Interests-Sensitive Data Dissemination Protocol for
Vehicular Ad-hoc Networks
Amal El-Nahas
The German University in Cairo, New Cairo city, Egypt
Abstract. Vehicular ad-hoc network (VANET) is a potential player in an intel-
ligent transportation system that would increase road safety as well as road con-
fort. In VANETs, vehicles excahnge road-related information either through an
established infrastructure, which is costly, or through their collaboration when in
common transmission range which we adopt in this paper. Information dissem-
ination is realized through broadcasting, thus an intelligent selection technique
should be deployed to decrease the traffic load caused by unnecessary rebroad-
casting. In this paper, we propose an interest-aware data dissemination protocol
that periodically exploits the current neighbors ´ınterests to select the proper set
of data to be broadcasted. The proposed approach is structure-less and imposes
minimum overhead on the communication bandwidth. The protocol is evaluated
through simulation experiments and rResults obtained demonstrate that this ap-
proach maximizes the number of relevant data reports received by the vehicles,
especially if a certain data type is more popular than the others.
1 Introduction
Vehicular ad-hoc networks (VANETs) have emerged as a result of the increased number
of vehicles capable of wirelessly interconnecting through their onboard radio commu-
nication devices, thus forming an ad hoc network on the fly. Moreover, DSRC standard
(Dedicated Short Range Communication), developed by IEEE, provides vehicular ad
hoc networks with large bandwidth. Seven non-overlapping channels, each of 27 Mbps
bandwidth, can be used for data dissemination. Only one channel is dedicated for safety
messages while the rest can be used for other road-related services.
Data dissemination is performed either through vehicle to infrastructure communi-
cation or vehicle to vehicle communication. While the former requires the existing of
an infrastructure in form of road side units which imposes additional cost and delay,
the latter is purely based on the ability of vehicles within common transmission range
to communicate. Multi-hop transmission is needed in order for the data to reach farther
vehicles.
The easiness and self configuring nature of VANETs enabled a broad range of in-
formation applications ranging from road safety to journey comfort to appear. Vehicles
collect and exchange information, in form of data reports, for traffic intensity, services
along the road, weather conditions, free parking places and others. Thus, each vehicle
can be a report producer, a report receiver, or both at the same time. It has been shown
Elnahas A. (2009).
Interests-Sensitive Data Dissemination Protocol for Vehicular Ad-hoc Networks.
In Proceedings of the 3rd International Workshop on Intelligent Vehicle Controls & Intelligent Transportation Systems, pages 57-65
Copyright
c
 SciTePress

that VANETs have a highly dynamic topology due to its highly mobile nodes. Con-
sequently, vehicular ad hoc networks tend to be often sparsely connected. Thus, data
reports received by a vehicle are most likely to be stored for a while before being re-
transmitted when encountering a new neighbor. Limiting the number of stored reports
and selecting the most important ones for transmission is a challenge that attracted re-
search lately. Intelligent dissemination protocols should be adopted to decide which
reports to store and rebroadcast later so as to efficiently share the wireless bandwidth as
well as decrease the amount of unwanted messages received by the drivers.
Data dissemination has long been studied for mobile users with short range wire-
less communication forming a mobile ad-hoc network (MANET). Various protocols
have been suggested in the literature, the simplest of all is one that relies on flood-
ing. Each moving node broadcasts data to all its neighbors until either covering the
whole network or reaching the maximum number of hops. This uncontrolled simple
flooding approach leads to increasing the number of unnecessary data retransmission,
causing what is known by the broadcast storm that results in inefficient bandwidth uti-
lization and severe congestion, as observed in [4]. Consequently, a constrained flooding
approach should be implemented. Different improvements over the basic flooding ap-
proach have been proposed in the literature that either control the time when to rebroad-
cast, or apply rules to decide whether to rebroadcast or not. A comprehensive survey
can be found in [1,2,3]. Relying on the observation that travelers tend to have individ-
ual preferences in the type of content of data reports they would prefer to receive, we
propose in this paper incorporating drivers ’ interests in the selection of data reports to
be broadcasted. Reports are assumed to belong to one of predefined service categories.
