IEEE NS2 PROJECT@ DREAMWEB TECHNO SOLUTION

Interference-Based Topology Control Algorithm
for Delay-Constrained Mobile Ad Hoc Networks
Xin Ming Zhang, Member, IEEE, Yue Zhang, Fan Yan, and
Athanasios V. Vasilakos, Senior Member, IEEE
Abstract—As the foundation of routing, topology control should minimize the interference among nodes, and increase the network
capacity. With the development of mobile ad hoc networks (MANETs), there is a growing requirement of quality of service (QoS) in
terms of delay. In order to meet the delay requirement, it is important to consider topology control in delay constrained environment,
which is contradictory to the objective of minimizing interference. In this paper, we focus on the delay-constrained topology control
problem, and take into account delay and interference jointly. We propose a cross-layer distributed algorithm called interference-based
topology control algorithm for delay-constrained (ITCD) MANETs with considering both the interference constraint and the delay
constraint, which is different from the previous work. The transmission delay, contention delay and the queuing delay are taken into
account in the proposed algorithm. Moreover, the impact of node mobility on the interference-based topology control algorithm is
investigated and the unstable links are removed from the topology. The simulation results show that ITCD can reduce the delay and
improve the performance effectively in delay-constrained mobile ad hoc networks.
Index Terms—Delay, interference, mobile ad hoc networks (MANETs), topology control algorithm
Ç
1 INTRODUCTION
WITH the increasing attention and development in
mobile ad hoc networks (MANETs), there is a grow-
ing demand for applications that require quality of service
(QoS) provision, such as voice over IP (VoIP), multimedia,
real-time collaborative work. Different applications often
have different QoS requirements in terms of bandwidth,
packet loss rate, delay, packet jitter, hop count, path reliabil-
ity and power consumption [1]. Real-time application is one
of the particularly useful application directions of MANETs,
especially VoIP applications, where there is a strict require-
ment of delay.
In order to guarantee the QoS requirement in terms of
delay, some researches explored the delay incurred in a for-
warding node or a routing path. In [2], [3], [4], [5], [6], delay
is defined as the transmission delay of a packet. Then, Xie
et al. [7] found that in many cases the queuing delay takes a
significant portion of the total delay over a hop. A path,
which contains many packets in queue of the nodes and
with short transmission delay on links, could have a larger
delay than the one, which has less packets in the queue at
nodes but longer transmission delay. And the larger the
number of the intermediate nodes between the source and
destination pair is, the larger the potential delay is.
However, in wireless ad hoc networks, the impact of
channel contention from neighbors must also be considered.
Because of the limited channel source, access delay and col-
lision are generated at nodes. If one of the nodes on a path
can not acquire channel in a long period for contention, it
may lead to massive packet drops and higher packet drop-
ping rate. Processing delay and propagation delay which
change in microseconds are much shorter than transmission
delay, contention delay and queuing delay which change in
millisecond. Therefore, the end-to-end (E2E) delay should
consider: transmission delay over intermediate links, con-
tention delay caused by nodes’ contention for the shared
channel and queuing delay induced at each intermediate
node due to queuing policy or severe channel conditions.
Topology control is to dynamically change the nodes
transmission range in order to maintain connectivity of the
communication graph, while reducing energy consumption
and/or interference that are strictly related to the nodes
transmitting range. A good topology not only can provide a
better service for routing layer, but also can save energy,
increase network capacity and satisfy the QoS requirements.
The previous topology control algorithms [8], [9], [10], [11],
[12], [13], [14] mainly focused on the interference constraint.
And how to employ topology control to reduce delay is not
fully researched by those works. An alternative way to
reduce the E2E delay is to increase the transmission power
of a certain node in a path, so that the transmission range of
the node is increased and thus the hops between the source
and destination are reduced. Transmission delay may be
decreased due to the reduction in hops; and the sum of
the queuing delay along a path is also decreased because
the number of the intermediate nodes is decreased. Thus,
increasing the transmission power may reduce the E2E
delay. However, it may cause more interference to other
nearby active receiving nodes, excessive contention to
nearby potential sending nodes, which may incur more
X.M. Zhang, Y. Zhang, and F. Yan are with the School of Computer
Science and Technology, University of Science and Technology of
China, Hefei 230027, P.R. China.
E-mail: xinming@ustc.edu.cn, {yuezhang, yanfan}@mail.ustc.edu.cn.
A.V. Vasilakos is with the National Technical University of Athens,
Athens, Greece. E-mail: vasilako@ath.forthnet.gr.
Manuscript received 25 June 2013; revised 8 May 2014; accepted 16 June
2014. Date of publication 18 June 2014; date of current version 2 Mar. 2015.
For information on obtaining reprints of this article, please send e-mail to:
reprints@ieee.org, and reference the Digital Object Identifier below.
Digital Object Identifier no. 10.1109/TMC.2014.2331966
742 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 14, NO. 4, APRIL 2015
1536-1233 ß 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

retransmissions. And retransmission means the increase of
E2E delay. Therefore, reducing delay and minimizing inter-
ference are two conflicting goals, and it is necessary to
jointly consider a tradeoff between them. Thus, the problem
of interference-based topology control with delay-constraint
is studied.
In addition, the mobility imposes a great impact on the
interference-based topology control algorithm and the E2E
delay. First, we need an appropriate interference-based
topology control algorithm for mobile ad hoc networks. The
vast majority of researches on topology control focused
only on reducing the power of each node to save energy
and reducing the network interference. Most of the algo-
rithms yield a minimally connected topology, which is
prone to suffer frequent link breakages in a mobile network.
Link breakages result in retransmissions and packet loss,
and deteriorate the network performance. Some recent
works have shown that mobility causes incorrect informa-
tion in terms of link availability [8], [14]. Consequently,
topology control algorithm may exploit the mobility to
reduce the frequency of link breakages. Second, the E2E
delay is particularly impacted by the mobility of the nodes
in a path. In mobile ad hoc networks, it is necessary to con-
sider the delay caused by the mobility of nodes. If a node
has a lower mobility, the impact of mobility on delay could
be ignored. If a node has a higher mobility, the node may
move out of the sender’s transmission range quickly so
that the link is unstable and prone to break. Once the link
breaks, the transmission delay will become infinite.
The main contributions of this paper are:
1) We concern about the relationship of delay and
interference in MANETs and make a good tradeoff
between reducing delay and minimizing interfer-
ence. By balancing the influence of delay and inter-
ference through adjusting the transmission power of
nodes, topology is controlled to satisfy both the
delay constraint and interference constraint.
2) The delay in our work fully considers the character-
istics of MANETs and takes the transmission delay,
the contention delay and queuing delay into account,
which is different from other QoS topology schemes.
We propose a simple but effective balance algorithm
to transform the delay constraint for a path into
delay constraints at intermediate nodes, and design
a balance factor in the algorithm which considers
both actual transmission delay and estimated delay
so that it could adapt to the different links dynami-
cally and control topology at a proper time. We fur-
ther divide links into stable links and unstable links.
If the duration of a link is greater than the delay con-
straint at the transmit node and each intermediate
node, the link will be selected as a candidate for-
warding link, otherwise it will be removed.
3) We implement an interference-based topology con-
trol algorithm for delay-constrained mobile ad hoc
networks. By inserting a particular field into the
routing packet during the routing discovery proce-
dure, the delay information for the topology control
algorithm is provided. Then we control the transmis-
sion power of node to minimize the interference and
satisfy the delay requirement according to the delay
information provided by the delay model. Our
topology control algorithm adjusts the transmission
power considering the the Signal to Noise Ratio
(SINR) threshold to enable the successful reception
of data packets at receiving node, thus the former
connection will not be changed.
The remainder of this paper is organized as follows:
Section 2 introduces the related previous work. Section 3
proposes an interference-based topology control algorithm
for delay-constrained mobile ad hoc networks. Section 4
presents simulation parameters and the performance eval-
uation results of the proposed algorithm. Section 5 con-
cludes the paper.
2 RELATED WORK
With the development of special applications like real-time
services, streaming video and so on, it is essential to realize
efficient QoS support in MANET [1]. Metrics used in QoS
include bandwidth, delay, hop count, and path reliability,
power consumption and service coverage. In order to
achieve the above vision, it is important to consider QoS
routing in the QoS architecture. QoS routing can determine
packets forwarding based on the actual network connection
according to the QoS requirements. Delay requirement is
one of the particularly useful QoS requirements for mobile
ad hoc networks. Many QoS routing protocols which
consider the end-to-end delay as a QoS measure have been
proposed. Draves et al. [2] defined the delay as the transmis-
sion delay of the packet. But it did not contain the queuing
delay. If there are massive packets in the queue waiting for
transmission since a node cannot transmit multiple packets
simultaneously, the queuing delay may take a significant
portion of the total delay. The delay in [7] contained the
queuing delay and the transmission delay. But it did not
explicitly consider the effects of channel contention. The
contention of the channel can cause access delay and colli-
sion at Medium Access Control (MAC) layer.
Because of the direct coupling among different layers, the
traditional layered design is not sufficient for multihop wire-
less networks. Zhang and Zhang [1] pointed out that the
physical layer affects the MAC and routing decisions by
changing its transmission power and rate. The MAC layer
can schedule and allocate the wireless channel, and eventu-
ally determine the available bandwidth and the packet
delay, which will then affect the link or path selection in the
routing layer. The routing layer selects the transmission
path for data packets, which will change the contention level
at the MAC layer, and accordingly the parameters at the
physical layer. Thus, the mutual impact in different layers
should be considered and it is necessary to consider all the
controls across different layers jointly to optimize the overall
performance while meeting the QoS requirements. There-
fore, cross layer design of congestion control, routing algo-
rithms with QoS guarantees is one of the most challenging
topics in wireless networking [15], [16]. Here, cross layer
design does not mean eliminating the advantages of layer-
ing, it means keeping some form of separation, while allow-
ing layers to actively interact, appears to be a good
compromise for enabling interaction between layers without
ZHANG ET AL.: INTERFERENCE-BASED TOPOLOGY CONTROL ALGORITHM FOR DELAY-CONSTRAINED MOBILE AD HOC NETWORKS 743

