Content uploaded by Tarik Taleb
Author content
All content in this area was uploaded by Tarik Taleb on Jul 30, 2014
Content may be subject to copyright.
Design Guidelines for a Network Architecture
Integrating VANET with 3G & Beyond Networks
Tarik Taleb∗and Abderrahim Benslimane+
∗NEC Europe Ltd., Germany. +University of Avignon, France
{talebtarik,benslimane}@ieee.org
Abstract— Vehicle Ad Hoc Networks (VANETs), based on
IEEE 802.11p, and 3G & beyond networks are characterized
by their high data transmission rates and wide range commu-
nication, respectively. This paper presents an architecture that
integrates between the two, making advantage of the features
of each. Design guidelines pertaining to vehicle clustering and
gateway management are defined. The former aims for enhancing
the link stability within the VANET, whereas the latter sustains
inter-connectivity of the VANET with the backhaul 3G & beyond
network. Simulations are carried out using NS2 to evaluate
the performance of the integrated network architecture and
encouraging results are obtained in terms of high data packet
delivery ratio, reduced control packet overhead, and reduced
packet drop rate.
I. INTRODUCTION
Along with recent and ongoing advances in the area of
wireless communications, a wide plethora of wireless tech-
nologies is emerging, defining different types of networks. The
IEEE 802.11-based Wireless Local Area Networks and the
3G Cellular Networks are two notable ones. The enhanced
version of IEEE 802.11 networks, which is IEEE 802.11p,
forms the standard for Wireless Access for Vehicular Envi-
ronments (WAVE). It operates at a frequency of 5.9 GHz,
divided into 7 channels, each operating at a frequency of 10
MHz. It provides a high data transmission rate, ranging from
6Mbps to 27 Mbps and a short-range radio communication
of approximately 300 meters. On the other hand, Universal
Mobile Telecommunication Systems (UMTS), as a 3G &
beyond cellular network technology, operates with a frequency
range of around 2 GHz. The UMTS dedicated channel offers
a peak downlink data rate of 2 Mbps and a peak uplink data
rate of 384 Kbps, whereas the UMTS High Speed Data Packet
Access (HSDPA) radio service offers peak downlink data rates
of 7.2Mbps and uplink rates of 2Mbps. UMTS offers a wide
range of communication of around 8 to 10 km per Base Station
(BST).
As an attempt to couple the high data rates of IEEE 802.11p
and the wide communication range of 3G networks, this paper
envisions an integration of IEEE 802.11-based Vehicular Ad
hoc Networks (VANETs) with UMTS. The need for such
integrated architecture stems from the numerous optimality
challenges associated with the inter-connectivity of VANETs
with static roadside infrastructure gateway units (i.e., using
Dedicated Short Range Communication (DSRC)). These chal-
lenges are mainly attributable to the dynamic, infrastructure-
less topology and the multi-hop nature of communication of
VANETs. An efficient integration of IEEE 802.11p and UMTS
network interfaces on vehicles would contribute to better data
access services to vehicles and would significantly reduce the
issue of dead spots in UMTS.
For the support of reliable and stable inter-vehicular com-
munication and internet connectivity to VANETs, two impor-
tant concerns need to be exhaustively addressed. They are
namely vehicle clustering and gateway management. In this
paper, gateways refer to vehicles that link VANET with the
3G/UMTS network. The present paper addresses these con-
cerns in the envisioned VANET-UMTS integrated architecture
and delineates the methodology of dynamic clustering and
adaptive gateway management.
The remainder of this paper is organized in the following
fashion. In Section II, we survey some clustering and gateway
management schemes. Section III introduces the envisioned
architecture, followed by a description of the adopted clus-
tering operation and the adaptive mobile gateway manage-
ment scheme. The performance of the architecture and the
introduced mechanisms is evaluated in Section IV. The paper
concludes in Section V.
II. RELATED WORK
In the area of vehicular communications, there has been
a plethora of research work. In [1], the authors proposed a
new protocol, which selects a route with the longest lifetime
to connect VANET nodes to the wired network by using the
characteristics of vehicular movements. This paper considers
vehicles to be stationary or mobile, but the gateways to
be purely stationary. The authors use two metrics, namely
Link Expiration Time (LET) between adjacent vehicles and
Route Expiration Time (RET) between vehicles and gateways.
