www.tjprc.org editor@tjprc.

org

International Journal of Computer Networking,
Wireless and Mobile Communications (IJCNWMC)
ISSN(P): 2250-1568; ISSN(E): 2278-9448
Vol. 4, Issue 2, Apr 2014, 53-64
TJPRC Pvt. Ltd.

ROUTING SCHEMES ALGORITHMS FOR DISTRIBUTED CLUSTERING IN TWO
TIERED SENSOR NETWORK
RAJU M. SAIRISE
1
, SANGITA M. POKHALE
2
& UMESH WAGHMARE
3

1,2,3
Department of Electronics & Telecommunication Engineering, Maharashtra, India
1
Yadavrao Tasgaonkar College of Engineering & Management, Karjat, Raigad, Maharashtra, India
2,3
Babasaheb Naik College of Engineering, Pusad, Maharashtra, India

ABSTRACT
In the past few years, the use of higher power relay nodes as cluster heads in two-tiered sensor networks have
been proposed, to achieve various objectives including improved network lifetime. These relay nodes may form a network
among themselves and route data towards the base station. In such a model, the lifetime of the network is determined
mainly by the lifetime of these relay nodes, which, in turn, is directly affected by the data communication scheme.
Two-tiered sensor networks, where higher-powered relay nodes are used as cluster heads, have been proposed recently for
designing sensor networks. Assigning sensor nodes to clusters, in an energy efficient way, is known to improve the lifetime
of such networks.
In this paper we have proposed an efficient distribued algorithm for assigning sensor nodes to clusters in
two-tiered networks, using both single-hop and multi hop routing schemes. Our distributed clustering strategy allocates
sensor nodes to clusters, based on limited local information only. We have also compared our approach to a number of
existing heuristics recently proposed in the literature and have shown, through simulations, that our approach consistently
outperforms current heuristics. In summary, the quality of the solutions obtained using our approach is comparable to those
obtained using an ILP formulation, but the solutions can be generated very quickly, making it suitable for practical-sized
networks with hundreds of sensor nodes.
KEYWORDS: Relay Nodes, Heuristics, ILP
I. INTRODUCTION
A sensor network is an interconnection of low-powered, battery operated nodes, equipped with sensor(s),
processors, memory and wireless communication devices [1], [2], [3], [4].
A sensor network performs tasks through the collaborative efforts of a large number (hundreds or even thousands)
of sensor nodes that are deployed within the sensing field [1]. Energy conservation in a sensor network to maximize its
lifetime is a major research topic. One approach is to introduce some special nodes, known as relay nodes, in sensor
networks [5], [6] to be used as cluster-heads in two-tier hierarchical networks [6]. Each sensor node in a two-tier network
communicates its data, over relatively short distances, to some relay node, Within the range of the sensor node. Each relay
node communicates this data to some data collection center (often called a base station), either directly (single-hop
communication), or through one or more relay nodes (multi-hop communication). These relay nodes can be provisioned
54 Raju M. Sairise, Sangita M. Pokhale

