HAQR: THE HIERARCHICAL ANT BASED QOS AWARE ON-DEMAND ROUTING FOR MANETS

David C. Wyld (Eds) : ICCSEA, SPPR, CSIA, WimoA - 2013
pp. 193–202, 2013. © CS & IT-CSCP 2013 DOI : 10.5121/csit.2013.3520
HAQR: THE HIERARCHICAL ANT BASED
QOS AWARE ON-DEMAND ROUTING FOR
MANETS
Debajit Sensarma 1
and Koushik Majumder 1
1
Department of Computer Science & Engineering, West Bengal University of
Technology, Kolkata, INDIA
debajit.sensarma2008@gmail.com koushik@ieee.org
ABSTRACT
A Mobile Ad Hoc Network (MANET) is a collection of wireless mobile devices with no pre
existing infrastructure or centralized control. Supporting QoS during routing is a very
challenging task. Clustering is an effective method for resource management regarding network
performance, routing protocol design, QoS etc. In real time various types of nodes with different
computing and transmission power, different rolls and different mobility pattern may exist.
Hierarchical routing provides routing through this kind of heterogeneous nodes. In this paper,
HAQR, a novel ant based QoS aware routing is proposed on a three level hierarchical cluster
based topology in MANET which will be more scalable and efficient compared to flat
architecture and will give better throughput.
KEYWORDS
MANET, Ant Colony Optimization, Hierarchical Routing, QoS Routing.
1. INTRODUCTION
A Mobile Ad Hoc network (MANET) [1] is a dynamically formed wireless network by the
mobile nodes. In this network nodes can move randomly. Due to the node mobility network is not
stable and topology is not fixed. So, stable routing is a very essential part in MANET. It is
possible to form a cluster of nodes which produces communication hierarchy. There are some
advantages of this type of hierarchical routing. Many of the contemporary ad hoc networks are
heterogeneous in nature. Mobile devices of this kind of network are equipped with different
communication capabilities with respect to frequency band, battery power, radio range, data rate
etc. E.g. in military networks, soldiers, tanks and command posts works in different interfaces.
Hierarchical routing makes the protocol more scalable. Furthermore, control overhead, routing
overhead decreases with the hierarchical routing scheme in comparison with flat routing and this
type of routing keeps the routing table size smaller in comparison with flat routing scheme.
Failure is isolated in hierarchical network topology. So, overhead of route maintenance also
decreases.
In this paper, clustering technique is used to organize mobile nodes in small cluster to provide a
hierarchical network structure of heterogeneous mobile nodes in MANET. This algorithm takes
advantages of Ant Colony Optimization (ACO) [3]. This proposed routing scheme decreases the
overhead and improves the overall performance of MANET.

