Minimizing energy consumption within wireless sensors networks

F. Shebli, I. Dayoub and J.M. Rouvaen

University of Valenciennes (IEMN, UMR CNRS 8520)
Le Mont Houy,BP311, 59313-Valenciennes Cedex, France

ABSTRACT
The use of sensor networks should keep developing, mainly in such fields as
scientific, logistic, military or healthcare applications. However, sensor size
represents a significant limitation mainly in terms of energy autonomy and
therefore of life period, for the batteries have to be too tiny. This is the reason why
intensive research is being conducted nowadays on how to control sensor energy
consumption within a network, taking communications into account as a priority.
For this purpose we propose a method to calculate energy consumption within
linear wireless sensor networks, according to the data flow rate, the number of
nodes and the distance between them. Furthermore, we have succeeded in reducing
energy consumption within linear sensor networks made up with nodes featuring
differing data flow rates.

Keywords: Wireless sensors networks, Clustering, Energy consummation,

Routing..

1 INTRODUCTION neither replaceable nor rechargeable. It may be

partially fed by energy generating units such as
Wireless sensor networks are a novel technology photo-voltaic cells. As it is small, it provides a
emerging from embedded system, sensor technology limited quantity of energy of the order of 1 to 2J by
and wireless networks. The rapid deployment, self- node (acquisition). Thus, it limits the sensor life time
organization and fault tolerance characteristics of and influences the overall network working process.
wireless sensor networks make them a very
promising sensing technique for military, Energy is a factor of outmost importance in WSNs.
environmental and health applications [1]. However To increase network lifetime, energy must be saved
finite, generally irreplaceable power sources in in every hardware and software solution composing
sensor node limit lifetime of the whole system. Some the network architecture. According to the radio
researches have proven that sensor node expends model proposed in [5], data communication is
maximum energy in data communication [2]. A responsible for the greatest weight in the energy
sensor network is a distributed system made up with budget when compared with data sensing and
a large number of small sensors, equipments, low processing. Therefore, it is desirable to use short-
power transmitters-receivers, without a central range instead of long-range communication between
processing unit. One of the major problems in these sensor nodes because of the transmission power
networks consists in reducing energy consumption to required. In most WSN scenarios, events can be
a minimum in such a way as to maximize a network sensed by many source nodes near the phenomenon
life time [3]. Recent advances in the IC technology of interest and far away from the sink nodes. Then,
make it possible to produce micro-sensing devices the use of short-range communication leads
that are equipped with processing, memory and obligatorily to data packets being forwarded through
wireless communication capabilities. intermediate nodes along a multi-hop path [2].

In Wireless Sensor Networks, WSNs, nodes are This article develops a model for the multi-hop
untethered and unattended. They are distributed communication in a linear array of nodes. Energy
across an area of interest and communicate among consumption in the various multi-hops scenarios has
themselves in multiple hops, building an ad-hoc been analyzed and optimized. This study uses a
network. Nodes have limited and non-replenishable detailed model for the energy consumed by the radio
energy resources. There are special nodes named link of each node and analyzes two topologies: the
sink (or gateway) nodes, that are responsible for first one with equidistant nodes hop and the other
processing and storing the information collected by one with optimal spaces between the last nodes. The
the network [4].The battery is an essential article is organized as follows.
component in data acquisition. In general, it is

Ubiquitous Computing and Communication Journal 1

 Figure 1: Energy consumption model for communications.

A model of basic energy consumption and transmission, eamp amplification, k the message
communication topology is presented in section 2. length, d the transmitter/receiver distance and α a
The analysis of multi hops is performed in section 3 factor describing attenuation. To receive a message
for different routing models and the optimal of k bits, the receiver then consumes:
transmission range is analyzed in section 4. In
section 5 the results based on the previous analyses α

