You are on page 1of 4

ALGORITHMS FOR ENERGY EFFICIENT RECONSTRUCTION OF A PROCESS WITH A MULTIHOP

WIRELESS SENSOR NETWORK


Felipe da Rocha Henriques*, Lisandro Lovisolo, Marcelo Goncalves Rubinstein
*Celso Suckow da Fonseca Federal Center of Technological Education (CEFET/RJ), Petr opolis, Brazil
Program of Graduate in Electronic Engineering (PEL), University of State of Rio of Janeiro (UERJ), Brazil
*henriquesfelipe@telpet.com.br, lisandro@uerj.br, rubi@uerj.br
ABSTRACT
In this work, a multihop Wireless Sensor Network (WSN) is
employed to monitor a eld, modeled as a process (, , ).
In order to extend the lifetime of the network, we propose two
algorithms for energy-efcient reconstruction of the monitored
process. The reconstruction of the process is done in sink node, with
samples that it receives from each sensor node. Both algorithms
explore the variation rate of the eld to manage the necessity of
communication by sensor nodes, aiming at reducing the amount of
transmissions. Furthermore, nodes can sleep between transmissions
to save energy. Simulations are done, and results show a signicant
increase in the network lifetime, compared to a WSN without any
energy saving method. The algorithms are evaluated with respect
to the reconstruction error of the eld being sensed and network
lifetime increase.
Index Terms Wireless Sensor Networks, Energy, Reconstruc-
tion.
I. INTRODUCTION
Recent advances in microelectronics and wireless communica-
tions made it possible to develop and deploy low cost, low energy
consumption and tiny sensors. These sensors can be used as nodes
in a Wireless Sensor Network (WSN) [1]. A WSN is a special
kind of an ad hoc network and can be applied in areas such as
medicine, with remote monitoring of patients and their biometric
data; military, with monitoring of forces; industrial automation; and
sensing of interest regions, like a forest [2].
In this work, a WSN is considered to sense a eld, modeled
as a process that depends on the spatial coordinates

and

of
sensor nodes, and time . Each sensor node takes samples of the
monitored process and, eventually transmits these measurements
to a sink node. The main objective is to make an energy-efcient
reconstruction of the monitored process. Energy efciency involves
improving the network autonomy, by increasing its lifetime. In this
work, it is considered that the network lifetime is the time until the
energy of the rst node ends [3].
The study of methods that lead to energy saving in a WSN is
an important issue. In [4], a survey of energy saving methods for
WSNs is presented, including a taxonomy of some energy saving
schemes. According to [1], communication (i.e., transmission and
reception) is the task that spends more energy in a WSN. This
means that it maybe advantageous to process data, in order to
decide which measurements have to be transmitted.
In this work, we propose two algorithms for energy conservation
in a multihop WSN. In the proposed algorithms, we intent to reduce
the amount of transmissions of each sensor node, using the variation
rate of the monitored process at the sensor location. The more rapid
is this rate, more transmissions are required by the nodes. Moreover,
nodes can sleep between transmission, in order to save more energy.
The decision whether or not to transmit and sleep is taken locally
by each node individually, that is, in a distributed and decentralized
fashion. The monitored process is reconstructed in the sink node,
using the samples received from sensor nodes.
This work is structured as follows: in Section II, the algorithms
for energy conservation are presented; Section III presents the
energy model used in this work and simulation aspects; in Section
IV, the results obtained are presented; nally, conclusions are
discussed in Section V.
II. PROPOSED ALGORITHMS FOR ENERGY
CONSERVATION IN A WSN
The presented algorithms run directly in the application layer of
sensor nodes, and aim at saving energy of sensor nodes by using
two strategies: i) To reduce the amount of transmissions, and ii)
to put nodes in a sleeping mode between transmissions. In this
work, we consider a multihop WSN, in which each node is an
information source, when it measures samples from the monitored
process; and also a router (relay), when it has to forward packets
from its neighbors.
We assume that a node

has an inactivity period

, during
which the node does not measure, process, receive or transmit.
Before

enters the sleeping mode, it veries if it has neighbors


that use it as a router. If this does not occur,

sleeps for

seconds. If there are neighbors that use it as a router, then

sleeps
for

= (

),
,#
, (1)
and each

represents the inactivity period of each neighbor of

. The sleeping period reduction factor (0 < < 1) is used


to increase the probability of

being awake to forward packets


from its neighbors.
II-A. Algorithm 1
In Algorithm 1, a node

only transmits samples that have a


percentage variation between the current measured one and the last
transmitted one that are larger than a predened threshold . Thus,
after a node transmits the sample

(), it only transmits the sample

( + ) if:

( + )

()

()
. (2)
If

transmits

() and

(+) at instants

() and

(+),
it calculates its inactivity period

by eq. (3). The ow of


this algorithm is presented in Algorithm 1. In Algorithm 1,
978-1-4673-4900-0/13/$31.00 c 2013 IEEE
represents the current instant of measurement or transmission,

is the energy of node

, and we consider that nodes initially


have an inactivity period of 0.1 seconds, the timebase used in the
simulations. Moreover, line 3 of Algorithm 1 indicates that it runs
while node

has energy, in consonance with the original concept


of network lifetime. (The same parameters are also considered in
the setup of Algorithm 2.)

