Content uploaded by M. Arif Khan
Author content
All content in this area was uploaded by M. Arif Khan on Jan 12, 2015
Content may be subject to copyright.
A multi-hop cross layer decision based routing for VANETs
Sabih ur Rehman •M. Arif Khan •Tanveer A. Zia
ÓSpringer Science+Business Media New York 2014
Abstract In recent years, vehicular ad-hoc networks have
emerged as a key wireless technology offering countless
new services and applications for the transport community.
Along with many interesting and useful applications, there
have been a number of design challenges to create an
efficient and reliable routing scheme. A conventional
design approach only optimizes routing schemes without
considering the constraints from other network layers. This
may result in an under-performing routing mechanism. In
this paper we present the design of a multi-hop cross-layer
routing scheme that utilises beaconing information at the
physical layer as well as queue buffer information at
medium access control layer to optimise routing objectives.
In particular, the proposed scheme integrates channel
quality information and queuing information from other
layers to transmit data. Using simulations as well as ana-
lytical studies we have presented results of our proposed
scheme and have done a thorough comparison with exist-
ing approaches in this area. The results highlight better
performance of the proposed cross-layer structure as
compared to other conventional single layer approaches.
Keywords VANETs Cross-layer design Channel
quality Routing Queuing information Multi-hop
wireless networks
1 Introduction
Vehicular ad-hoc networks (VANETs) have now become a
backbone for intelligent transport systems (ITS) by pro-
viding many life saving and comfort related applications
and services. ITS mainly focuses on establishing the
deployment of advanced wireless technologies such as
VANET in order to provide safety critical and emergency
related applications [1]. It is therefore important to make
sure that wireless communication networks responsible for
dissemination of such information, are reliable, robust and
efficient. In order to achieve these goals, it is essential to
design a vehicular network that must be reliable under the
odd circumstances where they are needed most. Many
efforts were made to optimise the network layers individ-
ually and are presented as the attractive solutions [2–4].
However, with the emergence of complex applications and
innovative physical communication structures, these pro-
posed solutions may not remain as optimal as promised.
Furthermore, many applications impose stringent QoS
requirements which may not be met by the existing con-
ventional VANET design solutions.
Routing in VANET is different to the traditional MA-
NET routing because of highly dynamic and ever changing
topologies in the former. Few protocols such as
DYMO [5], DSR [6] and AODV [7] that were earlier
designed for MANET environment have been tested on
VANET as well [8,9]. The challenge however remains as
how to reduce the delay associated with passing the
information from one node to another [10]. Most of the
routing protocols in VANET are closely linked with the
topology being used in the network architecture and the
performance deviates whenever there is a change in net-
work topology. Routing in VANET can be classified into
five major categories namely as Ad-hoc Protocols,
S. ur Rehman (&)M. A. Khan T. A. Zia
School of Computing and Mathematics, Charles Sturt
University, Wagga Wagga, Australia
e-mail: sarehman@csu.edu.au
M. A. Khan
e-mail: mkhan@csu.edu.au
T. A. Zia
e-mail: tzia@csu.edu.au
123
Wireless Netw
DOI 10.1007/s11276-014-0874-z
Location Based Routing Protocols, Cluster Based Proto-
cols, Broadcast Protocols and Geocast Protocols [11].
However, routing in each of these categories optimises
parameters such as end-to-end delay and packet delivery
ratio without considering that whether a wireless channel
can support the transmission or a particular node has suf-
ficient space in its buffer to store the packet for the duration
of processing time. While not considering these parame-
ters, the source node may face retransmissions request from
other nodes making the network congested [12].
This paper has threefold contributions. In first contri-
bution, we propose a cross-layer paradigm for efficient
routing in VANET. The detailed discussion on this pro-
posed approach and its parameters is given in Sect. 4.In
the second contribution, we propose an algorithm to actu-
ally create the routing mechanism based on the cross layer
approach as discussed earlier. The third contribution of the
paper is to probabilistically characterise the nodes inclu-
sion and exclusions within the transmission range of the
source node. In order to verify the performance of proposed
algorithm, we run a number of different simulations sce-
narios. The main algorithm is then compared with two well
known routing algorithms for VANET, GPSR [12] and
PROMPT [13]. It is clear from the results that the proposed
algorithm has better performance in each scenario simu-
lated. The rest of the paper is organised as follows: in
Sect. 3, we describe in detail VANET system model
including the wireless channel and queuing models which
are utilised in the proposed scheme. In Sect. 2, we provide
some related work. In Sect. 4, we provide detailed
description and analysis of the proposed cross-layer
approach. Section 5presents numerical analysis of the
proposed algorithm under different simulation scenarios
and compare the results with other known routing algo-
rithms. Finally Sect. 6, concludes the paper.
2 Related work
In VANET environment a lot of research has been focused
towards designing a robust routing scheme [14–16]. One of
the key challenges within the design of robust routing
scheme for VANET is to control the dynamic topology of
associated nodes. This is a common issue within an ever
changing topology structure environment such as wireless
sensor networks as highlighted by Li et al. [17]. To tackle
topological changes in the network, authors in [18] have
addressed the problem of channel assignment within
stringent wireless networks that can cause network parti-
tion and in turn link failures. Authors have shown that their
approach can be extended to the case of uneven traffic load
within the network, a scenario common in vehicular
communication environment. Initially routing protocols
that were developed for mobile ad-hoc networks (MA-
NETs) were applied to VANETs conditions and were
considered to have shown reasonable results. GPSR [12]is
one of the well known routing schemes that was initially
designed for MANETs. In GPSR greedy forwarding
approach is used to forward the packet from source to
destination. Routing strategies in stringent networks like
VANETs can give optimum results by utilizing a multi-
layer design approach as opposed to the traditional single
layer architecture. Categorically due to the nature of
communication involved in VANETs, these network are
also classified as special group of networks known as delay
tolerant networks (DTNs) [19]. A good literature in rela-
tion to this is provided in [20]. In this paper, authors have
highlighted some key issues in the design of a robust
communication delivery system and have identified a set of
useful guidelines to tackle those issues in such an envi-
ronment. Considering these design aspects, the authors
in [21] have presented a directional routing and scheduling
scheme for VANET that uses well known optimization
techniques based on constraints of other multi-layer met-
rics such as delay and bandwidth. Using extensive simu-
lations, authors have presented a complete performance
analysis of proposed approach that shows better perfor-
mance as compared to other known approaches. The
challenge of information dissemination within a multi-hop
vehicular network can also resemble a typical structure of
vehicle routing problem (VRP) that has received lot of
attention from researchers. VRP aims to achieve informa-
tion delivery routes with minimum cost within the entire
network. This problem has also been discussed in [22]
where the authors present an efficient protocol structure to
achieve an increase in the network lifetime without vio-
lating packet delay constraints. A similar approach has also
been presented by Li et al. [23], where authors have pro-
posed a reliable multicast protocol to address various per-
formance constraints within the wireless networks. Using
simulations, the authors have compared their proposed
approach with well known protocols in this area. The
results show a significantly better performance as com-
pared to its predecessors. Another novel approach in rela-
tion to data dissemination in a multi-hop wireless
environment has been adopted by authors in [24]. In this
scheme, authors have proposed a prediction based data
collection method at each node using double queue
mechanism. The authors have shown by experimental
results that the proposed methodology successfully reduces
redundant communication by synchronizing predicted data
at both ends of communication.
