Content uploaded by Mahmoud Abdulwahab Alawi
Author content
All content in this area was uploaded by Mahmoud Abdulwahab Alawi on Jan 19, 2016
Content may be subject to copyright.
International Conference on Computer and Communication Engineering (ICCCE 2012), 3-5 July 2012, Kuala Lumpur, Malaysia
978-1-4673-0479-5/12/$31.00 ©2012 IEEE
CLUSTER-BASED MULT-HOP VEHICULAR COMMUNICATION
WITH MULTI-METRIC OPTIMIZATION
Mahmoud A. Alawi, Rashid. A. Saeed, Aisha A. Hassan
Department of Electrical and Computer Engineering, IIUM
Jalan Gombak, 53100, Kuala Lumpur, MALAYSIA
Abstract-Dead spot and out of coverage problem are
serious problems especially in the rural and suburban
areas where the network infrastructure is not deployed. In
this paper, a new end to end multi hop relay scheme for
vehicular communication is proposed to extend the
coverage or/and relaxing frequent handoff process. The
scheme allows vehicles to continue connected to
infrastructure network i.e. UMTS. Relay discovery, relay
selection and gateway selection are discussed and
examined. The proposed scheme can be implemented on
top of any VANET routing protocols. AODV and DSDV
were used as underlying protocol to evaluate our scheme.
An integrated simulation environment combined of
VanetMobiSim and NS2 is used to simulate our proposed
scheme. The simplified gateway selection shows the best
performance on top of AODV compare to DSDV.
Keywords: VANET, heterogeneous network, UMTS, multi
hop relay scheme
I. INTRODUCTION
Heterogeneous wireless network is the one of the hot
research topics in wireless domain. It supports the
wireless devices with the capability to interoperate in
more than one wireless technology seamlessly. A good
example is the combination of Universal Mobile
Telecommunication System (UMTS) network and
IEEE802.11 based vehicular network (VANET) to form
more robust network in terms of speed and coverage area
[1]. VANET possesses many unsolved problems which
hinder the promising of intelligent transportation system
(ITS) to bring the safety and entertainment in
transportation system. Firstly, frequency network
disconnection caused by high speed of the vehicles
makes the inter connection time between V2V and V2I
to be very small due to the unpredictable motion of
vehicles which is difficult to predict when the vehicles
will change their speed and cause unexpected network
disconnection. Secondly is the limitation of the network
coverage especially in rural and suburban areas whereby
the infrastructure network is not well deployed. Lastly,
VANET signaling uses broadcast mechanism to
disseminate the data and control signals, this easily
causes flood to the network which results to extra
overhead and create network scalability problems[2] [3].
In [2][4][6] problems of establishment and
maintaining a connection to the Internet were discussed
and new protocols were proposed. The protocols use the
lifetime prediction of the link to select the best route to
the gateway. The longer life time the better the route in
the sense that, it reduce frequent link breakage which
have high impact on Quality of Service (QoS). A multi
hop relay mechanism to connect VANET to the
Infrastructure network was studied in [4]. The work
involves experimental evaluation of using relay vehicle
when the source vehicle moves towards the base station
(Node B) and when it moves away from the base station.
In [1], Clustering- based Multi-metric adaptive
mobile Gateway Management mechanism (CMGM) was
proposed. CMGM allows minimum number of vehicles
to connect to UMTS network by allowing normal
vehicles to use the GWC to communicate with the
UMTS network. Clustering mechanism was used to
cluster the gateway candidates, which depends on
direction of movement, UMTS-RSS (Received Signal
Strength), and IEEE802.11p transmission ranges. GWCs
closed to the cluster center are always selected as Cluster
head. Normal vehicles use cluster heads to communicate
to UMTS network. Three metrics are used for mobile
gateway selection; mobility speed, UMTS RSS and link
stability. However, this algorithm is complex in terms of
cost of creating and maintaining the clusters for rapid
environment as in VANET. In addition to that, the
signaling traffic and time that is used for clustering
formation and cluster head selection is larger than the
time of data traffic exchange.
