FFDA

A Tree based Energy Aware Data Aggregation Protocol In Wireless Sensor

Networks

Hamed Inanlou, Komail Shahmir Shourmasti, Hooman Marjani and Nima Attaran Rezaei

Computer Engineering Faculty, Islamic Azad University of Qazvin, Nokhbegan Blvd, Qazvin, Iran

Keywords: Wireless Sensor Network, Data Aggregation, Energy Efficient Routing, Tree based Routing.

Abstract: In wireless sensor networks (WSN's), data aggregation is used to increase energy efficiency by means of

eliminating redundancy and forwarding the collected abstract data of sensor nodes toward the sink. One of

the most important challenges in WSN is to keep the remaining energy of nodes high and balanced to

achieve longer system lifetime. In this article we propose an energy efficient data aggregation protocol

named FFDA (Feed Forward Data Aggregation) for constructing the spanning tree, we also represent a new

parameter called EAT (Energy After Transmission). This protocol considers EAT as the main parameter to

select a node as root for spanning tree in the beginning of each round of data aggregation. Using this new

parameter the remaining energy of nodes remain more balanced thus the first node die is delayed

significantly and it also improves the system's lifetime as indicated by simulation results.

1 INTRODUCTION

A wireless sensor network consists of some nodes

constructing a wireless network in order to sense the

environment and collect the sensed data for

processing and analyzing.

Each node consists of several parts including

radio antenna, memory, processor, sensor and a

power resource. Some of these resources have

constraint, specially the power resource which

usually is a small embedded battery. These nodes are

deployed in the environment often randomly and

stationary. In monitoring applications, when an

event occurs in the environment, the wireless sensor

network is triggered. Then all the nodes which

sensed that event capture the relevant data and send

their data to a base station (BS).

"Since wireless communications consume

significant amounts of battery power, sensor nodes

should spend as little energy as possible receiving

and transmitting data" (Lindsey and Raghavendra,

2002, p. 1). It's been a challenge to find a way to

decrease the energy consumption of nodes through

the process of data transmitting to get a longer

system lifetime from these limited resources. One of

the possible solutions to this problem is a process

called data aggregation.

Data aggregation is any

process in which information is collected and

expressed in an abstract form. The idea is to

combine all the data coming from different sources,

eliminating redundancy, minimizing the number of

transmissions and thus saving energy (Mohajerzadeh

and Yaghmaee and Eskandari, 2008).

In the case of tree based data aggregation, at first

nodes constitute a spanning tree based on the applied

protocol, then each node aggregates data received

from other nodes along with its own sensed data and

sends it to its parent. "This achieves a large

reduction in the energy dissipation, as computation

is much cheaper than communication" (Heinzelman

and Chandrakasan and Balakrishnan, 2000, p. 2).

We may continue the process of aggregation

based on the constructed spanning tree or we can

change it each round depending on the application

and distance of BS to the field. When BS is far away

from the field, a few transmissions from a node to

the BS could run its power out completely. therefore

considering the computation energy is a lot less than

long distance data transmission, we may prefer to

change the root and so the spanning tree each round.

In this article we introduce a tree based data

aggregation protocol FFDA (Feed Forward Data

Aggregation) in which we used EAT (Energy After

Transmission) factor as a new parameter to construct

the spanning tree. In section 2 we describe some of

Inanlou H., Shahmir Shourmasti K., Marjani H. and Attaran Rezaei N. (2010).

FFDA - A Tree based Energy Aware Data Aggregation Protocol In Wireless Sensor Networks.

In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 98-102

DOI: 10.5220/0002995100980102

 SciTePress

the proposed tree based data aggregation protocols

and describe their strengths and weaknesses. In

section 3 we explain our protocol FFDA in detail

along with the factor EAT and in section 4 we

evaluate the performance of FFDA algorithm and

compare it with some of recently proposed

algorithms.

2 RELATED WORKS

There are several tree-based aggregation algorithms

and each of them considers one parameter as the

main parameter to determine the aggregation tree’s

root and then construct the spanning tree based on

the selected root. Some of them select the node with

highest remaining energy among all nodes as root

while others consider the shortest distance to BS as

the main parameter. Moreover, many different

methods and parameters are used to construct the

spanning tree, either as a single factor or as a

combination of factors.

