Cooperative Infrastructure Discovery Through V2X Communication

Naourez Mejri∗

Email: mejri.naourez@cristal.rnu.tn∗CRISTAL Lab., ENSI

Campus Univ. Manouba,

2010, Manouba, Tunisia.

Farouk Kamoun∗†

Email: frk.kamoun@planet.tn†ESPRIT, School of Engineering,

El Ghazala Technopole,

2035, Ariana, Tunisia.

Fethi Filali‡

Email: filali@quwic.com‡QU Wireless Innovations Center

Doha, Qatar.

Abstract—Opportunistic applications like software or digitalmaps updates and vehicle diagnostic reporting may requirelimited Intermittent Internet access. Hence, those applicationscan benefit of the existing Infrastructure already deployed.Moreover, the knowledge of the location of these APs, especiallythe ones offering free access, can help gaining time and makemore accurate decision when data has to be transmitted. Inthis paper, we propose a Cooperative Infrastructure DiscoveryProtocol, called CIDP. It allows vehicles to gather informationabout encountered APs through direct communications withthe infrastructure (Infrastructure to Vehicle(I2V)/Vehicle toInfrastructure(V2I) communications) and mainly exchange itwith opportunistically encountered vehicles (Vehicile to Vehicle(V2V) communications). We studied the performances of CIDPthrough simulation. Results show that it improves the discoveryprocess of APs thanks to the dissemination of the Infras-tructure information cooperatively between equipped vehiclesthrough V2V communications. Then, in order to reduce theinduced overhead, we proposed two variants of CIDP, CIDP-D (CIDP with Density consideration) and CIDP-DH (CIDP-Dwith History consideration) in which the broadcast decisionsare dynamically taken based on the neighboring vehiclesencountered during a set of time frames. Further simulationsshowed that, with a judicious configuration of its parameters,CIDP-D can provide results similar to those of CIDP withlower overhead. However, with CIDP-DH, the best results areobtained with a single frame which is probably due to the highdynamism of the network making the knowledge of previoushistory irrelevant.

Keywords-Infrastructure discovery; opportunistic communi-cations; V2V2I communications

I. INTRODUCTION

There is a wide range of potential applications for

VANETs motivating the ever growing interest granted to

these networks. In these applications, vehicles can have

data to exchange among them in a peer to peer fashion

(such as multimedia content exchange) or have to deliver

collected data to a final destination. This task can be accom-

plished either by using neighboring vehicles as data relays

(through V2V communication) or by using the encountered

predeployed access points offering free anonymous access

(through V2I communication). For example, in the case of

vehicular sensor networks (VSNs [3]), heterogenous sensors

collect data during the vehicle journey based on the task they

are designed to serve. The set of collected data needs then

to be forwarded to the control center for further processing

and acting decisions. To reach the final sink, the vehicle can

connect to one of the free encountered APs in its range.

Another potentials applications such as software or digital

maps updates, vehicle diagnostic reporting and asynchronous

mail transfer can also benefit of this spontaneously deployed

Infrastructure.

In this paper, we present a Cooperative Infrastructure Dis-

covery Protocol (CIDP) which allows autonomous discovery

and exchange of up-to-date information about the available

access points in a specific region (a city for example).

Hence, vehicles will be able to make more accurate routing

decisions based on this information. It is worth mentioning

that the proposed CIDP protocol can be used for the discov-

ery of communication units integrating V2X communication

technologies under development in several standardization

bodies, initiatives, and projects. In particular, it will be

possible to deploy CIDP as a complementary module in the

architecture proposed by IEEE (WAVE - Wireless Access in

Vehicular Environment) to discover Road-Side Units (RSU)

or in those suggested by ETSI ITS technical committee

and ISO CALM working group to discover deployed ITS

roadside stations. In the context of the ETSI ITS communi-

cation architecture, CIDP can be one of the modules of the

facilities layer and contribute on the building of the Local

Dynamic Map (LDM) by adding the APs information in

the digital map. In the context of the under-development

V2X communication architectures, CIDP would use one of

the available Service Channels (SCH) and not the Control

Channel (CCH) reserved for critical safety V2X applica-

tions. The remainder of this paper is organized as follows.