Cross-layer based solutions to optimize the routing
decision have been tried and tested in wireless sensor
environment especially to achieve a range of QoS met-
rics [26,27]. A good literature summarizing the challenges
Wireless Netw
123
faced in designing a state-of-the art routing metrics for
wireless networks in presented in [28]. Understanding
these issues, authors in [29] have presented a design to
improve energy efficiency in a wireless network by
reducing routing re-computations. Using simulations
experiments, it has been shown that the proposed approach
is effective in reducing routing overheads. Another detailed
literature highlighting one of the key challenges in the
design of a generic mobile networks is articulated by Wang
et al. [30] as the availability of limited energy resource
within mobile nodes. A similar method has been proposed
by authors in [25] using a data aggregation method. Using
compressed sensing procedure that forecasts a complete
recovery of signal at the receiver based on prediction
techniques has been adopted in this approach. Authors have
tackled the problem of energy consumption in wireless
networks through compressed data aggregation and have
shown an improvement in results using their proposed
method. However, energy conservation does not pose to be
an issue within the vehicular networks due to large physical
size and structure of the vehicle nodes.
A detailed survey of cross-layer design in VANETs has
also been presented in [31]. The paper has shown four dif-
ferent cross-layer optimisation approaches namely M1, M2,
M3 and M4. The authors have discussed cross-layer design
challenges such as requirement analysis, VANET specific
constraints and PHY–MAC layer constraints in VANETs.
In [32], authors have presented a cross-layer design approach
for ad-hoc networks under Rayleigh Fading channel condi-
tions using physical layer and medium access control
parameters together. They have concluded through analytical
and simulation studies that the cross-layer approach enhances
system optimality in terms of throughput, control overheads
and packet drop ratio. It is shown that without considering
cross layer approach, the system requires a large number of
control packets and has a higher packet drop ratio and poor
system performance. The authors have also shown that with
cross-layer approach in ad-hoc networks, system performance
can be increased and certain guaranteed QoS requirements
can also be met. Yang et al. [33] have analysed a distributed
cross layer optimisation scheme for multi-hop wireless net-
works where different nodes can cooperate with each other.
They used network utility maximisation technique to optimise
flow control, routing, scheduling and relay assignment for
multi-hop wireless cooperative networks. They have shown
that their proposed cross-layer and graph theoretical approach
gives advantages in terms of convergence, system throughput
and the performance of scheduling algorithm. In [34], authors
have presented a stochastic characterization of information
propagation in vehicular ad-hoc networks in a highway sce-
nario. Authors have characterized traffic into a number of
traffic streams where vehicles in the same stream have the
same speed distribution while the speed distribution in
different stream is varied. The paper analytically studies the
expected propagation speed of vehicles and has shown that the
information propagation speed can be significantly enhanced
by exploiting the existence of even a small number of vehicles
travelling with a different speed or in opposite direction. A
position based routing protocol using a cross layer optimisa-
tion concept for VANETs is proposed by authors in [35].
They have used weighted SINR and MAC frame error rate to
improve the efficiency of the proposed routing protocol using
two-ray wireless channel model. They have compared the
performance of their algorithm with GPSR and found that the
proposed CLWPR algorithm has better performance in
comparison. In [36], authors have attempted to design a cross
layer routing strategy for vehicular communication by mod-
ifying the well known ad-hoc on-demand multipath distance
vector routing (AOMDV) protocol. The authors have com-
pared existing AOMDV protocol with their proposed routing
protocol in which decisions are based using retransmission
counts metric. Simulation results show an improved perfor-
mance for the proposed protocol for both the sparseand dense
VANET environments. In [37], authors present detailed
architecture of a cross layer based routing scheme for coop-
erative VANETs. The authors present a cross layer solution
that makes routing decisions based on link capacity and
adjusting the probability of connectivity at MAC layer
accordingly. The implemented architecture also looks into
some key metrics such as fairness and cost associated with
each communication point. Using analytical and simulated
results, authors have put forward a complete architecture for a
VANET environment. Another cross-layer design approach
for routing in VANET has been presented in [13]. Using
extensive simulation results, authors have presented a prom-
ising solution of a delay aware routing protocol that utilizes
communication path information using beacon messaging in
vehicular networks.
3 System model
Let us describe the complete system model used in the
proposed algorithm. The algorithm takes information from
physical layer as well as from MAC layer forcing us to
define a system model at both the physical and MAC
layers. We will define a generic VANET mobility model,
wireless channel model and queuing model for the entire
system.
3.1 VANET model
We consider a VANET model having Vvehicles such as
v¼1;...;Vand moving with a certain velocity uon a
stretch of road which resembles a freeway/highway model
as shown in Fig. 1. We assume a two-way traffic where
Wireless Netw
123
each direction of the road has three lanes for traffic. The
vehicles are statistically deployed at the beginning fol-
lowing a homogeneous Poisson distribution with density q.