For example, Espan protocol (Lee and Wong,

2005) selects the node with the highest remaining

energy within the entire network as tree’s root, then

each node selects the closest neighbor to root as its

parent and if there are more than one neighbor with

same distance then the node with highest remaining

energy is selected as parent. So distance and

remaining energy are two parameters used in Espan

protocol to construct the aggregation tree while

distance has higher priority. This way it is possible

that a node with low remaining energy level is

selected as parent because of short distance to root,

therefore after data aggregation is done, this node

loses its energy quickly because of high traffic

passing through it and it leads to incomplete

coverage and system failure.

In LPT (Lee and Wong, 2005) nodes with higher

remaining energy are chosen as parents in

aggregation tree in order to increase the lifetime of

nodes with high traffic. This way a node far away

from root may be selected as parent because of its

high remaining energy levels but it will be drained

quickly due to long distance transmission energy

consumption.

In the proposed algorithm by Mohajerzadeh et al.

(2008) they had introduced an efficient algorithm in

which nearly most of the problems we described in

Espan and LPT are solved, and the simulation

results has shown their algorithm has better

performance than Espan and LPT algorithms in

terms of both first node die and system's lifetime.

But there is a big disadvantage in their protocol that

causes an unbalanced energy consumption

throughout the network. In presented algorithm by

Mohajerzadeh et al. (2008) the main factor to choose

a node as root for spanning tree is residual energy of

nodes, so a node with highest remaining energy

(HRE) is selected as root.

In the following section we describe the

drawbacks of HRE factor and introduce the new

factor EAT, then we will explain the proposed

protocol.

3 PROPOSED ALGORITHM

As mentioned earlier, in the beginning of each round

we have to select a node as root for spanning tree,

and then construct the tree based on selected root

until every single node is covered. In (Mohajerzadeh

et al., 2008), LPT and many other tree based

protocols, the node with highest remaining energy is

selected as root. while this method causes an

unbalanced network in terms of energy, we replaced

it with the factor EAT as a new parameter to keep

the remaining energy of nodes more balanced

through the process and thus prolong the lifetime of

the system and delay the first node die. In the

following section we describe the factor EAT in

details and then we represent the complete protocol.

3.1 EAT Factor

Suppose 5 nodes named N1 to N5 with initial energy

of 1300, 1300, 1800, 1600 and 1700 are respectively

positioned in a field. Distances between nodes and

BS are 16, 18, 15, 20 and 10 respectively. One of

these nodes is selected as root depending on which

factor we use to determine the root node. Then all

the nodes constitute a spanning tree to gather the

captured data and send the data towards the root,

where the root sends the aggregated data to BS in

order to analyze.

In figure 1 it is shown how the BS and nodes are

positioned in the field. Using the proposed algorithm

in Mohajerzadeh et al. (2008) the node number 3 is

selected as root since it has the highest remaining

energy among nodes. This may seem logical since

sending packets to BS drains the energy of root node

specially when BS is positioned far away from field,

but this is not the best way to select the root amongst

the nodes because it ignores the distances between

nodes and BS in root selection.

Therefore we introduce the new factor EAT and

replace the HRE factor with it. We calculate EAT

FFDA - A Tree based Energy Aware Data Aggregation Protocol In Wireless Sensor Networks

for each node and then select the root using this new

factor, then we compare the results to show how

EAT factor helps to keep the average remaining

energy of nodes higher and also how it makes the

remaining energy of nodes more balanced within the

network.

Figure 1: physical placement of nodes and base station.

We assume the aggregated data is a 10 bit packet

which will be sent from the root to BS. To calculate

the required amount of energy for a node to send a

packet to BS, we use formulas that will be described

later on section 4. To calculate the EAT for each

node, we need to calculate the required amount of

energy to send a 10 bit packet to BS from the node

and then subtract it from residual energy of the node.

For node number one, sending a 10 bit packet to

BS consumes a total energy of 1024, so the

remaining energy will be 276. Therefore EAT for

node number one is 276. We calculate the EAT

factor for each node before we actually send any

data to BS and before constructing the spanning tree.