Section II gives an overview of existing research works in

the context of vehicle-to-Internet communications. We give

the details of CIDP protocol in Section III to show how

CIDP is different and/or complementary to these efforts. We

also describe the different transmission modes. Performances

of CIDP are presented and discussed in Section IV. Then, its

variants CIDP-D and CIDP-DH are described and analysed

in Section V. Finally, Section VI concludes this paper and

outlines future works.

978-1-4244-8435-5/10/$26.00 ©2010 IEEE

II. RELATED WORKS

Previous research works already dealt with 802.11 access

point mapping. In fact, there are several websites that

provide maps of WiFi access points. Some popular examples

are WiFiMaps [19], JIwire.com [15] and FON Maps [13].

These websites give the location and characteristics of the

encountered APs. However, they are limited to specific

regions (WiFiMaps for zones in the US) or for specific

hardware (FON Maps only locates FON APs). Besides,

in such solutions, data is constructed through war-driving

results uploaded by independent users which impacts the

accuracy of the collected data and also the frequency of

update. The showed maps can become outdated quickly by

the time the driver consults them.

The same problem of accurate data and update frequency

arises in the case of research studies such in Intel Place Lab

projects [14] where a database of up to 30000 802.11b APs

in different US cities is stored and maintained.

Other works also considered the issue of selecting the best

access point among a set of available ones. Some of these

techniques are based on passive scan for beacon signals the

AP willingly broadcasts. It then chooses the non encrypted

one with the strongest signal strength. However, conducted

studies in [4] showed that this technique does not perform

better than choosing an AP at random. In fact, the signal

strength showed to be an insufficient predictor of AP quality

since other factors impact also on this choice. In fact, the

selected AP can belong to a payed service to which the

end user needs to subscribe. Besides, ”open” access APs

may also be using a controlled access list by MAC address

filtering. Moreover, it can refuse granting a valid IP address

through DHCP to allow Internet access thus blocking traffic

from not allowed terminals.

In Virgil [5], the authors consider these factors. They launch

a set of tests to determine the APs characteristics and even

collect information that can be later used for QoS require-

ments. To fully accomplish those tests, the authors in Virgil

consider that the subject trying to connect to the candidate

APs has limited or no mobility meanwhile these tests in

order to find the best AP allowing him Internet connection.

However, if we consider vehicles speeding in highways, the

connectivity time with the encountered APs will be reduced

and thus can compromise fulfilling appropriately all the tests

proposed in Virgil.

To analyze and adapt APs scanning with vehicles’ mobilities,

other works such as [9] and [8] investigated through real

experimentations to which level such hazardous encounters

between vehicles and APs can offer reliable services to

potentials clients. In Cabernet [8], the authors proposed

using encountered open 802.11 APs for a one-hop delivering

data. Therefore, they proposed two components: Quickwifi

to reduce required time to establish end-to-end connection

with APs and CTP to improve throughput over TCP. Authors

in [9] also suggested that using APs caching can improve

spent time to associate with an existent AP. The results

and improvements proposed in these works can be used in

conjunction with our proposal CIDP in order to enhance the

connectivity between vehicles and existing Infrastructure.

In fact, our focus in this paper is the development of a coop-

erative protocol which allows the exchange of infrastructure

information between vehicles. We aim at knowing if the

AP offers free anonymous access(i.e whether the vehicle

could obtain an IPv4 address from the access point DHCP

server and then get an intermittent Internet connection) so

that it can be considered as a data relay candidate. This

task is achieved in a completely distributed and transparent

manner towards the driver and his driving habits in opposite

with required wardriving for previously mentioned works.

Moreover, our proposal enables frequent data updates in

order to keep accurate fresh information in the AP wireless

infrastructure databases (WIDs) since they will be updated

each time new APs are detected or when new information

are received from other vehicles.