& Umesh Waghmare

Impact Factor (JCC): 5.3963 Index Copernicus Value (ICV): 3.0

with higher energy, as compared to sensor nodes. However, like the sensor nodes, the relay nodes are also battery operated
and therefore, power constrained.
When the battery of a relay node, in a hierarchical architecture, is totally depleted, the sensor nodes which are
transmitting to this relay node will no longer be able to send their data to the base station. In such a situation, if the network
design does not provide fault tolerance, an entire region, currently handled by the failed relay node, becomes effectively
this may cause some other relay nodes, which are now required to carry the load of some additional sensor node reassigned
to them, to dissipate their energy at a higher rate and quickly become totally depleted. The effectiveness of a clustering
scheme depends on a number of factors such as the distribution of the sensor nodes, the number and the locations of the
relay nodes and the specific routing strategy used. In the single-hop model, where each relay node transmits directly to the
base station, it may be more effective to assign fewer sensor nodes to clusters which are further away from the base station,
rather than distribute the sensor nodes equally among the different clusters. In many applications, it may not be realistic to
assume that each relay node is aware of the positions and loads on all other relay and sensor nodes in the entire network,
or that there is a centralized controller to coordinate the activities of the nodes. Therefore, it is important to develop
distributed algorithms that can operate using local information only.
In the multi-hop data transmission model, [6], [11], [13] relay nodes, acting as cluster heads, not only transmit
data gathered from the sensor nodes in their respective clusters but also forward data from other relay nodes towards the
base station. Hence, the relay nodes form a network among themselves and use multi-hop paths for routing data to the base
station. It is convenient to view such a network as a two tiered network of sensor nodes and relay nodes. Each sensor node
transmits its data directly to its cluster head, a relay node, usually provisioned with a higher power. The sensor nodes form
the lower tier of the network. The relay nodes form a network for use in their communication paths to the base station,
using the multi-hop routing. This network of relay nodes forms the upper tier of the network. We note that single- hop
routing can be considered as a special case of multi-hop routing, where each relay node transmits its data directly to the
base station, i.e., the base station is the next-hop neighbor for every relay node in the network. In this paper, we present a
distributed algorithm for load balanced clustering to maximize the lifetime of a two-tier network using multi-hop routing.
Our algorithm is designed to handle the general case of multi-hop routing, and can be easily modified to accommodate
single hop routing.
As in [2], [7], the objective of our algorithm is to form appropriate clusters of sensor nodes with the relay nodes
acting as cluster heads. However, our approach takes a more comprehensive view that considers the cardinality of each
cluster as well as factors such as the routing scheme and the energy dissipation for transmitting and receiving data.
We also directly maximize the network lifetime, rather than optimize a secondary objective such as reducing the
differences in the cardinalities of the clusters [2], [3]. This makes our algorithm much more effective compared to existing
load balanced clustering techniques. An important feature of our algorithm is that it is quite fast and can quickly generate
efficient solutions for networks with hundreds of sensor nodes, with little overhead.
We have evaluated our algorithm, using simulation, for two cases - the single-hop case and the multihop case,
with a number of different networks and have compared the results with several commonly used heuristics and a clustering
scheme based on an Integer Linear Program (ILP) [7]. The ILP based clustering scheme is not distributed and we use it
only for comparison. The results show that our algorithm clearly outperforms the heuristics and the results are very close to
Routing Schemes Algorithms for Distributed Clustering in Two Tiered Sensor Network 55

www.tjprc.org editor@tjprc.org

the optimal solution. The remainder of the paper is organized as follows. In Section II, we have briefly reviewed some
relevant background material. In Section III we have presented our algorithm. We have discussed and analyzed the
experimental results in Section IV and have concluded with a critical summary in Section V.
2. MAIN ADVANTAGES OF CLUSTERING ROUTING ALGORITHMS IN WSNS
There exist various advantages in clustering routing algorithms compared with flat routing ones in WSNs.
The main advantages of clustering routing algorithms are surveyed as follows.
Increase of Scalability
In clustering routing scheme, sensor nodes are divided into a variety of clusters with different assignment levels
[19][20].
Decrease of Load
For clustering topology, all cluster members only send data to cluster heads, and data aggregation is performed at
the cluster heads, which help to dramatically reduce transmission data and save energy. [19][20].
Enhancement of Robustness
Compared with flat routing method, clustering routing scheme makes it more convenient for network topology
control and responding to network changes comprising node increasing, node mobility and unpredicted failures, and etc,
Alleviation of Collisions
In clustering routing scheme, a WSN is grouped into clusters and data communications between sensor nodes
comprise two modes, i.e. intra-cluster and inter-cluster, respectively for data collection and for information dissemination.
Reduction of Delay
In flat routing WSNs, data transmission is performed hop by hop usually using the method of flooding. In
contrast, in clustering based WSNs, only cluster heads perform the task of data transmissions from one cluster to another
one.
II. Review of Literature and the Network Model Used
In a hierarchical architecture, the sensor network is organized as a number of clusters and each sensor node
belongs to only one cluster. Some nodes are treated as cluster heads and have some additional responsibilities
(e.g., data gathering, data aggregation and routing) compared to the regular nodes. In the single-hop data transmission
model (also called the direct transmission energy model (DTEM)) [4], [8] the cluster heads send data directly to the base
station. Relay nodes, provisioned with higher energy, can be deployed in sensor networks to act as cluster heads [2], [6],
[7], [9]. In the multi-hop data transmission model, each relay node uses a multi-hop path to send data to the base
station [7]. A. Network Power Model The dominant factor in power consumption in sensor networks is the power needed
for wireless communication. In the first-order radio model discussed in [4], which we have used here, energy is dissipated
at a rate of 1 (2) nJ/bit for transmitting (receiving) the data. The transmit-amplifier also dissipates amount of energy to
transmit one bit of data over unit distance.
56 Raju M. Sairise, Sangita M. Pokhale