Communication with gateways is pro-active and gateway
handover is also addressed. Whilst LET and RET shall be
used in our paper as well, as a metric for gateway selection,
discussed in Section IV, communication with the gateways
is not purely pro-active. A clustering approach with a risk-
aware collaborative vehicular collision avoidance system is
devised in [2]. In this work, vehicles are clustered based on
their velocities, their direction of movement, and inter-vehicle
distances. Additionally, a risk-aware Media Access Control
(MAC) protocol is designed to increase the responsiveness of
the system by associating an emergency level with each vehicle
in its corresponding cluster. Though risk-aware collision-
avoidance is beyond the scope of our paper, our clustering
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
mechanism is also metric-based, as discussed in Section IV. A
stable routing protocol to support ITS services is proposed
in [4]. This paper addresses the issue of path disruptions
caused by vehicles’ mobility. Vehicles are grouped according
to their movement directions to ensure that vehicles of the
same group establish stable single and/or multi-hop paths
while moving together. It uses the vehicles’ characteristics to
predict a link-breakage event prior to its occurrence. Within
the same clusters, communications among the vehicles are
carried out over stable paths with acceptable LET values. In
our paper, direction of movement is used in clustering to select
the minimum number of optimal gateways per direction.
Regarding gateway selection, a wide library of research
work has been conducted in the recent literature. In [7], an
adaptive gateway management mechanism for multi-hop B3G
networks is proposed. In this research work, the authors use
multi-attribute decision making theory and simple additive
weighting (SAW) techniques to select an adequate gateway
based on residual energy, UMTS signal strength and mobility
speed of the gateway candidates. In case the current serving
gateway loses its optimality, the authors proposed a multi-
metric gateway migration approach for handing over the
responsibilities of the serving gateway to a newly-elected
one. Though our paper also deals with gateway handover,
battery is not a constraint in vehicles. Moreover, VANET
cannot exhibit random movement of vehicles like MANET. In
[8], an adaptive distributed gateway discovery mechanism for
hybrid wireless networks is introduced. The proposed gateway
discovery method is hybrid; combining both reactive and pro-
active approaches. It defines a limited Gateway Advertisement
(GWADV) zone within which the gateways periodically prop-
agate advertisement messages. It is equal to the number of
hops, corresponding to the TTL value, adaptively selected by
the gateway. As this reduces both delay and overhead, it has
been adopted in our mechanism too, pertaining to a VANET
scenario. In [9], the author proposes an initiation algorithm
for intersystem (i.e., 2G GSM and 3G UMTS) handover,
based on a combination of the geographical location of mobile
terminals and absolute signal strength thresholds. The paper
aims at improving the usage efficiency of network resource
by considering the threshold for the distance metric between
mobile terminals, in addition to signal strength.
III. PROPOSED VANET-3G INTEGRATED
NETWORK ARCHITECTURE
A. Architecture Description
Fig. 1 portrays the envisioned architecture, considering a
scenario of two different tracks over a particular road (e.g.,
highway), with a track for each direction. The key components
of the architecture are IEEE 802.11p-based VANET vehicles,
a UMTS Node B and the main components of the UMTS core
network. Communication over the VANET network is multi-
hop and on a peer-to-peer basis. VANET is linked to UMTS
via selected VANET mobile gateways using the Universal
Terrestrial Radio Access Network (UTRAN) interface.
Fig. 1. An example scenario of the envisioned VANET - UMTS integrated
network architecture.
The main purpose of our paper is to select only a minimum
number of vehicles to communicate with the UMTS network
as gateways.
Referring to the architecture shown in Fig. 1, the VANET
region under the coverage of UMTS BST, where the UMTS
Received Signal Strength (RSS) is intense, is termed as the
3G active region. The vehicles, equipped with both the IEEE
802.11p and UMTS interfaces, lying within or moving into
the 3G active region, are called Gateway Candidates (GWCs).