Neighbors ´ınterest in each category is locally computed at each vehicle. Most certainly,
exploring the continuously changing neighbors ´ınterests without imposing extra over-
head is not trivial. However, our protocol uses the periodic transmission of the beacon
messages generated by the medium access control protocol for interests advertisement.
The proposed protocol is evaluated through simulation experiments and proved to max-
imize the number of relevant information received by the vehicles.
The rest of the paper is organized as follows. Section 2 presents the different dissem-
ination approaches suggested before. Section 3 presents an overview of the proposed
interests-sensitive dissemination algorithm. In section 4, the simulation experiments
and the analysis of the obtained results are discussed. Finally, section 5 concludes the
paper with proposals for future work.
2 Data Dissemination Approaches
In general, data dissemination approaches proposed for ad-hoc networks considered one
or more of the three main resource constraints in MANETs, namely communication
bandwidth, energy consumption and storage [5]. Controlled flooding were suggested to
decrease the number of repetitive retransmissions, thus efficiently use the bandwidth as
well as decrease the energy consumption on the mobile devices. Control is performed
either by the sender or the receiver. Receiver-based approaches proved to perform well
in MANETs. In [6], data is forwarded to all nodes within a specific area defined by
the sender. Each receiving node decides whether to rebroadcast the data or not based

on its physical location with respect to the specified area. An improvement over this
geographic-based approach were proposed in [7] where a content-based forwarding
approach (CBF) is defined. In this approach, the next node for forwarding the message
is selected by all neighboring nodes based on their actual position when receiving the
message using a contention process. The farthest node that most likely would reach the
destination is the best one considered for forwarding the data. Other variants exist in
the literature. A comprehensive survey can be found in [1, 3].
Although the above techniques succeeded in reducing traffic load, they did not take
into consideration the node ’ s content requirements. A more selective approach should
be used to intelligently select the set of data reports to be forwarded to prevent users
from receiving unwanted messages. In [8], Wolfson et al. proposed a spatio-temporal
selection approach that is based on the data novelty probability. The novelty probability
of a data report reflects how new, and hence useful, this report is for the recipients
based on its generation time and distance to generation location. As time or distance or
both increases, report ages and eventually disappears. Although this approach proved to
be efficient in terms of throughput and response time, it did not consider the individual
users interests in the novelty probability. An autonomous gossiping approach for ad hoc
networks were proposed in [9] where information is sent only to neighbors interested
in receiving it. Each node advertises its profile that defines its interest. In addition, each
data item maintains its own profile. Based on nodes and data item profiles, data items
decide whether to replicate to a better node, migrate or do nothing.
However, it is worthy to note that VANETs have unique and challenging features
that do not exist in MANETs. Examples of such features are the highly dynamic topol-
ogy, highly mobile nodes, time critical responses, and insensitivity to energy consump-
tion and computation power that are considered unlimited As a result, data dissemina-
tion protocols specifically designed for VANETs have been proposed. In [10], Tonguz et
al. propose using the traffic density as well as the covered distance to decide whether to
retransmit or not. In a dense area, only a subset of cars needs to rebroadcast. Moreover,
as distance increase between the source of information and the node, the frequency of
broadcasting is decreased. A similar approach, but taking time into consideration, was
proposed in [11] where nodes receive data and store it for later retransmission. Only
fresh data is rebroadcasted. Combining both distance and time in a relevance function,
which extend the idea of Wolfson previously proposed for MANETs to VANETs, is
introduced in [12, 13]. AutoCast in [14] uses a probabilistic flooding, that depends on
neighborhood size. Individual interests in data disseminated were taken into consid-
eration in some recent work. In [15], messages selected are based on their benefit to
expected recipients. The benefit depends on the message context, vehicle context and
information context. A different approach in [16] is based on a pull model where a node
uses an utility-based approach to determine which data to pull upon meeting another
node.