& Umesh Waghmare

Impact Factor (JCC): 5.3963 Index Copernicus Value (ICV): 3.0

The energy loss/bit due to channel transmission at a distance d is computed using the formula dq, where q is the
path loss exponent, 2 q 4, for free space, using short to medium-range radio communication [5]. Therefore, the energy
dissipated to receive b bits is given by ERx (b) = 2b and the energy dissipated to transmit b bits over a distance d is given
by, ETx (b, d) = 1b + bdq. The transmission power dissipated by a sender node to transmit each bit of data to a receiver
node increases rapidly with the increase of the distance between the sender and the receiver [1], [2], [4]. It is possible that
some relay nodes dissipate energy at higher rates due to more load and/or longer transmission distances compared to other
nodes.
This uneven energy dissipation among the relay nodes may lead to the faster death of some relay nodes in the
network due to the depletion of the batteries of these nodes, assuming that the initial energy provisioning for all relay
nodes are equal. This may affect the functionality of the sensor networks, as the inoperative node(s) will not be able to
perform their assigned functions. This may even cause the network to lose its usefulness, even though many other nodes in
the network still retain power. Clustering of nodes in a sensor network is a well researched field [2], [4], [7], [10].
Most clustering protocols use factors, such as the cluster ID or the degree of connectivity to form clusters. If a relay node j
is within the transmission range r of sensor node s, we will say that j covers s. The problem of clustering is illustrated in
Figure 1. Since the sensor nodes in the shaded region are covered by relay nodes A, B and C, one or more of these sensor
nodes may be assigned to any of the clusters corresponding to A, B or C. Depending on the routing scheme and the energy
dissipation of relay nodes A, B and C, one assignment may be more advantageous than the others. The goal of our
clustering algorithm is to assign each sensor node to an appropriate cluster in a way that extends the lifetime of the
network.

Figure 1: Sensor Nodes in Overlapping Coverage Area
In [6], the authors have focused on minimizing the number of higher powered relay nodes (acting as cluster heads)
such that all sensor nodes are covered by the relay nodes. This work does not discuss the assignment of sensor nodes to
clusters. In [2], the authors have provided a heuristic that attempts to minimize the variance of the cardinality
(defined as the number of sensor nodes associated with the cluster) of each cluster in the system to increase the system
lifetime. The clustering problem has also been addressed in [7] where an optimal solution is obtained using a centralized
approach employing an ILP. In [17], a distributed clustering scheme that minimizes the total energy spent in the system
due to the communication of data has been studied. In [18], a hybrid approach for clustering has been proposed where
cluster heads are selected randomly with a consideration to the residual energy, and then sensor nodes are assigned to
cluster such that the communication cost is minimized.
Routing Schemes Algorithms for Distributed Clustering in Two Tiered Sensor Network 57