After this, the vehicles follow a Random Walk mobility
model which means that vehicles can vary their spatial
location(lane) as well as their speed. As shown in Fig. 1,
source vehicle smoving rightward wants to communicate
with destination vehicle dmoving in left (opposite) direc-
tion. The first step for sis to send a beaconing message for
all vehicles in its communication range. The wireless
channel from sto each vehicle vis represented by hvas
shown in the figure. Each node in the transmission range of
scalculates its dand sends it back to s. The node sthen
decides either to choose the dor rnodes. If rnode is
selected, the similar process is repeated at every rnode
until the destination node dis reached. The overall objec-
tive of this model is that a source node scan transmit its
packets to destination node dsuccessfully either using a
direct communication or multi-hop propagation process.
3.2 Wireless channel model
Let us consider the wireless propagation channel model for
the communication between different nodes. We assume that
each node has a single antenna for both transmitting and
receiving purpose. This models our wireless link between
two vehicles as a single input single output (SISO) wireless
channel. A more sophisticated and advanced antenna system
such as multiple input and multiple output (MIMO) can also
be used at the vehicle transceiver for better transmission and
reception of signal [38]. However, this makes the overall
system model more complicated, therefore, in this paper, we
will assume only SISO link between different nodes. We
consider a time slotted system where a transmission occurs
during a particular time interval called a time slot. We denote
the wireless channel by hvrepresenting the SISO link
between nodes sand vwhere v¼1;...;V. This wireless
channel hvfollows Nakagami distribution model
hvNakagamiðm;XÞwith mean value of mand variance X.
This model gives more realistic VANET physical RF
channel modeling compared to Two-Ray or Shadowing
models [39,40]. The probability density function (pdf) of hv
describing the wireless channel characteristics can be given
as chapter 5 p. 102 of [41].
fhvðhv;m;XÞ¼ mm
CðmÞXmhm1
vexp m
Xhv
;ð1Þ
where Cis the Gamma function,mdenotes the Nakagami
distributionm-parameter and Xrepresents the average
channel gain.
Let us consider that source node sbroadcasts a beaconing
signal which is received by all nodes within the transmission
range Rsof the source node. Each node receiving the beac-
oning signal from scan send three different types of channel
information back to the source node. First and the simplest
feedback that a node can transmit back to the sis its channel
strength denoted by the channel norm, i.e. khvk2. For this
channel information, the node is not required to know about
the channel information of other nodes. The second channel
information that a node can transmit back to sis the signal to
noise ratio (SNR) given as
SNRv¼hvPv
nv
:ð2Þ
The third and more comprehensive feedback can be
obtained using SINR and this is the approach we have
adopted in this paper. Each node calculates its Signal to
Fig. 1 VANET model used in
this paper
Wireless Netw
123
Interference Plus Noise Ratio (SINR) and sends it back to
the source node through an error free and non-delay
feedback channel. This SINR is calculated as follows:
SINRv¼hvPv
nvþPV
j¼1;j6¼vhjPj
;ð3Þ
where Pvis the transmit power and nvis the noise of node v
respectively. The noise is assumed to be additive white
Gaussian (AWGN). It is also assumed that each node has a
perfect knowledge of its own and neighbouring nodes
wireless channels. The wireless channel is assumed to be
independent and identically distributed (i.i.d) for each time
slot. Once sknows d, it can make its routing decision based
on this information. The SINR of each neighbour in fCNg
is d¼½d1;d2;...;dV. We also assume that each node can
transmit with a certain allowed transmit power following
the total transmit power constraint given as PV
v¼1PvP,
where P is the total allowed transmit power. As mentioned
by Agarwal and Kumar [42], another important metric to
understand and include in VANET routing decisions is the
wireless channel capacity, denoted here by Cvthat is
defined as:
Cv¼log2ð1þSINRvÞbps=Hz:ð4Þ
This capacity is defined as the actual data rate that a node
can transmit packets with. It is important to know this
supported data rate in the routing scheme to avoid a
transmission failure and hence reducing the control
overheads.
3.3 Queuing model
The routing algorithm proposed in this paper uses channel
and queuing information together to make a better routing
decision. For this purpose, we consider that each node has a
buffer with the maximum queue capacity q0and it can store
the packet for a very short period of time known as per-hop
delay and denoted by b. During this time the node decides
where to transmit the next packet. As per the available
technology the per-hop delay b, has a typical value of
4ms[34]. In order to avoid the packet drop and to increase
the overall packet drop ratio, the node should have space in
its buffer to store at least one packet for the duration of b.If
the transmitting node doesn’t have this information and it
sends a packet to the next node while next node doesn’t
have space in its buffer to store the packet, then in this
scenario the next node will discard that packet. This will
also result in the transmission failure and hence increase
the retransmission in the network. we assume that each
node has the perfect knowledge of its buffer capacity and
buffer state at particular time slot and sends this informa-
tion back to the source node with the d. This queuing
information of node vis called as qv. For the sake of
brevity and to avoid complex queuing analysis, we assume
that each node uses simple FIFO mechanism to store and
take the packets from the queue.
4 Proposed cross-layer approach
The main objective of our design is to provide an optimal
routing scheme for VANET that meets all the design
requirements and QoS variables by considering the infor-
mation from other layers that will effect the design of
routing scheme. The proposed cross-layer design paradigm
for vehicular ad-hoc network is pictorially explained in
Fig. 2. In traditional OSI layered architecture, each layer is
optimised by using its own set of variables. Whereas in the
cross-layer approach, Fig. 2(a), considering the initial
parameters from the physical layer, the upper layers are
optimised accordingly, also known as Bottom-Up
approach. We have subdivided the OSI layer model based
on the IEEE WAVE protocol [43,44] into three main sub-
layers namely as PHY–MAC,Network and Application
layers for vehicular communication as shown in Fig. 2(b).
We take PHY–MAC layer parameters such as SINR, data
Wireless Netw
123
rate and channel dynamics into network layer to calculate
communication range.
Each node can possibly send three types of feedbacks to
the source node based on the physical layer parameters.
These feedbacks are simple channel gain measured at each
node, the signal to noise ratio and signal to interference
plus noise ratio. In order to calculate the first two param-
eters the node requires only its own channel information
where as in calculating the third parameter, the node
requires channel information from other nodes as well. We
assume that different nodes can communicate with each
other and while doing so they share their channel infor-
mation with each other. It is expected that the nodes at the
border of the transmission range of the source node will
have low SINR and by using some weighting factor, its
SINR can be enhanced so that the node is included in the
selection process. However, we argue that as a result of this
weighting, the source node may choose a weak wireless
link for transmission that cannot support the transmission
successfully. Therefore, it is important to include the
realistic values of SINR in the decision making.