eliminating the layering principle. Xue and Ekici. [15], Li
and Eryilmaz [16] mainly solve the problem of scheduling
algorithm for end-to-end delay constrained network to reach
a high throughput. Xue and Ekici [15] proposed an algo-
rithm to achieve guaranteed throughput while satisfying
QoS requirements and guaranteeing that all actual queue
backlogs are deterministically upper-bounded. Li and Eryil-
maz [16] proposed a queueing architecture to exploit the
new degree of freedom of choosing service discipline for dif-
ferent arrival process. However, in MANETs, applications
may have traffic to be sent at anytime, it is hard to know
these information in prior. Moreover, different paths chosen
by routing strategy will lead to different end-to-end delay.
And topology is the foundation of routing, it offers certain
paths to be selected by a routing protocol. Thus, a good
topology control may greatly help the routing protocol to
select an appropriate path to meet the E2E delay
requirement.
Topology control is one of the most important mecha-
nisms in wireless ad hoc and sensor networks. The main
purpose of topology control is to dynamically change the
nodes transmitting range to provide connectivity for wire-
less networks while meeting some given requirements,
including power consumption, interference, broadcast,
QoS, antennas, and reliability [17].
Power control [12] and backbone [13] construction are two
major research directions about topology control. Power con-
trol [12] is a NP hard problem in ad hoc networks. The prob-
lem of assignment of transmitting range is the idealized
solution of power control. The problem of minimum con-
nected dominating set is the idealization of backbone con-
struction [13], which can be considered as the research of
sleep scheduling scheme in wireless sensor networks. How-
ever, neither assignment of transmitting range nor minimum
connected domination can be regarded as a specific
definition of the problem of topology control. We need to
minimize the interference during topology controls [9], [10],
[11]. Interference-based topology control can combine exist-
ing power control with backbone construction preferably.
Burkhart et al. [9] contradicted the idea with establishing a
simple model of interference, and proved that a great many
former topology control algorithms were not low interfer-
ence with the definition. They also showed that all the neigh-
bors can interfere with the communication node, and
defined the number of all neighbors covered by the range of
the two communicating nodes as the interference of the path.
After that, the study of topology control [8], [10], [11], [14] in
recent years starts to regard interference control as another
objective of topology control, instead of implicitly lower the
interference.
Previous topology control algorithms [8], [9], [10], [11],
[12], [13], [14] have only focused on the interference con-
straint, it is also important for a topology control algorithm
to meet the QoS requirement. Some researches have been
carried out in this field. Jia et al. [18] focused on the problem
of energy efficient QoS topology control. They proposed an
algorithm to find the network topology that all traffics can
be routed and the maximal node power is minimized by
incrementing node power to connect two nodes that have
the shortest euclidean distance among the unconnected
node-pairs and then checking if the traffics can be routed on
the topology constructed. QoS requirements included band-
width and delay. However, the traffic demands are
assumed to be known in prior, interference and node mobil-
ity are not taken into account, which are important factors
that affect the network performance in MANETs. Chou and
Suen [19] proposed a residual-hop-count estimation based
on nodes’ available bandwidth, and then adjusted the
power level of each node by checking the residual-hop-
count. Tang et al. [20] studied interference-aware topology
control and QoS routing in multi-channel wireless mesh
networks, they computed the channel assignment for the
nodes by going through the edges in a K-connected sub-
graph of the network in a non-increasing order of link
potential interference values to make the topology interfer-
ence minimum, and then seeked routes for QoS connection
requests with bandwidth requirements. The transmission
power of a node is fixed.
Additionally, the impact of mobility on the topology con-
trol algorithm and delay should be investigated. Burri et al.
[8] removed unreliable and redundant links from the net-
work while guaranteeing the connectivity of the network.
However, the mobility factor was not considered. In
practice, the link quality between neighboring nodes fluctu-
ates significantly over time especially when the nodes are
mobile. Some topology control algorithms with consider-
ation of mobility were also presented [14]. One- and two-
hop topology control algorithm (OTTC) and one- and two-
hop topology control algorithm for mobile nodes (AOTTC)
[14] were proposed on the basis of [8]. The nodes compute
their transmission radius on the basis of their one- and two-
hop neighbors’ information. Each node orders its one-hop
neighbors in an ordered list and then the ordered lists are
exchanged between the neighbors as [8]. In OTTC, the
nodes are stationary and AOTTC considers the mobility.
However, the reorganization phase in [8] or [14] is applied
in response to neighbor changes. The neighbors’ changing
takes large amount of overhead. Study has shown that
mobility causes incorrect information in terms of link avail-
ability and view consistency [14]. Delay is also affected by
mobility of a node. Delay has been studied under different
mobility models [21], [22]. Mammen and Shah [21] studied
the maximal throughput scaling and the corresponding
delay scaling in a random mobile network with restricted
node mobility and particular mobility restriction does not
affect the delay scaling. Zhu et al. [22] showed that a high
throughput can be achieved at the cost of very high delay
in the mobile network and the network with a low delay
has a low throughput. Tradeoff between throughput and
delay in mobile ad hoc networks is provided by controlling
nodes’ mobility. Here, we only need to study the impact of
mobility on the delay and make the algorithm satisfy the
delay constraint under the mobility environment, especially
the link connect time, which has been studied in our previ-
ous work [23].
3 INTERFERENCE-BASED TOPOLOGY CONTROL
ALGORITHM WITH DELAY CONSTRAINT
In this section, we describe the interference-based topology
control algorithm for delay-constrained mobile ad hoc net-
works. The purpose of this work is to design a cross-layer