As a matter of fact we use this factor to choose a

node as root for spanning tree. Therefore we choose

the node with highest EAT as root instead of the

node with highest remaining energy. Although node

number five has less energy than node number 3, a

shorter distance to BS results in a higher EAT 1300,

and we select it as the root for the spanning tree.

As you can see in table 1, after selecting the root

with two different parameters, EAT and HRE, total

energy of network is 7300 for EAT mode and 6800

for HRE mode. Total energy of system is the sum of

remaining energy of all nodes in the network.

Furthermore with EAT factor we have a more

balanced network in terms of energy.

In short, we can describe EAT as remaining

energy of each node if it is selected as root and sends

the aggregated data to BS. By using the EAT factor

we delay the first node die in the network by means

of keeping the remaining energy of nodes more

balanced and also keep total remaining energy of

nodes higher through the process.

Table 1: EAT and Highest Remaining Energy.

Node

number

Initial

energy

Distance

to base

EAT

Remaining

energy

(EAT root

selection)

Remaining

energy

(HRE root

selection)

1 1300 16 276 1300 1300

2 1300 18 4 1300 1300

3 1800 15 900 1800 900

4 1600 20 0 1600 1600

5 1700 10 1300 1300 1700

3.2 FFDA Protocol

In our proposed protocol, we consider the distance to

root as second parameter in root selection, therefore

if there are more than one node with the highest

EAT amongst nodes, the node with shorter distance

to BS is selected as root. This way the total energy

of network remains higher. We also used the

"maximum child number" parameter in order to

prevent a node with high density of nodes in its

neighborhood to get all of them as child. This

parameter could be either prefixed or variable.

The next parameter we used in proposed

algorithm is average path's energy (APE) first

introduced by Mohajerzadeh et al. (2008). APE is

the sum of remaining energy of nodes within a path

(up to the root) divided by the path’s length. When a

node wants to choose a parent, if there are more than

one option then the node with highest APE is

selected. Likewise if a node which is already in

spanning tree could select a new parent with higher

APE (while the new parent has not reached to

maximum child number limit), it does change its

parent to the new parent with a higher APE. By

using the factor APE we prevent a node with high

remaining energy to be selected as parent while it is

far away from root, APE combines both "Remaining

Energy" and "Distance to Root" parameters in a

single parameter.

4 SIMULATION RESULTS

We have evaluated the performance of our proposed

algorithm and then compared it with proposed

algorithm by Mohajerzadeh et al. (2008) which

outperformed LPT and Espan protocols.

WINSYS 2010 - International Conference on Wireless Information Networks and Systems

100

4.1 Simulation Model

We conducted the simulation in two different

simulation scenarios to evaluate the performance of

proposed protocol in different conditions. In first

mode, nodes are deployed in a 25m*25m square area

with the BS positioned in [0,0]. Number of nodes is

variable from 100 to 400 and the initial energy of

each node is a random variable from 80mJ to 100mJ.

Radio range of all the nodes are limited to 5 meters

during constructing the spanning tree while they are

capable of sending data to BS when they are selected

as root for spanning tree. Plus the maximum child

number is prefixed to 5. In second mode, the field is

a 40*40 meters square area with the BS in [20,60] to

evaluate the performance of the protocol in long

distance transmissions. Because the BS is located far

from field in second mode we consider the initial

energy of nodes higher than first mode as a variable

from 120mJ to 150mJ.

We assume all the nodes sense the environment

and send their data in each round of data

aggregation. All the nodes are homogenous and

stationary.

Energy consumption is calculated with following

formulas:

= C

(1)

= C

K dist

(2)

= C

(3)

Where ER and EA are the required energy for data

receiving and data aggregating for node i

respectively. ET

and dist

are required transmission

energy and distance between nodes i and j

respectively. In first order radio model there is also

an energy consumption for running the transmitter

circuitry though we ignored it in equation 2 because

of its minor effect on the result. C1, C2 and C3 are

prefixed values determined based on energy and

aggregation models in (Zhang and Yu and Chen,

2005), also the values are described in details for

first order radio model in (

Stanislava and Heinzelman,

2009)

. And K is the length of receiving and

transmitting packet by nodes. We assume input and

output data length are equal to 2000 bit.