www.tjprc.org editor@tjprc.org

Network Model
In our networks, we assume that a) all nodes are stationary after deployment, b) the locations of both the sensor
nodes and the relay nodes are known beforehand and c) the relay nodes are aware of the location of the base station.
If the locations are not known, each node may be equipped with a GPS system, which will enable the node to report its
current location. The GPS system will be turned on briefly [2], [3], [7] for initial configuration only. The average amount
of data generated by each sensor node is also known. In this paper we assume that each sensor generates an equal amount
of data.
However, we note that our scheme can be easily modified to accommodate sensors with varying amount of data.
We further assume that the placement strategy applied during the deployment phase of the network ensures proper
coverage of each sensor node (i.e., each sensor node is able to send its data to at least one relay node) and that the
network is connected. We are using the non-flow-splitting model of data routing, where each relay node has only one
transmitting station and can transmit to one single node only (either the base station or some relay node). This approach
relieves the relay nodes from performing complex routing functions and/or costly packet level power control for nodes that
are equipped with a single transmitter [12], [13]. The flow-splitting approach is also not suitable for use in conjunction
with directional antennae, which has been shown to improve the performance of wireless sensor and ad hoc networks [14],
[15]. In our model, data gathering is proactive (i.e., data is collected and forwarded to the base station periodically,
following a predefined schedule). In the case of multi-hop routing, in addition to the data generated by its own cluster,
each relay node also relays any data it receives from neighboring relay nodes.
We refer to each period of data gathering as a round [9], [11], [12]. In every round of data gathering, each relay
node gathers the data it receives from its own cluster and transmits that data, either directly to the base station
(for the single hop model) or forwards the data towards the base station using a multihop path (for the multi-hop model)
[7]. We will measure the lifetime of the network by the number of rounds the network operates from the start, until one
relay node depletes its energy completely and ceases to function. In [5] this is called the Nof- N metric to measure the
network lifetime. Assuming equal initial energy provisioning in each relay node, Nlifetime, the lifetime of the network,
using the N-of-N metric, is defined by Nlifetime = _Einitial/FMAX_ where Einitial denotes the initial energy of each relay
node and FMAX is the maximum energy dissipated by any relay node in a round. In such a model, it is easy to see that
maximizing the lifetime is equivalent to minimizing the maximum energy dissipation of any relay node in a
round exactly what we have done in our distributed algorithm and the ILP formulation.
III. Distributed Algorithm for Clustering
In this section we present an algorithm for distributed load balanced clustering for the multi-hop network model.
We assume, in the following discussions, that all sensor nodes have a transmission range of r, each relay node has a
transmission range of at least 2r and that this range is sufficient to transmit its data to the next node in the path to the base
station. In the distributed algorithm discussed here, our network has n sensor nodes, m relay nodes and one base station.
We assign a label of i, 1 i n to each sensor node, j, n < j m+n to each relay node, and n + m + 1 to the base station.
We will use S (R) to denote the set of labels for the sensor nodes (relay nodes) and use Cj, j R, to denote the set of sensor
nodes currently belonging to the jth cluster, with the relay node having ID j as the cluster head. We will use Uj to denote
58 Raju M. Sairise, Sangita M. Pokhale