III. CIDP OVERVIEW

There are already deployed solutions aiming at discover-

ing and listing the available Infrastructure as listed in Section

II. However, in our solution we aim at having a distributed

knowledge of information related to each wireless equip-

ment built basically through V2V communication besides

to passive scanning. Each vehicle will fill its own database

dynamically and collect up-to-date information about the

existing Infrastructure. This task will be achieved in a

completely transparent manner towards the driver. Hence,

a tourist entering an unfamiliar city or region can use CIDP

to get information about the available free APs without any

required Internet connection or complicated requirements.

In fact, the assumptions made for CIDP are very simple

and realistic. First, we assume that there are no energy or

storage constraints which is a common assumption made for

vehicular networks. Besides, vehicles are supposed to have

exact knowledge of their geographical location. This can be

easily achieved by the use of in-built location devices such

as GPS antennas. The position of vehicles will help later

for a better estimation of the location of encountered access

points. In fact, the knowledge of the GPS coordinates of the

APs is not necessary for CIDP operations since what would

be important is the visibility area of these APs in the streets

which can be estimated by knowing the positions of vehicles

at the time when they discovered these APs. In this paper,

we are not interested in the way access points are passively

scanned. The reader should refer to [11] for further details

on how this can be accomplished. In this paper, we focus on

CIDP operating mode which is centered around the possible

ways of exchanging messages between vehicles to build a

global Wireless Infrastructure Database (WID) containing

the information about the wireless access points. We briefly

overview the content of this database in the next paragraph.

A. Wireless Infrastructure Database

Each entry of this database contains two main parts.

The first one is reserved for the AP properties such as

its BSSID, its ESSID, the frequency channel number field

(which depends on the AP supported 802.11 variant: a/b/g/p)

indicating on which channel the AP is operating and whether

the Dynamic Frequency Selection (DFS) option is enabled

or not. Note that if DFS is not enabled, when data is to

be transmitted, the vehicle will not have to scan again all

the channels. There is also the AP estimated position which

can be approximated with triangulation methods [6] using at

least the positions and signal strength information of three

of the vehicles that detected it. The time when the AP has

been spotted for the first time is also stored. If not updated,

obsolete entries are removed by coupling this field with

the expire duration one. Other values such as signal level,

noise level and quality of the signal (computed with both

signal and noise levels values) are also stored to depict the

quality of the signal being received from the AP. Depending

on whether the AP uses security mechanisms or not, extra

fields with information about the IP, the default gateway,

and the DNS address(es) obtained whenever a vehicle suc-

cessfully connect to the AP using DHCP (Dynamic Host

Configuration Protocol) can be added. Otherwise, it will be

filled with information about the used security mechanism.

Storing these information can help reduce the overall time

to associate with previously encountered APs. Experimental

results showing improvements obtained thanks to IP caching

can be found in [9].

The second one stores data related to the vehicle which

originally scanned the AP 1. Among these fields, there is

the identity of the vehicle, its velocity and its position.

B. Announcing access points

When meeting, vehicles can exchange information gath-

ered about APs in two main fashions: periodically or at

demand respectively referred as unsolicited and solicited

announcements. These two categories are detailed in the

following subsections.

1) CIDP unsolicited announcements: Unsolicited an-

nouncements designate periodic broadcast messages sent by

vehicles to cooperatively help each other update their WID

databases with fresh and new information about available

APs. We defined three different kinds of unsolicited an-

nouncements namely position, time and type-based broad-

casts which can be used by applications looking for a

temporary (limited) Internet connectivity. In the first type,

the vehicle will broadcast the information about only the

APs located in a specific zone. The second one serves to

1The vehicle that broadcasted the APs data may be different of the onethat scanned it.

limit broadcasted data to that inserted/updated after a specific

time. This type can be used to increase the freshness degree

of the WID data since only new infrastructure information

will be circulating in the network. Finally, the type-based

broadcast restricts the broadcasted information to only APs

that for instance offer free access without security protection.