A run time round is defined as the process of

collecting the data from all nodes in the network and

transferring the aggregated data to BS. The energy

consumed by each node in each round is as follows:

Leaf nodes consume the required energy to

transmit a packet to their parent. Intermediate nodes

consume the total energy required to receive

transmitted data by their child nodes, aggregating

received data and transmitting the aggregated data to

their parents. The root node consumes the energy

required to receive transmitted data by its child

nodes, aggregating received data and transmitting

the aggregated data to the BS.

The network lifetime is expressed as the rounds

passed until the point in which the number of

remaining alive nodes are less than 80% of the

number of the nodes in the first place.

4.2 Simulation Results

In this section we evaluate the performance of

proposed algorithm in different simulation

environments. when an event occurs in the

environment all nodes constitute the spanning tree

and send the captured data to the root using the

aggregation tree.

We repeated the simulation twenty times and

calculated the average values to get accurate results.

In figure 2, first node die round time is compared

between proposed algorithm in this paper and

proposed algorithm by Mohajerzadeh et al. (2008) in

first simulation environment. Using the factor EAT

keeps the remaining energy of nodes more balanced,

therefore the first node die round time is delayed

significantly regardless of the density and number of

deployed nodes.

Figure 2: First node die round comparison.

In figure 3 system's lifetime is compared between

the two mentioned algorithms in second simulation

environment. The affect of using the EAT factor is

related to the number of deployed nodes when

analyzing the system's lifetime. When the density of

nodes is low and the distances between them are

high the effect of the EAT factor is more noticeable.

FFDA - A Tree based Energy Aware Data Aggregation Protocol In Wireless Sensor Networks

101

Figure 3: System lifetime comparison.

5 CONCLUSIONS

One of the most challenging parts in wireless sensor

networks is to decrease the energy consumption of

nodes to delay the system failure. Two important

factors that have a direct effect on system failure are

the energy of network and the equality of remaining

energy of nodes in the network. In this paper we

introduced an energy aware protocol in which we

used a new factor called EAT. Using this factor as

the main factor to select the root for aggregation tree

the remaining energy of nodes remain more

balanced. therefore the system failure is delayed. As

we evaluated EAT factor in simulation section, it is

shown that HRE factor can be replaced with EAT in

similar protocols which use the highest remaining

energy as the main factor to select the root node.

REFERENCES

Heinzelman, W. R. Chandrakasan, A. and Balakrishnan,

H. (2000). Energy-Efficient Communication Protocol

for Wireless Microsensor Networks. Proceedings of

the 33rd Hawaii International Conference on System

Sciences, 8, 8020.

Lee, M. and Wong , V. W. S. (2005). An Energy-Aware

Spanning Tree Algorithm for data aggregation in

wireless sensor networks. IEEE Pacific RIM

Conference on Communications, Computers and

signal Processing(PACRIM), 300-303.

Lee, M. and Wong, V. W. S. (2005). LPT for Data

Aggregation in Wireless Sensor Networks. Global

Telecommunications Conference, GLOBECOM '05.

IEEE, 5, 6 pp. - 2974.

Lindsey, S. and Raghavendra, C. S. (2002). PEGASIS:

Power-Efficient Gathering in Sensor Information

Systems. Aerospace Conference Proceeding, 3, 1125-

1130.

Mohajerzadeh, M. Yaghmaee, M. H. and Eskandari, Z.

(2008). Tree Based Energy Efficient and Congestion

Aware Routing Protocol for Wireless Sensor

Networks. 11th IEEE Singapore International

Conference on Communication Systems (ICCS2008).

1707-1711.

Stanislava, S. Heinzelman, W. B. (2009). Cluster head

election techniques for coverage preservation in

wireless sensor networks. Ad Hoc Networks, 7(5),

955-972

Xiaowei, Z. Xiaohai, Y. and Xiangqun, C. (2005). Inter-

Query Data Aggregation in Wireless Sensor Networks

international conference on Wireless

Communications, Networking and Mobile Computing

Proceedings. 2, 930-933.

WINSYS 2010 - International Conference on Wireless Information Networks and Systems

102