194 Computer Science & Information Technology (CS & IT)
The paper is organized as follows: Section 2 describes the related works. In section 3 assumptions
of this algorithm is given. Section 4 describes the design of the proposed protocol. Section 5
explains the proposed HAQR protocol. Finally section 6 concludes the paper.
2. RELATED WORKS
Several cluster based hierarchical routing protocols has been proposed. In [4] a two layer
heterogeneous network has been proposed. It provides the internet access to the lower layer nodes
but it does not provide communication between lower layer nodes. The upper layer nodes are
gateways to provide the internet access to the lower layer nodes. In [5], an end-to-end QoS aware
routing in physically hierarchical Ad hoc network has been proposed. It considered the QoS
metrics bandwidth and node speed but other QoS metrics does not taken care of. In [6], HAODV
is proposed which is an extension of AODV. Here nodes are heterogeneous in nature and operates
on Wi-Fi or Bluetooth. It is a reactive routing and the advantage it helps to reduce the control
overhead. But the disadvantage there can be lacks of scalability as every time number of nodes
increases, new route discovery initiated. In [7], a dynamic adaptive routing protocol (DARP) has
been proposed. It is same as [6] and suffers from lacks of scalability. In [8], a weight based
adaptive clustering has been proposed. It tries to improve routing in heterogeneous MANET
using Global Positioning System (GPS). So, by knowing the user mobility pattern accurately, the
routing efficiency is increased. But when no GPS is available it is not very useful. In [9], a
heterogeneous routing protocol based stable routing has been proposed. It is a hybrid cluster
based routing protocol and combines AODV and DSDV for intra cluster and inters cluster
routing. But disadvantage is there can be a delay in AODV route discovery process.
3. ASSUMPTIONS
The following assumptions are taken into consideration to construct the protocol:
1. Here, a three level hierarchical cluster model is considered.
2. Ant Colony Optimization is used here for cluster head selection. Cluster heads are selected
based upon the weight value of the nodes. In each layer same procedure is followed. Each
cluster head can only have one hop neighbours and neighbours can communicate with each
other.
3. A node can communicate maximum with three levels based on the transmission power and
battery power. Level-0 nodes can communicate with only level-0 nodes. Level-1 nodes can
communicate with level-1 and level-0 nodes and level-2 nodes can communicate with all
three level nodes.
4. Here, hierarchical addressing is used for each node and based on the transmission power
nodes are represented with different shapes. Details of this presentation are depicted in
section 5.1.
5. In each level QoS constrains are taken care of and Ant Colony Optimization is used for both
intra and inter cluster routing.
4. DESIGN OF PROPOSED PROTOCOL
4.1 LOGICAL TOPOLOGY LEVELS
Fig. 2 illustrates the network architecture of the proposed routing protocol. The architecture is
designed as follows- Each clusters are leveled with ‘C’ followed by a topology level at which the
cluster is formed, followed by cluster head id of that cluster. For example, C0.A1 denotes level-0
cluster having A1 as cluster head. The nodes having single interface are denoted by white circles

Computer Science & Information Technology (CS & IT) 195
and represented by single digit (1, 2, 3 etc.). These nodes are formed only in bottom level.
Multiple interface nodes are represented by node’s name (e.g. A1, A2 etc.) followed by a digit
denoting the node’s interface where it operates. Nodes with interfaces indicated by triangles can
operate in three levels (e.g. A1.2, A1.1, and A1.0) and nodes denoted by dark circles can
communicate with lower levels (e.g. A3.1, A3.0).
Fig.2. Architecture of proposed three layer hierarchical structure
4.2 CLUSTER HEAD SELECTION AND CLUSTER FORMATION
In cluster head election process for each cluster, Cluster head is selected based on the weight
function, pheromone value, probability. According to [10] cluster head can be selected based on
combined weight metric of the node. Cluster head is selected with weight factors: Battery power,
Node Connectivity, Node Mobility and Distance. The algorithm is described as follows:
Algorithm 1: Weight Calculation Algorithm.
Begin
1. Find connectivity ‘c’ for each node which is the number of neighbours of each node. Find the
remaining energy, ‘e’ for each node.
2. Compute the mobility m for each node which is the running.
3. Compute the sum of distances d with all its neighbours for each node.
4. Calculate the combined weights weighti as
Member node (node with 1 interface)
Cluster head (node with 2 interfaces)
Connection between inter label clusters
Connection between intra label clusters
Cluster head (node with 3 interfaces)

weighti = w1*ci + w2*ei - w3*mi + w4*di
Here w1+w2+w3+w4=1.
ci = Connectivity of node i. i.e. number of nods that can directly communicate with the
given node within the transmission range of each other.
ei = Battery power of node i. i.e. remaining power of that node.
mi = Mobility of node i.
di = Distance of each node from i. i.e. sum of distance of the node from all its
neighbours.
End
Algorithm 2: Cluster Head Selection Algorithm.
Begin
1. Each node finds its neighbours and builds its neighbourhood table.
2. A set of clusters are formed from nodes and its neighbours.
3. Each node calculates its weight by calling the weight calculation algorithm given above.
4. At first, in a cluster, a node is selected as a cluster head randomly.
5. In each iteration, a node is selected as a cluster head based on probability of the neighbour
nodes. The probability of each node to be selected as cluster head is:
Pi =
∑=
n
k
k
i
1
τ
τ
(1)
6. Each time a node is selected as the cluster head, the pheromone value is updated according to
the following formula:
iτ = (1- ρ ) iτ + ρ (weighti) (2)
Here ρ is the pheromone evaporation factor (0< ρ <1).
7. Continue step 5-6 for each node in the cluster until all the nodes in the network are covered
(i.e. a node is a cluster head or falls within the range of an existing cluster head).
8. The node, whose weight value and the pheromone value is greater than its neighbour and
greater than threshold, is selected as cluster head and it sends message to all its neighbours.
9. On receiving a message, all neighbour nodes unicast an acknowledgment message to the
selected cluster head.
10. For each cluster continue the steps 3-9.
End
5. PROPOSED HAQR ALGORITHM
This is a three level hierarchical cluster based routing scheme which utilizes the Ant Colony
Optimization. This is also a QoS aware routing with parameters: delay, bandwidth, energy and
link expiration time. It takes care of QoS constrains in each layer. Here 5 kinds of packets are
used. Route_Ant is used by a node when it wants to know if the destination is a member of its
cluster head or not. Request Knave_Ant and Reply Knave_Ant are used for intra cluster routing.
Request King_Ant and Reply King_Ant are used for inter cluster routing. This algorithm has two
phases: Route discovery phase and route maintenance phase.