CIDP also provides the possibility to combine two or all

these types in unsolicited announcements.

Obviously, the efficiency of CIDP thanks to unsolicited

announcements broadcasting is highly correlated with dif-

ferent factors such as the movement pattern and density

of the vehicles. For instance, high vehicles speed may

compromise the exchange of the data as in this case the

inter-vehicle connectivity time may be not enough to allow

vehicles to exchange a high data volume. Besides this factor,

vehicles density is also an important factor that influences

the efficiency of CIDP since the likelihood of getting and

spreading more updates among neighbors will increase for

high number of vehicles in a specific geographic area which

in return would help CIDP to converge quickly and properly.

Moreover, the broadcast interval, which defines the duration

between two successive broadcasts of unsolicited announce-

ments, has to tuned effectively since it would impact the

frequency and quality of updates of WID’s entries.

The value of this interval can be dynamically tuned accord-

ing to rush hours and vehicle expected scheduled journey

that can be based on vehicles traffic history. Furthermore,

it can also be configured by taking into consideration the

mobility and density of encountered vehicles.

The effect of these factors will be investigated through

extensive simulations in Section IV.

2) CIDP solicited announcements: Sometimes, a vehicle

can require a precise information about the wireless infras-

tructure in a specific zone (for example, a tourist’s vehicle

can demand the full list of APs located in a first-time visited

city). In this case, unsolicited announcements can not be

sufficient for rapidly handling this request. Hence, CIDP

allows a vehicle to send an explicit infrastructure broadcast

request to its neighbors. In return, only vehicles which have

the data matching this request have to answer. Although

these replies can be unicasted to the originator vehicle,

we prefer broadcasting them in order to allow vehicles in

the neighborhood to graciously update their own WIDs.

The replies issued to answer CIDP request messages are

called solicited announcements. It is clear that one CIDP

infrastructure information request may result in more than

one solicited announcement as the replies generated from

different vehicles may not contain the same content. Vehicles

overhearing at least one solicited announcements containing

similar information they are willing to send have to cancel

their sending process. A distance-based congestion control

mechanism, where a backoff-time inversely proportional

to the distance from the originator vehicle is used. This

mechanism would contribute on the increase of the efficiency

Figure 1. The global CIDP flowchart.

of CIDP by avoiding collisions between solicited announce-

ments. For solicited announcements with different content,

the originator vehicle has to merge the gathered wireless

infrastructure information.

Similarly to unsolicited announcements, solicited announce-

ments can also be position-based, time-based, or type-based

according to the information the user applications require to

have at the time of generating the request.

Figure 1 shows how each of these described CIDP messages

is processed. Upon receiving either solicited or unsolicited

announcements, a vehicle has to check the possible updates

if its WID. This operation can be achieved either by adding

entries of new announced APs or by updating the content

of one of more fields of an existing AP. Besides, in the case

when a vehicle is preparing a solicited announcement while

it receives an unsolicited announcement, it has to consider

fresher announced information if its solicited announcement

message. Furthermore, when a request message is received,

a vehicle will check if it has some request-matching entries

and prepare a solicited announcement to be broadcasted.

Further details on the types and formats of the messages can

be found in [2].

In the rest of the paper, we focus on the study of only the

unsolicited announcement and investigate how to tune the

broadcast periodicity to get better performances.

IV. CIDP PERFORMANCE ANALYSIS

We implemented the CIDP protocol in NS-2.33 [16]. For

simulation purposes, we used random yet realistic mobility

0

20

40

60

80

100

30 60 90 120 150 180

Pe

rce

nta

ge

of

dis

cove

red

AP

s

Broadcast Interval Duration (in Seconds)

Transmission Range 100mTransmission Range 300mTransmission Range 500m

Figure 2. Impact of variable transmission range of vehicles on CIDPperformances in the case of 250 vehicles.

traces. These traces were generated using TraNS [18] which

is a GUI tool that integrates traffic and network simulators

(NS-2 and SUMO [17]) to simulate realistic simulation

scenarios.