distributed topology control algorithm under both the
interference constraint and the delay constraint. A cross-
layer information exchange among the physical layer, the
link layer and the network layer is required. Distributed
power control is running on the physical layer to constrain
the delay incurred by transmission delay over intermediate
links, contention delay caused by nodes’ contending for
the shared channel and queuing delay induced at each
intermediate node due to queuing policy or severe channel
conditions.
First, we establish delay models for a path and interme-
diate node respectively, and then, we design a cross-layer
distributed topology control algorithm under both the inter-
ference constraint and the delay constraint. Finally, the
impact of mobility on the interference-based topology con-
trol algorithm and the delay are investigated.
3.1 Delay Model
3.1.1 Delay Model of a Path
The end-to-end delay contains transmission delay over
intermediate links, contention delay caused by nodes’ con-
tention for the shared channel and queuing delay induced
at each intermediate node due to queuing policy/or severe
channel conditions. The transmission delay is the time for
successful transmission, which is defined as the period from
the instant that a packet is transmitted for the first time to
the instant that it is either successfully transmitted or
dropped after a predefined number of retransmissions. The
contention delay is the access delay. And each retransmis-
sion will cause new access delay. The contention delay is
determined by the contention window (CW) size of node,
and the contention window size reflects the contention level.
The queuing delay at an intermediate node can be inter-
preted as the interval between the time a packet reaches the
node and the time this packet is to be transmitted.
For a path P : n1; n2; . . . ; ni; . . . ; nN ðN ! 1Þ, a packet is
sent from node n1 to nN according to path P. LðiÞðiþ1Þ is the
transmission delay of link between node i and node i þ 1.
Let Ci and Qi denote the contention delay and queuing
delay at node i, respectively. The total delay DP contains
the contention delay and the queuing delay at each node
and the transmission delay of links on the path P.
DP ¼
XNÀ1
i¼1
ðLðiÞðiþ1Þ þ Ci þ QiÞ: (1)
3.1.2 Delay Model of an Intermediate Node
If the data is transmitted successfully in the first attempt,
transmission delay of link between node i and node i þ 1 is
LðiÞðiþ1Þ ¼
L
B
þ DIFS þ TACK þ SIFS; (2)
where L denotes the packet length and B is transmission
data rate, DIFS stands for the Distributed Inter-Frame
Spacing, SIFS stands for the Short Inter-Frame Spacing
and TACK represents the transmission delay of acknowl-
edge frame.
However, transmission failure happens frequently, and
retransmission delay should also be considered. We
represent the retransmission times by the expect transmis-
sion count (ETX) [24]. The ETX of a link is calculated using
the forward and reverse delivery ratios of the link. The for-
ward delivery ratio df is the measured probability that a
data packet successfully arrives at the recipient; the reverse
delivery ratio dr is the probability that the ACK packet is
successfully received. The expected probability is df Ã dr
when a transmission is successfully received and acknowl-
edged. A sender will retransmit a packet that is not success-
fully acknowledged. As each attempt to transmit a packet
can be considered a Bernoulli trial, the expected number of
transmissions is ETX ¼ 1
df Ã dr
. Based on the expected num-
ber of transmission above, the expected transmission time
can be estimated by
EðLðiÞðiþ1ÞÞ ¼ ETX Ã LðiÞðiþ1Þ: (3)
In ad hoc networks, nodes contend for the shared chan-
nel, which causes access delay and collision at MAC layer.
To derive the contention delay of the link, we need to find
an explicit relation between the following two variables: the
attempt probability Pa that node i transmits in any virtual
slot and the conditional collision probability Pc of node i
given that at least a transmission attempt is made.
Pa ¼
2ð1 À 2PcÞ
ð1 À PcÞðCWmin þ 1Þ þ PcCWminð1 À ð2PcÞm
Þ
; (4)
where m ¼ log2ðCWmax
CWmin
Þ is the maximum number of back-off
stages. [25] claims that by re-deriving the Pc, Eq. (4) can be
applied to multihop wireless networks.
Before a node transmits, it performs physical carrier
sense. If a sender detects at least one other node transmit-
ting in the carrier sense area, it will encounter a collision.
However, the effective set of contending nodes is larger
than that, which includes virtual nodes accounting for the
effect of accumulative signals by multiple nodes outside the
carrier sense range. In [25], the effective set of contending
nodes jCSijeff is equal to a
aÀ2 jCSij. Thus, the conditional col-
lision probability can be represented by
Pc ¼ 1 À ð1 À PaÞjCSijeff ; (5)
where jCSij is the number of nodes in the carrier sense area
of sender i except itself and a is path-loss index. The
collision probability is Pcol ¼ Pa Ã Pc. If node i does not
receive an ACK within an interval of SIFS after the data
frame is transmitted, it knows that a collision occurs, then it
doubles CW and starts a back-off timer, uniformly distrib-
uted between ½0; CW À 1Š. Let tsk, tck be the time duration of
a success transmission and a collision for the kth access
channel, respectively. Time duration QAk caused by conten-
tion can be expressed by
QAk ¼ ð1 À PcÞtsk þ Pctck
tsk ¼ DIFS þ CWk þ RTS þ SIFS
þ CTS þ SIFS
tck ¼ DIFS þ CWk þ RTS;
(6)
where CWk ¼ 2kÀ1
CWmin.

The delay incurred by contending channel should take
the retransmission times into account. However, the backoff
times a sending node has to wait in retransmissions are cor-
related and not independent. With the channel access occur-
ring to conflict continuously for k À 1 times, the probability
for kth channel access is Pk ¼ PkÀ1
col Pa. Thus, the expected
delay for node i to access channel is represented by
EðCðiÞðiþ1ÞÞ ¼
Plog2m
k¼1 QAk Ã Pk
Plog2m
k¼1 Pk
: (7)
In the contention delay, packets collision and retransmis-
sions are taken into account. Thus, if there are some other
transmissions using the same link, the contention delay will
increase.
The queuing delay takes a significant portion of the
total delay over a hop. The current popular method for
calculating the queuing delay is based on Queuing The-
ory. If we choose the M/M/1/K queue [26] as our queu-
ing model, each node will be an M/M/1/K queue system.
We can define ¼
Pk
j¼1
wj
RTTj
to be the aggregated packet
arrival rate of the node where k is the number of active
sessions, R T Tj is the round-trip time and wj is the send-
ing window. And we can obtain the service speed of node
mi ¼
1
EðLðiÞðiþ1ÞÞ þ EðCðiÞðiþ1ÞÞ
; (8)
where EðLðiÞðiþ1ÞÞ is the transmission delay of node i and
EðCðiÞðiþ1ÞÞ is the contention delay of node i, which are
respectively obtained in Eqs. (3) and (7). Therefore, we can
calculate the queue length Lqi by using queuing theory and
obtain the the queuing delay of a node as
EðQðiÞðiþ1ÞÞ ¼
LqðiÞ
mi
: (9)
It needs to limitedly ensure the Poisson arrival and the
Exponential distributed service to use M/M/1/K. If we
take the more general G/G/1/K queue into consideration
for widely use, each node will be an G/G/1/K queue sys-
tem. Bolch et al. [27] gave a detailed analysis of G/G/1 and
the solution in theory. The average queue length of node is
LqðiÞ ¼ ri
1À ^ri
, where ri ¼ Ã ei
mi
and ^ri ¼ expðÀ 2ð1ÀriÞ
C2
Ai
Ã ri þ C2
Bi
Þ, ei
refers to the visit ratio of node i, CAi and CBi refer to the
coefficients of the inter-arrival and service times respec-
tively. However, it is difficult to obtain the variables and the
statistics in this model. Therefore, we choose the autoregres-
sive model AR(p) [28] to predict the future state based on
the historical status. By using the AR(p) model, we can
obtain the queuing delay at a certain time as follows:
qt ¼ kq
0 þ kq
1qtÀ1 þ kq
2qtÀ2 þ Á Á Á þ kq
pqtÀp þ sq
t ; (10)
where p is a nonnegative integer and kq
p 2 are parameters
of the AR(p) model, j ¼ 1; 2; . . . ; p. sq
t is a white noise
sequence with mean zero and variance of a fixed value and
is independent of qtÀj. Queuing delay qt at a certain time is
determined by historical values and white noise. Finally, we
can obtain the queuing delay of node i: Qi ¼ qt.
3.1.3 Delay Constraint at an Intermediate Node
From the above equations, the delay constraint for a path is
transformed into delay constraints at intermediate nodes.
However, it is an extremely hard problem to partition the
end-to-end delay constraint into each node accurately. We
define a max delay Dmax at intermediate nodes, which is
similar to the max transmit power. Obviously, Dmax is
determined by the information of the real-time requirement
Treal and the number of hops n. Treal is given by the require-
ment of applications. As to n, we use a prediction method
to estimate the number of hops. Singh and Dutta [29]
adopted the prediction model AR(p) in [28] to estimate the
hop count of the next routing path by historic record of
number of hops. We use a similar prediction method with
correction under unsteady state to estimate the path hop
count before a routing discovery. According to the AR(p)
model, we can obtain the estimation of the ith hop count by
the last p hop counts as below:
hi ¼ kh
0 þ kh
1 hiÀ1 þ kh
2 hiÀ2 þ þkh
p hiÀp þ sh
i ; (11)
where hi is the estimated ith
routing hops, and hiÀ1 Á Á Á hiÀp
are the last pth
measured historic routing hops respectively.
kh
i is the ith
parameter of the AR(p) about hops. sh
i is the
white noise in the prediction. With Yule-Walker equation,
we can solve the value of hi.
To verify the accuracy of the prediction method, we eval-
uate the predicted value and the measured value of hop
count with the parameters in Table 2 in Section 4. The num-
ber of nodes is set to 100, and there are 10 connections with
random sources and destinations. The hop count of a rout-
ing path is obtained by a routing discovery in the AODV
[30] routing protocol every 1 second. Suggested by [29], the
value of p in AR(p) model is set to 3. We calculate the devia-
tion between the predicted values and the measured values,
and count the statistics of the absolute value of the devia-
tion. Fig. 1 shows the average prediction deviation for each
connection with confidence level 95 percent. In the predic-
tion method, the difference between the estimated hop
Fig. 1. Average error of estimation for different connections.