In this paper we used hybrid metrics of the link life
time to determine the stability of the link between the
source vehicle and the relay vehicle along with best
route capacity selection to avoid overload and hence
bottleneck in one gateway. The remainder of this paper
is organized as follows; section II reviews the related
work. Section III introduces the simplified gateway
selection through multi hop scheme, while section IV
deals with the simulation and performance evaluation of
the proposed scheme and section V concludes the paper.
In our work we proposed a simplified mechanism to
formulate a cluster and gateway selection by formulating
clusters only at the UMTS cell edges. This done by
using received signal strength (RSS) threshold as a
cluster formation indicator, in addition to use the best
available route capacity for gateway selection.
22
II. SIMPLIFIED GATEWAY SELECTION (SGS)
SCHEME
Here we proposed simplified gateway selection
(SGS) to enable vehicles to continue connecting to
Internet or other mobile service (i.e. location map
download). Also we assumed that each vehicle has two
interfaces one to connect to infrastructure mode which is
UMTS UTRAN interface and another one is to connect
to the ad-hoc mode (IEEE802.11), both of these
interfaces must be active all the time. Our system model
shown in Figure 2, involves vehicles moving on the
highway into two opposite directions. All vehicles
access the UMTS network via UMTS Node B over the
UMTS radio interface (UTRAN) [1][5]. The vehicles
will experience different signal strength (RSS) from
Node B. All vehicles connect directly to Node B if they
are inside of the coverage (using UMTS UTRAN
interface) and once UMTS RSS fall to certain predefined
threshold the vehicle will use it IEEE802.11 interface to
find the partner to act as relay to connect to the UMTS
BST.
Fig 2: System model and scenario
A. Scheme overview
All the vehicles in the UMTS coverage zone use
UMTS UTRAN interface to connect to UMTS network.
When the UMTS receive signal strength (RSS) degrades
below the threshold it means that the vehicle is nearly to
move outside the UMTS cell area to handoff to other cell
or enter to the dead spot zone.
When vehicle receive RSS less than the RSS
threshold, the vehicle send broadcast message using
IEEE802.11 interface to search for relay to connect to
UMTS network. The source vehicle starts to broadcast
Relay request message (RReq) asking for the vehicle
which its UMTS-RSS is greater than the predefined
threshold (γ). All intermediate vehicles (IV) upon the
receiving of the RReq compute their received signal
strength and compare to the one sent in RReq, if
received signal strength is greater than the one
mentioned in RReq, the vehicle identified itself as one of
the relay candidates and unicast the relay respond
message (RRep) to the source vehicle. The source
vehicle can receive numbers of RRep but choose the best
one according to our proposed algorithm of selecting the
best relay (in section D). When this relay loses its
optimality, it will apply the same algorithm to select
another relay and continue to help the source vehicles to
forward their packets otherwise it will notify the source
vehicle the need of performing the handoff to new relay
and the source vehicle will broadcast again the RReq to
discover other relay vehicles. Figure 3 shows the flow
chart of our proposed scheme.
Fig. 3. Flow chart of the relay selection algorithm
B. Relay Discovery
After detecting the need of finding the relay, the
source vehicle (SV) uses IEEE802.11 interface to
broadcast the relay request packets (RReq). To mitigate
the broadcast storm, RReq will include time to live
(TTL) value that specifies the number of hops to search
for the relay that can assist the source vehicle (SV) to
connect to the UMTS network. Upon receiving RREq,
the intermediate vehicle will employ algorithm1 to
decide if it is the right candidate to be the relay. When
the intermediate vehicle receives RReq, it checks the
TTL value and sequence number (SeqNo) of the received
23
RRep, if the TTL value is greater than one (TTL>1) and
the sequence number (SeqNo) is greater than any
prevision RReq packet the intermediate vehicle will
compare its UMTS-RSS with one specified in RReq,
otherwise it will discard the packet.