( + )

()
2
. (3)
Algorithm 1
1: n 1
2:

0.1
3: while

> 0 do
4:

measures

()
5: if n = 1 then
6:

transmits

()
7:

()
8:

() transmission instant
9:

()
10: else
11: if

()

then
12:

transmits

()
13:

()
14:

() transmission instant
15:

()
2
16:

()
17: if

has packets to forward from # neighbors then


18:

transmits packets from its N neighbors


19:

= (
1
,

),
,#
20: end if
21:

sleeps for

seconds
22:

wakes up after

seconds
23: end if
24: end if
25: n n+1
26: end while
The sink node uses a zero order interpolator to reconstruct the
monitored process, i.e., one considers that the process does not vary
between the transmitted/received samples.
II-B. Algorithm 2
In Algorithm 2, each node

, when computing the

,
considers how the sink reconstructs the process from the measure-
ments that it receives. In doing so, one may impose an additional
constraint that is keeping the reconstruction error () to be smaller
than a given predened threshold , that is, . One considers
that the sink node uses a rst order interpolator to reconstruct the
process, i.e., the process varies linearly.
Suppose that a sensor node

transmits its measured samples


to a sink node

generates a vector that contains the collected


measurements, s

= [

(1),

(2), . . . ,

()], and a vector


with the measuring instants, t

= [

(1),

(2), . . . ,

()]. At
the sink node, the reconstruction of the process is done from
the available information, a subset of the measured set vector,
generated from the received samples s

and from their corre-


sponding instants t

, i.e., s

= [

(1),

(2), . . . ,

()] and t

= [

(1),

(2), . . . ,

()]. The sink node uses s

and t

to
reconstruct the monitored eld.
In this algorithm, we use a linear estimation: The last two
samples are considered to estimate a future transmission. Sensor
nodes can estimate a future measurement that will be transmitted

( + 1) (and that is assumed to be received at sink node) such


that , from measured samples, and from transmitted samples.
The goal is the sensor nodes to be capable of estimating the time
interval between transmissions, in order to sleep and save energy.
We consider the following notation: For last two measured
samples and instants of measurements, we have

( 1) and

( 2),

( 1) and

( 2); for last two transmitted


samples and instants of transmissions, we have

( ) and

(),

() and

(). We dene the indexes


and for transmission, because not all measured samples
are transmitted. Furthermore, we dene the following variations:

( 1)

( 2), (4)

( 1)

( 2), (5)

( )

( ), (6)

( )

( ), (7)
and variation rates:

, (8)

. (9)
For this model, we consider that the transmitted and received sets
are equal. Thus, we want that the percentage variation between the
next transmission

( + 1) and the real value

( + 1) is less
than a given threshold (),

( + 1)

( + 1)

( + 1). (10)
Using a rst order interpolation, we can estimate

( + 1) as
follows:

( + 1) =

( ) +

()

( )), (11)
with

given by eq. (8). It is assumed that

( + 1) denotes the
next sample to be measured, thus

( + 1) =

(). As

()
is not available, this value is estimated using the variation rate of
the monitored process,

(), dened by:


() =

( 1) +

()

( 1)), (12)
with

dened by eq. (9). By this way, eq. (10) turns:

( ) +

()

( ))

()

(). (13)
Replacing eq. (12) in (13), we obtain:

( )+

()

( )) (14)
(

( 1) +

()

( 1)))

( 1) +

()

( 1)).
By solving inequation (14), the inactivity period of node

can be estimated as the time interval between the last transmitted


sample and the future one.