The rest of the vehicles, that do not lie in the 3G Active
Region, are not equipped with the UMTS interface, or do
not have their UTRAN interfaces enabled are called Ordinary
Vehicles (OVs). Among the gateway candidates, a minimum
number of Cluster Heads (CHs) per direction are elected as
optimal gateways (GWs) using different metrics. The number
of gateways is required to be minimum, so as to avoid bottle-
neck at the UMTS BST and save UTRAN resources. Gateway
candidates are grouped into clusters using dynamic clustering
mechanism, and only the selected gateways will have their
3G UTRAN interfaces activated. However, the IEEE 802.11p
interface is enabled and activated on all the VANET vehicles.
B. Dynamic Clustering Operation
For the sake of effective relaying of messages and enhanced
stability of inter-vehicular links, we cluster vehicles using
three metrics, namely the direction of vehicles’ movement,
UMTS Received Signal Strength, and the IEEE 802.11p Wire-
less Transmission Range. The per-moving direction clustering
is initially carried out relative to the moving direction of
vehicles in the Cartesian space and then relative to the position
of the UMTS BST. In each cluster formed in the first step,
vehicles are then grouped into two sub-clusters: vehicles
moving towards the BST (CL1.1 and CL2.1 in Fig. 1) and
vehicles moving away from the BST (CL1.2 and CL2.2 in
Fig. 1).
For further refinement of the clustering operation, we use
the UMTS Received Signal Strength (RSS). Indeed, vehicles,
in each sub-cluster formed in the first step, that are equipped
with both the UTRAN and IEEE 802.11p network interfaces
and lying within or moving into the 3G active region, would
receive intense UMTS signal intensity (i.e., greater than a
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
specific Signal Strength threshold SSTh), and hence, will
together form a single gateway candidate sub-cluster. These
vehicles are called gateway candidates (GWCs), as shown in
Fig. 1, and their UTRAN interface is enabled. The rest of the
vehicles behave as Ordinary Vehicles (OVs).
The pen-ultimate stage in clustering is based on the IEEE
802.11p wireless transmission range of gateway candidate
vehicles: a pair of gateway candidates, whose inter-vehicular
distance is less than or equal to their IEEE 802.11p transmis-
sion range, form a new sub-cluster or join an existing one
(i.e., if one of the gateway candidates is already a member of
a cluster). The transmission range of a GWC is determined as
follows:
R=Tr·(1 −υ)(1)
where, Trdenotes the maximum IEEE 802.11p transmission
range and υreflects the wireless channel fading conditions in
the current location.
After the clustering operation, the final stage is to elect a
Cluster Head (CH) for each cluster. A CH is in charge of
initiating communication and controlling the flow of signaling
messages among GWCs within the cluster. The border edge
GWCs in each cluster are to be identified. A GWC having
no neighbour before it, is the leading edge GWC and the
GWC with no neighbour behind it, is the tail edge GWC.
To determine the CH, the leading edge GWC broadcasts its
position in the opposite direction of its movement and the tail
edge GWC broadcasts its position in the same direction of its
movement. Each GWC, receiving both these messages, shall
compute its relative distance from these two edge GWCs. The
GWC, whose relative distances to both edge GWCs are almost
equal, is closest to the centre of the cluster and deems itself to
be the Cluster Head. Computation of TTL
cis then followed
by CH. It is set to the maximum number of hops from the CH
to the tail or to the leading edge GWCs. The TTL
cvalue is
used to restrict the flow of signaling messages pertaining to a
particular cluster within the same cluster.
C. Adaptive Mobile Gateway Management
The Adaptive Mobile Gateway Management mechanism
consists of three mechanisms, namely “multi-metric mobile
gateway selection”, “gateway handover” and “gateway discov-
ery/advertisement” mechanisms. The gateway selection mech-
anism is used to select the minimum number of adequate
gateways to optimally communicate with the backhaul UMTS
network. It is based on the Simple Additive Weighting (SAW)
technique using metrics such as the mobility speed of the CH,
its UMTS RSS, and the stability of its link with the source
vehicles (i.e., LET and RET metrics [1]).
