Energy Efficient Spanning Tree for Data Aggregation

in Wieless Sensor Networks

Zahra Eskandari Mohammad Hossien Yaghmaee AmirHossien Mohajerzadeh

Department of Computer Department of Computer Engineering, Department of Computer
Engineering, Ferdowsi Ferdowsi University of Mashhad Engineering, Ferdowsi
University of Mashhad Lane Department of Computer Science and University of Mashhad
e-mail: [email protected] Electrical Engineering, West Virginia e-mail: [email protected]
University, Morgantown, WV 26506
e-mail: [email protected]

Abstract—Wireless Sensor Networks (WSN) consist of some network lifetime, data transmission must be decreased, that
nodes that have limited processing power, memory and energy means we must minimize number of transmitted packets.
source. These constraints cause the algorithms that presented in
this field focus on these constraints. Data aggregation is any Energy consumption for transmission has direct correlation
process in which information is gathered and expressed in a with distance between sender and receiver [2]. In order to
summary form. Data aggregation has been put forward as an decrease the energy consumption, routing in these networks is
essential paradigm for wireless routing in sensor networks. The often performed in multi-hop architecture that causes energy
idea is to combine the data coming from different sources, efficiency and increase network scalability [3]. In this
eliminating redundancy, minimizing the number of transmissions architecture sensor nodes sense data from their environment,
and thus saving energy. For this purpose, sensor nodes must then transmit them to the intermediate nodes called relay
form aggregation tree, then forward sensed data to the root of nodes. These nodes transmit data to the next nodes in the path
this tree. Data is aggregated in intermediate nodes and the results toward the sink node. In this architecture intermediate nodes
are sent toward the root. In this paper we propose an energy must transmit not only the received data but also their local
aware algorithm for construction the aggregation tree. The sensed data toward the sink.
proposed algorithm considers both the energy and distance
parameters to construct the tree. Simulation results show that the During data transmission surly collision occurs. Therefore
proposed algorithm has better performance in terms of energy nodes must be scheduled properly. Using scheduling
efficiency and number of failed nodes which increases the mechanisms, we can solve the collision problem of data
network lifetime. transmission in wireless sensor network, but this solution
causes more delay in the network. Furthermore, transmission of
Keywords- Wireless Senssor Networks;Data Aggregation; this volume of data decrease relay node’s energy quickly and
Energy Aware Routing;Spanning Tree causes energy of these nodes to be finished.
Data aggregation is a good solution for these problems. In
I. INTRODUCTION
this case, intermediate nodes aggregate data, remove
Wireless sensor networks consist of some nodes that have redundancy in data and then transmit result to the next node in
limited processing capability, small memory and low energy the path toward sink. As shown in figure 1, because nodes are
source. These nodes are equipped with the sensors that sense deployed densely in the environment, the regions that the nodes
physical phenomenon such as light, pressure, temperature and sense, overlap one another and surly redundant data will be
humidity from their environment. These nodes are deployed generated [4].
randomly and often densely in the environment. In monitoring
applications, sensor nodes sense data from the environment Data aggregation is performed during routing in wireless
periodically and then transmit them to a base station which is sensor networks. Sensor nodes transmit data to the next node in
called sink node [1]. Thereby data transmission consumes the path and intermediate nodes aggregate data and then
node’s energy based on transmission distance. In most wireless transmit the result toward the sink. Finding the route from
sensor networks, the energy source of the node is limited and several nodes to the sink in a way that maximizes the shared
cannot be charged. In these networks, the energy of nodes that path and redundancy removing is one of the main objectives in
transmit data decrease quickly, so after some rounds of these routing protocols [4].
transmission, the power of nodes runs out and network can’t
sense data from the environment. Therefore to maximize