3 Interests-Sensitive Data Dissemination Algorithm
The interests-sensitive dissemination algorithm we propose is a structure-less algorithm
that does not rely on any existing infrastructure. It is based on the same model as in [12,

13] to send and receive data reports, but augmented with an interest level component
that collects neighbors ´ınterests using the MAC layer single hop beacon messages. We
assume that data reports belong to one of predefined service categories based on their
content. Categories may represent traffic data, parking service, weather information,
and many others. Each vehicle has its own interest in each of those categories based on
their content. Each vehicle has its own interest in each of those categories. The vehicle
model and protocol description are presented below.
3.1 Vehicle Model
Fig.1. Vehicle Model.
We consider a network formed by a number of moving vehicles modeled as in Fig-
ure 1. Each vehicle is equipped with a GPS receiver for localization, integrated sensors
to collect road-related data, such as road surface condition, speed, light intensity or
others, as well as a wireless radio communication interface to communicate with other
vehicles. Each moving vehicle is modeled to have unlimited processing power but lim-
ited storage capacities. A local database is maintained at each vehicle that is limited in
size to M reports, where M is a configurable parameter. The interest level component
calculates and stores the interest level of current neighbors for each category. In the ex-
ample shown in Table 1, 65 percent of the current neighbors are interested in receiving
traffic related reports, 25 percent are interested in weather conditions, while 10 percent
are interested in information about availability of parking places. This information is
collected from the beacon messages periodically generated by each vehicle and saved
in the node ´s neighbor table.
Table 1. Interests Levels for 3 categories.
Category Interest level
Traffic conditions 0.65
Weather conditions 0.25
Parking places 0.1

3.2 Interests-Sensitive Algorithm
As mentioned before, a vehicle can generate, send or receive data reports. A data report
is composed of the following tuples:
carid, report-id (ID), generation time (t), generation location in x (x), generation loca-
tion in y (y), category-id (CAT)
where category-id represents the service category the record belongs to. Report gen-
eration occurs upon detection of variations in the vehicle sensed data, like change in
speed, road surface, or others. In this case, a report is generated and added to the ve-
hicle local database either for immediate or later transmission. Sending data reports
occurs upon either detecting a new vehicle in the neighborhood, or having a new data
report generated that needs to be sent immediately. In this case, the vehicle checks its
local database and selects the most relevant reports to be broadcasted. The relevance
of the reports is based on the current neighbors ´s interest level that is calculated for
each category (i) by dividing the number of nodes interested in i, (nodes
i
), by the total
number of neighboring nodes, as follows.
intLevel(i) =
P
nodes
i
P
neighbouringN odes
(1)
The number of nodes interested in each category is calculated from the information
saved in the neighbors table updated with the recipient of each beacon message, while
the interest level is calculated upon transmitting a data report.
Lastly receiving data reports occurs when in range with any of the neighboring
vehicles transmitting. The received set of reports is checked against the stored one and
new reports are then added to the database. Figure 2 illustrates a simplified pseudo code
for the proposed algorithm.
4 Simulation Methodology and Experimental Results
To further prove our concept, we simulated the behavior of the proposed protocol using
Vsim, a VANET simulator created in the University of Ulm, Germany [17]. The simu-
lator used combines both a road traffic simulation with a communication simulation as
discussed below. In the following subsections, we present the simulation methodology,
then the analysis of the results obtained.
4.1 Traffic Model
Traffic simulation is based on the traffic model of Nagel and Schreckenberg [18] where
vehicles are generated randomly from the roads endpoints, heading to randomly chosen
destinations. Their velocity and position are updated every 100 msec taking into consid-
eration the rules for changing lanes and the behavior at intersections. In our experiment,
a single bidirectional road model was used for testing. Vehicles are generated from both
ends and move in opposite directions.