count and the measured hop count is at most 1 with confi-
dence level 95 percent.
With Treal and n above, we could make a rough estima-
tion of Dmax as Dmax ¼ Treal
n . However, even on the same
path, the transmission delay of different links are different.
It doesn’t mean a path is not preferred if the transmission
delay of a link along the path is larger than Dmax, because
transmission delay of other links may be less than Dmax
which may let the total path delay DP less than the time
requirement Treal. Thus, setting Dmax to a fixed value is not
rational. Considering these, we make a balanced algorithm
to adjust the Dmax so that it could adapt to different links
dynamically. When a source node needs to send packets, an
initial Dmax is set to Treal
n and we define a balance factor tb
which has an initial value of 0. During the transmitting of
the packet, the actual transmitting time Ti between node i
and i þ 1 is recorded, this time indicates the actual period
that the packet traveled from node i to i þ 1 and it can be
obtained from the time stamp in the packet. We compare Ti
with Dmax, if Ti is less than Dmax, it means the actual trans-
mission on this link is completed while meeting the con-
straint of Dmax, and the time is less than the constraint
delay. By catching on this, we know that this transmission
could make a margin for the downstream link so that when
the transmission time of downstream link is greater than
Dmax but the difference is under the margin, we could still
forward in a low transmission power. Then the balance fac-
tor tb is set by tbþ ¼ Ti À Dmax and Dmax is updated by
Dmax þ ¼ tb. If Ti is greater than Dmax, that means the actual
transmitting time upstream exceeds the delay requirement
and we need to make a topology control we will present in
Section 3.2. After the balance, next node could get a new
Dmax which considers both actual transmitting of upstream
links and the estimated constraint. Dmax will affect when
topology control is made, i.e., control prematurely may
increase interference of other nodes and waste the limited
energy, control too late may let the path could not meet the
real-time requirement. By balancing Dmax dynamically, we
will make the topology control at a proper time.
3.2 Topology Control
The objective of the topology control algorithm is to mini-
mize the power consumption while satisfying both the
interference constraint and the delay constraint. The delay
calculation has been given in Section 3.1. We need to control
the transmission power of a node to satisfy the delay
requirement according to the delay information provided
by delay model. Then, interference-based topology control
algorithm is represented in the following part and the delay
constraint is appropriately added.
The total power could be minimized when satisfying the
constraint on the SINR at each receiver. Assumed that there
are m simultaneous transmission links, which affect each
other’s transmission, i.e., each has interference from others.
Given that ith ongoing transmission, let node si and node ri
be the sender and the associated recipient, respectively.
And DTsi
represents the delay for the packets to be transmit-
ted at node si. It is the sum of delay obtained from Eqs. (3),
(7) and (9). The SINR threshold to enable the successful
reception of data at receiving node ri is ri
. The available
SINR at receiver ri is SINRri . Psiri is the transmission power
from node si to node ri and A represents the set of active
nodes. Pmax is the maximum transmitted power. Dmax is set
to Treal
n . If the transmit power is large, the number of hops n
is small, and vice versa. Therefore, transmit power is related
to Dmax. The topology control problem can be formulated as
a constrained optimization problem:
min
X
si;ri2A
Psiri ; (12)
subject to
DTsi
Dmax
0 Psiri Pmax
SINRri
! ri
8

:
: (13)
Eq. (13) represents the delay constraint and the inter-
ference constraint, respectively. The transmitted power
of each node can not be larger than the maximum
transmitted power. And we also assume that a node can-
not transmit and receive at the same time and a node is
not allowed to receive packets from multiple nodes
simultaneously.
The expected value of SINR by taking radio propagation
aspects into account which is the multi-hop characteristic of
the network at receiver r is as follows:
SINRr ¼
Psr Á asr
2
À
PI
r þ sr
2
Á
dsr
b
; i ¼ 1; 2; . . . ; N; (14)
where asr
2
is the fading coefficient. dsr is the distance
between node s and node r and the signal strength decays
exponentially with respect to distance by b. sr
2
is the ther-
mal noise at the receiving node r. We ignore the thermal
noise in our simulation since it is ignorable comparing to
interference signal. PI
r is the multiple interference at node r.
In the networks, the transmission from node s to node r is
interfered by other nodes’ simultaneous transmissions, and
can be expressed as follows:
PI
r ¼ C Á
X
s06¼s
Ps0r Á a2
s0r
db
s0r
; (15)
where C is constant coefficient. s0
is a transmitter other than
the intended transmitter s, and Ps0r represents the transmis-
sion power of node s0
.
We focus on how to minimize the interference through a
distributed algorithm that uses local information. When
node s adjusts its transmit power, it affects the interference
among other receivers, vice versa, node r also suffers from
collective interference from other senders. To solve this
problem, we employ the distributed power control scheme
in [31] to achieve the optimal power assignment under
interference constraints
Ptþ1
sr ¼ Pt
sr Á r=SINRr; while Pt
sr Pmax; (16)
where SINRr can be obtained by feedback from the receiver.
For example, under 802.11 MAC protocol, the SINR informa-
tion can be inserted into the ACK frame. The sender can
adjust its transmission power according to this information.