978-1-4244-2390-3/08/$25.00 ©2008 IEEE 1

 II. RELATED WORKS
During past few years, different aggregation algorithms
have been proposed for wireless sensor networks. In [10],
finding the optimal aggregation tree witch is an NP-hard
problem, has been considered. Data-centric routing has been
modeled and its performance has been compared with
traditional end-to-end routing schemes. The authors of [10]
examined the impact of source-destination placement and
communication network density on the energy costs and delay
associated with data aggregation. In [11] by considering the
location and remaining energy of all sensors and sink, a set of
trees for collecting data from all the sensors and sending them
Figure 1. Sensors region overlapping to the sink , has been constructed. Results show that the
lifetime is maximized. In [12] the sink saves global network
Transmission of aggregated data instead of raw data causes state and then by considering link cost, in centralized form,
intermediate nodes consume less energy. Furthermore it constructs minimal cost tree. In clustering algorithm given in
decreases collisions and delays for retransmission. Note that [13] after partition network into clusters with almost 20 or 40
data aggregation itself consumes energy [5]. Furthermore, nodes inside each cluster, cluster’s members construct
intermediate nodes must wait to receive data from all their aggregation tree and transmit data to cluster head. After
neighbors and then do aggregation. This waiting causes delay, aggregation, cluster heads transmit aggregated data to the sink
but the consumed energy and delay are negligible. As in one hop or multihop manner [14]. DCTC [15], dynamically
described in [6] the energy consumption for data processing is constructs an aggregation tree for mobile target tracking,
less than the necessary energy for data transmission. Also by dependant on target location, subset of nodes participate in tree.
considering the collision, retransmission delay and the needed Espan [8] is an energy-aware spanning tree algorithm. It is a
energy for retransmitting this data volume is certainly more distributed protocol and facilitates the sources within an event
than retransmission delay and consumed energy to transmit the region to perform data aggregation. In Espan, the source node
aggregation result [7]. In routing with aggregation, we must which has the highest residual energy is chosen as the root.
construct aggregation spanning tree [8]. The spanning tree is a Other source nodes choose their corresponding parent node
tree contains all network nodes and doesn’t have any loop. among their neighbors based on the residual energy and
distance to the root. One of the most important problems of this
When an event is sensed by sensor nodes, relevant data
algorithm is that nodes with least distance to root maybe
must be forwarded to sink. The sink is the root of spanning tree
selected as parent by many nodes. So these nodes consume
and all the other nodes that sense event, construct the tree. Each
their energy quickly and then they will be failed. In LPT [16]
intermediate node aggregates this data with data sent by its
after selection node with most energy as root, each node selects
child and then transmits this aggregated data to its parent. This
neighbors with most energy as parent.
procedure continues until data arrives to the sink.
To decrease energy consumption, all nodes are usually III. PROPOSED ALGORITHM
initially in sleep mode [9]. When an event is detected, nodes
that have sensed event, construct this aggregation spanning tree As mentioned earlier, to construct aggregation tree, the
and forward data to the sink based on this tree. To evaluate the Espan protocol first selects a node with highest remaining
aggregation tree, in the intermediate nodes the consumed energy within all network nodes as tree’s root. Then each node
energy for sending data from the leaves to the sink must be selects closest neighbors to root as its parent. If there are
considered. Furthermore the tree’s delay which is equal to the multiple neighbors with equal distance, the node which has
tree’s depth should be considered too. In addition to these most remaining energy is selected as parent. As Espan protocol
parameters, scheduling mechanism and queuing delay can be considers distance as main parameter and remaining energy as
considered as aggregation tree evaluation parameters. second, the network coverage is not high; because in some
cases the nodes with low remaining energy are selected as
In this paper, to construct aggregation tree we present a new parent. After local aggregation and data transmission, the
energy efficient algorithm. To decrease number of failed nodes remaining energy of these nodes is finished quickly. This
and to increase the network lifetime, we consider both causes the node failure and network can not coverage region
remaining energy and distance parameters. Each node selects a completely.
node that has the most energy within neighbors as its parent.
Furthermore, the distance from this parent to the root must be LPT aims to prolong the lifetime of the sources which
reasonable. transmit data reports periodically. In LPT, nodes which have
higher residual energy are chosen as the aggregating parents.
The remainder of the paper is organized as follows: in These parents may have higher distance to root and cause more
section 2, we review some existing algorithms for aggregation energy consumption. LPT does not consider the distance
tree construction. The proposed algorithm is explained in parameter in parent selection.
section 3. In section 4 using computer simulation we evaluate
the performance of the proposed algorithm under different Unlike the Espan algorithm, the proposed algorithm
scenarios. Finally, section 5 concludes the paper. considers the remaining energy in each node as the main