If the UMTS-RSS of the intermediate vehicle is
smaller than the RSS threshold () it will reduce TTL
value by one and forward the packet to all its neighbors
except from source vehicle (SV) of the RReq. When the
vehicle found that its UMTS-RSS is greater then, it will
create the relay respond packet (RRep) and send back to
the SV using the same chain of nodes specified in RReq
packet. The SV selects the relay based on the certain
policy and starts to handoff vertically and connects to
the UMTS network in relaying mode.
C. Relay Selection parameters
Most of the prevision works are mainly focused on
the number of hops and route life time (RLT) as the
main parameters for optimal relay or gateway selection
[2][4][6], however using only these parameters will not
guarantee the best selection of the relay or gateway. In
this paper, multi metric relay selection algorithm is
proposed which is based on the three metrics, UMTS
received signal strength (RSS), route life time (RLT) and
available relay capacity (λ).
1. UMTS RSS
In VANET, the nodes are moving in high speed
which causes frequent change of the UMTS received
signal strength. Normally, when the vehicle moves
towards the Node Bits RSS is increasing and if it is
moving away from the Node Bits RSS is degraded.
However, the faster a vehicle moves towards or away
Node B, the faster will be the increase or decrease in its
UMTS RSS.
The UMTS of the vehicles at the time is
computed as in [1].Suppose the velocity of the vehicle at
time is and at the time −1 its velocity was
so the current is calculated using equation 1,
which will be equal to the prevision RSS which is
plus the function that denote the variation of
the UMTS receive signal strength i.e. 1+ ||
∞.
To compute the UMTS RSS when the vehicles moves
toward or a way of the BST Node B we use equation 1.
Where ∞ is constant that defines the rate of variation of
UMTS signal strength
= ±1+||
∞ (1)
2. Route Lifetime
Rapid changes of the nodes position in VANET
results frequent route failure [1][2][5]. Thus, it is very
important to know the route lifetime (RLT) in order to
select the best route which can stay longer without
break. Hence, good prediction of the route lifetime can
significantly reduce the repairing of existing route or
reconstructing of the new route. We adopted the route
lifetime prediction algorithm proposed in [2].
When the RRep packet is created, the Relay vehicle will
insert its location as well as its velocity, and set the route
life time to infinity (RLT=∞). Again, when the
intermediate node receives the RRep, it extracts the
information from the packet (location and velocity) and
calculates the link life time (LLT) by using;
= ||
|| (2)
Where if R is the range of IEEE802.11 radio, is the
absolute distance between adjacent nodes and and
is the velocity of the adjacent nodes.
If the value is smaller than the route life time
mentioned in the RRep, the life time field in the packet
is changed to the new value otherwise the current value
is kept unchanged. In this way, the route lifetime value
continues to change until the RRep packet reaches the
SV. RRep packet contains the route predicted lifetime
(RLT) which is the minimum life time of all the links
(equation 3). Figure 4, shows the process of route link
life time prediction
RLT= min {∑
} (3)
Fig. 4. Shows the process of predicting the route life time
3. Available route Capacity ()
In the relay scenario, if the selected relay serves a
number of vehicles more than its capability, it can result
in higher packet loss and delay due to the congestion on
that selected relay. Also in multi hop scenario, a multiple
24
route can have some vehicles in common in the route
from the source vehicle to relay or gateway vehicles.
This Also causes bottleneck problem on that route or on
the relay node buffer. To overcome these problems, the
source vehicle (SV) selects the route based on the
available route capacity .
Available route capacity can be defined as the
minimum available load capacity at any node [7][5],
including intermediate nodes and the gateway node.