A. Traffic and network model

The network is modeled as a square grid consisting of

10*10 edges. Each edge of a length of 300 meters. The

number of existing APs is set to 20. The number of vehicles

in the network varies from 200 to 600 vehicles to study the

impact of different densities on CIDP performances. The

simulation ends when all vehicles reach their destinations.

B. Simulation results

We describe and discuss the results obtained with CIDP

unsolicited announcements broadcast exchange when vary-

ing a set of parameters such as transmission range and

density.

1) Impact of variable transmission range: First, we study

the impact of varying the vehicles’ transmission range on

CIDP performances. As depicted in Figure 2, with 250

vehicles traveling with transmissions ranges varying between

a 100, 300 and 500 meters, we notice that increasing the

vehicles’ transmission range highly improves CIDP perfor-

mances even for low densities. In fact, with a trasmission

range equal to 500m, more than 90% of deployed APs were

detected compared to less than 30% when the range is fixed

at 100 m with variable broadcast interval duration. This is

due to the fact that when increasing the transmission range,

more vehicles will be able to overhear more broadcasts from

other vehicles for longer time.

In the following subsections, the vehicles’ transmission

range is fixed to 300m.

2) Impact of CIDP broadcasting interval: As depicted in

Figure 3, the percentage of discovered APs is the highest

for the smallest periods i.e broadcast intervals. In fact, the

more the broadcast is frequent, the more vehicles will have

the oppourtunity to overhear other vehicles diffusing the

0

10

20

30

40

50

60

70

80

90

100

30 60 90 120 150 180 210 240 270 300

Pe

rce

nta

ge

of

dis

cove

red

AP

s

Broadcast Interval (in Seconds)

200 vehicles400 vehicles600 vehicles

Figure 3. CIDP performance.

information they collected about the Infrastructure. More-

over, when the broadcasts get spaced, we notice that the

percentage of discovered APs naturally decrease especially

for lower densities (i.e the scenario with 200 vehicles).

3) Effect of the V2V communication: To explain the

obtained high percentages of APs, we separate in Figure

4 the percentage of APs discovered thanks to V2V com-

munication and those directly scanned in the case of 200

vehicles2. We remark that the highest percentage of the

APs is discovered thanks to the V2V periodic exchanges.

We also notice that the larger is the interval separating two

consecutive broadcasts, the more the discoveries are made by

direct scanning of the APs themselves. This is due to the fact

that the vehicles will communicate less with their neighbors

causing the exchanges between them to get scarcer. Hence,

the APs discovered will become basically limited to those

that the vehicle directly came across itself.

4) CIDP broadcasting overhead: Intuitively, the more

vehicles in the network, the more messages will be ex-

changed between them and hence the overhead is dependant

of the vehicles densities as shown in Figure 5. Besides, the

overhead is dependant of the broadcast periodicity: it is the

lowest for longer broadcast intervals and becomes higher

when the broadcast intervals get shorter. To summarize,

there is a compromise to make when choosing the broadcast

interval in order to keep a good visibility of the existing

Infrastructure in the network with reduced overhead.

In the next section, we investigate how we can reduce this

generated overhead while maintaining good performances of

CIDP.

V. TUNING BROADCAST PERIODICITY BASED ON

NEIGHBORING VEHICLES

In the previously described version of CIDP, vehicles

broadcast periodically without any consideration of their

surroundings. This can cause extra non beneficial overhead

2Similar results were obtained for the other scenarios with higherdensities

0

10

20

30

40

50

60

70

80

90

100

30 60 90 120 150 180 210 240 270 300

Pe

rce

nta

ge

of

dis

cove

red

AP

s

Broadcast Interval (in Seconds)

Total discoveredDirectly discoveredV2V discovered

Figure 4. V2V communication impact on the Infrastructure discovery.

