(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 2, 2010
Clustering in Mobile Ad hoc Networks: A Review
Meenu Chawla
Jyoti Singhai
J L Rana
Department of CSEMANIT, Bhopal, Indiaranajl@manit.ac.in
Abstract
—
Mobile Ad-hoc Networks (MANETs) are futurewireless networks consisting entirely of mobile nodes thatcommunicate on-the-move without base stations. Nodes in thesenetworks generate user and application traffic and carry outnetwork control and routing functions. Dynamic and randomtopologies lead to rapidly changing connectivity and networkpartitions. This dynamic nature along with bandwidth and powerconstraints together pose new problems in network scalability,network control, especially in the design of higher level protocolssuch as routing, and in implementing applications with Quality of Service requirements. Hierarchical routing provides a means totackle the above mentioned problems in large scale networks.Clustering is the process of building hierarchies among nodes inthe network. In this approach an ad hoc network is partitionedinto group of nodes called as clusters. This paper presents areview of the different clustering algorithms and the criterion onthe basis of which each of them takes the clustering decisions.Keywords- Mobile Ad-hoc networks; clustering; clusterheadselection.
I.
I
NTRODUCTION
In hierarchical routing the nodes in the network aredynamically organized into partitions called clusters, and thenthe clusters are aggregated again into larger partitions calledsuper clusters and so on. The nodes geographically close toeach other form a cluster. Each cluster elects a leading nodecalled the cluster head which acts as a coordinator for thecluster. The nodes connected to more than one cluster arecalled gateway nodes and act as relays between clusters.Dividing a network into clusters helps maintain a relativelystable network topology. Clustering makes network moremanageable. Cluster size is controlled through the radiotransmission power.Cluster based algorithms are among the most effectiverouting algorithms due to their scalability[1,2,26]. Clusteringoutperforms other routing algorithms in case of large networks.As all inter-cluster routing in such a scenario is through thecluster head, it is therefore more burdened than its membersand tends to be a bottleneck in the system if not chosenappropriately. The objective of any clustering algorithm is topartition the network into several clusters which is the focus of current literature in this area.Several algorithms have been suggested for clustering andclusterhead selection. A number of clustering algorithms havebeen proposed, some very simple[3,4,5] and some with a viewof optimally utilizing the critical parameters[6,7,8,9,10,14]of ad hoc networks. A review of the clustering and cluster headselection algorithms is being done in this paper.II.
R
EVIEW OF
C
LUSTERING
A
LGORITHM
A.
Lowest ID algorithm
The Lowest-ID algorithm [3, 16] is the simplest clusteringalgorithm .In this algorithm every node in the network has aunique identifier (ID). Nodes periodically broadcast their ID in
“hello messages”. Each node compares the IDs of its neighbors
with its own ID, than a node having lowest ID decides tobecome a cluster head. The algorithm takes following steps:1. Every node broadcast its own ID periodically inHello message.2. All nodes receive hello messages from theirneighboring nodes and match their IDs then the node havinglowest ID is elected as cluster head.3. The node, which can hear broadcast from two clusterhead, is than becomes gateway node.In this algorithm there is no limit to the member nodes thata cluster can have. No network related parameter is given anyconsideration in selection of clusterhead, and hence theperformance of such networks is of random and unpredictablenature.
B.
Highest Degree Algorithm
The Highest-Degree heuristic [3,4] takes into account thedegree of a node, i.e. the number of its one-hop neighbors.Each node periodically broadcasts its degree value. A nodewith the highest value of degree in its neighborhood is selectedas the cluster head and its neighbors join it as cluster members.The procedure is repeated with the remaining nodes until eachnode is assigned to a cluster. Any tie is broken by the lowest id
criterion. This heuristic doesn‟t put any upper bound on
thenumber of nodes in a cluster, consequently the cluster headbecomes highly overloaded leading to performancedegradation. As the network topology changes, this approachcan result in a high turnover of cluster heads. This is becausewhen the highest connectivity node drops even one link due tonode movement, it may fail to be re-elected as a cluster head.These are two most popular criteria to partition mobilenodes. Both these algorithms do not provide any quantitativemeasure of cluster stability.
Corresponding author: Meenu Chawla
293http://sites.google.com/site/ijcsis/ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 2, 2010
C.
Least Cluster head Change Algorithm
k-CONID [5] combines the two approaches HighestDegree and LowestID. Connectivity is considered as a primaryand lower ID as a secondary criterion for selecting clusterheads. The algorithm considers at most k hop neighbours of anode for cluster head selection. At the beginning of thealgorithm, a node starts a flooding process in which aclustering request is send to all other nodes. In the Highest-degree heuristic, node degree only measures connectivity for 1-hop nieghbours. k-CONID generalizes connectivity for a k-hop neighborhood. Thus, when k = 1 connectivity is the sameas node degree.Each node in the network is assigned a pair: dID = (d, ID).
d is a node‟s connectivity and ID is the node‟s identifier. A
node is selected as a cluster head if it has the highestconnectivity. In case of equal connectivity, a node has clusterhead priority if it has lowest ID. Every node broadcasts itsclustering decision only after all its k-hop neighbors withhigher value of (degree, id) pair have broadcast their clusteringdecision.Although each node determines one cluster, clusters mayoverlap. This means that a node can belong to all clusterswhose cluster head is at most k-hops distance from the node.Nodes that belong to more than one cluster become gatewaynodes.