begin
while (true) {
If (change in sensed data > threshold) {
compose report;
add to localDB;
}
TransmitData();
ReceiveData();
}
end.
TransmitData() {
if (new neighbor || timer expires) {
for each category (i)
calculate interest level;
sort database descendingly;
transmit top R records;
}
ReceiveData() {
if (receive report from neighbor){
for each report i in localDB {
if (receivedreport == report.i)
discard;
}
insert in localDB
}
Fig.2. Pseudo-code of the algorithm.
4.2 Communication Model
In this model, vehicles are communicating using 802.11 standard, where every 100 ms
each vehicle broadcasts a beacon message to exchange its state with the surrounding
neighborhood. A beacon message (HELLO message) is used by each node to build its
own neighbor table. Each HELLO message is of length 105 bytes: 25 bytes for the
header and 80 bytes for the data. The message header contains the car-ID, generation
time, (x,y) coordinates of the vehicle and a list of its categories of interests. We limited
our model to only 4 categories, as discussed below. The transmission range is set to
500m.
4.3 Experimental Results and Analysis
We consider a single road with vehicles generated from both ends in opposite directions.
Data reports are generated by only 20 percent of the vehicles, which represent an in-
jection rate of 0.2. Only four categories for data reports were defined in our simulation:
traffic condition, road services, weather and no preferences. Each vehicle interest is se-
lected randomly amongst those categories with different probabilities. We conducted
two experiments, one with uniform distribution amongst different categories by setting
the all probability values to 25 percent. In the second experiment, we simulated the

scenario where 50 percent of vehicles were interested in traffic conditions, 20 percent
in weather conditions, 20 percent in available gas stations and 10 percent in available
parking places, as in table 2. Those values are tuning parameters that can be adjusted.
Table 2. Interests Levels for 4 categories.
Category Interest level
Traffic conditions 0.5
Weather conditions 0.2
Gas stations 0.2
Parking places 0.1
Data report has a fixed size of 100 bytes and the local database size is fixed to 200
reports. The simulation experimented were conducted for a total simulation time of 30
minutes that is divided into steps, each of 100 msec length. The performance measure
chosen for evaluation is the percentage of relevant reports received by the vehicles. It is
calculated as the percentage of the number of relevant reports out of the total number of
receivedreports. In Figures 3 and 4, the average percentage value for vehicles belonging
to the same category obtained by applying our approach is plotted against the basic
approach were relevance is not considered.
Fig.3. Interest level per category, equiprobable interests.
Figure 3 represents the case where all categories have the same interest probability.
It is clear that our approach has no benefits over the basic one as all categories are
the same. As in Figure 4, when a certain category is of more interest to most of the
vehicles, the improvement is clear for those vehicles. The percentage of relevant report
received approaches 99 percent, while the rest of the vehicles experience decrease in
their percentage. This is due to the selection process applied that selects only the top 10
relevant reports from the local database for transmission.
In order to enhance the performance of our approach, a better selection technique
could be applied. The 10 reports selected for transmission should be selected from the 4

Fig.4. Interest level per category, category 4 most popular.
different categories with respect to the percentage of their interest level at transmission
time. In this case, and according the the values chosen in table 2, the ten selected data
reports will consist of five reports belonging to category 1, two reports from category
2, two reports from category 3 and one report from category 4. Applying this technique
is currently being investigated.
5 Conclusions
In this paper, we presented how we can substantially increase the percentage of relevant
data reports received using a selective dissemination protocol that considers the vehi-
cles ´ındividual interests. With the application of our interest-sensitive protocol, up to 99
percent of the reports received were of interest to the users belonging to the most pop-
ular service category.Our experiments and results proved our concept. However more
investigation needs to be conducted. Currently, we are testing the improved selection
procedure to include non-popular categories as well. Furthermore, a city model is used
for testing the effect of the city traffic on the overall performance.