()

( ). (15)
The ow of this algorithm is presented in Algorithm 2.
Algorithm 2
1: n 1
2:

0.1
3: while

> 0 do
4:

measures

()
5: if n = 1 then
6:

transmits

()
7: else
8:

transmits

()
9:

calculates

using linear estimation


10: if

has packets to forward from # neighbors then


11:

transmits packets from its # neighbors


12:

= (
1
,

), #
13: end if
14:

sleeps for

seconds
15:

wakes up after

seconds
16: end if
17: n n+1
18: end while
III. ENERGY MODEL AND SIMULATION ASPECTS
The energy model used in this work is a state-based model, in
which nodes may operate in two states: Inactive or active. The
inactive state (sleep mode) is an energy saving mode. The active
state is composed by four operation modes: Measuring, processing,
transmission, and receiving. The proposed energy model takes into
account the packet payload size, and it is based on [5], an empirical
energy model, obtained using the TELOS commercial hardware [6],
in which it is observed that the energy consumption and the packet
payload size are linearly related (in the transmission mode).
The energy consumption

of a node can be estimated, as


a function of the period of time in which the node stays in the
different operation modes.

) +

)
+

) +

), (16)
in which

and

are, respectively, the cumulative


sum of intervals in which a node remains in inactive and active
states, and in measuring, processing, receiving, and transmitting
operation modes. If a node is active, there is an increment in its
energy consumption, depending on the task it is performing. The
associated consumptions

of each one of the states and modes


are parameters that are presented in Table I.
The simulations were performed in TrueTime 1.5 [7], a sim-
ulation environment based in MatLab/Simulink and the network
standard considered was the IEEE 802.15.4 [8].
For simulations, we considered monitoring a smooth process
modeled as surface (, , ). The process, used to evaluate the
estimator is described by eq. (17), in which

and

represent
the location of the monitored phenomenon, and

it is the location
of its maximum;

and

represent the connement of the


sensed phenomenon in space; and is a constant.
(, , ) =
[

)
2
2
2
+
(

)
2
2
2

)
2
2
2
+ . (17)
Table I. Static parameters of the simulations Cons. refers to
energy consumption.
Node initial energy (J) 2.00
Transmission power (dBm) -5
Reception sensibility (dBm) -66
Radio range (m) 40

: Inactive state Cons. (mJ/s) 1.80

: Active state Cons. (mJ/s) 10.00

: Measuring mode Cons. (mJ/s) 18.00

: Processing mode Cons. (mJ/s) 18.00

: Rx mode Cons. (mJ/s) 62.40

: Tx mode Cons. (mJ/s) 58.62


Payload size (Byte) 1
In this work, a multihop communication model is considered, in
which a routing protocol is used, in order to forward packets node-
to-node from sources to the sink. The Ad-hoc On Demand Distance
Vector (AODV) [9] routing protocol is considered in simulations.
For the scenario, a WSN with fteen nodes was used to sense a
8080 region, in which sink node is positioned in the middle
at the right and sensor nodes are randomly positioned. In each run
of simulation, the position of sink node is xed, and the positions
of sensor nodes are randomly sorted.
IV. SIMULATION RESULTS
This section presents the results obtained. Each simulation was
run ten times, and a 95% condence interval for the mean is used
in the graphs, represented by vertical bars. Thresholds () of 0.1%,
1.0%, 5.0% and 8.0% were considered. The following parameters
were considered, for the monitored process:

=
40;

= 20; and = 5. Moreover, we consider


= 0.5 in the simulations.
Figure 1 shows the Cumulative Distributed Function (CDF) of
the reconstruction error of the monitored process, for different
thresholds, for Algorithm 1 and Algorithm 2. It represents the
probability of error () being less than a given threshold , i.e.,
( < ). In this gure, solid lines show the reconstruction
error with Algorithm 1, and the same metric for Algorithm 2
is presented by dashed lines. It can be observed an increase in
the reconstruction error with the increment of , as expected. This
increment of allows a larger percentage variation between the
measured sample and the last one transmitted.
As discussed in Section II, a constraint imposed by Algorithm 2
is that the reconstruction error should be smaller than the threshold.
It can be seen in Figure 1, that this constraint is satised. The same
condition is not observed in simulations with Algorithm 1. Thus,
Algorithm 2 leads to smaller reconstruction error than Algorithm
1 does.
Figure 2 shows the increase in the lifetime of the WSN, and
the decrease in the amount of transmissions, in function of .
These results were obtained comparing our proposal to a network
without any kind of energy management, in which nodes simply
take measures and transmit them periodically at each 0.1 seconds.
It can be observed that the increase of leads to the reduction in
the transmissions and to the increase in the network lifetime. To
increase the threshold means that nodes only transmit measured
samples for a greater percentage variation, because a larger re-
construction error is allowed. Furthermore, nodes may have larger
sleeping periods, and they transmit fewer measurements.
We also observe that Algorithm 1 leads to a grater increase in
the lifetime of the network than Algorithm 2 does. This is so,
because the rst algorithm allows larger reconstruction error, and
this behavior is compatible with a tradeoff between the energy
consumption and the reconstruction of the process. Thus, for
different applications, it can be preferred to have a controlled
reconstruction of the monitored process at the expenses of the
lifetime of the network. However, as can be noted, both algorithms
lead to a signicant energy saving.
0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Reconstruction error (%)
R
e
c
o
n
s
t
r
u
c
t
i
o
n

e
r
r
o
r

C
D
F


= 0.1%
= 1%
= 5%
= 8%
= 0.1%
= 1%
= 5%
= 8%
0 0.02 0.04 0.06 0.08 0.1
0
0.2
0.4
0.6
0.8
1