D.
(α,t) Cluster Framework
McDonald and Znati[6] have proposed a framework fordynamically organizing mobile nodes in a MANET into
clusters which has been called the (α,t)
-cluster framework. Theapproach is to maintain topology which allows for optimalrouting in face of low mobility and efficient routing if nodemobility is high. Here the focus is on mathematicalcharacterization of the probability of link and path availabilityas a function of a random walk based mobility model [21]. In
the ( α , t) approach
it is attempted to provide an effectivetopology that adapts to node mobility.Path availability is a random process which is determined
by the mobility of nodes that lie along a certain path. In the ( α ,
t) approach paths are evaluated by two system pa
rameters, αand t. α establishes a lower bound on the probability that agiven cluster path will remain available for a time t. α controls
cluster stability while the role of t is to manage cluster size fora given level of stability.The actions taken by the clustering algorithm depend uponthe information given by the routing and network-interface
layer protocol. Each node in the network is given a node‟s
cluster identifier number (CID) and makes use of a timer
named α timer. This timer establishes the m
aximum time t forwhich a node guarantees that paths will be available to each
cluster destination with probability = α.In the (α, t) algorithm, clusters which satisfy the (α, t)criteria are maintained. The (α, t) criteria is accomplished if the
probabi
listic bound α on the mutual availability of paths
between nodes in a cluster exists over a specified interval of time t. Therefore, the algorithm applies prediction of node
mobility as criteria for cluster organization. The (α, t) algorithm
characterizes the probability of link and path availability as afunction of a random walk mobility model.The algorithm is designed to take appropriate actions upontopological changes. A topological change requires that nodes
revaluate the (α, t) criteria. The docume
ntation that supportsthis clustering approach presents the pseudo code for five
important topological changes that determine the (α, t) cluster
algorithm: Node activation, link activation, link failure, node
deactivation and α timer expiration.
It has been shown that the (
,t)-cluster strategy has beeneffective in terms of adapting to node mobility, achieving nodestability in face of mobility and protocol efficiency.
E.
MOBIC
Basu et.al [10] proposed a variant of Lowest-ID algorithm,MOBIC, which is similar in execution to the Lowest-IDalgorithm except that the mobility metric is used as a basis of cluster formation instead of ID. MOBIC uses a new mobilitymetric; Aggregate Local Mobility (ALM) to elect CH. ALM iscomputed as the ratio of received power levels of successive
transmissions by transmitting periodic „hello‟ messages,
between a pair of nodes. This gives a measure of relativemobility between neighbouring nodes. Each node thencalculates aggregate local mobility metric M value bycalculating the variance (with respect to zero) of the entire setof relative mobility samples of all its neighbours. The nodewith lowest value of M becomes clusterhead.The main drawback of this algorithm is that it uses signalstrength as a measure of node mobility. However, because of noise, obstacles, variation in battery power, etc, weight basedon variation in signal strength may not be accurate, so stabilityof a node can not be evaluated clearly. Although mobility isone of the most important factors that can affect the stability of a clusterhead, there are other equally critical parameters thatneed to be considered for stable clusterhead selection.Considering only a single parameter will not give desiredstability in all types of scenarios. Also the algorithm here is justlooking at the stability of the clusterhead alone, and not at thestability of the complete network. To ensure stability of theentire network, consideration for stability of gateway nodes isof importance
F.
MobDhop:
A distributed clustering algorithm called MobDhop[9] hasbeen reported which partitions an ad hoc network into d-hopclusters based on a mobility metric. The objective of formingd-hop clusters is to make the cluster diameter more flexible.MobDhop is based on mobility metric and the diameter of acluster is adaptable with respect to node mobility. Thisclustering algorithm assumes that each node can measure itsreceived signal strength. In this manner, a node can determinethe closeness of its neighbors. Strong received signal strengthimplies closeness between two nodes. The MobDhop algorithmrequires the calculation of five terms: the estimated distancebetween nodes, the relative mobility between nodes, thevariation of estimated distance over time, the local stability,and the estimated mean distance. A node calculates its
294http://sites.google.com/site/ijcsis/ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 2, 2010
estimated distance to a neighbor based on the measuredreceived signal strength from that neighbor. Relative mobilitycorresponds to the difference of the estimated distance of onenode with respect to another, at two successive time moments.This parameter indicates if two nodes move away from eachother or if they become closer.The variation of estimated distances between two nodes iscomputed instead of calculating physical distance between twonodes. This is because physical distance between two nodes isnot a precise measure of closeness. For instance, if a node runsout of energy it will transmit packets at low power acting as adistanced node from its physically close neighbor. Thevariation of estimated distance and the relative mobilitybetween nodes are used to calculate the local stability. Localstability is computed in order to select some nodes as clusterheads. A node may become a cluster head if it is found to bethe most stable node among its neighborhood. Thus, the clusterhead will be the node with the lowest value of local stabilityamong its neighbors.MobDhop is executed in three stages as follows:
Discovery stage: At the initialization of the network, twohop clusters are to be formed in this stage. For this thenodes exchange hello messages periodically whichincludes the local stability value of the node(initialized tiinfinity). After a discovery period in which nodes acquirecomplete knowledge of their neighbour nodes, each nodecomputes its local stability value and broadcasts it forinformation to its neighbours. Node with lowest value of local stability becomes cluster head and its local stabilityvalue is the group stability (GS). If a node can hearmessages from a node that belongs to a different cluster,it becomes a gateway node. If not, it becomes a clustermember.