In this topology control algorithm, when the estimated
delay is less than the delay constraint, transmission power
is minimized as shown in Fig. 2. For a node s, the original
transmission radius is R. As the minimizing of transmis-
sion power, the communication range is reduced to R1.
However, our transmission power is minimized according
to the SINR threshold to enable the successful reception of
data packets at the farthest one-hop neighbors (node r in
Fig. 2) and the available SINR at receiver, which means the
transmission power is minimized while maintaining the
former connection. When the estimated delay is larger than
the delay constraint, transmission power is increased to
reduce the hops of an end-to-end transmission. Since trans-
mission power is increased, the communication range is
increased, which indicates that the former neighbors of the
sender will still be the one-hop neighbors of the sender,
and the sender can have more neighbors due to the
increase of communication range. This is an important con-
sideration of topology control, because in MANETs, multi-
ple parallel sessions emerge frequently, which may lead to
the situation that a certain link is crossed by several differ-
ent sessions or data flows in MAC layer, that is, the collec-
tive behavior. And if a link is adjusted by other sessions
according to their requirements, it may affect the former
transmission on this link. To solve this problem, our topol-
ogy control algorithm adjusts the transmission power con-
sidering the SINR threshold to enable the successful
reception of data packets at receiving node, thus the former
connection will not be changed. Due to the simultaneous
transmissions, adjusting the transmission power of a node
may affect the interference around other nodes, and, thus,
a node may suffer from the aggregative interference from
other transmitters in the network. To minimize the interfer-
ence, we employ the distributed power control algorithm
[31], which estimates the interference using the actual
SINR of the receiver to adjust transmission power with a
constant iterative algorithm. The convergence of the adjust-
ing scheme was proved in [31] (as shown in Lemma 1).
Then, if the formulated delay satisfies the real-time
requirement, topology is controlled with only consideration
of interference; if not satisfied, both delay constraint and
interference constraint need to be considered. Let PD
sr repre-
sent the minimized transmission power to satisfy the delay
constraint, PI
sr represent the minimized transmission power
to satisfy the interference constraint, and PI
sr Pmax.
1) If PD
sr Pmax, the requirement for delay is too strict,
the existing maximum transmitted power cannot
meet the requirement. Other measures can be taken,
such as using better hardware.
2) If PD
sr Pmax, and PI
sr PD
sr , which means if the
delay constraint is met, the interference constraint
will not be met. Then, the transmission power is
adjusted to PI
sr. This is the global optimal transmis-
sion scenario, and this will lead to a more suitable
network environment.
3) If PD
sr Pmax, and PI
sr PD
sr , which indicates if the
interference constraint is met, the delay constraint
will not be met. Then, the transmission power is
adjusted to PD
sr . The global optimal transmission sce-
nario can not be found. Both the interference con-
straint and the delay constraint are met, but the
power consumption is not the minimum. Thus, our
algorithm can be transformed into the adjustment of
transmission power. And it is a simple power control
process at the physical layer.
3.3 Mobility
Delay and topology control algorithm have been studied
under the assumption that all nodes in the networks are sta-
tionary. Then the impact of mobility should be investigated.
First, our algorithm focuses on reducing a network inter-
ference to improve capacity while keeping connectivity and
delay requirements. Most of the nodes have minimum con-
nectivity. Frequent link breakages are prone to happen in
the mobile environment. The poor links which are easily
broken should be moved from the topology to reduce the
effects of mobility.
Second, if a receiving node moves around in a small area in
the transmission range of the sender in a lower speed, delay at
receiver is only determined by transmission delay, contention
delay and queuing delay. If the node has a higher mobility,
node may move out of the sender’s transmission range
quickly. The link between sender and receiver is unstable and
prone to break. Once the link breaks, the transmission delay
will become infinite. Thus, delay is also affected by mobility.
Let TMr denote the connect time between sender s and
recipient r before r moves out of the transmission range of
node s and DTr be the delay for packets transmitted between
s and r. If DTr TMr , transmission can be completed before
the link breakage; If DTr TMr , the link between s and r is
unstable. Before the reception of the packet at node r, the
link has been broken up. We call this unstable link as poor
link. And when DTr ¼ TMr , it is also considered as a poor
link. To avoid frequent link breakages and reduce the impact
of mobility on the interference-based topology control algo-
rithm, the unstable poor links are moved from the topology.
Fig. 3 illustrates the relative motion between sender s and
recipient r. Point S represents the position of node s at time
t1, Points R1, R2, R3 represent the relative positions of node
r at time t1, t2, t3, respectively. And D1, D2, R are the relative
distances between s and r. Distance between two nodes can
be calculated by using a radio propagation model [32].
Fig. 2. Topology control of the sender node s.

Node r is moving at a relative velocity of ~v as given in Fig. 3.
The position of node r at time t1 and t2 is known. But the
time t3 when node r moves out of the sender’s transmission
range is not known. We first state them using the law of
cosines as the following equations, which were derived by
Zhang et al. in [23] (Eqs. (6), (7) and (8) in [23]).
D2
1 þ v2
ðt2 À t1Þ
2
À 2D1 Á vðt2 À t1Þ Á cos u ¼ D2
2
D2
1 þ v2
ðt3 À t1Þ
2
À 2D1 Á vðt3 À t1Þ Á cos u ¼ R2
:
(
(17)
Then together with the equation SSR1R3
¼ SSR1R2
þ SSR2R3
where the term S represents the area of a triangle, the con-
nect time (TMr ¼ t3 À t2) can be obtained by solving the
equations at time t2.
Our proposed interference-based topology control algo-
rithm for delay-constrained mobile ad hoc networks is
shown in Algorithm 1 and Table 1 shows the parameters
notation.
We prove ITCD can converge to a stable operating point
in the following. Given that power control has power con-
straints Cmin and Cmax, node i could adjust its transmission
power within the range Cmin CðiÞ Cmax.
Lemma 1 (31). The distributed power control algorithm derived
from Eq. (16) can converge to the unique optimal power
assignment, with kCt
À CÃ
k at most after Ts ¼
½logðCmax=Þ=logðSINRÃ
=bÞŠ iterations, where Ct
¼
ct
ð1Þ; . . . ; ct
ðnÞ is the power assignment at slot t, CÃ
is
the optimal power assignment, SINRÃ
is the maximum achiev-
able SINR and is a predefined accuracy level.
Theorem. The algorithm ITCD can converge to a stable operat-
ing point.
Proof. First, we will prove that it means CITCD has a lower
bound CÃ
. 1) For the start, from statements 4-5 of algo-
rithm ITCD, we have C1
ITCD ¼ C1
. Then, after statements
12-15, we can have C1
ITCD ¼ c1
ð1Þ; . . . ; ð1 þ wÞc1
ði1Þ;
. . . ; ð1 þ wÞc1
ðikÞ; . . . ; ð1 þ wÞc1
ðisÞ; . . . ; c1
ðnÞ where
at 1th slot node i1 . . . is go in if-then statements 12-15 and
increase the transmission power by factor 1 þ w. Thus, it’s
obvious that C1
C1
ITCD. 2) Given that we have Ct
Ct
ITCD after t phases or slots. Then, after statements 4-5, we
have Ctþ1
that ctþ1
ðiÞ ¼ ct b
SINRri
¼ bda
i
P
j6¼i
ctðjÞ
da
j
, for dis-
tributed power control algorithm. Also, we can derive that
ctþ1
ITCDðiÞ ¼ ct
ITCDðiÞ b
SINRri
¼ bda
i
P
j6¼i ct
ITCDðiÞ =da
j . Thus,
we have Ctþ1
ITCD ! Ctþ1
. Through induction and analysis,
we have Ct
ITCD ! Ct
for each slot. It means CITCD has a
lower bound CÃ
. Second, the statement 13 implies that
Ctþ1
ITCD ! Ct
ITCD, which means that Ct
ITCD is descending tu
In conclusion, we claim that our algorithm ITCD is
converging.
Algorithm 1. ITCD
Path P: S; . . . ; i À 1; i; i þ 1; . . . ; R;
Ppre ¼ Pmax;
1: Forwarder i selects stable links which satisfy
DTr TMr ;
2: while (Psr Pmax) and (DTsr Dmax) do
3: {minimizing the power consumption while
satisfying the interference constraint}
4: SINRr ¼ Psr Á asr
2
=ððPI
r þ s2
rÞdsr
b
Þ;
5: Psr ¼ Psr Á r=SINRr;
6: {adjusting Dmax with the balancing factor tb}
7: if (Ti Dmax) then
8: tbþ ¼ Dmax À Ti;
9: Dmaxþ ¼ tb;
10: end if
11: {linkðs; rÞ can not meet the requirement for delay,
DTsr Dr
s and increase the transmission range}
12: if (DTsr Dmax) or (Ti Dmax) then
13: Psr ¼ minðPsrð1 þ wÞ; PpreÞ;
14: Ppre ¼ Psr;
15: end if
16: end while
Fig. 3. Connect time between s and r.
TABLE 1
Parameters Notation
Treal Delay constraint given by the requirement of applications
TMr Connect time between sender s and receiver r
before r moves out of the transmission range of node s
DTr Delay for packets transmitted between s and r
Psr Transmission range of node s to node r
Pmax Maximum transmission range of all nodes
Ppre A temporary variable to adjust the Psr
DTi1i2
The delay on linkði1; i2Þ
Dmax Delay constraint
SINRr Available SINR at a receiver r
a2
sr
Fading coefficient
PI
r
Multiple interference at r
s2
r
Thermal noise at r
dsr Distance between sender s and receiver r
r The SINR threshold to enable the successful
reception of data at receiving node r
Ti Actual transmission time between node i and i þ 1
tb Balance factor to adjust Dmax
Dpa End-to-end delay on path P.