5.1. MATHEMATICAL MODEL
For mathematical analysis MANET is represented by a connected undirected graph. Let G (V, E)
represents the mobile ad hoc network. Here V denotes the set of network nodes and E denotes the
set of bidirectional links. QoS metrics with respect to each link e∈E is delay (e), bandwidth (e),
link expiration time (e). With respected to node n∈V, it is delay (n), energy (n). Another QoS
metric considered here is hop count. It is important because multiple hops are used for data
transmission in MANET. So, it is necessary to find paths with minimum hops. The main
motivation of this proposed algorithm is to find path from source to destination which will satisfy
the QoS requirements such as delay, bandwidth, energy, link expiration time.
Let, path (i, j) or R is entire path from node i to j where QoS constraints have to satisfied.
From an arbitrary node i to an arbitrary node j, delay, bandwidth, energy, link expiration time and
hop count is calculated as-
delay (path (i, j )) or D (R) = ∑ ∈ ),(
)(jiPe
edelay + ∑ ∈ ),(
)(jiPn
ndelay
where, delay (e) is the transmission and propagation delay of the path(i,j) and delay (n) is the
processing and queuing delay of node ‘n’ on path(i, j).
bandwidth(path(i,j)) or B(R)= ),(min jiPe∈ {bandwidth(e)}
where, bandwidth (e) is the available bandwidth of that link on path(i, j).
link expiration time or T(R) = ),(min jiPe∈ { link expiration time (e)}
where, link expiration time (e) is the expiration time of a link in route R.
energy (path (i, j)) or E (R) = ),(min jiPn∈ { energy (n)}
where, energy (n) is the residual energy of node ‘n’ on path(i, j).
hop count (path (i, j)) or HC (R) = Number of nodes in the path.
5.2. CALCULATION OF PHEROMONE
Ant deposits pheromone during traversal of the link for finding a route. The quantity of
pheromone it deposited on each link (i, j) along the route R is noted by ji,τ∆ and it is a function
of global quality of route R. It is expressed by the following equation-
ji,τ∆ =
HCD
TEB
RHCRD
RTRERB
λλ
λλλ
)()(
)()()(
+
++
(3)
Here λB, λE, λD, λHC and λT are the weight factors which indicate the relative significance of the
QoS parameters during pheromone update on path (i, j). The quantity of the deposited
pheromone is defined only after finding the route.
5.3. CALCULATION OF PATH PREFERENCE PROBABILITY
Path Preference Probability is calculated in each intermediate node as well as source node upon
receiving of Reply Knave_Ant or Reply King_Ant.
Suppose current node i receives Reply Knave_Ant or Reply King_Ant from node j for destination
d, then the Path Preference Probability is calculated as-