0

100

200

300

400

500

600

700

800

900

1000

30 60 90 120 150 180 210 240 270 300

Nu

mb

er

of

sen

t b

yte

s/s

Broadcast Interval (in Seconds)

200 vehicles400 vehicles600 vehicles

Figure 5. The Generated overhead by V2V communication.

especially in the case where there are no vehicles in range

at the broadcast time. Thus, in an attempt to reduce such

overhead, we propose a variant of CIDP with consideration

of the vehicles’ surroundings when making decision to

broadcast or not.

A. CIDP with Density consideration (CIDP-D)

In CIDP-D, vehicles will store the number of newly

encountered vehicles during the current frame. At the end

of each time frame, each vehicle will compute the ratio R

such as in Equation 1 where nb new is the number of the

new vehicles encountered and nb total is the total number

of vehicles encountered during the frame duration (the old

neighboring vehicles still in range from the previous frame

duration and the new ones encountered during this frame).

R =nb new

nb total(1)

If the computed ratio R is beyond a certain threshold value,

then the vehicle decides to broadcast the collected data. If

not, it will wait for another frame before deciding again

in the same manner whether it should broadcast or not.

With this method, the broadcasts will be more effective.

They will not occur if there are not enough new vehicles

nearby. The vehicle can have neighbors from the previous

0

10

20

30

40

50

60

70

80

90

100

20 40 60 80 100 120 140 160 180

Pe

rce

nta

ge

of

dis

cove

red

AP

s

Frame Duration (in Seconds)

Threshold=0Threshold=0.5Threshold=1

Figure 6. Impact of threshold values on CIDP performances.

frame but those vehicles probably have already received

the APs information in previous broadcasts. Consequently,

the overhead caused by the V2V communication should

decrease.

1) Impact of the threshold parameter on CIDP-D perfor-

mances: We measured the impact of the threshold value on

the overall overhead and on the Infrastructure discovery pro-

cess. We notice in Figure 7, that the overhead significantly

decreases especially for the lower frame durations when

considering higher threshold values. However, we notice that

with threshold values above 0.5, CIDP (the upper curve)

outperfroms CIDP-D and the percentage of discovered APs

by CIDP-D decreases as described in Figure6. In fact, a

threshold value equal to 0.5 means that at least the half of

the current neighbors weren’t in the vehicle’s transmission

range during the previous broadcast. This condition implies

a high dynamism in the whole network which can’t be easily

met especially in realistic life. In fact, unless vehicles are

moving in opposite directions (the case of two way lane

or at intersections) or at very different speeds (at highways

for example), they will stay in each other range for a while

before leaving. Hence, the ratio computed by vehicles will

not easily reach or surpass threshold above 0.5. To conclude,

the threshold value selection is a key criteria for ensuring a

better information exchange.

In the next section, we investigate the impact of keeping

a history of encountered vehicles when making broadcast

decision.

B. CIDP with Density History consideration(CIDP-DH)

Motivated by the rendered results of CIDP-D, we wanted

to see whether considering the history of the vehicles’

encounters will enhance the Infrastructure discovery process.

To accomplish this, the vehicles make decisions, by compar-

ing the ratio computed among a set of frames nb frames as

described in Equation 2 and the threshold value. ai is a

weight factor such as∑nb frames

i=1ai = 1 and nb newi is

the number of new vehicles encountered during the frame

number i. Finally, nb total is the sum of the vehicles

0

100

200

300

400

500

600

700

800

900

20 40 60 80 100 120 140 160 180

Nu

mb

er

of

sen

t b

yte

s/s

Frame Duration (in Seconds)

Threshold=0Threshold=0.5Threshold=1

Figure 7. CIDP-D overhead with different threshold values.

encountered during each frame and the vehicles that were

already in range at the first frame.