4 PERFORMANCE EVALUATION
4.1 Simulation Environment
In order to evaluate the performance of the proposed
ITCD protocol, we compare ITCD with the conventional
AODV [30] protocol and the energy-efficient and delay-
constrained routing protocol [33] (for convenience, we
rename the protocol as EEDCR) which is to find an
energy-efficient path with explicit delay constraint. By
adjusting the transmission power in data packet transmis-
sions, EEDCR can select the optimal path with minimized
cost of links (path loss).
Algorithm 2. AODV Equipped with ITCD
Path P: S; . . . ; i À 1; i; i þ 1; . . . ; R À 1; R;
1: Sender S:
2: if a generated packet has a delay constraint Treal then
3: Estimate the hops number n to the destination
node R;
4: Dmax ¼ Treal=n;
5: Call the ITCD algorithm;
6: Insert Dmax into the header of RREQ packet;
7: Broadcast the RREQ packet;
8: else
9: Broadcast the conventional RREQ packet;
10: end if
11: if Sender S receives an RREP packet then
12: Choose the path with the least Dpa for routing;
13: Send data packets with the selected path;
14: end if
15:
16: Forwarder i:
17: if Forwarder i receives a new RREQ packet with
Dmax inserted then
18: Call the ITCD algorithm;
19: Dpaþ ¼ DTiÀ1;i
;
20: Rebroadcast the RREQ packet;
21: else if Forwarder i receives an RREP or data packet
then
22: Forward the RREP or data packet according to its
routing table;
23: end if
24:
25: Receiver R:
26: if Receiver R receives an RREQ packet then
27: Dpaþ ¼ DTRÀ1;R
;
28: Send an RREP packet to the source S with Dpa;
29: end if
We modify the source code of AODV in NS-2 (v2.34) to
implement our proposed protocol as shown in Algorithm 2.
After the breakage of a link, a route procedure may fail, there
are two solutions to deal with the link breakage: the first one
is that the node which detects the link breakage forwards a
route request (RREQ) packet according to the delay constraint;
the second one is that the node which detects the link break-
age forwards a route error (RRER) packet to the source node,
and then the source node forwards a new RREQ packet. We
need to clarify both ITCD and EEDCR do not use additional
control packets but insert necessary information into the
header of the RREQ or RREP packets, both ITCD and EEDCR
have the same routing discovery scheme as that of AODV.
Simulation parameters are as follows. The Distributed Coordi-
nation Function (DCF) of the IEEE 802.11 protocol is used as
the MAC layer protocol. The radio channel model follows a
Lucent’s WaveLAN with a bitrate of 2 Mbps. The topology
size is 1,000 Ã
1,000 m. VoIP is one of the most important appli-
cations with delay constrained, which is usually treated as
constant bit rate (CBR) [34], [35]. We consider CBR data traffic
and randomly choose different source-destination connec-
tions. Every source sends four CBR packets whose size is 512
bytes per second. Thus, the send window size is 512 Ã
4 bytes.
The mobility model is based on the random waypoint model
in a field of 1,000 m Â 1,000 m. In this mobility model, each
node moves to a random selected destination with a random
speed from a uniform distribution [1, max-speed]. After the
node reaches its destination, it stops for a pause-time interval
and chooses a new destination and speed. In order to reflect
the network mobility, we set the max-speed to 5 m/s and set
the pause-time to 0. The simulation time for each simulation
scenario is set to 200 seconds. In the results, each data point
represents the average of 20 trials of experiments. The confi-
dence level is 95 percent, and the confidence interval is shown
as a vertical bar in the figures. The detailed simulation param-
eters are shown in Table 2.
The experiments are divided to three parts, and in each
part we research the impact of one of the following parame-
ters on the performance of routing protocols:
Number of nodes. We vary the number of nodes from
50 to 300 in a fixed field to research the impact of dif-
ferent network density. In this part, we set the num-
ber of CBR connections to 15.
Number of CBR connections. We vary the number of
randomly chosen CBR connections from 10 to 20
with a fixed packet rate to research the impact of dif-
ferent traffic load. In this part, we set the number of
nodes to 150.
Delay constraint. We vary the delay constraint from
40 to 140 ms in a fixed field to research the impact of
delay constraint. In this part, we set the number of
nodes to 150, the number of CBR connections to 15.
TABLE 2
Simulation Parameters
Simulation Parameter NS-2(v2.34)
Simulation Time 200 s
Topology Size 1,000 m Ã
1,000 m
Max power 0.8 W
Carrier sense threshold 6.30957e-12
Noise floor 7.96159e-14
SINR of data capture 10
Min speed 1 m/s
Max speed 5 m/s
Pause time 0 s
Traffic Type CBR
Packet size 512 bytes
Max delay Treal
n
Bandwidth 2 Mbps

4.2 Performance with Varied Number of Nodes
Fig. 4 measures the average end-to-end delay of CBR pack-
ets received at the destinations with increasing network
density. Average delay is defined as the average delay of a
successfully delivered CBR packet from the source node to
the destination node. In MANETs, inappropriate transmis-
sion power will increase the delay. If transmission power is
too large, it will incur too many channel contentions, which
increases the backoff timer in MAC layer, so as to increase
the delay. On the other hand, small transmission power
may reduce the interference and contentions, however, it
will increase the number of hops, which increases the queu-
ing delay of nodes in a path. AODV does not adjust the
transmission power. When the transmission power is too
large, it will incur channel contentions, thus increase the
E2E delay. In EEDCR, contention is intensified in the route
discovery due to redundant broadcast of the RREQ packets.
In ITCD, transmission power is adjusted according to the
expected delay and the delay constraint. When the expected
delay is less than the delay constraint, transmission power
is minimized to reduce the interference, when expected
delay is larger than the delay constraint, transmission
power is increased to reduce the transmission hop so as to
reduce the E2E delay. The ITCD protocol decreases the
average end-to-end delay to balance the routing hops and
the network interference. On average, the end-to-end delay
is reduced by about 33.9 percent in the ITCD protocol when
compared with the conventional AODV protocol. Under the
same network conditions, the delay is reduced by about
22.4 percent when the ITCD protocol is compared with the
EEDCR protocol.
Fig. 5 shows the packet delivery ratio with increasing
network density. In ITCD, when the estimated delay is less
than the delay constraint, transmission power is adjusted to
minimize the interference while maintaining the former
transmission. Interference is reduced, the packet decoding
failures are reduced, thus, the packet loss probability is
reduced. Moreover, the estimated delay fully considers the
transmission characteristics of MANETs. Packets collisions
caused by simultaneous transmissions are taken into
account. When the transmission condition is not good
enough, which may make a transmission overtime, topol-
ogy is controlled to provide other choices. However, in
EEDCR, transmission power is adjusted according to the
cost of links (path loss) while inappropriate transmission
power in AODV may increase the interference and conten-
tions. Hence, the ITCD protocol can increase the packet
delivery ratio. On average, the packet delivery ratio is
improved by about 23.2 percent in the ITCD protocol when
compared with the conventional AODV protocol. And in
the same situation, the ITCD protocol improves the packet
delivery ratio by about 19.6 percent when compared with
the EEDCR protocol.
4.3 Performance with Varied Number of CBR
Connections
Fig. 6 measures the average end-to-end delay of CBR pack-
ets received at the destinations with increasing traffic load.
In ITCD, transmission power is minimized while keeping
the connectivity and packet collisions are taken into
account. Mobility is also considered to remove unstable
links in the topology. Thus, topology is controlled to
reduce interference and avoid the frequent packet colli-
sions. On average, the end-to-end delay is reduced by
about 34.8 percent in the ITCD protocol when compared
with the conventional AODV protocol. Under the same
network conditions, the delay is reduced by about 36.6 per-
cent when the ITCD protocol is compared with the EEDCR
protocol. When the traffic load is heavy, the inappropriate
Fig. 4. Average end-to-end delay with varied number of nodes.
Fig. 5. Packet delivery ratio with varied number of nodes.
Fig. 6. Average end-to-end delay with varied number of CBR
connections.