Pijd=
∑ ∈ iNk ijdikdikdikdikdik
ijdijdijdijdijdij
TEBD
TEBD
654321
654321
][].[].[].[].[][
][].[].[].[].[][
αααααα
αααααα
ητ
ητ (4)
Here α1, α2, α3, α4, α5 and α6 are the tunable parameters which control the relative weights of
pheromone trails, hop count, bandwidth, energy and link expiration time respectively.
Ni is the set of neighbors of i and k is the neighbor node of i through which a path to destination is
known.
The relative metrics are calculated from source i to destination d via j as-
Dijd
d))(i,delay(path
1
=
ijdη
d))ath(i,hopcount(p
1
=
Bijd= bandwidth (path(i,d))
Eijd = energy (path (i,d))
Tijd = link expiration time (path (i,d))
Now, source as well as neighbors has multiple paths from source to destination. The path with
higher Path Preference Probability is selected for the data transmission.
5.4. ROUTING IN PROPOSED HIERARCHICAL NETWORK
Algorithm 3: Route Discovery Phase
BEGIN
Suppose Source S wants to communicate with destination D.
/* Intra-cluster routing */
Step 1: S searches in its neighbour table to see if D is a neighbour of S.
Step 2: If S finds that D is its own neighbour, then it directly sends data packets to D.
Step 3: If S finds that D is not its neighbour, it unicast the Route_Ant packet to its cluster head
(i.e. CH(S)).
Step 4: CH(S) searches in its member table to see if D is a member of it.
Step5: If CH(S) finds that D is its own member, then it unicast the Route_Ant to S by setting
flag=1.
Step 6: After receiving Route_Ant from cluster head, S starts finding route to D which will
satisfy the required QoS constrains delay, bandwidth, energy, link expiration time and hop count.
Step 7: It first consults with its route cache, if an unexpired route exits to D, and then the route
with higher path preference probability is used for transmitting data.
Step 8: If cache contains no unexpired route, S initiates a Request Knave_Ant to the destination
through all its neighbours.
Step 9: While travelling to the destination the Request Knave_Ant collects delay, bandwidth of
each link and energy, link expiration time of each node.
Step 10: When the Request Knave_Ant reaches the destination it will converted to Reply
Knave_Ant and forwarded towards the original source. It will take same path as the Request
Knave_Ant but in reverse direction.
Step 11: For every Reply Knave_Ant reaching an intermediate node or source node, node can
find the delay, bandwidth, energy and link expiration time to the destination and node calculate
Path Preference Probability. If it is better than the requirements, and then the path is accepted and
stored in the memory.
Step 12: The path with higher Path Preference Probability will be considered as best path and
will be selected for transmitting data.

/* Inter-cluster routing in the same network or region */
Step 13: IF D is not a member of CH(S), it will unicast Request_Ant to its cluster head (CH
(CH(S))) in level-1hierarchy.
Step 14: If CH (CH(S))) finds D in its own member table but not in its cluster, as well as D is in
same network or region, then it will unicast the Route_Ant to the CH(S) and will set flag=1.
Step 15: After receiving Route_Ant from cluster head, CH(S) starts finding route to cluster head
of D (i.e. CH (D)), which will also satisfies the QoS requirements.
Step 16: At first, it consults with its route cache, if unexpired paths exists to CH (D)), then the
path with better Path Preference Probability is selected for data transmission.
Step 17: If cache contains no unexpired route, then CH(S) initiates a Request King_Ant to
destination through all its neighbour.
Step 18: The Request King_Ant also collects the delay, bandwidth, energy, link expiration time
and hop count of each links and each nodes respectively.
Step 19: When Request King_Ant reaches the destination cluster head, it will be converted to
Reply King_Ant and also forwarded to the CH(S) by the same path as Request King_Ant but in
reverse direction.
Step 20: Intermediate cluster nodes calculate the Path Preference Probability and if the Path
Preference Probability is better than requirements, then it is stored in the node’s cache.
Step 21: The path with higher Path Preference Probability is selected for data transmission and
the data is transmitted directly to the destination through its cluster head (CH (D)).
Step 22: If CH (CH(S)) finds D in its own member tables the member of its cluster, then the
communication starts directly through that cluster head [ i.e. CH CH(S) have at least 2 interfaces].
/* Inter-cluster routing in the different network or region*/
Step 23: If CH (CH(S)) finds D in its member table but D is in another network, then it will
forward the Request Knave_Ant packet to the corresponding cluster member node with required
QoS constraints in level-2. Then the member node unicast the request to the desired destination’s
cluster head and finally when the Request King_Ant packet reaches the destination, it sends the
Reply King_Ant packet to the original source by the same path but in reverse order.
Step 24: If CH (CH(S)) does not find D in its member table, it unicast Request packet to cluster
head (i.e. CH (CH (CH(S)))) and if D is in its member list, it sets the flag to 1 and sends to
CH(CH(S)) and route discovery continues like previous process.
END
Algorithm 4: Route Maintenance Phase
There are three cases in maintenance: single node leaving the cluster, single node joining the
cluster, the whole cluster moving together.
BEGIN
Maintenance of level-0 cluster.
Case 1: Leaving of a node in level-0 cluster.
Cluster head periodically will send the beacon to the member node and the member nodes send
acknowledgement to the cluster head. If a cluster head does not receive any reply from its
member within a specific period of time, then it can detect leaving of member from the cluster.
Again, if the cluster member cannot receive any beacon from the cluster head, then it can detect
leaving of its cluster head from the cluster.