2
parameter. It also considers the distance as second parameter.
In the proposed algorithm, each node selects a node with most
energy within its neighbors as parent. If there are some
neighbors with equal energy, a neighbor with least distance
will be selected.
By using this strategy, a node with low remaining energy
can be alive more than that of Espan protocol. This increases
the lifetime of the network and supports better coverage. To
provide fairness in energy consumption, in addition to the
reminding energy and distance, we consider the third parameter
which is the maximum number of children. In the proposed
algorithm, each node could have a predetermined maximum
number of children. By using this parameter, we can provide
fairness in the energy consumption between all network nodes.
We determine this parameter based on nodes density and
network’s energy average.
In fact the proposed algorithm might select a node with
higher energy but farther from root as its parent, while Espan
selects a node that is closer to root and may has lower Figure 2. The spanning trees of different algorithms a) connectivity graph,
remaining energy. Furthermore, the second major difference b) Espan’s tree, c) LPT’s tree, d) proposed algorithm’s tree
between the proposed algorithm and Espan is that in the
proposed algorithm to provide fairness each node has a The details of the proposed algorithm are as below:
predefined number of children. The proposed algorithm is a Define:
distributed algorithm which doesn’t need to save global
information about entire network. Each node maintains Sr: A message which is sent by root and contains only one
information about its neighbors in its routing table. This makes field named root's ID.
the proposed algorithm more scalable. Furthermore, in the Sn: A message which is sent by each node n and contains
proposed algorithm routing is done in a multihop manner. three fields: node's energy, node's distance_to_root,
As the transmission power is directly related to the path_energy and node's ID.
distance, to prevent high power consumption, the proposed Initialization:
algorithm uses the average path’s energy as a new parameter.
This parameter is calculated as the sum of residual energy of For all nodes n in network set:
each node among the path divide to the path length. When a n.parentID=n
node wants to select a new parent, a node with the highest
path’s energy is chosen. n.parent's_distance_to_root=’
Unlike the LPT algorithm, the proposed algorithm prevents n.parent_energy=0
of selecting a parent with high remaining energy, but far Calculate max_child for each node base on density and
distance to the root. We use the average of path’s energy to network energy
balance energy and distance parameter while the algorithms
given in [8, 16] use only one of them as main parameter and n.children_number=0
the other in the next priority.
Protocol:
An example which helps us to understand the details of the
1. A node with highest remaining energy is selected as
proposed algorithm is given in figure 2. In this figure, a
the root node.
wireless sensor network with 8 different sensor nodes is shown.
The remaining energy of nodes 1, 2, 3, 4, 5, 6, 7 and 8 are 2. The root node broadcasts the Sr message to all its
equal to 10J, 2J, 8J, 3J, 6J, 8J, 7J and 9J, respectively. Suppose neighbors. This message conations the root’s ID.
that node 8 wants to select its parent. Using Espan algorithm,
node 4 will be selected as parent of node 8 which has minimum 3. Each root's neighbors n which receives this message,
distance to root, while in the proposed algorithm node 5 which select the root node as its parent and set:
has more average path’s energy is selected as the parent of a. n.parentID=root's ID
node 8. The average path’s energy of node 8 in the proposed
and Espan algorithms are equal to 7.5 and 6, respectively. In b. n.diatance to root=1
figure 2, for Espan, LPT and proposed algorithms, the spanning c. if (n.children_number < n.max_child)
trees are shown. As shown in figure 2(c), LPT’s tree has longer
path which causes more energy consumption. i. node n calculates path_enrgy
and broadcasts the Sn message
to all its neighbors