are presented. MAC protocols are discussed in

Elinear (i ) = [(2 * i − 1)(eelec + ieamp d i ] (1)
section 6 and finally the conclusion is presented in
section 7. 2.2 Communication topology
We adopt a simple linear topology to discuss
2 ENERGY MODEL AND COMMUNICATION communication mode, as shown in figure 2 in order
TOPOLOGY to simplify the analysis. At the left of the Figure 2,
there are n sensor nodes arranged at intervals of r.
2.1 Energy model The base station is on the right end. Using single hop
A sensor uses its energy in order to carry out mode, each node directly communicates with the
three main functions: acquisition, communication base station; by contrast, each node communicates
and data processing. with the closest neighbor in multi-hop mode, nodes
1. Acquisition: the energy consumed to carry out the route data destined ultimately to the base station
acquisition is generally negligible. Nevertheless, it through intermediate nodes.
varies in considerable proportions depending on the
type of monitoring being carried out.
2. Communication: It consumes more energy than
any other task. It covers the communications in terms
of emission and reception. Figure 1 presents a
transmission system model and the rules applied to
controlled energy consumption [6].
3) Data processing: The energy consumed for the Figure 2: Multi-jump linear model featuring equal
calculation operation is very low as compared with distances between the nodes.
the communication energy. The energy needed to
transmit 1 KB over a 100m distance is approximately
equivalent to the energy necessary to carry out 3 3 ENERGY CONSUMPTION IN THE
million instructions at a speed of 100 million MULTIHOP SENSOR NETWORKS
instructions per second (MIPS). This level might be Using equations (1) and (2), the energy
much dependent on the circuitry installed in the consumed during data transmission from the source
nodes and the features requested. towards the destination going through intermediate
nodes aligned in a row is written as:
α
Elinear (i ) = [(2 * i − 1)(eelec + ieamp d i ] (1)
n
[ 
Elinear = k ∑ 2 * eelec + eamp (d i )α  ] (3)
where Eelec represents energy consumed in  i =1 

Ubiquitous Computing and Communication Journal 2

Elinear is minimum when all the di are equal to D/n, Example 1: A sensor network made up with 20 nodes,
when the number of hops is at its optimal value: of which 20 nodes simultaneously detect an event,
applying equation (6) leads to NPack =n-i+1
1
 D   2 * eelec  α Example 2: A sensor network made up with 20 nodes,
nopt =  => d =   (4)
char  eamp (α − 1)  of which 15 nodes simultaneously detect an event
 d char    and 5 simultaneously detect five events, applying
equation ([6) leads to NPack =n-i+2.
The optimal number of hops depends on the
propagation loss coefficient α and the transmitter and 3.2 Total energy cost for the system
receiver parameters [9,1]. By replacing dchar in From the average number of transmitted packets,
statement Elinear, we obtain the following relation: the average energy consumed within the linear
sensor network may be computed as:
 2 * nopt * eelec *α 
opt
Elinear = m − eRC  (5) Elinear− bit = −neRC +
 α −1 
∑[(e ) ] (9)
n

TC + eRC + eTA (di )α * NPack

In the linear sensor network, two possibilities arise i =1
for the probability of event occurrence:

1. First case: uniform number of data flow rate 4 OPTIMAL TRANSMISSION RANGE
constant among the nodes. If each node detects the Bhardwaj et a1. [10] have shown that energy
same number of events (that is the necessity to consumption within the multi-hop networks can be
convey information towards the destination), then we reduced by adjusting the distances between the n
can suppose that each node receives an event nodes which transmit the signal. But, if each node of
simultaneously. This means this node should take (n- the linear chain transmits data, this will be
1) information packets originating from nodes impossible. The distribution of the distances di
upstream, in addition to the transmission of its proper between the nodes can be analyzed in order to
packet [7], [8]. minimize the energy consumption in this particular
case. For all the nodes involved in the transmission
2. Second case: data flow rate variable between scenario, the total energy consumption is presented
nodes. If each node does not detect the same number in equation (8).
of events, the average number of packets transmitted
within the linear sensor network has to be calculated. Elinear will now be minimized using the
n

3.1 Packets relayed

constraint D = ∑i=1
d i , equivalent to the
following formula:
If xi stands for the number of events detected by
a given node, the number of packets relayed by this
[  n
] 
n
node is given by: L = eTA ∑ N pack (d i ) α − λ  ∑ d i − D  (10)
i =1  i =1 
Pi = n − i + xiα (6)
where λ is the Lagrange's multiplier. Equating the
Let us assume the existence of an underlying partial derivates of L with respect to the di to zero, it
supervision protocol, in charge of building and follows that:
updating the routing tables for nodes. Then the mean
number of events seen by each node is: ∂L
= eTA .α .N pack ( d i ) α −1 − λ = 0 (11)
n ∂d i
ne = [ ∑ xi ] / n (7) 1
i =1  λ  α −1
di =   (12)
The average number of transmitted packets NPack at  e .α.N 
node i may be estimated as:
 TA pack 

n
N Pack = n − i + neα (8) Using the condition D = ∑ d i the value of λ can
i =1
where n is the number of nodes within the linear be obtained from equation (12). Thus, for α = 2 the
sensor network and i the node under consideration. values for di are found to be:

Ubiquitous Computing and Communication Journal 3

 D The cases where the di are equidistant or not are
di = (13) again considered here. Using equation (14), the
 n 
 ∑ (1 i ) N pack results for optimal distances between nodes given in
 i =1  table I are obtained.