transmission power will lead to large interference and the
simultaneous transmissions will lead to frequent collisions.
By adjusting the transmission power, EEDCR may reduce
the interference, however, RREQ packets collisions will
also increase the end-to-end delay. That is why the perfor-
mance of EEDCR degrades dramatically when the number
of CBR connections increases.
Fig. 7 shows the packet delivery ratio with increasing
traffic load. As the traffic load increases, the packet drops
of the conventional AODV protocol dramatically increase
with the increase of traffic load. Both the EEDCR and ITCD
protocols increase the packet delivery ratio compared with
the conventional AODV protocol, because both of them
adjust the transmission power, which may reduce the inter-
ference. However, in EEDCR, contention is not taken into
account while in ITCD, the transmission characteristics are
fully considered. Moreover, node mobility is also consid-
ered to strengthen topology. On average, the packet deliv-
ery ratio is improved by about 23.7 percent in the ITCD
protocol when compared with the conventional AODV pro-
tocol. And in the same situation, the ITCD protocol
improves the packet delivery ratio by about 19.6 percent
when compared with the EEDCR protocol.
4.4 Performance with Varied Delay-Constraint
Fig. 8 represents the packet delivery ratio of the three
routing protocols under different delay constraints. In this
experiment, delay constraint is considered for all the three
protocols. When the data packets are overtime, they will
be discarded. The delay constraint increases from 100 to
200 ms. The packet delivery ratio of ITCD enhances
approximately to 77.3 and 61.5 percent respectively, com-
pared with that of AODV and EEDCR. Algorithm ITCD
considers the contention delay, and controls the topology
to reduce the contention and interference. Thus, its packet
delivery ratio is higher than other protocols. The looser
the delay constraint is, the larger the packet delivery ratio
of ITCD, AODV and EEDCR is. The delay-constrained
mobile ad hoc networks have a delay requirement. As the
conditions are loosed, the packet delivery that is along the
invalid path becomes effective. This can increase the
packet delivery ratio.
Fig. 9 shows the normalized routing overhead of the
three routing protocols under different delay constraints.
Control packets include Hello, RREQ, RREP, and RRER
packets. The normalized routing overhead is defined as the
ratio of the size of control packets to the size of all the data
packets which are successfully transmitted to the destina-
tions under a given delay constraint. As shown in Fig. 9,
EEDCR yields the largest routing overhead among all the
three protocols, since redundant RREQ packets should be
transmitted in its routing discovery. For the ITCD protocol,
more data packets can reach the destinations successfully in
the given delay constraint as the interference and contention
are reduced. Furthermore, node mobility is taken into
account to avoid frequent link breakages and more stable
links can be selected. As a result, the ITCD protocol incurs
less routing overhead than AODV.
5 CONCLUSION
In this paper, we propose an interference-based topology
control algorithm for delay-constrained mobile ad hoc net-
works. The objective of the topology control algorithm is to
adjust the transmission power to minimize interference,
which is contradictory to the requirement of delay con-
straint. When transmission power is increased to reduce
the delay, which increases the number of neighbors covered
by the transmission range and causes more interference
from other active nodes in the network. Therefore, we
make a tradeoff between reducing delay and minimizing
Fig. 7. Packet delivery ratio with varied number of CBR connections.
Fig. 8. Packet delivery ratio with varied delay-constraint.
Fig. 9. Normalized routing overhead with varied delay-constraint.