For any particular routing scheme, that is usually
implemented at network layer, the main set of variables to
optimise includes the latency and packet drop ratio (PDR).
A routing scheme continues transmitting packets by
assuming that the connection between transmitter and
receiver is established by inspecting acknowledgement
messages. In case of unsuccessful packet delivery without
having knowledge of packet delivery failure, transmission
continues and creates congestion in the transmission
medium. This increases end-to-end delay in the network
resulting in an inefficient routing. This inefficiency in
routing can be improved if the network layer has infor-
mation of successful transmission from the physical layer.
In the following paragraph, we describe the required
information from the physical layer (PHY) for the design
of optimal routing scheme in the network layer in vehicular
communication.
Physical layer (PHY) is mainly responsible for the actual
transmission of data among moving vehicles. The perfor-
mance of PHY layer depends greatly on the condition of
wireless channel. Therefore, the performance of VANET also
depends on the condition of dynamic wireless channel. At
PHY layer, data rate (throughput) is the main variable that can
be utilised at upper layers. Data rate mainly depends on signal
strength [such as Signal to Interference plus Noise Ratio
(SINR)], available bandwidth, transmit/receive power and
wireless channel dynamics with respect to time. By adjusting/
tuning the routing parameters in accordance with the MAC
and physicallayer variables, an overall efficient system can be
designed [45].
4.1 Main algorithm
The proposed algorithm sends packet to the destination and if
the destination is not reachable it finds the next best hop node
(a) (b)
Fig. 2 Cross-layer approach for
VANETs
Wireless Netw
123
and sends the packet to it. In order to set the limitation of the
simulation run, we define Mto limit the number of iterations.
At the first iteration i.e. m=1, the source node screates a set
of common neighbourhood fCNgby sending the periodic
beaconing signal as receiving dfrom other nodes. Once
fCNgis created using the routine ‘‘CN ROUTINE’’ shown in
Algorithm 2, the source node sdetermines whether the d
node is present within fCNg. If the destination dbelong to
fCNg, then the packet is directly delivered to dand the
algorithm terminates. However, if the destination node dis
not a member of fCNg, then the algorithm finds the next best
hop node rthat is the best possible candidate for packet
transmission. The algorithm sorts din descending order such
that d1[d2[;dV. Note that greater value of dvmeans
that a node is closer to the source node swhereas a smaller
value of dvmeans node is away from the source node s. The
first index of sorted dis selected as the next hop node nand
ssends packet to n. The function TimeStamp calculates the
total time of transmission for the packet from node sto node
n. The algorithm also keeps track of hop count defined as
function HCOUNT for the purpose of counting number of
hops for successful transmission from sto d. The algorithm
increments the value of mand repeats the iteration until it
finds dor it reaches to M. The complete procedure is high-
lighted in Algorithm 1 and the flow diagram of this algorithm
is shown in Fig. 3.
Fig. 3 Flow diagram of the proposed scheme
Wireless Netw
123
An interesting question to ask is that what is the prob-
ability that a particular node cannot be included in the fdg
that means to find the Prfdvd0jd0gin other words we
can also ask that what is the probability that a particular
node is out of the transmission range of s. In order to
answer this question, let us consider the probabilistic space
Sdefined as
S¼d1d0[d2d0[[dVd0¼[
V
v¼1
dvd0;ð5Þ
and
did0\djd0¼; suchthat i 6¼ j:ð6Þ
Let Xbe any event in S. Now the probability that a
particular node is in or out of the transmission range of
source nodes can be calculated as
PrfXg¼X
V
v¼1
PrXjdvd0
fgPrdvd0
fg;
¼X
V
v¼1
PrX\dvd0
fg
:ð7Þ
and
Prdv[d0
fg
¼1X
V
v¼1
PrX\dvd0
fg
:ð8Þ
4.2 Common neighbour routine
In order to create the fCNg, the main algorithm invokes
the CN routine for each iteration. The CN Routine has
two special cases. Case I, only considers CQI while Case
II, considers both CQI and QI to create the fCNg. Let us
consider total number of vehicles V;d0as the CQI
threshold, q0as the maximum queue size of each node
and qas the current status of the queue of each node. In
case I, the CN Routine uses ddefined in step 4 of the
algorithm for vehicle vto decide whether the vehicle is
within the transmission range of s.Ifdv[d0then the dv
is stored in fdgelse if dvd0then that node is declared
to be out of transmission range of s. This procedure is
repeated for v¼1;...;V. In case II, each vehicle sends dv
and qvback to the s.Ifdv[d0and qv\q0then that
vehicle is included in d. This condition makes sure that
each node has a sufficient empty space in its buffer to
store the packet before it is retransmitted to the next node.
In order for the node not to be included in fdg, the
algorithm makes sure that dvd0kqvq0. The fdgcal-
culated either using case I or case II is then passed to the
main algorithm.
5 Numerical results
In this section, we present numerical results on the perfor-
mance of the proposed main algorithm presented in Algo-
rithm 1. We also compare the results of proposed algorithm
with GPSR and PROMPT algorithms. In our simulations we
consider a highway model with two lanes and a simulation
area of 300 m 920 m and each algorithm constructs its
fCNgfrom the same distribution of vehicles. We assume an
equal transmit power of 10 dB and d0¼0:30 for each node.
Note that the parameter d0has a variable range and depends
on the simulation scenario under consideration. The simu-
lation parameters are summarised in Table 1. The selection
of these parameters is based on the realistic modeling of
VANET architecture. The selection of value of vehicle
density here is more appropriate for the sparse traffic envi-
ronment such as highway scenario. In addition to this, the
transmission power and its corresponding value of dhas been
selected to implement a key attribute of VANET and that is
there is no power constraint as opposed to other wireless
network environments. This also makes the calculation easy
with the loss of generality in the results. We also assume that
each node has sufficient queue size such that the packets
being received can be stored at the node without the need of
discarding them until their transmission to next node. The
first step is that the source node creates fCNgbased on the
one of the criteria according to the particular algorithm. It
then checks if the destination is present in the fCNg. If the
destination is not found, the next best hop node is selected
based on various criterion. Then the next neighbour node
performs the step 1. In the simulations, it is made sure that
fCNgof the current neighbour node has at least one node
present. This will make sure that the packets are successfully
routed to the next available node. The total number of nodes
Valso vary from one scenario to another scenario.