In this paper, the first vehicular source broadcasts a Gateway
Solicitation (GWSOL) message within the VANET, using
the TTL value (TTL
s) as discussed below. In the gateway
selection mechanism, the UMTS RSS and RET are metrics
with positive criterion (i.e., more optimality with increase in
value). As far as the mobility speed metric is concerned, if
the direction of movement is towards the BST, the criterion
is positive; whereas if the movement is away from the BST,
the criterion is negative (i.e., less optimality with increase
in value). Given the fact that a hybrid gateway discovery
mechanism is employed, every GWC belonging to a cluster
tends to know information about its CH, using TTL
c. Hence,
it is sufficient for the GWSOL to reach a GWC of any cluster
to get information about its CH, instead of reaching the CH.
As shown above, the source elects the CH with the maxi-
mum weight as its Gateway (GW). Each metric of the CH has
its own threshold value. After a time instance Δt, if another
vehicle becomes an active source for communicating with the
UMTS BST, that source checks if the UMTS RSS of the serv-
ing gateway and its RET with the gateway are greater than the
respective threshold values. If yes, the active source uses the
same GW for communicating with the UMTS BST. Otherwise,
the source selects another new GW from the remaining CHs of
the other clusters, by applying the same “multi-metric mobile
gateway selection” approach. Additionally, a CH can directly
receive the GWSOL message, if it forms a sub-cluster of itself,
with no neighbor.
Gateway handover is performed to elect a new gateway
and handover the responsibilities of the serving gateway to it,
when the serving gateway starts losing its optimality. A loss
of optimality is accounted when either the value of the UMTS
RSS of the serving gateway or its RET with any one of the
vehicular sources goes below the respective threshold values.
Handover, thus, aims to sustain the inter-connectivity of the
integrated network to pursue the data transaction. In this case,
the serving gateway broadcasts METRIC REQUEST packet
within the VANET. Some of the CHs/GWCs which receive
this packet respond, transmitting the metric values of their
respective CHs. Of course, there is no common GWC between
any two clusters. However, a GWC/CH of another cluster can
receive the metric information through OVs as intermediate
vehicles, especially in case they traverse through the clusters.
As a result of dynamic clustering, GW, which was the CH
when it got elected, may not be the CH at another instance,
as it may become a part of a new cluster. If its optimality,
with respect to its sources, is affected, then the CH of the
cluster in which the GW is present, may also respond to
the METRIC REQUEST broadcast by the GW. The serving
gateway elects one or more CHs with the maximum weights
with respect to each of its vehicular sources. The vehicular
sources are informed of the new gateway(s) by hybrid gateway
discovery mechanism, as discussed below and the vehicles
communicate to the backhaul UMTS network using the newly-
elected GWs from then on.
In this paper, we employ a hybrid gateway discovery
mechanism: we integrate the periodic pro-active Gateway Ad-
vertisement (GWADV) and the on-demand reactive Gateway
Solicitation (GWSOL). The gateway broadcasts its GWADV
message within the cluster using TTL
cand every GWC within
this cluster gets information about the GW. In case the CH
is not the Gateway, it then broadcasts Cluster Advertisement
(CA). As stated above, the sources need to communicate to at
least one GWC within a reachable cluster to seek information
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
Fig. 2. Performance of the three protocols in terms of data packet delivery
for different numbers of vehicular sources in VANET.
Fig. 3. Performance of the three protocols in terms of control overhead for
different number of vehicular sources in VANET.
about the CH/GW. Furthermore, once it seeks information
about the GW, it should broadcast the information among
the OVs in the VANET so that each OV gets information
about the GW. Consequently, the TTL value of the source
(TTL
s) should be the maximum of these hop distance values
(i.e., maximum of hop distances between source and at least
one GWC, and between source and the first OV (OV1)inthe
VANET). It should be also noted that the nearest GWC will
be at one-hop distance from the last OV (OVn). Therefore,
TTL
sis computed as:
TTL
s=MAX(d(s,OV1)
Rs
,d(s,OVn)
Rs
+1) (2)
where, OV1and OVndenote the leading edge ordinary vehicle
and the tail one, respectively. d(s,m)denotes the distance
between a source vehicle sand a destination vehicle m.Rs
is the wireless transmission range of the source vehicle s.