During the relay discovery phase, when the relay vehicle
receives the RReq, before replies with the RRep
message, it computes its available capacity. Suppose the
maximum load capacity of a node is and the
current traffic load handled by is then the
available load capacity of the node is at , it is
computed as follows :
= −− (4)
The current traffic load ( ) can be calculated using
equation 5, where is the current traffic load on the
node that is relaying traffic from traffic sources and
P and denote the average packet arrival rate and
average packet size of the traffic from source and b is
back off value which protects the vehicles from being
overloaded.
=∑
(5)
The overall residual load capacity of route i is
computed as in equation 6 where j is vehicles in a route
including relay vehicle.
= min {∑
} (6)
D. Multi Metrics Relay Selection Algorithm
As we have mentioned in the searching phase in
section 2.4, the source vehicle will receive a number of
RRep from different relays which fulfill the condition
mentioned in the request message (RSS>). Upon
receiving the number of RRep, the source vehicle
applied our multi metric relay selection algorithm which
is based on Simple Additive Weighting (SAW)
technique to select the best relay.
E. Relay Handoff Mechanism
In SGS scheme, when the relay vehicle losses its
optimality (UMTS-RSS< or RLT<), the source
vehicle has to handoff its connection from one relay to
another. The Relay vehicle will assist the source vehicle
to perform the handoff if its RLT with source vehicle is
above the Threshold. The Relay vehicle will start to find
another best relay and forward the incoming data to new
relay and notify the source vehicle about the new relay.
But when the route life time between the relay and
source vehicle is near to finish the relay vehicle, it will
notify the source vehicle on the need for performing the
handoff process.
The source vehicle will start to search for the new
relay .When the source vehicle succeeds to find another
best relay it will perform handoff to new relay and
notifies the old relay in order to release the resource
allocated for that source vehicle.
Begin Multi Metric Relay Selection Algorithm
A source vehicle broadcast the RReq using 802.11 interface
asking for the vehicle UMTS-RSS > .
When the RReq message receiving by a vehicle
If UMTS-RSS [VEHICLE]> then
send the RRep contain three metrics(UMTS-
RSS, RLT , )
Ignore duplication RReq from the same source.
else
Continue to broadcast the message to the
neighbor vehicles.
end if
When source vehicle Receive RRep
Calculate the scale metric Si
for each metric Mi of the P/R, where 1<i<3 do
If Mi [CRITERION] is POSITIVE then
=−
−
else
if Mi [CRITERION] is NEGATIGE then
= −
−
end if
end if
end for
The source vehicle calculates the weight of each Relay by
/ =∑(
[]∗ )
The source determines the Relay with the maximum weight
and selects it as the GATEWAY.
End of Multi metric relay selection algorithm
III. PERFORMANCE EVALUATION
In this section effectiveness of SGS scheme is
evaluated by using the combination of VanetMobiSim
[8] and NS2 [9]. VanetMobiSim is used to generate the
vehicular mobility model to be used in evaluating SGS
scheme while NS2 is used to measure the network
performance of SGS scheme. The vehicle mobility
model generated by VanetMobiSim is passed to NS2 to
define the node movement and speed. SGS scheme can
be implemented in any VANET routing protocol, we
compare SGS scheme performance on top of reactive
(i.e. Ad-hoc on demand routing protocol (AODV)[10])
and proactive (i.e. Destination sequence distance vector
(DSDV)[11]) routing protocols .In addition to that pure
AODV protocol and AODV with SGS were also
studied their performance VANET-UMTS scenario.
25
A. Simulation Environment
In this research NS2.33 version was used, patched with
NSMIRICLE [12] to support multiple interfaces in
wireless node. The mobility file obtained from the
VanetMobiSim was used in NS2 to define the node
motion during the simulation. Our simulation involves
one Base station (Node B) which is connected to the
UMTS network and the number of vehicles which use
that Node B to connect to the UMTS network as
illustrated in Figure 2. Table I and Table II shows the
parameter used to configure VanetMobiSim and NS2
respectively.