Merging stage: The two-hop clusters established in thediscovery stage are expanded by a merging process. Amerging process can be initiated by a nonclustered nodethat requests to join its neighbouring clusters or when twoneighbouring gateways request to merge their clusters.The merging is allowed only if the two merging criterionas stated are fulfilled. First condition ensures that thevariation of estimated distance between two mergingnodes should be less than or equal to the minimum valueof group stability of the two clusters. Second conditionstates that the mean distance between two gatewaysshould be less than or equal to the higher value of estimated mean distance of the two clusters. This is toensure that the distance characteristics of the clusters aremet.
Cluster maintenance stage: A cluster maintenance stage isinvoked when topology changes occur due to eitherarrival of a new node or a node leaving the network.When a node switches on it will begin the mergingprocess as described in order to join a cluster. When anode which is a clusterhead switches its immediateneighbours begin the discovery process as described sothat a new cluster head can be selected. During theperiod when the nodes are without a clusterhead(clusterhead election period) the two hop neighbournodes initiate a merging process and join otherclusterheads if the merging criterion is met.This algorithm also suffers from the same drawback asMOBIC. The algorithm here is just looking at the stability of the clusterhead alone, and not at the stability of the completenetwork. To ensure stability of the entire network,consideration for stability of gateway nodes is of equalimportance. If gateway nodes are highly mobile thenintercluster routes will break frequently leading to frequent re-routing causing high routing overhead. Also only the mobilitycriterion is taken into account for determining the stability of the network. Other parameter such as remaining battery powerof the node is an important parameter which affects the stabilityof the network and should be considered.
G.
DMAC
The Distributed and Mobility-Adaptive Clustering(DMAC) [7] algorithm provides a generalized solution forclustering framework. Nodes are assigned weights based on
nodes‟ mobility
-related parameters. The weights express howsuitable a node is for the role of cluster head given its own
current status. The bigger a node‟s weight, the more suitable it
is for the role of cluster head. This implies that, when due tothe mobility of the nodes two or more cluster heads becomeneighbors, those with the smaller weights have to resign andaffiliate with the now bigger neighboring cluster head. DMACovercomes a major drawback found in most clusteringalgorithms. A common assumption that is presented in mostalgorithms is that during the set up time nodes do not movewhile they are being grouped into clusters. Normally, clusteringalgorithms partition the network into clusters and only after thisstep has been accomplished, the non mobility assumption isreleased. Afterwards, the algorithm tries to maintain the clustertopology as nodes move. In real ad hoc situations thisassumption can not be made due to the constant mobility of nodes. Therefore one important feature of DMAC is that nodescan move, even during the clustering set up.During the algorithm execution it is assumed that each node
has a weight (a real number = 0) and an ID (node‟s unique
identifier) associated to it. The weight of a node representsnode mobility parameters. A node chooses its own role (clusterhead or ordinary node) based on the knowledge of its currentone hop neighbors. A node becomes a cluster head if it has thehighest weight among its one-hop neighbours; otherwise it joins a neighbouring cluster head.During execution of this algorithm, every node hasknowledge of its ID, its weight as well as its neighbors ID and
its neighbor‟s weight. DMAC is a message driven algorithm
(except during the initial procedure). Two types of messagesare used: If a node joins a cluster it sends out a Join messageand if it becomes a cluster head it sends a CH message. A nodedecides its own role once all its neighbors with bigger weightshave decided their roles.DMAC executes five procedures at each node: an initroutine, a link failure procedure, a new link procedure, aprocedure depending upon the reception of a CH message anda procedure depending upon the reception of a Join message.
295http://sites.google.com/site/ijcsis/ISSN 1947-5500
Search History:
Result 00 of 00
00 results for result for
p.
Clustering in Mobile Ad hoc Networks: A Review
Mobile Ad-hoc Networks (MANETs) are future wireless networks consisting entirely of mobile nodes that communicate on-the-move without base stations. Nodes in these networks generate user and application traffic and carry out network control and routing functions.…
(More)
4.5K
Reads
36
Readcasts
4
Embed Views
Recommended
More From This User
Notes
Load more