Table 1: OPTIMAL DISTANCES BETWEEN NODES.

Figure 3: Linear and non linear sensor networks.

In [8, 12, 13], the energy consumption per packet

sent consumed at node i for linear arrays of equally
spaced nodes discussed is computed. In our approach
the energy consumption per packet sent consumed at
node i for linear arrays with unequal data flow rate
ca be calculated by the following equation:

Elinear (i) = [(2 * N pack − 1). eelec +

α (14)
N pack . eamp . d i ]

5 SIMULATIONS AND RESULTS

A sensor network made up from 30 nodes
distributed over a 1000 m distance is simulated here,
with parameters eelec = 50 nJ/bit, eamp = 100 pJ/bit/m2,
α=2,using MatlabTM. Let us notice that this implies a
33.33 m distance between nodes in the equidistant
case.

-6
x 10 Total energy consumption -- equidistant nodes
7
Npack=(n-i+1)->Total energy = 86.5mJ
6 Npack=(n-i+2)->Total energy =92.55mJ
Npack=(n-i+3)->Total energy = 98.25mJ
Npack=(n-i+4)->Total energy = 103.95mJ
5
Npack=(n-i+5)->Total energy = 109.65mJ
E n ergy (m J )

0
0 5 10 15 20 25 30
Number of nodes
Figure 4: Energy consumed with transmission distances. Fig. 4 presents the energy by each node in a sensor
network with equidistant nodes, is obtained using

Ubiquitous Computing and Communication Journal 4

equation (14) and with all di fixed to 33.33 m, for, smaller for optimally spaced nodes than for equally
respectively, uniform or non-uniform mean numbers spaced ones.
of events detected by each node. It is worthwhile
noting that the total energy consumption when each 6 DISCUSSION ON MAC PROTOCOLS
node doesn't detect the same number of events Shelby et Al [10, 11] described the impact of
(variable flow rates) is slightly higher than that in the MAC protocols on the energy model used in this
uniform number of events bottom curve case. paper, namely Nonpersitent CSMA , S-MAC [9] and
NanoMAC. Fig. and Fig7 presents Transmission and
Fig. 5 presents the amount of energy consumed by Receive energy consumption model for nanoMAC.
each individual node within a sensor network
featuring optimal distances shown in table I between Nonpersitent CSMA [10, 11, 6, 12] is a well known,
the nodes, obtained from equation (14), for uniform relatively well performing MAC protocol in almost
or non-uniform number of detected events. any scenario. It gives the worst-case energy
consumption that any sensor MAC protocol should
It must be noticed that the total energy consumption outperform. S-MAC is the current sensor MAC
within the network has been reduced by 38% when benchmark protocol which is used to highlight some
Npack=(n-i+1), 26% when Npack=(n-i+2), 21% of the faults of traditionally designed sensor MAC
when Npack=(n-i+3), 17% when Npack=(n-i+4) and protocols.
15% when Npack=(n-i+5) following the approach
proposed in section IV

-6
x 10 Total energy consumption - optimally spaced nodes
7
Npack=(n-i+1) ->Total energy = 70.31mJ
Npack=(n-i+2) ->Total energy = 81.03mJ
6
Npack=(n-i+3) ->Total energy = 91.58mJ
Npack=(n-i+4) ->Total energy = 99.8mJ
5 Npack=(n-i+1) ->Total energy = 107.64mJ
E n er g y (m J)

2
Figure 6: Transmission energy consumption model for
nanoMAC.
1
0 10 20 30 They compared these to nanoMAC, a protocol
Hop number designed to operate in a sensor networking
Figure 5: Energy consumed with optimal transmission environment. They have analytically investigated a
distances. cross layer energy consumption model with realistic
radio transceiver characteristics, three MAC
Comparing figures 4 and 5,it is evident that the protocols and a linear network model suitable for
nodes far from the sink consume significantly less many sensor network protocols in steady state. Based
energy for optimally spaced nodes than for equally on this analysis, they have discovered many
spaced nodes. This is in fact due to their smaller interesting results that relate to single hops vs. Multi-
separation. The consumed energy tends to decrease hop communications and MAC protocol features.
with hop number in figure 5 in figure 4, thanks to a
lower number of packets to be transmitted. 1) When a realistic radio model is applied for a
sensor network, they discovered that, assuming
However the distance between nodes is an increasing feasible transmission distances, single-hop
function of hop number, which tends to increase the communications can be more efficient than multi-
energy consumption. So it is surprising that, starting hop in the energy perspective.
at the some hop number, the consumed energy
sharply increases and this for few last nodes. 2) A well designed sensor MAC protocol has a
However the overall energy consumption tends to be behavior similar to the ideal MAC protocol, but the