Acknowledgements
Our thanks to Prof. Hans-Peter Grossmann and his group, especially Bernhard Wiegel
from the University of Ulm, Germany, for their valuable support and stimulating dis-
cussions.
References
1. Williams, B., Camp. T.: Comparison of Broadcasting Techniques for Mobile Ad Hoc Net-
works. ACM MOBIHOC (2002) 194–205
2. Krishnamurthy, P.: Information Dissemination and Information Assurance in Vehicular ad-
hoc Networks: A Survey. Proc. of 3rd annual iConference (2008)
3. Hanbali, A., Ibrahim, M., Simon, V., and Endre: A Survey of Message Diffusion in Mobile
Ad-hoc Networks. Proc. of International Conference on Performance, InterPerf (2008)

4. Ni, S., Tseng, Y., Chen, Y., Sheu, J.: The Broadcast Storm Problem in a Mobile Ad hoc
Network. Fifth Annual ACM/IEEE International Conference on Mobile Computing and Net-
working (1999) 152–162
5. Wolfson, O., Xu, B., Tanner, R.: Mobile peer-to-peer dissemination with resource con-
straints. 8th International conference on mobile data management, MDM (2007)
6. Ko, Y., Vaidya, N.: Flooding-based geocasting protocols for mobile ad hoc networks. Mobile
Networks and Applications (2002) 471–480
7. Fussler, H., Widmer, J., Kasemann, M., Mauve, M., Hartenstein, H. :Contention-based for-
warding for mobile ad hoc networks. Elsevier´s ad hoc networks (2003) 351–369
8. Luo, Y., Wolfson, O., Xu, B.,: A spatio-temporal approach to selective data dissemination in
mobile peer-to-peer networks. 3rd International conference on wireless and mobile commu-
nication, ICWMC (2007)
9. Datta, A.: Autonomous gossiping: a self-organizing epidemic algorithm for selective infor-
mation dissemination in wireless mobile ad-hoc networks. International conference on se-
mantics of a networked world , ICSNW (2004)
10. Tonguz, O., Wisitpongphan, N., Bai, F., Mudalige, P., and Sadekar, V. : Broadcasting in
VANET. Proceedings of the 2007 Workshop on Mobile Networking for Vehicular Environ-
ments. (2007) 7–12
11. Dornbush, S., Joshi, A.: StreetSmart Traffic: Discovering and disseminating automobile con-
gestion using vanets. 65th IEEE Vehicular Technology Conference (2007) 11–15
12. Xu, B., Ouksel, A., Wolfson, O.: Opportunistic resource exchange in inter-vehicle ad-hoc
networks. MDM (2004)
13. Zhong, T., Wolfson, O.: Disseminating real-time traffic information in vehicular ad hoc net-
works. IVS (2008)
14. Wegener, A., Hellbrck, H., Fischer, S., Schmidt, C., Fekete, S.: AutoCast: An adaptive data
dissemination protocol for traffic information systems. In VTC (2007) 1947–1951
15. Eichler, S., Schroth, C., Kosh. T., Strassberger, M., : Strategies for context adaptive message
dissemination in vehicular ad hoc networks. V2VCOM (2006)
16. Boldrini, C., Conti, M., Passarella, A.:Context and resource awareness in opportunistic data
dissemination. AOC (2008)
17. Gunter, Y., Grossmann, H., Khalifa, W., Salah, W., Karam, O.: Simulator for inter-vehicle
communication based on traffic modeling. International Conference on Intelligent vehicles
(2004)
18. Nagel, K., Schreckenberg, M. : A cellular automation model for freeway traffic. Jornal of
Physique I, (1992) 2221–2229
19. Wischhof, L., Rohling, H. : On utility fair broadcast in vehicular ad-hoc networks. 2nd inter-
national workshop on intelligent transportation, Germany. (2005)
20. Adler, C., Eigner, R., Schroth, C., Strassberger, M. :Context-adaptive information dissemi-
nation in vanets: maximizing the global benefit. (2006)