Fig. 1. Reconstruction error CDF of the monitored process with
Algorithm 1 and Algorithm 2.
0 1 2 3 4 5 6 7 8
0
500
1000
1500
(%)
L
i
f
e
t
i
m
e

i
n
c
r
e
a
s
e

(
%
)

Algorithm 1
Algorithm 2
0 1 2 3 4 5 6 7 8
0
50
100
(%)
T
r
a
n
s
m
i
s
s
i
o
n

d
e
c
r
e
a
s
e

(
%
)

Algorithm 1
Algorithm 2
Fig. 2. Lifetime increase and decrease in the transmissions .
Table II shows the packet delivery ratio in function of , for both
algorithms. This metric evaluates the connectivity of the network.
It is dened as the ratio between the amount of received and
transmitted packets. In the simulations, condence intervals for
the means between 0.5% and 1.0% were obtained. We can see
that by increasing the threshold, there is a reduction in packet
delivery ratio. In Algorithm 2, nodes transmit more packets than
in Algorithm 1. This may lead to an increase in the probability of
collisions in the network, affecting the network connectivity more
than in Algorithm 1.
V. CONCLUSIONS
In this work, we propose two algorithms for energy conservation
in WSNs. The algorithms aim at reconstructing a process (, , )
Table II. Packet delivery ratio .
0.1% 1.0% 5.0% 8.0%
Algorithm 1 98% 97% 88% 83%
Algorithm 2 89% 88% 87% 87%
with energy efciency. This means that sensor nodes explore the
variation rate of the monitored process to reduce the amount of
transmissions, in order to increase the lifetime of the network.
Furthermore, nodes may enter in an inactivity mode between the
transmissions to save energy.
It can be observed that Algorithm 1 can reach a larger energy
conservation, but without keeping the reconstruction error less
that the predened threshold. In Algorithm 2, this constraint is
respected, while obtaining signicant increases in the network
lifetime.
VI. ACKNOWLEDGMENT
This work has been supported by FAPERJ, CAPES and CNPq.
VII. REFERENCES
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci,
Wireless Sensor Networks: A Survey, Computer Networks, vol.
38, pp. 393422, 2002.
[2] A. Mainwaring, J. Polastre, R. Szewczy, and D. Culler, Wireless
Sensor Networks for Habitat Monitoring, in ACM International
Workshop on Wireless Sensor Networks and Applications, Septem-
ber 2002, pp. 8897.
[3] Z. M. Wang, S. Basagni, E. Melachrinoudis, and C. Petrioli, Ex-
ploiting Sink Mobility for Maximizing Sensor Networks Lifetime,
in Proc. 38th Annual Hawaii International Conference on System
Sciences (HICSS05), Hawaii, January 2005, pp. 36.
[4] V. K. Sachan, S. A. Imam, and M. T. Beg, Energy-efcient
Communication Methods in Wireless Sensor Networks: A Critical
Review, International Journal of Computer Applications, vol. 39,
no. 17, February 2012.
[5] Y. Panthachai and P. Keeratiwintakorn, An Energy Model for
Transmission in Telos-Based Wireless Sensor Networks, in
Proceedings of the 4th International Joint Conference on Computer
Science and Software Engineering, 2007.
[6] J. Polsatre, R. Szewczyk, and D. Culler, Telos: Enabling Ultra-
Low Power Wireless Research, in Proceedings of the 4th Interna-
tional Symposium on Information Processing in Sensor Networks,
2005.
[7] A. Cervin, D. Herinksson, B. Lincoln, J. Eker, and K.-E. Arz en,
How Does Control Timing Affect Performance?, IEEE Control
Systems Magazine, vol. 23, no. 3, pp. 1630, 2003.
[8] P. Baronti, P. Pillai, V. Chook, S. Chessa, A. Gotta, and Y. F. Hu,
Wireless Sensor Networks: a Survey on the State of Art and the
802.15.4 and ZigBee Standards, Computer Communications, vol.
30, no. 7, pp. 16551695, May 2007.
[9] C. Perkins, E. Belding-Royer, and S. Das, Ad Hoc On Demand
Distance Vector Routing (AODV), in RFC 3561, July 2003.

You might also like