R =

∑nb frames

i=1ai ∗ nb newi

nb total(2)

1) Impact of the number of frames in CIDP-DH: To

see how the history can affect the Infrastructure discovery,

we considered a network consisting of 200 vehicles where

we varied the number and duration of frames on the APs

discovery. Figure 8 shows the obtained results for both

the threshold values 0.2 and 0.5. As illustrated in Figure

8(a), we notice that with a small threshold value results

are nearly similar for the different number of frames. In

fact, this threshold value is low and can be easily reached

(since each vehicle will encounter new neighbors coming

from other directions or intersections) and thus the vehicles

will broadcast frequently. However, when the threshold value

equals 0.5, the best results are obtained when considering

only one frame. The more frames are considered, the re-

sults worsen (see Figure 8(b)). This confirms our previous

interpretations. To decide broadcasting, at least half of

the vehicles should not have been encountered previously.

Obviously if this condition is not easily met for a single

frame, it will be harder to satisfy when computing the ratio

with consideration of more frames.

As already mentionned, we remark in Figure 9 that

the overhead is highly correlated with the percentage of

discovered APs. Hence, the overhead generated by vehicles

which consider only one frame is higher than this of two or

five frames.

In the following, we study the version of CIDP-DH with

only two frames where the ratio R is computed as described

in Equation 3.

R =αnb new1 + (1 − α)nb new2

nb total(3)

We measured the effect of varying the α values on the

performances. Figure 10 outlines the results obtained. We

notice in Figure 10(a), with a frame duration of 30 seconds,

0

10

20

30

40

50

60

70

80

90

100

30 60 90 120 150 180 210 240 270 300

Pe

rce

nta

ge

of

dis

cove

red

AP

s

Frame duration (in Seconds)

Number of frames = 1Number of frames = 2Number of frames = 5

(a) Threshold=0.2

0

10

20

30

40

50

60

70

80

90

100

30 60 90 120 150 180 210 240 270 300

Pe

rce

nta

ge

of

dis

cove

red

AP

s

Frame duration (in Seconds)

Number of frames = 1Number of frames = 2Number of frames = 5

(b) Threshold=0.5

Figure 8. Impact of varying the number of frames on CIDP-DHperformances.

0

10

20

30

40

50

60

30 60 90 120 150 180 210 240 270 300

Nu

mb

er

of

sen

t b

yte

s/s

Frame Duration (in Seconds)

Threshold = 0.5

Number of frames = 1Number of frames = 2Number of frames = 5

Figure 9. CIDP-DH overhead with different numbers of frames.

that with small threshold values (threshold=0.2), the alpha

has practically no impact on the percentage of discovered

APs. In fact, with this threshold, the broadcast is frequent

enough and hence we have good performances. However,

when the threshold becomes higher, we remark that the best

performances are when α = 0 then with α = 1. Those casesare when we consider only one frame to decide whether to

broadcast or not.The worst results are when we uniformly

consider the number of vehicles discovered in both frames

(α = 0.5) especially with high threshold values. Similar

results are pictured in Figures 10(b) and 10(c) respectively

with 400 and 600 vehicles in the network.

VI. CONCLUSION

In this paper, we proposed a new protocol CIDP (Coop-

erative Infrastructure Discovery Protocol) for gathering the

wireless infrastructure information through vehicular coop-

eration. It offers the advantage of easy and accurate update of

access points database. Different information dissemination

capabilities (solicited and unsolicited) have been defined in

order to allow vehicles to communicate and exchange their

knowledge about the discovered APs. Simulation results

showed that the V2V broadcast interval and the vehicles

density have an expected impact on the performance of

CIDP. Of course, CIDP’s performances are the best with

higher vehicle density which help AP information spreading

to all vehicles. Nevertheless, this condition can be easily met

in real daily traffic scenarios. Besides, CIDP still performs

well even for lower densities.

Moreover, in order to reduce the generated overhead, we

proposed two variants of CIDP, CIDP-D and CIDP-DH.

We found that CIDP-D decreases the overhead thanks to

considering the vehicles’ surroundings when deciding to

broadcast. We also found, with CIDP-DH, that keeping an

history of the encountered vehicles is only prejudicious for

only short duration with higher thresholds since the vehicle

environment can change quickly and unpredictably. Thanks

to CIDP and its variants, vehicles can have an overview of

the available APs in the zone to be entered soon and hence

can make more appropriate routing decisions. Future works

will focus on this routing issue.