interference. First, the problem of minimizing the power
consumption while satisfying the interference constraint is
solved by iteration. Then, the transmit power is increased
to meet the delay constraint. The proposed algorithm con-
trols the topology to satisfy the interference constraint, and
increases the transmit range to meet the delay requirement.
The simulation results show that ITCD can reduce the delay
and improve the throughput performance effectively in
delay-constrained mobile ad hoc networks.
ACKNOWLEDGMENTS
The authors would like to thank the editors and the anony-
mous reviewers for their valuable comments and sugges-
tions. They would also like to thank Jingjing Xia, Leyi Wu,
Kaiheng Chen, Haitao Zhu, Bo Yang, Xulei Cao, Caifang Li,
and Hui Zhang for their helpful discussions. This work was
supported in part by the National Natural Science Founda-
tion of China under Grant 61379130, Grant 61073185, Grant
60673171, and Grant 61232018, and by the Anhui Provincial
Natural Science Foundation under Grant 11040606M139.
REFERENCES
[1] Q. Zhang and Y. Q. Zhang, “Cross-layer design for QoS support in
multihop wireless networks,” Proc. IEEE, vol. 96, no. 1, pp. 64–76,
Jan. 2008.
[2] R. Draves, J. Padhye, and B. Zill, “Routing in multi-radio, multi-
hop wireless mesh networks,” in Proc. ACM 10th Annu. Int. Conf.
Mobile Comput. Netw., 2004, pp. 114–128.
[3] Y. Yang, J. Wang, and R. Kravets, “Designing routing metrics for
mesh networks,” in Proc. IEEE Workshop Wireless Mesh Netw., 2005.
[4] G. Sharma, R. Mazumdar, and N. B. Shroff, “Delay and capacity
trade-offs in mobile ad hoc networks: A global perspective,”
IEEE/ACM Trans. Netw., vol. 15, no. 5, pp. 981–992, Oct. 2007.
[5] P. Li, C. Zhang, and Y. Fang, “Capacity and delay of hybrid
wireless broadband access networks,” IEEE J. Sel. Areas Com-
mun., vol. 27, no. 2, pp. 117–125, Feb. 2009.
[6] X. Zhu, P. Li, Y. Fang, and Y. Wang, “Throughput and delay in
cooperative wireless networks with partial infrastructure,” IEEE
Trans. Veh. Technol., vol. 58, no. 8, pp. 4620–4627, Oct. 2009.
[7] F. Xie, X. M. Zhang, J. F. Guo, and G. L. Chen, “A delay oriented
adaptive routing protocol for mobile ad hoc networks, (in Chinese
with English abstract,” J. Softw., vol. 16, no. 9, pp. 1661–1667, 2005.
[8] N. Burri, P. V. Rickenbach, R. Wattenhofer, and Y. Weber,
“Topology control made practical: Increasing the performance of
source routing,” in Proc. 2nd Int. Conf. Mobile Ad-hoc Sens. Netw.,
2006, pp. 1–12.
[9] M. Burkhart, P. von Rickenbach, R. Wattenhofer, and A. Zollinger,
“Does topology control reduce interference,” in Proc. ACM 5th Int.
Symp. Mobile Ad Hoc Netw.Comput., 2004, pp. 9–19.
[10] D. M. Blough, M. Leoncini, G. Resta, and P. Santi, “The k-neighbors
approach to interference bounded and symmetric topology control
in ad hoc networks,” IEEE Trans. Mobile Comput., vol. 5, no. 9,
pp. 1267–1282, Sep. 2006.
[11] J. Kim and Y. Kwon, “Interference-aware topology control for low
rate wireless personal area networks,” IEEE Trans. Consum. Elec-
tron., vol. 55, no. 1, pp. 97–104, Feb. 2009.
[12] A. Muqattach and M. M. Krunz, “A distributed transmission
power control protocol for mobile ad hoc networks,” IEEE Trans.
Mobile Comput., vol. 3, no 2, pp. 113–128, Apr.–Jun. 2004.
[13] S. C. Wang, D. S. Wei, and S.-Y. Kuo, “An SPT-based topology
control algorithm for wireless ad hoc networks,” Comput. Com-
mun., vol. 29, no. 16, pp. 3092–3103, 2007.
[14] M. Kadivar, M. E. Shiri, and M. Dahghan, “Distributed topol-
ogy control algorithm based on one- and two-hop neighbors’
information for ad hoc networks,” Comput. Commun., vol. 32,
no. 2, pp. 368–375, 2009.
[15] D. Y. Xue and E. Ekici, “Delay-guaranteed cross-layer scheduling
in multihop wireless networks,” IEEE/ACM Trans. Netw., vol. 21,
no. 6, pp. 1696–1707, Dec. 2013.
[16] R. G. Li and A. Eryilmaz, “Scheduling for end-to-end deadline-
constrained traffic with reliability requirements in multihop
networks,” IEEE/ACM Trans. Netw., vol. 20, no. 5, pp. 1649–1663,
Oct. 2012.
[17] P. Santi, “Topology control in wireless ad hoc and sensor
networks,” ACM Comput. Surv., vol. 37, no. 2, pp. 164–194, 2005.
[18] X. H. Jia, D. Y. Li, and D. Z. Du, “QoS topology control in ad hoc
wireless networks,” in Proc. IEEE Conf. Comput. Commun., 2004,
pp. 1264–1272.
[19] C. F. Chou and H. P. Suen, “Topology-control-based QoS routing
(TLQR) in wireless ad hoc networks,” in Proc. IEEE 17th Int. Symp.
Pers., Indoor Mobile Radio Commun., 2006, pp. 1–5.
[20] J. Tang, G. L. Xue, and W. Y. Zhang, “Interference-aware topology
control and QoS routing in multi-channel wireless mesh
networks,” in Proc. ACM 6th Int. Symp. Mobile Ad Hoc Netw. Com-
put., 2005, pp. 68–77.
[21] J. Mammen and D. Shah, “Throughput and delay in random wire-
less networks with restricted mobility,” IEEE Trans. Inf. Theory,
vol. 53, no. 3, pp. 1108–1116, Mar. 2007.
[22] X. Zhu, P. Li, Y. Fang, and Y. Wang, “Throughput, delay, and
mobility in wireless ad hoc networks,” in Proc. IEEE Conf. Comput.
Commun., 2010, pp. 1–9.
[23] X. M. Zhang, F. F. Zou, E. B. Wang, and D. K. Sung, “Exploring the
dynamic nature of mobile nodes for predicting route lifetime in
mobile ad hoc networks ,” IEEE Trans. Veh. Technol., vol. 59, no. 3,
pp. 1567–1572, Mar. 2010.
[24] D. S. J. De Couto, D. Aguayo, J. Bicket, and R. Morris, “A high-
throughput path metric for multi-hop wireless routing,” in Proc.
ACM 9th Annu. Int. Conf. Mobile Comput. Netw., 2003, pp. 134–146.
[25] Y. Yang, J. C. Hou, and L. Kung, “Modeling the effect of transmit
power and physical carrier sense in multi-hop wireless
networks,” in Proc. IEEE Conf. Comput. Commun., 2007, pp. 1–9.
[26] L. Takacs, Introduction to the Theory of Queues. London, U.K.:
Oxford Univ. Press, 1982.
[27] G. Bolch, S. Greiner, H. Meer, and K. S. Trivedi, Queueing Net-
works and Markov Chains: Modeling and Performance Evaluation
with Computer Science Applications. Hoboken, NJ, USA: Wiley,
2006, pp. 463–465.
[28] J. Hamilton and J. Douglas, Time Series Analysis. Princeton, NJ,
USA: Princeton Univ. Press, vol. 2, 1994.
[29] J. Singh and P. Dutta, “The temporal effect of mobility on path
length in MANET,” Int. J. Wireless Inf. Netw., vol. 19, no. 1, pp. 38–
48, 2012
[30] C. Perkins, E. Belding-Royer, and S. Das, “Ad hoc ondemand dis-
tance vector (AODV) routing,” RFC 3561, Internet Society, Jul.
2003.
[31] Z. Lotker, M. parter, D. Peleg, and Y. Pignolet, “Distributed powe-
firstr control in the SINR model,” in Proc. IEEE Conf. Comput. Com-
mun., 2011, pp. 2525–2533.
[32] T. K. Sarkar, Z. Ji, K. Kim, A. Medouri, and M. Salazar-Palma, “A
survey of various propagation models for mobile communication,”
IEEE Antennas Propag. Mag., vol. 45, no. 3, pp. 51–82, Jun. 2003.
[33] S. H. Lee and D. H. Cho, “On-demand energy-efficient routing for
delay-constrained service in power-controlled multihop cellular
network,” in Proc. IEEE 60th Veh. Technol. Conf., Fall 2004,
pp. 3100–3104.
[34] J. F. Lee, J. M. Chen, W. Liao, H. H. Lee, and M. C. Chen, “A prac-
tical cross-layer QoS mechanism for voice over IP in IEEE 802.11 e
WLANs,” in Proc. IEEE Multimedia Expo, 2007, pp. 1067–1070.
[35] D. Kumar, Y. Ryu, and H. Jang, “Quality of service (QoS) of voice
over MAC protocol 802.11 using ns-2,” in Proc. ACM Int. Work-
shop Commun. Des. Eval. Cultural Ecological Multimedia Syst., 2008,
pp. 39–44.

Xin Ming Zhang received the BE and ME
degrees in electrical engineering from the China
University of Mining and Technology, Xuzhou,
China, in 1985 and 1988, respectively, and the
PhD degree in computer science and technology
from the University of Science and Technology of
China, Hefei, China, in 2001. Since 2002, he has
been with the faculty of the University of Science
and Technology of China, where he is currently
an associate professor with the School of Com-
puter Science and Technology. From September
2005 to August 2006, he was a visiting professor with the Department of
Electrical Engineering and Computer Science, Korea Advanced Institute
of Science and Technology, Daejeon, Korea. His research interest
includes wireless networks. He has published more than 60 papers in
wireless ad hoc and sensor networks. He is a member of the IEEE.
Yue Zhang received the BS degree in computer
science and technology from Anhui Normal Uni-
versity, Wuhu, China, in 2010, and the ME
degree in computer science and technology from
the University of Science and Technology of
China, Hefei, China, in 2013. His research inter-
est includes wireless networks.
Fan Yan received the BS degree in electrical
engineering and information science from the
University of Science and Technology of China,
Hefei, China, in 2011. He is currently working
toward the PhD degree in the School of Com-
puter Science and Technology, University of Sci-
ence and Technology of China. His research
interest includes wireless networks.
Athanasios V. Vasilakos is a visiting professor
at the National Technical University of Athens,
Greece. He has authored or coauthored more
than 200 technical papers in major journals and
conferences, and author/coauthor of ﬁve books
and 20 book chapters. He was a general chair
and technical program committee chair for many
international conferences. He also was or is an
editor or/and guest editor for many technical jour-
nals, such as the IEEE Transactions on Network
and Service Management, IEEE Transactions on
Systems, Man, and Cybernetics-partB, IEEE Transactions on Informa-
tion Technology in Biomedicine, and the IEEE Journal on Selected
Areas in Communications special issues of May 2009, January 2011,
and March 2011, and the ACM Transactions on Autonomous and
Adaptive Systems and the IEEE Communications Magazine. He is the
founding editor-in-chief of the International Journal of Adaptive and
Autonomous Communications Systems and the International Journal of
Arts and Technology. He is also a general chair of the Council of Com-
puting and Communications of the European Alliances for Innovation.
He is a senior member of the IEEE.
For more information on this or any other computing topic,
please visit our Digital Library at www.computer.org/publications/dlib.

IEEE NS2 PROJECT@ DREAMWEB TECHNO SOLUTION

More Related Content

What's hot (19)

Viewers also liked (12)

Similar to IEEE NS2 PROJECT@ DREAMWEB TECHNO SOLUTION (20)

Recently uploaded (20)

IEEE NS2 PROJECT@ DREAMWEB TECHNO SOLUTION