Case 1.1: If an internal node moves out of a cluster, then the cluster head deletes the member
node entry from the member table.
Case 1.2: If a cluster head moves out of a cluster, nodes that do not belong to any cluster
started the cluster head selection procedure.
Case 2: Joining of a node in level-0 cluster.
When a cluster head starts receiving an acknowledgement from a new node, then it adds this new
member node to the member table.
Case 3: The whole cluster moving together.
When a cluster head comes in the transmission range of another cluster head at level-0, then a
new cluster head selection procedure started.
Case 4: Level-1 change caused by one node leaving level-0 cluster.
Case 4.1: Leaving node is the level-1 cluster head.
When level-0 cluster head cannot get any beacon from its cluster head, then a new cluster head
selection procedure started in level-1 as well as in level-0 cluster.
Case 4.2: Leaving node is not the level-1 cluster head.
When level-1 cluster head cannot find any beacon from a member node which is a cluster head of
level-0 cluster, it deletes its corresponding node entry from its member table and a new cluster
selection procedure starts at level-0.
Case 5: Level-1 change caused by new cluster head joining in level-0 cluster.
When a new cluster head selected in level-0, it is added to the member table of the cluster head of
level-1 which is in the same network or region.
Case 6: Level-2 change caused by one node leaving level-1 cluster.
Case 6.1: Leaving node is the level-2 cluster head.
In this case a new cluster head selection procedure is invoked. Here, if the leaving node has
two interfaces (i.e. level-2 and level-1), then new cluster head selection procedure is occur in
level-2 cluster and level-1. But if it has 3 interfaces, then new cluster head selection procedure
occurs in three levels.
Case 6.2: Leaving node is not the level-2 cluster head.
In this case, the corresponding node entry is removed from the cluster head member table.
Here, if the leaving node has two interfaces (i.e. lavel-2 and level-1), then new cluster head
selection procedure is occur only in level-1 cluster and if it has 3 interfaces, then new cluster
head selection procedure occurs in level-2 and leve1-1 cluster.
Case 7: Level-2 change caused by new cluster head joining in level-1 cluster.
When a new cluster head selected in level-1, it is added to the member table of the cluster head of
level-2 cluster.
END