& Umesh Waghmare

Impact Factor (JCC): 5.3963 Index Copernicus Value (ICV): 3.0

the set of sensor nodes that are covered by relay node j but have not been allocated to any cluster yet. When the algorithm
is over, i) S = _jR
Cj (and hence, Uj = , j R), and ii) Cj Ck = , j, k R, j = k. We consider relay nodes j, k R to be
neighbors if the distance between the relay nodes is 2r or less. This ensures that a sensor node can communicate with both
relay nodes j and k only if j and k are neighbors. Henceforth we will call the graph G = (V,E), the relay node graph,
where V is the set of all relay nodes in the network and there is an edge (j k) E if relay nodes j and k are neighbors.
We will say that a sensor node s S is essential to the cluster Cj of a relay node j R if only j covers s. The clustering
decision is taken by each relay node after taking into account only the situations in neighboring relay nodes.
Our experiments show that such decisions, based on local information only, still give excellent results. The central idea of
the algorithm is to start with each cluster Cj containing sensor nodes which are essential for relay node j. Each relay node
keeps information about its neighborhood and periodically broadcasts needed information (e.g., the maximum energy that
each relay is aware of) to its neighbors. Relevant information percolates through the entire network of relay nodes through
these periodic broadcasts.
The relay nodes gradually add sensor nodes to their respective clusters iteratively in a way that increases the
worst-case energy dissipation of the relay nodes (i.e., the value of FMAX) as little as possible. The algorithm has three
steps setup, initialization and cluster formation. For node j, the process of forming cluster Cj terminates when every
sensor node covered by j is allotted to some cluster, so that Uj = . During all the three steps, when relay node j is taking
decisions about cluster Cj and broadcasting information about itself, all other relay nodes in the neighborhood of j must be
in a quiescent state, passively listening to the message broadcast by j and updating their respective databases. This ensures
that two relay nodes never include the same sensor node in their respective clusters.
To achieve this, it is necessary to assign a color to each relay node j R, using a distributed graph-coloring
algorithm [16] on relay node graph G, so that if two relay nodes are neighbors, they are assigned two different colors.
Using any coloring algorithm [16], let a set P of colors, be used for coloring G. We have not included the details of the
coloring algorithm and assume that each relay node j R is aware of its color cj P when the algorithm starts.
A. Algorithm for Distributed Clustering in Multi-hop Networks (ADC-M) In our model, we assume that the network uses
multi-hop paths for routing data to the base station and the routing scheme is known. In a multi-hop network, relay node j,
in general, communicates using a multi-hop path j j1 j2 . . . jh m + n + 1, using some intermediate relay
nodes j1, j2, . . . , jh (h < m) to the base station n +m+ 1. Let Ej denote the total energy currently required by relay node j
and j the energy required to send the data corresponding to a single sensor node from j to the next relay node in the
multi-hop path from j to the base station. If a sensor node is now added to cluster Cj , the energy of the relay nodes
(j, j1, j2, . . . , jh) in the above multi-hop path become (Ej + j,Ej1 + j1,Ej2 + j2, . . . , Ejh + jh) respectively.
We will use Rj to denote the maximum of (Ej + j,Ej1 + j1,Ej2 + j2, . . . , Ejh + jh). In ADC-M (Algorithm 1,
given below), if a sensor node s is currently not allocated to any cluster and s is covered by two relay nodes j and k, the
algorithm includes s in cluster Cj only when Rj Rk. Relay node j, in general, is also an intermediate node in a number
( 0) of paths from other relay nodes to the base station. In other words, a number of relay nodes, say w1, w2, . . . , w_, use
j as the next node in their respective paths to the base station. We call nodes w1, w2, . . . , w_ as the predecessors of j.
Using any scheme (for instance, based on the labels of the relay nodes w1, w2, . . . , w_), we can always order the
Routing Schemes Algorithms for Distributed Clustering in Two Tiered Sensor Network 59