3
 4. When each node k (except the root) receives Sn the resulting aggregated data to the root. Network lifetime is
message, the following operations are done: defined as the longevity of network in terms of number of
rounds performed before the failure of 5% of the total nodes of
If(sender’s Path energy average > Parent’s path the tree [18].
energy average )
THEN B. Performance Evaluation
Node k selects the sender as its new parent In this subsection, using simulation results, we evaluate the
performance of the proposed algorithm. We produce some
Else events in the simulation network. When a node detects an
Node k doesn’t change its parent. event, forwards sensed data to the root node using aggregation
tree.
As mentioned above, the proposed algorithm selects a node
with the most remaining energy as the root. After that other All nodes that sense an event and forward their data to the
nodes which are root’s neighbors and have one hop distance to root, consumes energy for only data transmission (ETij).
the root, select this node as their parent and start to broadcast Intermediate nodes consume energy for both data receiving and
their information. Nodes that receive these messages, select data transmission (ERi and ETij, respectively). Furthermore,
their parents as described in the above algorithm. This they consume energy for data aggregation (EAi).
continues until all alive nodes connect to the tree. At the first simulation trial, we set the value of N to 400.
After each simulation round, we measure the number of alive
IV. SIMULATION RESULTS nodes. Figure 3, shows the results of the first trial. As the
proposed algorithm, selects the nodes with high remaining
In this section, using computer simulation, we evaluate the
energy, so the node with low energy remains longer time in the
performance of the proposed algorithm and compare it with
network, therefore the number of alive nodes are more than that
Espan algorithm.
of Espan and LPT algorithms.
A. Simulation Model
The simulation network environment is considered as a
25m*25m square area. The number of nodes (N) is varied
between 50 and 500. The initial energy of each node is a
random variable between 800J and 1000J. The communication
range of all nodes is set to 4 meter. We assume that all nodes in
network can sense occurred events and participate in
aggregation process.
Suppose at each node i the variables ERi and EAi represent
the required energy for data receiving and data aggregation,
respectively. Furthermore, suppose distij and ETij are the
distance and required transmission energy between node i,j,
respectively. Using the energy model given in [2], the
following equations are obtained:

ERi = C1 * K
(1)

ETij = C 2 * K * dist ij2 Figure 3. Number of alive nodes at N=400

(2)
As mentioned before, energy efficiency is a main objective
EAi = C3 * K of routing and aggregation algorithms in wireless sensor
(3)
networks. In figure 4, the average hop count is plotted versus
Where C1, C2 and C3 are three fixed variables which are number of nodes. These values represent average depth of tree,
determined based on energy and aggregation models in [17]. so, the deeper tree means data transmit via longer path, and this
We assume that the transmission power is directly related to causes more delay and also more energy consumption. As
the squared distance, this hold for free space, but the exponent shown in figure 4, unlike the proposed algorithm the LPT
heavily depends on the communication medium [18]. transmits data via longer path which causes that energy
As described in [8] if aggregation function be simple, the consumption of the proposed algorithm be less than that of
energy consumption for data aggregation is negligible. We LPT algorithm.
assume that the input data length is equal to the output data
length. In the above functions, K represents the length of
transmitted and received data packet.
In this paper, a run time round is defined as the collection
of one data unit from every node in the network and delivering

4
 construct aggregation tree. Simulation results show that the
proposed algorithm has better performance than existing
algorithms. Simulation results confirm that the proposed
algorithm decreases number of failed nodes and provides
higher network lifetime.