REFERENCES

[1] A. Akella and G. Judd and P. Steenkiste and S. Seshan,SelfManagement in Chaotic Wireless Deployments,11th AnnualInternational Conference on Mobile Computing and Network-ing, MobiCom, 2005.

[2] N. Mejri, F. Filali and F. Kamoun, A Cooperative Infrastruc-ture Discovery Protocol for Vehicle to Internet OpportunisticCommunications, ACS/IEEE Future Trends on Ad-hoc andSensor Networks Workshop, FT-ASN’10, 2010.

[3] P.Bellavista and E. Magistretti and U. Lee and M. Gerla,Standard Integration of Sensing and Opportunistic Diffusionfor Urban Monitoring in Vehicular Sensor Networks: theMobEyes Architecture, IEEE International Symposium, 2007.

[4] K. Matsuzawa and K. Mase and Y. Hirano and S. Kajita,Experience Map Creation by Virtual WLAN Location Esti-mation, International Symposium on Wearable Computers,ISWC, 2006.

0

10

20

30

40

50

60

70

80

90

100

0 0.2 0.4 0.6 0.8 1

Pe

rce

nta

ge

of

dis

cove

red

AP

s

α value

Threshold=0.2Threshold=0.5Threshold=0.7Threshold=0.8

(a) 200 vehicles

0

10

20

30

40

50

60

70

80

90

100

0 0.2 0.4 0.6 0.8 1

Pe

rcen

tag

e o

f d

isco

vere

d A

Ps

α value

Threshold=0.2Threshold=0.5Threshold=0.6Threshold=0.8

(b) 400 vehicles

0

10

20

30

40

50

60

70

80

90

100

0 0.2 0.4 0.6 0.8 1

Pe

rce

nta

ge

of

dis

cove

red

AP

s

α value

Threshold=0.2Threshold=0.5Threshold=0.6Threshold=0.8

(c) 600 vehicles

Figure 10. Impact of varying alpha values on CIDP performances withdifferent threshold values.

[5] A. Nicholson and Y. Chawathe and M. Chen and B. Nobleand D. Wetherall, Improved access point selection, 4th Inter-national Conference On Mobile Systems, Applications AndServices, Mobisys, 2006.

[6] A. Roxin and J. Gaber and M. Wack and A. Nait-Sidi-Moh, Survey of Wireless Geolocation Techniques, GlobecomWorkshops,2007.

[7] S. Srinivasan, Opportunities and Challenges in Unlicensed-Band Networks, Wireless/Mobile Planning Group Workshop.

[8] J. Eriksson and H. Balakrishnan and S. Madden, Cabernet:vehicular content delivery using WiFi, ACM Mobicom, 2008.

[9] V. Bychkovsky and B. Hull and A.n Miu and H. Balakrishnanand S. Madden, A Measurement Study of Vehicular InternetAccess Using In Situ Wi-Fi Networks, ACM MOBICOM,pp50–61, 2006.

[10] Y. Qian and N. Moayeri, Design Secure and Application-Oriented VANETs,IEEE VTC’2008, 2008.

[11] I. Amdouni and F. Filali, Intelligent strategies of accesspoint selection for vehicle to infrastructure opportunisticcommunication,IEEE VNC’2009, 2009.

[12] COMeSafety: Communication for eSafety,Http://www.comesafety.org.

[13] FON maps, Http://maps.fon.com.

[14] Intel research Seattle, place lab: a privacy-observant locationsystem,Http://placelab.org.

[15] JIWire, Wi-Fi access point locator, Http://jiwire.com.

[16] NS-2 simulator, Http://www.nsnam.isi.edu.

[17] SUMO, Simulation of Urban MObility,Http://sumo.sourceforge.net.

[18] TraNS, Http://www.trans.epfl.ch.

[19] WiFi Maps: War-driving maps and access points locator,Http://www.wifimaps.com.