www.tjprc.org editor@tjprc.org

predecessor nodes of each node so that we can associate a predecessor number k, 1 k _ for each wk, 1 k _.
We will use M to denote the maximum possible number of relay nodes which can be predecessors of any relay node, n0j
to denote the number of sensor nodes which are in cluster Cj , ni j to denote the number of sensor nodes whose data is
routed from the ith predecessor of j, 1 iM, nj to denote the total number of sensor nodes processed using relay node j.
We can compute nj using the formula nj M_ p=0 np j , L to denote the maximum possible number of relay nodes in the
path of any relay node to the base station, Mj to denote the maximum energy required by all relay nodes, as currently
known by node j, Before the algorithm ADC-M starts, the sensor nodes operate only once, when each sensor node
broadcasts a hello message notifying its presence.
The ADC-M given below describes the operations of relay node j. All other relay nodes carry out the same steps
in a synchronized manner. In the setup step (lines 1-8 of Algorithm 1), our intent is to i) determine which sensor nodes are
essential to j, ii) form the initial cluster Cj , iii) determine the values of Ej and n0j , based on the essential nodes in Cj .
In lines 2- 8, each relay node exchanges the sensor node coverage information with its neighbors. Each iteration (lines 3-7)
give some relay node in the neighborhood of j a chance to be active exactly once. After all the iterations in lines 2- 8 are
over, relay node j becomes aware of the distribution of sensor nodes in its neighborhood.
In line 9, the set Uj is the set of sensor nodes that are covered by j that are not essential. We compute the values of
Mj, nj and Rj , based on current information, namely Mj = Ej , nj = n0j and Rj = EjThe initialization step (lines 11-17) must
be repeated 2 L times, so that the data broadcast by the relay node furthest from the base station has a chance to
propagate all the way to the base station and then back again to that relay node. When relay node j becomes active
(lines 13), it determines the new value of nj , determines the new value of Ej , the energy to send data from nj sensor
nodes to the next relay node, using the formula Ej j nj , determines whether Rj (Mj ) needs to be changed since Ej
may have changed.
When relay node j is not active (line 15), it receives data (Ck,Mk,Rk, nk) if relay node k in its neighborhood is
active. This data is used to
Change the set Uj ,
Change the value of Mj, if Mk >Mj ,
Change the value of Rj, if k is the next node in the multihop path from j to the base station and Rk > Rj , change
the value of nrj , if k is the rth predecessor of j. The cluster formation step (lines 19 - 34) allots all sensor nodes to
clusters. Lines 19 - 34 are repeated until all relay nodes report that they are not aware of any unallocated sensor
nodes. Details of the distributed process of determining that there is no unallocated sensor node have been omitted
since it is tedious but straight-forward. Inside the while-loop, the lines 21 - 32 are repeated for all values of ,
0 2L. Every time = 0 and Uj = , relay node j allots a number. (ranging from 1 to some small, pre-
determined number w) of sensor nodes from Uj to its own cluster Cj and to clusters of neighboring relay nodes. In
line 24, j examines all sensor nodes in Uj and temporarily allocates, w (or |Uj | sensors, if |Uj | < w) unassigned
sensor nodes in its neighborhood to the clusters for itself and all its neighbors. In this temporary allocation, if
Rj Mj , j starts by allocating one sensor node to itself, without looking at the value of Rk for any neighbor k of j.
60 Raju M. Sairise, Sangita M. Pokhale

& Umesh Waghmare

Impact Factor (JCC): 5.3963 Index Copernicus Value (ICV): 3.0

The remaining allocations (or all the allocations, if Rj >Mj ), are based on the values of Rj,Mj and Rk, for all
neighbors k of node j.
A sensor node is temporarily allocated to cluster Ck, where k is a neighbor (or to Cj ) in this process if the value of
Rk (or Rj ) is the least of all relay nodes that covers this sensor node. In these temporary allocations, j will increase the
value of Mj , if the energy Rk > Mj (or Rj > Mj ). In line 25, relay node j permanently includes, in Cj, the sensor nodes it
has allotted to itself. When = 1, 2, . . . 2 L, the result of including additional sensor nodes to clusters, carried out when
= 0, in terms of energy needed for different relay nodes in the multi-hop path from each relay node to the base station is
determined in a distributed manner. In other words, the changes in the values of Mj , Ej and Rj for relay node j, for all j,
n + 1 j m + n, corresponding to these changes in the sizes of the clusters are determined iteratively as varies from 1 to
2L. When all nodes in Uj have been allotted to clusters, j no longer carries out lines 24 - 26 since Uj = . To show that
the algorithm terminates, we note that during each iteration of the outermost loop (lines 20-33), there is at least
1 unassigned sensor node that is assigned to a cluster. Let j be one of the nodes such that Uj = . For a given iteration of
the outermost loop, when = 0 and cj = p, in line 24, node j becomes active and ensures that, in its neighborhood, at least
1 and at most w sensor nodes, from Uj, will be assigned to neighboring clusters and the value of Mj is adjusted so that these
sensor nodes may be assigned to the neighbors. The new value of Mj will be broadcast in line 28. There are two cases to
consider - Case I: node j itself absorbs at least one of sensor nodes in Uj during line 25 in the current iteration. Case II:
node j does not allocate any sensor nodes in Uj to itself in the current iteration. For case I, at least 1 unassigned sensor node
in the network has been allocated to cluster Cj .
For case II, in the remaining part of this iteration of the outermost loop and until j becomes active again in the
next iteration of the outermost loop, each of the neighbors of j will become active exactly once. Let k be a neighbor of j
which was temporarily assigned at least one sensor node by j in the current iteration. Clearly, Uk = when j did this
temporary assignment. Further, when j assigned a sensor to k, the value of Rk is the least of all relay nodes that can absorb
the sensor node. If Rk >Mj before the assignment, j will set Mj = Rk and will broadcast this new value of Mj in line 28.
When k becomes active (i.e., = 0 and the value of p in the inner loop from lines 21 to 32 matches ck), for the first time
following j, the value of Mk is such that it is at least Rk so that 1) either k will absorb at least one sensor node belonging to
Uk, 2) or Uk has become in the period since j was active. In both cases, the number of sensor nodes in Uk has been
reduced by at least 1. Thus, in all cases, every time there is an iteration of the outermost loop, the number of unassigned
sensor nodes in the network will be reduced by at least one. Since the number of sensor nodes in the network is finite,
the operations in node j must terminate in at most as many iterations as the number of sensor nodes.
Algorithm for Distributed Clustering in Single-Hop Networks
The algorithm ADC-M, proposed in section III-A, assumes that the relay nodes form a network of their own and
use multihop paths to route data to the base station. In a single-hop or direct transmission model, each relay node collects
data only from the sensor nodes belonging to its own cluster and sends the data directly to the base station. Our ADC-M
algorithm is also able to handle such transmission model by appropriately setting the values of requisite parameters.
To accommodate the single-hop transmission model in ADC-M, for each relay node j R, we set the number of
predecessors of j to zero (and hence L = 0). The values of j, j R, which is the energy required by j to send the data
gathered from a single sensor node, is computed such that the next hop of each j is the base station. The rest of the
Routing Schemes Algorithms for Distributed Clustering in Two Tiered Sensor Network 61