IV. PERFORMANCE EVALUATION
The proposed Clustering-based Multi-metric adaptive mo-
bile Gateway Management mechanism (CMGM) is imple-
mented in the Network Simulator NS2.33 [10], using WAVE
[11] and NS-Miracle [12]. The performance of the integrated
network is evaluated in terms of Data Packet Delivery Ratio
(DPDR), Control Packet Overhead (CPO) and packet drop
fraction parameters. We used the AODV routing protocol
incorporating our proposed CMGM mechanism, and evaluated
its performance against AODV+ [14]] and DYMO [15]. For
simulation purposes, the threshold values of the metrics for
Gateway Migration are set to 25% of the initial values of the
metrics possessed by the vehicles, when they were elected as
GWs.
The graph, shown in Fig. 2, demonstrates the good per-
formance of the proposed CMGM in terms of higher DPDR,
compared to the other two protocols, and that is for different
numbers of vehicular sources in the VANET. The graph
indicates that regardless of the underlying protocol, DPDR
generally tends to decrease along with increase in the number
of sources. The curves show a negative trend: as the num-
ber of sources increases, the packet drops also subsequently
increase, especially when the gateway is on the verge of
losing its optimality. By handover, another gateway assumes
responsibility to proceed with the transactions. This explains
the good performance of CMGM. Fig. 3 shows increase in
CPO against the number of sources generating data. Though
this is generally the trend, CMGM over AODV shows less
CPO compared to the other protocols due to the fact that
only minimum number of adequate gateways are elected for
carrying on the transaction. Indeed, CMGM exhibits 12.07%
and 23.39% decrease in CPO compared to AODV+ and
DYMO, respectively.
In Fig. 4, we plot DPDR achieved by the three protocols
for different mobility speed variances of VANET vehicles.
Concerning our proposed CMGM, we consider both the case
when the selected gateway is moving towards the base station
(Positive Criterion) and when it is moving away from it
(Negative Criterion). In the figure, depending on the movement
direction of the gateway with respect to BST, our proposed
CMGM mechanism shows 18.79% and 2.96% improvement
in terms of DPDR over AODV, and 22.75% and 10.65%
improvement in DPDR over DYMO.
In the graph shown in Fig. 5, performance of CPO is eval-
uated against the IEEE 802.11p wireless transmission range
of vehicles. IEEE 802.11p transmission ranges of less than
225m may correspond to urban scenarios whereas transmission
ranges exceeding 250m may correspond to highway scenarios.
Intuitively, with short transmission ranges, many clusters of
small sizes may be formed. This leads to high CPOs as
indicated in Fig. 5. Short IEEE 802.11p transmission ranges
result also in frequent gateway handoffs and consequently loss
of in-flight packets during the handover process.
Fig. 6 emphasizes the importance of having an optimal
number of clusters. The packet drop fraction increases with
the increase in the number of clusters. This is because the
generation of control packets increases during the selection
of gateways among CHs. This may result in congestion
within the network resulting in wasteful consumption of
available bandwidth, as a result of which error messages
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
Fig. 4. Performance of the three protocols in terms of data packet delivery
for different mobility speed variances.
Fig. 5. CPO for different average IEEE 802.11p wireless transmission ranges.
are flooded within the network. AODV in CMGM shows an
improvement of 8.75% over AODV+ and 16.4% over DYMO
in integrated VANET-Internet network, as gateway handover
increases DPDR and hence, reduces the packet drop fraction.
V. CONCLUSION
In this paper, we introduced a network architecture that
integrates VANET with UMTS. To enable such an integrated
architecture, vehicles are clustered according to different met-
rics. A minimum number of adequate vehicles is selected
to serve as a liaison between VANET and UMTS. Gateway
management and selection is also performed in a dynamic
manner using different metrics.
The performance of the overall architecture was evaluated
using computer simulations and interesting results were ob-
tained. As future research direction, we would like to investi-
gate how QoS requirements can be reflected in the clustering
of vehicles and the selection of vehicle gateways.