TABLE I
VANETMOBISIM MOBILITY GENERATOR PARAMETERS
Mobility Feature
Parameters/Value
Type of road
Highway
Number of lanes
2 lanes
Maximum speed
Varies from 20 to 160km/hr
Number of vehicle
Varies between 10 to 50
Simulation period
500s
Stop time
1 to 3s
TABLE II
NS2 SIMULATION PARAMETERS
Parameter
Value
Simulation Area
9000 * 1500 ( )
Channel
Channel/WirelessChannel
Radio-propagation model
Propagation/TwoRayGround
Network Interface
Phy/WirelessPhyExt
MAC interface
Mac/802 11Ext
Routing protocol
AODV/DSDV
Interface queue
Antenna/.OmniAntenna
UMTS-RSS Threshold
-94 dBM
Base station Wireless
transmission range
8km
Application
FTP
Packet size
1000B
Initial TTL value
10
B. Simulation Result
In Figure 7, the packet delivery ratio of SGS scheme
implemented on top of AODV is higher compared to the
one implemented in DSDV. For lower speed of vehicles
AODV and DSDV show closed packet delivery ratio
but when the speed of the vehicles increase the packet
delivery ratio of SGS scheme on top of AODV
outperform its counterpart. It can be observed that the
above elaboration is true because DSDV is designed for
lower mobility speed and that is the reason, when the
mobility of the vehicles increases, the packet drop for
DSDV grows up which lead the packet delivery ratio to
become lower. It has been shown from the graph that,
the packet delivery ratio of SGS scheme on top of
AODV is almost 21% higher than the packet delivery
ratio of SGS scheme on top of DSDV. Therefore, it can
be concluded that SGS scheme performed better in terms
of packet delivery ratio if we are implementing on top of
AODV rather than DSDV.
Figure 8 indicates that the throughput of SGS scheme
shows negative trends, when the vehicle speed increases
the throughput goes down. In this respect, when the
speed of the vehicles reaches above 100km/hr, the
throughput decreased rapidly for both AODV and
DSVD as underlying protocols. However, it can be
observed that the throughput of SGS scheme on top of
AODV is higher compared to SGS scheme implemented
on top of DSDV. The results in Figures 7 and 8 indicate
that SGS performed well when implemented on top of
AODV rather than DSDV. For that reason, AODV
protocol is selected to be underlying protocol of SGS
scheme.
Fig. 7. Packet delivery ratio of SGS scheme on top of Reactive
and proactive routing protocol with variation of mobility
speed of vehicles.
In Figure 9 we compare the packet delivery ratio of
AODV with SGS scheme and pure AODV (AODV
without SGS scheme) with different vehicles mobility
speed. It can be seen that, the packet delivery ratio of
AODV with SGS is higher compared to the packet
delivery of AODV without SGS scheme.
Fig. 8. Throughput of SGS scheme on top of Reactive and
Proactive routing protocol with variation of mobility speed of
the vehicles.
AODV with SGS shows insignificant decrease in
packet delivery ratio for the vehicles mobility speed
below 120km/hr. However, when the speed of the
vehicles increased above 120km/hr, there is noticeable
decrease in packet delivery ratio. Above this speed,
frequency migration of one relay to another by source
vehicles is likely to occur. In contrast, pure AODV
26
cannot sustain itself in this highly dynamic environment
because as it is shown in the graph its packet delivery
ratio goes down continuously when speed of the vehicles
goes up.
The variation of throughput against the vehicle
mobility speed of AODV with SGS and the AODV
without SGS is shown in Figure 10.AODV with SGS
shows an improvement of 31.7% of throughput
compared to AODV without SGS. Pure AODV using
pure flooding mechanism for the gateway discovery
which results in high number of routing traffic packet to
be sent when the vehicles speed is high and fewer
channels to be used for send data packet, the
consequence of this is the falling of throughput of the
network in pure AODV scheme. In contrast, AODV with
SGS improve the flooding mechanism used in pure
AODV by restricting the flooding area using incremental
TTL value.