www.tjprc.org editor@tjprc.org

algorithm remains the same. In the following discussions, we shall denote the simplified Version of ADC-M that
incorporates the single-hop data Transmission model as ADC-S.
Algorithms
Algorithm 1 Algorithm for Distributed Clustering (ADC-M)
1: Node j determines the sensor nodes it covers.
2: for p P do
3: if cj = p then
4: j broadcasts the value of j and the list of sensors it can cover.
5: else
6: j receives, from its neighbor k, where ck = p, the list of sensors that k can cover.
7: end if
8: end for
9: Compute the initial value of Mj , Rj and nj .
10: for {1..2L} do
11: for p P do
12: if cj = p then
13: j computes and then broadcasts to its neighbors the values of Mj , Rj , Cj and nj .
14: else
15: j receives and processes, the values of Mk, Rk,
Ck and nk from neighbor k, where ck = p.
16: end if
17: end for
18: end for
19: while all sensor nodes are not allocated to clusters do
20: for {0..2L} do
21: for p P do
22: if cj = p then
23: if = 0 and Uj = then
24: j temporarily assigns, up to w sensor nodes from Uj , to the clusters for itself and all its neighbors.
25: j absorbs into cluster Cj the sensor nodes it has allocated to itself (if any).
26: n0j
|Cj |.
27: end if
28: j computes and then broadcasts to its neighbors the values of Cj , Mj , Rj and nj .
29: else
30: j receives and processes, the values of Ck, Mk, Rk and nk from neighbor k, where ck = p.
31: end if
32: end for
33: end for
34: end while
CONCLUSIONS
It is clear that these different clustering routing algorithms mentioned above are encouraging for improving the
performances of WSNs. However, some issues remain to be considered and there are still some open questions. First of all,
it remains a challenging problem how cluster formation is performed in heterogeneous WSNs where different types of
sensor nodes are deployed and each of them has different communication and processing capabilities. For load balanced
clustering in two-tiered sensor networks. An interesting and novel aspect of this research was to develop clustering
algorithms that can take into consideration the effect of different routing strategies and adapt the clusters accordingly.
The approach can lead to significant improvements over existing clustering algorithms, which do not consider the routing
schemes used for data transmission.
62 Raju M. Sairise, Sangita M. Pokhale