6.CONCLUSION
As MANET is dynamic in nature, so QoS provisioning is very difficult in this network. In this
paper we proposed an Ant based hierarchical on-demand routing. It is a three level cluster based
routing algorithm. It takes the advantages of both the ant colony optimization and cluster
architecture. It is also a power efficient routing. Node’s transmission power plays a very crucial
role for increasing routing stability. Unlike other routings, QoS constrains are taken care of each
layer. So, it is very efficient for real time communication with heterogeneous nature of nodes.
Scalability also increases in this routing scheme. Here, an efficient cluster formation is used for
handling the increased number of nodes. Besides this, it emphasize on cluster maintenance for
reducing the overhead and delays of the network.
In future we will simulate this protocol and will compare it with other hierarchical routing
algorithms. Also we will take new QoS metrics to provide better throughput for real time
communication among the heterogeneous nature of nodes.
ACKNOWLEDGEMENTS
The authors would like to thank West Bengal University Technology, West Bengal, India for the
supports and facilities provided to carry out this research. The authors also thank the reviewers
for their constructive and helpful comments.
REFERENCES
[1] J., Macker, S., Corsen, IETF Mobile Ad Hoc Networks (MANET) Working Group Charter,
http://www.ietf.org/html.charters/manet-charter.html.
[2] K. L., Pang (2006), The Comparison Study of Flat Routing and Hierarchical Routing in Ad Hoc
Wireless Networks, ICON '06. 14th IEEE International Conference, vol. 1.
[3] D., Sensarma, K., Majumder, (2012), “A Comparative Analysis of the Ant Based Systems for QoS
Routing in MANET,” Recent Trends in Computer Networks and Distributed Systems Se-curity,
Communications in Computer and Information Science, Vol. 335, pp. 485-496.
[4] C.F., Huang, H.W., Lee, Y.C., Tseng (2004), A two-tier heterogeneous mobile ad hoc network
architecture and its load-balance routing problem, Mob. Netw. Appl., vol. 9, no. 4, pp. 379–391.
[5] K. Y., Lee, J., B., Hwang, J. D., Ryoo (2005), End-to-end QoS routing in physically hierarchical
wireless ad-hoc networks, Vehicular Technology Conference.
[6] H., Safa, H., Artail, M., Karam, H., Ollaic, and R., Abdallah (2007), HAODV: a New Routing
Protocol to Support Interoperability in Heterogeneous MANET, Computer Systems and
Applications, AICCSA ’07. IEEE/ACS International Conference on, pp. 893–900, May 2007.
[7] E. H. K., Wu, Y.Z., Huang (2004), Dynamic adaptive routing for a heterogeneous wireless network,
Mob. Netw. Appl., vol. 9, no. 3, pp. 219–233.
[8] Y., Wang, H., Chen, X., Yang, D., Zhang (2007), Wachm: Weight based adaptive clustering for large
scale heterogeneous manet, Communications and Information Technologies, ISCIT ’07. International
Symposium on, pp. 936–941, Oct. 2007.
[9] K., Xu, M., Gerla (2002), A heterogeneous routing protocol based on a new stable clustering scheme,
MILCOM 2002 Proceedings, vol. 2, pp.838–843 vol.2, Oct. 2002.
[10] M., Chatterjee, S.K., Das, D., Turgut (2001), WCA: A Weighted Clustering Algorithm for Mobile
Ad hoc Networks, Journal of Cluster Computing (Special Issue on Mobile Ad hoc Networks).

Authors
Debajit Sensarma has received his B.Sc degree in Computer Science in the year
2009 from university of Calcutta, Kolkata, India and M.Sc degree in computer
Science in the year 2011 from West Bengal State University, Kolkata, India. He
obtained his M.Tech. degree in Computer Science and engineering from West
Bengal University of Technology, Kolkata, India, in the year 2013. He is now
pursuing his PhD degree from the department of Computer Science and Engineering,
University of Calcutta, Kolkata. He has published several papers in International
journals and conferences.
Koushik Majumder has received his B.Tech and M.Tech degrees in Computer
Science and Engineering and Information Technology in the year 2003 and 2005
respectively from University of Calcutta, Kolkata, India. He obtained his PhD degree
in the field of Mobile Ad Hoc Networking in 2012 from Jadavpur University,
Kolkata, India. Before coming to the teaching profession he has worked in reputed
international software organizations like Tata Consultancy Services and Cognizant
Technology Solutions. He is presently working as an Assistant Professor in the Dept.
of Computer Science & Engineering in West Bengal University of Technology,
Kolkata, India He has published several papers in International and National level
journals and conferences. He is a Senior Member, IEEE.

HAQR: THE HIERARCHICAL ANT BASED QOS AWARE ON-DEMAND ROUTING FOR MANETS

More Related Content

What's hot (19)

Similar to HAQR: THE HIERARCHICAL ANT BASED QOS AWARE ON-DEMAND ROUTING FOR MANETS (20)

More from cscpconf (20)

Recently uploaded (20)

HAQR: THE HIERARCHICAL ANT BASED QOS AWARE ON-DEMAND ROUTING FOR MANETS