Fig. 9. Packet delivery ratio of AODV with SGS scheme and
AODV without SGS scheme with variation in the mobility
speed of VANET vehicles
Fig. 10. Throughput of AODV with SGS scheme and AODV
without SGS scheme with variation in the mobility speed of
VANET vehicles.
IV. CONCLUSION
VANET–UMTS integration is the promising
architecture in the future. Vehicles may contain more
than one network interfaces which makes possible for
the vehicles to connect to either infrastructure mode or
ad hoc mode. This will help the Vehicles enjoy steady
Internet connection through UMTS network and at the
same time can communicate with other vehicles for
safety or entertainment information. In this paper, new
simplified scheme (SGS) is proposed to reduce dead spot
and increase the network coverage area in integrated
vehicle to UMTS network. The proposed SGS scheme
can be implemented below any VANET routing protocol
i.e. AODV and DSDV. In our future work, SGS scheme
will be compared with clustering based scheme proposed
in VANET-UMTS integration.
References
[1] Benslimane A., Talib T. and Sivaraj R., Dynamic
Clustering Adaptive Mobile Gateway Management in
integrated VANET-3G Heterogeneous Wireless Network,
IEEE Journal on Selected Areas in Communication vol
29. No. 3 2011.
[2] Vinod Namboodiri and Lixin Gao, “Prediction-based
routingfor vehicular ad hoc networks,”in Vehicular
Technology, IEEE Transactions on. Jul 2007, vol. 56(4),2,
pp.2332–2345.
[3] Fatma Hrizi and Fethi Filali, “On congestion-Aware
Broadcasting in V2X Networks.” EURECOM Mobile
Communication Department 2009.
[4] S.Bargi A., Benslimane and C.Assi., “A Lifetime based
routing for connect VANETs to the Internet” In Proc.of
IEEE International Symposium on World of
Wireless,pages 1-9,June 2009.J. Zhao, T. Arnold, Y.
Zhang, G. Cao, “Extending DriveThru Access byVehicle-
to-Vehicle Relay,” ACM VANET, 2008
[5] Mahmoud A. Alawi, Rashid A. Saeed, Aisha A. Hassan,
Othman Khalifa, “Simplified Multiple Metrics Gateway
Selection Mechanism For VANET To Infrastructure
Wireless Network”, The 6th international conference
SETIT 2012: Sciences of Electronic, Technologies of
Information and Telecommunications (SETIT2012), April
2012, Tunisia.
[6] S. Wan, J. Tang, and R. S.Wolff, “Reliable routing for
road side to vehicle Communications in rural areas,” in
Proceedings of IEEE International Conference on
Communications, 2008, pp.3017- 3021
[7] Safda H. and Sasase I, “Multiple End- to-End QoS Metrics
Gateway Selection Scheme in Mobile Ad hoc Networks,”
International Conference on Emerging Technologies 2009.
[8] Haerri J, Fiore M, Fethi F, Bonnet C. “VanetMobiSim:
generating realistic mobility patterns for VANETs”,
Institute Eurécom and Politecnico Di Torino, 2006.
Available at: http://vanet.eurecom.fr/
[9] The network simulator- ns2, http://www.isi.edu/nsnam/ns/
[10] C. Perkins and E. Royer. ”Ad hoc On-Demand
Distance Vector Routing.” In Proc. 2nd IEEE Workshop
on Mobile Computing Systems and Applications, USA,
Feb. 1999.
[11] C.E. Perkins and P. Bhagwat, “Highly dynamic
destination-sequenced distance vector routing (DSDV)
for mobile computers,” in Proc. ACM SIGCOMM 94,
London, UK, pp. 234-244 , Oct. 1994.
[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, France, Oct. 2007.
27