& Umesh Waghmare

Impact Factor (JCC): 5.3963 Index Copernicus Value (ICV): 3.0

REFERENCES
1. L. Robinson, Japans new mobile broadcast company: Multimedia for cars, trains, and handhelds. In Advances
Imaging, pp 18-22, July 1998.
2. J. Agre and L. Clare, An integrated architecture for cooperative sensing networks. computer, vol 33, No. 5,
pp 106-108, May 2000
3. Ossama Younis and Sonia Fahmy, HEED: A hybrid energy efficient, distributed clustering apparoach for Ad-hoc
sensor networks, IEEE Transactions on Mobile Computing, Oct.-Dec. 2004, Volume: 3, Issue: 4, p.p. 366 - 379.
4. W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, An application-specific protocol architecture for
wireless micro sensor networks, IEEE Transactions on Wireless Communications, vol. 1, no. 4, pp. 660670,
October 2002.
5. I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci. Wireless sensor networks: a survey.
Computer networks, vol. 38, pp. 393-422, 2002.
6. G. Gupta and M. Younis. Load-balanced clustering of wireless sensor networks. In IEEE International
Conference on Communications, vol. 3, pp. 1848-1852, 2003.
7. G. Gupta and M. Younis. Performance evaluation of load-balanced clustering of wireless sensor networks.
10
th
international Conference on Telecommunications, vol. 2, pp. 15771583, 2003.
8. W. Heinzelman, A. Chandrakasan, and H. Balakrishnan. Energy efficient communication protocol for wireless
micro-sensor networks. In HICSS, pp. 3005-3014, 2000.
9. J. Pan, Y. T. Hou, L. Cai, Y. Shi, and S. X. Shen. Topology Control for Wireless Sensor Networks. In Proc.
of International Conference on Mobile Computing and Networking, pp. 286-299, 2003.
10. J. Tang, B. Hao, and A. Sen. Relay node placement in large scale wireless sensor networks. Computer
Communications, vol. 29(4), pp. 490501, 2006.
11. Xu Y., Heidemann J. and Estrin D., Geography-informed energy conservation for ad hoc routing,
ACM MOBICOM, pp.7084, Jul.2001. Article (CrossRef Link)
12. Buttyan L. and Schaffer P., PANEL: Position-based aggregator node election in wireless sensor networks, in
Proc. of the 4th IEEE International Conference on Mobile Ad-hoc and Sensor Systems Conference, pp. 19.
Oct.2007. Article (CrossRef Link)
13. Luo H., Ye F., Cheng J., Lu S. and Zhang L., TTDD: Two-tier Data Dissemination in Large-scale Wireless
Sensor Networks, ACM Journal of Mobile Networks and Applications Special Issue on ACM MOBICOM,
pp.115, Nov.2003. Article (CrossRef Link)
14. Koutsonikola D., Das S., Charlie H.Y. and Stojmenovic I., Hierarchical geographic multicast routing for wireless
sensor networks, Wireless Networks, vol.16, no.2, pp.449466. Feb. 2010.
Routing Schemes Algorithms for Distributed Clustering in Two Tiered Sensor Network 63

www.tjprc.org editor@tjprc.org

15. Moufida Maimour, Houda Zeghilet and Francis Lepage, Cluster-based routing protocols for energy-efficiency in
wireless sensor networks, Sustainable Wireless Sensor Networks, InTech Open Access Publisher, 2010.
Article (CrossRef Link)
16. Poonguzhali, P.K., Energy Efficient Realization of Clustering Patch Routing Protocol in Wireless Sensors
Network, in Proc. of the 2nd International Conference on Computer Communication and Informatics,
Coimbatore, Tamilnadu, India, pp.1012, Jan. 2012. Article (CrossRef Link)
17. Wei D., Jin Y., Vural S., Moessner K. and Tafazolli R., An energy-efficient clustering solution for wireless
sensor networks
18. A.A. Abbasi and M. Youni, A survey on clustering algorithms for wireless sensor networks, Computer
Communications, vol.30, no.14-15, pp. 28262841, Oct. 2007.
19. Moufida Maimour, Houda Zeghilet and Francis Lepage, Cluster-based routing protocols for energy-efficiency in
wireless sensor networks, Sustainable Wireless Sensor Networks, InTech Open Access Publisher, 2010.