 2003.

[50] A. Lakas, M. Chaqfeh, A novel method for reducing road traffic congestion using vehicular communication, in: Proceedings of the 6th International Conference on Wireless Communications and Mobile Computing, 2010, pp. 16–20.

[51] A. Lakas, M. Chaqfeh, Geocache: sharing and exchanging road traffic informa-tion using peer-to-peer vehicular communication, in: Proceedings of IEEE 73rd Vehicular Technology Conference, VTC2011-Spring, Hungary, 2011.

[52] S. Al-Sultan, M.M. Al-Doori, A.H. Al-Bayatti, H. Zedan, A comprehensive survey on vehicular ad hoc network, J. Netw. Comput. Appl. (2013) 1–13.

[53] G. Karagiannis, O. Altintas, E. Ekici, G. Heijenk, B. Jarupan, K. Lin, T. Weil, Vehicular networking: a survey and tutorial on requirements, architectures, challenges, standards and solutions, IEEE Commun. Surv. Tutor. 13 (4) (2011) 584–616.

[54] R.S. Schwartz, H. Scholten, P. Havinga, A scalable data dissemination protocol for both highway and urban vehicular environments, EURASIP J. Wirel. Com-mun. Netw. 2013 (1) (2013), article 257.

[55] R.S. Schwartz, R.R.R. Barbosa, N. Meratnia, G. Heijenk, H. Scholten, A directional data dissemination protocol for vehicular environments, Comput. Commun. 34 (2011) 2057–2071.

[56] C. Liu, C. Chigan, RPB-MD: providing robust message dissemination for vehicu-lar ad hoc networks, Ad Hoc Netw. 10 (3) (2012) 497–511.



http://standards.ieee.org/findstds/standard/802.11p-2010.html





























































































Documents

A survey on data dissemination in vehicular ad hoc networks