5.1 GPSR: greedy perimeter stateless routing
GPSR is a distance based routing algorithm that has mainly
two packet forwarding modes: Greedy Forwarding, which
is used more generically wherever possible and Perimeter
Forwarding, that is mostly used in the regions where
Table 1 Simulation parameters
Parameter description Value
Number of vehicles (V)10
SNR ðdÞ10 dB
SNR threshold ðd0ÞVariable
Total transmit power ðPÞ1W
Transmission range (range) 20 m
Simulation area (m) 300 m 920 m
Wireless Netw
123
greedy forwarding can not be used [12]. In our imple-
mentation of GPSR, we only used Greedy Forwarding
mode of the algorithm. The source node in GPSR, creates
the fCNgbased on the distance from source to each node
within the transmission range. The source then searches for
destination in fCNg. If the destination is found, the data is
transmitted directly to the destination and the algorithm
terminates. However, if the destination is not found within
the current fCNg, the next best hop is selected based on the
maximum distance from the source. We argue that select-
ing the next hop node based only on the distance metric
may not result in the successful data transmission because
of wireless channel fading.
In the implementation of GPSR, the fCNgmay result in
null for a particular node. In this situation, that node will
not be able to forward its data because of all nodes being
out of range. In order to avoid this situation, we modify
GPSR in a way that a particular node will retain the data
arrived and will keep searching for the relay node such that
its fCNghas at least on next hop available. During all this
search process, the current node will keep an eye on TTL
of the packet. If TTL expires, the node request for the
retransmission from either from source or the previous hop
node.
5.2 PROMPT: cross-layer routing
PROMPT cross layer routing is based on the delay infor-
mation of each path from source to destination [31].
However, it is described that the knowledge of average
delay along a path is not sufficient to make correct path
selection decision because each source can additionally add
bundle of multiple packets with same path. Therefore, the
delay estimation in PROMPT is performed using five-tuple
statistical information of each path from source to desti-
nation. The best path with the minimum delay between
source and destination is then selected for transmission. In
our simulation we assume the equal packet arrival rate of
each node. We also assume that the relay nodes do not add
their packets with the data being transmitted from source.
This means once the packets from one node are success-
fully transmitted, then only the second node can start its
transmission. In order to simplify the simulation, each relay
node is modelled as a G/G/1 queuing system. Once the best
path is decided, the transmission between source and des-
tination is then performed successfully.
5.3 Simulations
5.3.1 Scenario I
In this scenario, we study the CN set formation and hence
the number of hops to reach the destination for the three
above mentioned algorithms. Note that all vehicles in our
simulation are moving only in one direction. For this
simulation it has been assumed that all vehicles are moving
in the same direction. Let us assume we have 10 vehicles
moving on a highway and node 2 wants to communicate
with node 9. This can be achieved with either a direct link
between nodes 2 and 9 or through a multi-hop scenario
depending on the formation of fCNg. Note that the value
of d0was kept at 0.5 in this simulation. The proposed main
algorithm establishes this link after three hops. In the first
hop source node searches for the neighbouring nodes and
creates a fCNgbased on d. The cardinality of first fCNgis
two. In this fCNgthe destination node (9) is not present,
therefore the source node selects the next best hop node
that is node 7 in this scenario. In the next step node 7
searches its possible neighbouring node and formulates its
fCNgwith the cardinality of four. Still the current relay
node can not find destination node in fCNg, so it selects
the next best hop node with is node 8. Now node 8 repeats
the above procedure and formulates its fCNg. In this
instance, the destination node is present in the fCNgof
current relay node therefore node 8 transmits data to the
destination node. Hence completing this communication in
three hops. A typical formulation of fCNgof nodes 2, 7
and 8 is shown in Table 2.
Now let us consider the same communication scenario using
GPSR algorithm. Note that in GPSR the formation of fCNgis
based on similar distribution of vehicles moving with the same
velocity and direction as that of main algorithm. At the first
step, the cardinality of fCNgis three and the members of this
setareshownintheTable 3. As the destination node can not be
foundinthefCNg, the source selects the best next hop node
based on the criteria given in GPSR algorithm description. The
next best hop node selected is node 4. Now the node 4 repeats
the same procedure and formulates its fCNg. The typical for-
mation of the fCNgof node 4 is also shown in the Table 3.
Again node 4 can not find the destination node in its fCNgand
selects the best next hop node for the data transmission. This
procedure continues until GPSR reaches to the destination node
after 9 hops. We found that due to the distance based fCNg
formation in GPSR, the destinationnodeisreachedinmore
number of hops as compared to other algorithms degrading the
performance of GPSR in this environment.
Table 2 fCNgformation for main algorithm
Node 2 Node 7 Node 8
Node ID dNode ID dNode ID d
5.0 0.5466 1.0 0.5201 3.0 0.5525
7.0 0.5938 5.0 0.5298 5.0 0.5216
––8.0 0.5626 8.0 1.1965
– – 10.0 0.5063 9.0 1.5270
Wireless Netw
123
Next we compare the performance of PROMPT cross-
layer routing algorithm in the similar environment. The
PROMPT algorithm is based on the path formation from
source to destination and then selecting the best possible
path with the minimum end to end delay. In this scenario,
the PROMPT algorithm formulates two possible paths
from source node 2 to destination node 9. The detailed path
information with next hop nodes and the total number of
hops to reach the destination for each path is shown in
Table 4. The first path consists of 7 hops, while second and
third paths consist of 6 hops each. Based on the delay for
each path, PROMPT algorithm selects the fourth path for
communication as it has the lowest number of hops. This
selected path is italicized in the Table 4.
A comparison of the three above mentioned algorithms
shows that the proposed main algorithm results in the best
performance by establishing communication between
source and destination nodes in a total of 3 hops. The
second best performance is achieved by PROMPT algo-
rithm. In this scenario, GPSR fails to reach the destination
within the given TTL of the packets.