REFERENCES
[1] S. Barghi, A. Benslimane, and C. Assi, “An Efficient Routing Protocol
for connecting Vehicular Networks to the Internet”, in Proc.10th IEEE
Int’l Symp. on a World of Wireless, Mobile and Multimedia Networks,
Kos, Greece, Jun. 2009.
Fig. 6. Packet drop fraction for different numbers of clusters.
[2] T. Taleb, A. Benslimane, and K.B. Letaif, “Towards an Effective Risk-
conscious and Collaborative Vehicular Collision Avoidance Systems”,
IEEE Transactions on Vehicular Technology, Nov. 2010. (to appear)
[3] A. Lo, J. Zhou, I. Niemegeers, “Simulation-based Analysis of TCP over
beyond 3G Cellular Multi-Hop Networks”, In Proc 17th IEEE PIMRC,
Helsinki, Finland, Sep. 2006.
[4] T. Taleb, E. Sakhaee, A. Jamalipour, K. Hashimoto, N. Kato, and Y.
Nemoto, “A Stable Routing Protocol to support ITS services in VANET
Networks”, IEEE Transactions on Vehicular Technology, Vol.56, No. 6,
pp.3337 - 3347, Nov. 2007
[5] M. Heddebaut, J. Rioult, J. P. Ghys, C. Gransart, and S. Ambellouis,
“Broadband vehicle-to-vehicle communication using an extended au-
tonomous cruise control sensor”, Meas. Sci. Technol., Vol. 16, No. 6,
pp. 1363-1373, Jun. 2005.
[6] L.E. Hodge, S.C. Ghosh, S. Hurley, R.M. Whitaker, S.M. Allen, “Cover-
age and Service Bounds for UMTS”, In Proc. International Conference
on Mobile Technology, Applications, and Systems, Yilan, Taiwan, Sep.
2008.
[7] R. Manoharan, S. Rajarajan, S. Sashtinathan, and K. Sriram, “A Novel
Multi-hop B3G Architecture for Adaptive Gateway Management in
Heterogeneous Wireless Networks,” in Proc. 5th IEEE WiMob 2009,
Marrakech, Morocco, Oct. 2009.
[8] U. Javaid, D.-E. Meddour, S.A. Mahmud, T. Ahmed, “Adaptive Dis-
tributed Gateway Discovery Scheme in Hybrid Wireless Networks”, In
Proc. IEEE WCNC, Las Vegas, NV, Mar. 2008.
[9] B. Singh, “An improved handover algorithm based on signal strength
plus distance for interoperability in mobile cellular networks”, In
Springer Netherlands Wireless Personal Communications, Vol. 43, No.
3, pp. 879 - 887, Apr. 2007
[10] K. Fall and K. Varadhan, “The NS Manual”. available at
http://wwwi.isi.edu/nsnam/ns/ns-documentation.html
[11] T. Murray, M. Cojocari, and H.Fu, “Measuring the performance of IEEE
802.11p using ns-2 Simulator for Vehicular Networks,” In Proc. IEEE
Int’l Conf. on Electro/Information Technology (EIT), Ames, IA, May
2008.
[12] N. Baldo, F. Maguolo, M. Miozzo, M. Rossi, and M. Zorzi, “ns2-
MIRACLE: a modular framework for multi-technology and cross-layer
support in network simulator 2”, In Proc. 2nd International Conference
on Performance Evaluation Methodologies and tools, Value Tools,
Nantes, France, Oct. 2007.
[13] C.E. Perkins and E.M. Royer. “Ad hoc On-Demand Distance Vector
Routing.” In Proc. 2nd IEEE Workshop on Mobile Computing Systems
and Applications, New Orleans, LA, USA, Feb. 1999.
[14] A. Hamidian, “A Study of Internet Connectivity for Mobile Ad hoc
Networks in NS2”, Masters Thesis, Faculty of Engineering, LTH at Lund
University, Feb. 2003.
[15] C. Sommer, and F. Dressler, “The DYMO Routing Protocol in VANET
Scenarios”, in Proc.66th IEEE VTC, Baltimore, MD, Sep. 2007.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.