REFERENCES
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci,“Wireless
Sensor Networks: A Survey, Computer Networks”, Computer Networks
Journal, 2002
[2] Stephanie Lindsey, Cauligi S. Raghavendra, “PEGASIS: Power-
Efficient GAthering in Sensor Information Systems”, IEEE 2002
[3] Jin zhu, symeon papavassiliou, stella kafetzoglou,jie yang, ”An Efficient
QoS-Constrained Data Aggregation and Processing Approach in
Distributed Wireless Sensor Networks”, IEEE 2006
[4] Bing Liang and Qun Liu, “A Data Fusion Approach for Power Saving in
Wireless Sensor Networks”, IEEE 2006
[5] Hong Luo, Jun Luo, Yonghe Liu, and Sajal K. Das, “Adaptive Data
Fusion for Energy Efficient Routing in Wireless Sensor Networks”,
Figure 4. Average hop count to root IEEE 2006
[6] emanuele toscano, orazio Mirabella, iucia lo bello, An energy-efficient
In figure 5, for three algorithms the average lifetime is real-time communication framework for wireless sensor networks,
plotted versus number of nodes. These results were obtained [7] yi hu, nuo yu, xiaohua jia, energy efficient real time data aggregation in
after 20 different simulation trials. As the proposed algorithm wireless sensor network, IWCMC 2006
consumes less energy and has more alive nodes, so its lifetime [8] Marc Lee and Vincent W.S. Wong,” An Energy-Aware Spanning Tree
is higher than Espan algorithm and LPT algorithm. Algorithm for data aggregation in wireless sensor networks”, IEEE 2005
[9] Fei Hu,Xiaojun Cao,Carter May,” Optimized Scheduling for Data
Aggregation in Wireless Sensor Networks”, ITCC 2005
[10] B. Krishnamachari, D. Estrin, and S. Wicker. ”The Impact of Data
Aggregation in Wireless Sensor Networks”, International Workshop on
Distributed Event-Based Systems, 2002.
[11] K. Kalpakis, K. Dasgupta, and P. Namjoshi, "Efficient algorithms for
maximum lifetime data gathering and aggregation in wireless sensor
networks", The International Journal of Computer and
Telecommunications Networking, 2003.
[12] S. Upadhyayula, V. Annamalai, and S. K. S. Gupta, “A lowlatency and
energy-efficient algorithm for convergecast,” EEE GLOBECOM 2003.
[13] O. Younis and S. Fahmy, “HEED: A hybrid, energy-
efficient,distributed clustering approach for ad hoc sensor
networks,”IEEE 2004
[14] Yuanzhu Peter Chen, Arthur L. Liestman and Jiangchuan Liu,” A
Hierarchical Energy-Efficient Framework for Data Aggregation in
Wireless Sensor Networks ”,IEEE 2006
[15] W. Zhang and G. Cao, “DCTC: Dynamic convoy tree-based
collaboration for target tracking in sensor networks,” IEEE 2004
[16] Marc Lee and Vincent W.S. Wong,” LPT for Data Aggregation in
Figure 5. Average lifetime comparison Wireless Sensor Networks”, IEEE GLOBECOM 2005
[17] Xiaowei zhang, xiaohai yu, xiangqun chen, “Inter-Query Data
Aggregation in Wireless Sensor Networks”, IEEE 2oo5
V. CONCLUSION [18] S. Upadhyayula, S.K.S. Gupta, Spanning tree based algorithms for low
latency and energy efficient data aggregation enhanced convergecast
One of the most important constraints in wireless sensor (DAC) in wireless sensor networks, 2006 Elsevier
networks is the energy consumption. To decrease the power
consumption; we can decrease the number of transmitted
packets. Data aggregation is any process in which information
is gathered and expressed in a summary form. Data aggregation
has been put forward as an essential paradigm for wireless
routing in sensor networks. The idea is to combine the data
coming from different sources, eliminating redundancy,
minimizing the number of transmissions and thus saving
energy. Data aggregation is an effective mechanism to save
energy in wireless sensor networks. In this paper we propose an
energy efficient algorithm to construct aggregation tree. The
proposed algorithm considers both energy and distance to