In another run of this simulation the three algorithms
establish communication between source node 1 and des-
tination node 10 in the total 10 node scenario (the detailed
CN sets are not shown due to space limitation). Note that all
the simulation parameters were kept similar to the previous
example. In this run, the main algorithm reaches the desti-
nation in 2 hops, GPSR reaches the destination in 21 hops
while PROMPT algorithm formulates 8 paths to reach the
destination and the best selected path has 6 minimum hops
to reach the destination. This strengthens our argument that
in a sparse traffic distribution, the main algorithm gives
better performance compared to its counter parts.
5.3.2 Scenario II
In this scenario we study the effect of number of hops of
each mentioned algorithm over a period of simulation run.
The simulation starts with a random selection of source and
destination for each scenario. Each algorithm is run and the
number of hops needed to perform multi-hop communi-
cation for particular source and destination combination is
performed. We selected 10 random vehicles in our topol-
ogy and a simulation area of 300 m 920 m is used to
define the road topology. The value of d0is kept at 0.50.
Figure 4shows a comparison of number of hops calcula-
tion over a selected value of simulation run for all three
algorithm. This figure clearly highlight the poor perfor-
mance of GPSR due to its dependency of establishing
fCNginformation using Euclidean distance as the only key
metric. In this scenario, proposed algorithm performs better
than both GPSR and PROMPT algorithms (Fig. 5).
Table 3 fCNgformation for GPSR
Node 2 Node 4
Node ID EDistance Node ID EDistance
1.0 2.2361 2.0 16.4924
3.0 1.0000 3.0 16.1245
4.0 2.8284 5.0 12.0416
– – 6.0 15.2971
–– 7.0 19.1050
Table 4 Path and hops information for PROMPT
Path ID Path structure Hops
12?3?4?5?6?7?8?97
22?3?4?5?6?7?96
32?3?4?5?7?8?96
42?3?4?5?7?95
Fig. 4 Number of hops to reach the destination
Fig. 5 Effect of vehicle density on the number of hops
Wireless Netw
123
5.3.3 Scenario III
In this scenario, we study the effect of number of hops
while varying vehicle density over the simulation area.
Each algorithm is run by selecting random source and
destination to initiate a multi-hop communication scenario
while varying the number of vehicles in a defined area
(stretch of road) for all simulations. The number of hops for
each individual scenario is then captured and the results are
plotted in Fig. 5. The proposed algorithm performs better
in each scenario as compared to its predecessors. Figure 5
also clearly highlights the under performance of GPSR in
low vehicle density conditions. In this study, primarily to
obtain optimal results for number of hops, we have set the
upper bound value of time to calculate number of hops at a
higher value for GPSR. The impact of this on to keep an
optimal value of TTL for desired packet transmission is
another area that will be explored later in the study.
In order to understand the variance in number of hops for
each algorithm in more detail we introduce another metric,
Coefficient of Variance (CoV) that is defined as the ratio of
standard deviation ðrÞto the mean ðlÞ. This metric indicates
that the algorithm with high variation has poor performance
compared to the algorithms with low fluctuations in data
sample. This result is shown in Fig. 6where normalised CoV
is plotted against each simulation run. For each algorithm,
CoV is normalised to the maximum value of respective
algorithm. It is evident from the figure that the PROMPT
algorithm has less variations compared to its counterparts,
hence performing better in almost all scenarios.
0 2 4 6 8 10 12 14 16 18 20
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Simulation Run
Normalised CoV
Main
GPSR
PROMPT
Fig. 6 Normalised CoV over
simulation run
Fig. 7 Hop distribution for all algorithms
Table 5 Distribution variables for three algorithms
Algorithm lr
Main algorithm 6.5727 6.9262
GPSR 6.8909 8.2770
PROMPT 4.4636 2.5308
Wireless Netw
123
It is evident from the number of hops distribution for the
three algorithms compared that the best followed distri-
bution is Gamma distribution. Figure 7shows the distri-
bution of hops data for all of the three mentioned
algorithms. In order to verify these distributions let us have
a closer look at the probability distribution function(pdf) of
the number of hops that can be written as follows.
fðx;a;bÞ¼ ba
CðaÞxa1ebx;for x [0and a;b[0;
ð9Þ
where Cis the Gamma function,arepresents the shape of
the distribution and bis the scaling factor.
Similarly the cumulative distribution for the data is
given as:
Fðx;a;bÞ¼Zx
0
fðu;a;bÞdu ¼ca;bxðÞ
CðaÞ;ð10Þ
where cis the lower incomplete gamma function. These
parameters are calculated using the lðxÞand rðxÞrespec-
tively. The plotted distribution as shown in Fig. 7has been
derived using the data tabulated in Table 5.
6 Conclusion
In this paper, we have presented an architecture of cross-layer
based routing, using beaconing for a multi-hop environment
in VANETS. We have verified the proposed structure for a
linear VANET design via simulations and have compared its
results with existing protocols using three key performance
scenarios. In the first scenario, we look closely into the route
formation using true channel conditions via channel quality
indicator and compare its validation with distance based and
pre-route selection algorithms. A generic example of node
distribution in this multi-hop environment is discussed in
which the proposed structure is found to establish a better
performance as compared to its counterparts. In the second
scenario, we study the effects of varying the vehicles density
in the network design. This approach clearly shows the under
performance of routing approaches that make decisions based
only on metric of Euclidean distance. To gather a deep insight
into the variation of hop counts we have derived distribution
patterns for each algorithm using statistically best fit
approach. In future work, we would like to use these obser-
vations to obtain optimal values of channel quality indicator
together with the queuing information for all vehicular
environments under the realistic channel conditions.
References
1. Watfa, M. (2010). Advances in vehicular ad-hoc networks: Devel-
opments and challenges—Intelligent Transport Systems, IGI Global.
2. Luo, Y., Zhang, W., & Hu, Y. (2010). A new cluster based
routing protocol for VANET. In Networks security wireless
communications and trusted computing (NSWCTC), second
international conference (Vol. 1, pp. 176–180).
3. Soares, R. B., & Loureiro, A. A. F. (2012). VCARP: Vehicular
ad-hoc networks context-aware routing protocol. In IEEE Sym-
posium on computers and communications (ISCC) (Vol. 0,
pp. 442–447).
4. Bakhouya, M., Gaber, J., & Wack, M. (2009). Performance
evaluation of DREAM protocol for inter-vehicle communication.