Ubiquitous Computing and Communication Journal 5

energy consumption is two orders of magnitude IEEE, Vol. 91, No. 8, pp.1247-1256, 2003.
higher. [3] Hill, M. Horton, R. Kling and L.
Krishnamurthy,"The Platforms Enabling
3) There are some inherent flaws in adapting existing Wireless Sensor Networks", Communications of
ad hoc MAC protocols to sensor networks. Idle the ACM, Vol. 47, No. 6, 2004.
listening and overhearing avoidance are important [4] Teixeira, I., de Rezende, J. F., and Pedroza,
factors as already discussed in other publications, A.C.P. - "Wireless Sensor Network: Improving
such as [14, 11], but also any listening that is not the Network Energy Consumption", in XXI
absolutely necessary, like listening for the RTS in S- Simpõsio Brasileiro de Telecomunicações -
MAC, decreases the energy performance of a sensor SBT'2004, Belém, Brazil, September 2004.
MAC. [5] W.R.Heinzelman, A. Chandrakasan, and H.
Balakrishnan, "Energy efficient communication
protocol for wireless microsensor networks",
Proceedings of the 33rd IEEE Hawaii
International Conference on System Sciences
(HICSS), pp. 1-10, Jan. 2000.
[6] W. Li. and C. G. Cassandras, "Sensor Networks
and Cooperative Control", European Journal of
Control, Vol. 11, No. 4-5, pp. 436-463, 2005.
[7] H. Karvonen, Z. Shelby, and C.A. Pomalaza-
R´aez, "Coding for Energy Efficient Wireless
Embedded Networks" ,Proceedings of the
International Workshop on Wireless Ad Hoc
Networks, May 31 - June 3, 2004, Oulu, Finland.
[8] J. Haapola, Z. Shelby, C.A. Pomalaza-R´aez,
and P. Mahonen,"Multihop Medium Access
Control for WSNs: An Energy Analysis Model",
Figure 7: Receive energy consumption model for EURASIP Journal of Wireless Communications
nanoMAC.
and Networking, Vol. 2005, No. 4, pp. 523-540,
September 2005.
7 CONCLUSION [9] Q Gao, K J Blow, D J Holding, I Marshall, and
In this work, it has been shown that the total X H Peng,"Radio Range Adjustment for Energy
energy consumption in a linear sensor network may Efficient WirelessSensor Networks", Ad-Hoc
be reduced (by 15% to 38%) using optimal spacing Networks, Vol. 4, No. 1, pp. 75-82, 2006.
between nodes. [10] M. Bhardwaj, A. Chandrakasan, and T. Garnett,
Upper bounds on the lifetime of sensor networks,
The model used suggests that the transmitter can Proceedings of the IEEE ICC, pp. 785-790, 2001.
dynamically adjust its transmission power, in such a [11] W. Ye, J. Heidemann, and D. Estrin, "Medium
way that the requested signal to noise ratio may be access control with coordinated adaptive
warranted at the receiver. This work will be sleeping for wireless sensor networks",
pursuited by studying the scenarios which determine IEEE/ACM Transactions on Networking, vol. 12,
the transmission power and finding the efficient data pp. 493-506, June 2004.
coding and modulation techniques for practical [12] J. Haapola, Low-power wireless measurement
applications of wireless sensor network (WSNs). system for physics sensors, Master Thesis,
Department of Physical Sciences, University of
Our analytical results will also be compared with the Oulu, Oulu, Finland, 2002, unpublished,
results of physical measurements. The influence of available online on
different routing models and protocols for sensor http://www.ee.oulu.fi/jhaapola/.
networks on energy consumption will also be [13] J. Haapola, Z. Shelby, C. Pomalaza-R´aez, and P.
considered. Moreover an extension to 2D array of M'ah'onen, Cross-layer energy analysis of multi-
sensors is also planned hop wireless sensor networks, in Proc. 2nd
European Workshop on Wireless Sensor
8 REFERENCES Networks (EWSN 05), pp. 3344, Istanbul,
Turkey, January February 2005.
[1] F.Akyildiz I F, Su W and Sankarsubramaniam Y [14] L. Kleinrock and F. A. Tobagi, Packet switching
2002 Wireless sensor networks: a survey in radio channels: Part ICarrier sense multiple-
Computer Networks 38 393-422 access modes and their throughput-delay
[2] C.-Y. Chong and S. Kumar,"Sensor networks: characteristics, IEEE Trans. Commun., vol. 23,
Evolution, opportunities, and challenges", Proc. no. 12, pp. 1400? 1416-1975.

Ubiquitous Computing and Communication Journal 6