In 1st International conference on wireless communication,
vehicular technology, information theory and aerospace elec-
tronic systems technology VITAE (pp. 289–293).
5. Chakeres, I., & Perkins, C. (2008). Dynamic MANET on-demand
(DYMO) routing. In Internet Draft (draft-ietf-manet-dymo-14).
6. Johnson, D. B., & Maltz, D. A. (1996). Dynamic source routing
in ad hoc wireless networks. In Mobile computing, ch. 5.
7. Perkins, C. E., & Royer, E. M. (1999). Ad-hoc on-demand dis-
tance vector routing. In Second IEEE workshop on mobile com-
puting systems and applications (pp. 90–100).
8. Sommer, C., & Dressler, F. (2007). The DYMO routing protocol
in VANET scenarios. In IEEE 66th vehicular technology con-
ference (VTC-2007) (pp. 16–20).
9. Harri, J., Filali, F., & Bonnet, C. (2006). Performance comparison
of AODV and OLSR in VANETs urban environments under
realistic mobility patterns. In 5th Annual mediterranean ad hoc
networking workshop, Med-Hoc-Net.
10. Li, F., & Yu, W. (2007). Routing in vehicular ad hoc networks: A
survey. IEEE Vehicular Technology Magazine,2(2), 12–22.
11. Rehman, S. U., Khan, M. A., Zia, T. A., & Zheng, L. (2013).
Vehicular ad-Hoc networks (VANETs)—An overview and
challenges. Journal of Wireless Networking and Communica-
tions,3(3), 29–38.
12. Karp, B., & Kung, H. T. (2000). GPSR: Greedy perimeter
stateless routing for wireless networks. In Proceedings of the
ACM/IEEE international conference on mobile computing and
networking (MobiCom).
13. Jarupan, B., & Ekici, E. (2010). PROMPT: A cross-layer posi-
tion-based communication protocol for delay-aware vehicular
access networks. Ad Hoc Networks,8(5), 489–505.
14. Seet, B. C., Liu, G., Lee, B. S., Foh, C. H., Wong, K. J., & Lee, K.
K. (2004). A-STAR: A mobile ad hoc routing strategy for
metropolis vehicular communications in networking 2004. In
Lecture notes in computer science (pp. 989–999). Berlin:
Springer.
15. Naumov, V., & Gross, T. (2007). Connectivity-aware routing
(CAR) in vehicular ad-hoc networks. In 26th IEEE international
conference on computer communications (INFOCOM07) (pp.
1919–1927).
16. Chen, C., Jin, Y., Pei, Q., & Zhang, N. (2014). A connectivity-
aware intersection-based routing in VANETs. EURASIP Journal
on Wireless Communications and Networking,1, 42.
17. Li, M., Li, Z., & Vasilakos, A. V. (2013). A survey on topology
control in wireless sensor networks: Taxonomy, comparative
study, and open issues. Proceedings of the IEEE,101(12),
2538–2557.
18. Cheng, H., Xiong, N., Vasilakos, A. V., Tianruo Yang, L., Chen,
G., & Zhuang, X. (2012). Nodes organization for channel
assignment with topology preservation in multi-radio wireless
mesh networks. Ad Hoc Networks,10(5), 760–773.
19. Vasilakos, A. V., Zhang, Y., & Spyropoulos, T. (2011). Delay
tolerant networks: Protocols and applications—Wireless net-
works and mobile communications. Boca Raton, FL: CRC Press.
20. Spyropoulos, T., Bin Rais, R. N., Turletti, T., Obraczka, K., &
Vasilakos, A. (2010). Routing for disruption tolerant networks:
Taxonomy and design. Wireless Networks,16(8), 2349–2370.
Wireless Netw
123
21. Zeng, Y., Xiang, K., Li, D., & Vasilakos, A. V. (2012). Direc-
tional routing and scheduling for green vehicular delay tolerant
networks. Wireless Networks,19(2), 161–173.
22. Yao, Y., Cao, Q., & Vasilakos, A. V. (2013). EDAL: An energy-
efficient, delay-aware, and lifetime-balancing data collection
protocol for wireless sensor networks. In Proceedings of IEEE
10th international conference on mobile ad-hoc sensor systems
(pp. 182–190).
23. Li, P., Guo, S., Yu, S., & Vasilakos, A. V. (2012). CodePipe: An
opportunistic feeding and routing protocol for reliable multicast
with pipelined network coding. In Proceedings of IEEE-INFO-
COM (pp. 100–108).
24. Wei, G., Ling, Y., Guo, B., Xiao, B., & Vasilakos, A. V. (2011).
Prediction-based data aggregation in wireless sensor networks:
Combining grey model and Kalman Filter. Computer Commu-
nications,34(6), 793–802.
25. Xiang, L., Luo, J., & Vasilakos, A. (2011). Compressed data
aggregation for energy efficient wireless sensor networks. In 8th
Annual IEEE communications society conference on sensor, mesh
and ad hoc communications and networks (pp. 46–54).
26. Yuan, J., Member, S., Li, Z., Yu, W., Li, B., & Member, S.
(2006). A cross-layer optimization framework for multihop
multicast in wireless mesh networks. IEEE Journal on Selected
Areas of Communications,24(11), 2092–2103.
27. Ghada, K., Li, J., & Ji, Y. (2012). Cross-layer design for topology
control and routing in MANETs. Wireless Communications and
Mobile Computing,12, 257–267.
28. Youssef, M., Ibrahim, M., Abdelatif, M., Chen, L., & Vasilakos, A.
V. (2014). Routing metrics of cognitive radio networks: A survey.
IEEE Communications Surveys & Tutorials,16(1), 92–109.
29. Chilamkurti, N., Zeadally, S., Vasilakos, A., & Sharma, V.
(2009). Cross-layer support for energy efficient routing in wire-
less sensor networks. Journal of Sensors,2009, 1–9. doi:10.1155/
2009/134165.
30. Wang, X., Vasilakos, A. V., Chen, M., Liu, Y., & Kwon, T. T.
(2011). A survey of green mobile networks: Opportunities and
challenges. Mobile Networks Applications,17(1), 4–20.
31. Jarupan, B., & Ekici, E. (2011). A survey of cross-layer design
for VANETs. Ad Hoc Networks,9(5), 966–983.
32. Pham, P., Perreau, S., & Jayasuriya, A. (2005). New cross-layer
design approach to ad-hoc networks under Rayleigh fading. IEEE
Journal on Selected Areas in Communications,23(1), 28–39.
33. Yang, S., Sheng, Z., McCann, J. A., & Leung, K. K. (2014).
Distributed stochastic cross-layer optimization for multi-hop
wireless networks with cooperative communications. IEEE
Transactions on Mobile Computing,13(10), 2269–2282.
34. Zhang, Z., Mao, G., & Anderson, B. (2014). Stochastic charac-
terization of information propagation process in vehicular ad hoc
networks. IEEE Transactions on Intelligent Transportation Sys-
tems, 15(1), 122, 135.
35. Katsaros, K., Dianati, M., Tafazolli, R., & Kernchen, R. (2011).
CLWPR: A novel cross-layer optimized position based routing
protocol for VANETs. In IEEE vehicular networking conference
(VNC) (pp. 139–146).
36. Chen, Y., Xiang, Z., Jian, W., & Jiang, W. (2009). A cross-layer
AOMDV routing protocol for V2V communication in urban
VANET. In Fifth international conference on mobile ad-hoc
sensor networks (pp. 353–359).
37. Zhou, L., Zheng, B., Geller, B., Wei, A., Xu, S., & Li, Y. (2008).
Cross-layer rate control, medium access control and routing
design in cooperative VANET. Computer Communications,
31(12), 2870–2882.
38. El-Keyi, A., ElBatt, T., Bai, F., & Saraydar, C. (2012) MIMO
VANETs: Research challenges and opportunities. In Interna-
tional conference on computing, networking and communications
(ICNC) (pp. 670–676).
39. Torrent-Moreno, M., Schmidt-Eisenlohr, F., Fubler, H., & Har-
tenstein, H. (2006). Effects of a realistic channel model on packet
forwarding in vehicular ad hoc networks. In IEEE wireless
communications and networking conference (WCNC) 2006 (Vol.
1, pp. 385–391).
40. Hassanabadi, B., & Valaee, S. (2014). Reliable periodic safety
message broadcasting in VANETs using network coding. IEEE
Transactions on Wireless Communications,13(3), 1284–1297.
41. Simon, M. K., & Alouini, M.-S. (2000). Digital communication
over fading channels. New York: Wiley.
42. Agarwal, A., & Kumar, P. R. (2004). Improved capacity bounds
for wireless networks. Wireless Communications and Mobile
Computing,4(3), 251–261.
43. Rehman, S. U., Khan, M. A., Zia, T. A., & Khokhar, R. H.
(2013). A synopsis of simulation and mobility modeling in
vehicular ad-hoc networks (VANETs). IOSR Journal of Com-
puter Engineering (IOSR-JCE), 15(2), 1–16.
44. Kenney, J. B. (2011). Dedicated short-range communications
(DSRC) standards in the United States. Proceedings of the IEEE,
99(7), 1162–1182.
45. Baszczyszyn, B., Paul, M., & Toor, Y. (2009). Maximizing
throughput of linear vehicular ad-hoc NETworks (VANETs) a
stochastic approach. In Wireless conference EW, 2009 (pp. 32–36).
Sabih ur Rehman received his
Bachelor in Electronic Engi-
neering from University of
South Australia, Adelaide,
Australia, in 1998 with Honors.
He is a Research Scholar and
Adjunct Lecturer in School of
Computing and Mathematics at
Charles Sturt University (CSU).
Prior to joining CSU, Sabih
worked in IT and telecommu-
nication industry for more than
10 years. His broader research
interests are in the areas of
wireless sensor networks,mobile
and ad-hoc networks, cloud computing and network infrastructures.
Currently, Sabih’s research is focused on Quality of Service (QoS),
cross-layer protocol architecture designing, wireless propagation
modeling using mathematical/stochastic models and performance
analysis. Sabih has worked as a reviewer for respected journals and
conferences. Sabih is also a member of Institute of Engineers Aus-
tralia (IEAust), Institute of Electrical and Electronic Engineers
(IEEE) and IEEE Computer Society.
M. Arif Khan is a Lecturer in
the School of Computing and
Mathematics, Charles Sturt
University, Australia. Arif has
earned his Ph.D. from Macqua-
rie University Sydney in 2010,
Master of Science in Engineer-
ing from GIK Institute of Engi-
neering Sciences and
Technology in 2002 and Bach-
elor of Engineering from Uni-
versity of Engineering and
Technology Lahore in 2000.
Arif is the recipient of Interna-
tional Macquarie University
Research Scholarship (iMURS) for his Ph.D., Research Scholarship
from ICT CSIRO Centre Marsfield Sydney and a competitive
Wireless Netw
123
Scholarship for his M.S. from GIK Institute. Arif’s research interests
are in Wireless Ad Hoc Networks, ICT security, MIMO Wireless
Communication and Physical Systems in ICT. Arif is a member of
Institute of Electrical and Electronic Engineers (IEEE).
Tanveer A. Zia is an Associate
Professor, Course Coordinator
for the Doctor of Information
Technology course, and Asso-
ciate Head of School at School
of Computing and Mathematics,
Faculty of Business, Charles
Sturt University, Australia. He
has earned his Ph.D. from the
University of Sydney, Master of
Interactive Multimedia
(MIMM) from University of
Technology Sydney, MBA from
Preston University USA, and
Bachelors of Science in Com-
puter Sciences from Southwestern University, Philippines. Tanveer’s
broader research interests are in ICT security. Specifically he is
interested in security of low powered mobile devices. He is also
interested in biometric security, cyber security, cloud computing
security, information assurance, protection against identity theft, trust
management, and forensic computing. He is serving on Technical and
Program Committees of several international conferences in his area
of research. He actively publishes in international conferences,
symposiums, workshops, and refereed journals. Tanveer is a Senior
Member Australian Computer Society and Certified Professional
(MACS Snr CP), Senior Member Institute of Electrical and Elec-
tronics Engineers (IEEE), Senior Member International Association
of Computer Sciences and Information Technology (IACSIT),
Member IEEE Computer Society, and Member Australian Informa-
tion Security Association (AISA).
Wireless Netw
123
