2007 International Conference on Sensor Technologies and Applications

Topology Optimization in Wireless Sensor Networks for Precision Agriculture

Applications
Katsalis Konstantinos
University Of Thessaly
Volos, Greece
kkatsalis@inf.uth.gr

Xenakis Apostolos
University Of Thessaly
Volos, Greece
axenakis@inf.uth.gr

Kikiras Panagiotis
University Of Thessaly
Volos, Greece
kikirasp@inf.uth.gr

Stamoulis George
University Of Thessaly
Volos, Greece
george@inf.uth.gr
required for a WSN to function. These sensors can be
programmed to record measures like temperature and
humidity. All the data which are collected from the
sensors, using a wireless multi-hop routing technology
[6], end up in a gateway which transfers them to the
end user through wireless network, internet or LAN
[1][7], as it is shown in figure 1.

Abstract
In this paper a new way to build a wireless sensor
network is proposed, which is based on measuring the
fields electrical conductivity, staying away from the
classic network grid implementation. Furthermore it is
explained how a typical WSN works, which are the
pros and cons and the technical characteristics, as
well as how electrical conductivity can influence the
decision to build the WSN topology and the advantage
of this approach comparing to the typical ones.

1. Introduction
Precision Agriculture refers to the use of an
information system for the within-field management of
crops [1][2][3]. This basically means to add the right
quantity of fertilizer to the right time and to the exact
location within certain cultivate extend of ground. This
means than every part of the crop is treated in different
way and not as a whole part. The use of precision
agriculture techniques gives agronomists the potential
to apply new and continuously developing
technologies which help to manage better the
production. Some of these technologies are GPS, GIS,
Remote Sensing, variable Rate Technology, Machine
Controls, Smart Sensor Arrays and WSN technology
[4][5].

Figure 1: A Wireless Sensor Network

1.2 The advantages from applying a WSN in

the field
The benefits which arise from the application of
precision agriculture technique come from the
precision in the irrigation quantity, the use of
chemurgy only in the appropriate field areas, the
control in the quantities of the fertilizer, the exact
definition of the semination and crop. More over the
use of appropriate quality of seed depending on the
field conditions, water control, the optimum quantity
of seed semination and spending less money on
agriculture scientists and consulting firms are factors
that WSN has a direct impact.
In addition with WSN the advantages are:

1.1 WSN Technology

A number of sensors which will be placed
appropriately and will cover the whole filed are

0-7695-2988-7/07 $25.00 2007 IEEE

DOI 10.1109/SENSORCOMM.2007.101

526

a zone. Moreover the shape of the zones depends on

which essence the agronomists measure and form the
variability in the zones. However, generally the zones
should be clearly detected.

Ability to observe for long periods of time crop

state.
Direct, exact briefing of the field state and ability to
interfere in case of an emergency.
Distant decision making.
Analytical information storage in order to create a
case record of the field crop.
Friendly Graphical User Interface with the
monitoring system.
Potential to make exact evaluation of new crop
methods and techniques.

2.2 Electrical conductivity and Veris System

The Electric Conductivity is a measurement of
easiness with which electric current goes through the
soil. It mainly depends from the presence of salts and
the constitution of soil in sand, clay, organic substance
and water. Regions with the same EC values belong in
the same territorial type with a great probability and
thus management maps construction can be done based
in EC measurements.
According to so far research, the measurements of
electrical conductivity are mainly influenced from
humidity. This means that each time the levels of
humidity change within the field the network should
not be re-designed because these levels change
respectively for the entire field and the management
zones remain the same as before and so the network
diagram is not affected either.
One of the most used systems for measurement of
electric conductivity is Veris 3100. This device comes
along with a GPS receptor so that each measurement is
recorded precisely and collected in a central unit. A
data file is produced for each area scanned with Veris
including each measurement along with the
coordinates of the point that the measurement was
taken. The data can be represented in a graphic mode
so that regions with similar characteristics can be
categorized, distinguished and clearly seen. This
requires processing of data from specialised software
of geographic information system (GIS).

2. Materials and Methods

Precision Agriculture is mainly based on the
management of the fields differentiality. The
differentiallity in the production is defined from the
variability in the field structure, the organic matter, the
level of saline and the level of water. In order to treat
the field in different ways, detailed management field
maps should be produced, on which this differentiallity
is depicted. The production of these management maps
give in an indirect way evidence of how the topological
diagram of the WSN should be built. These maps act
as a criterion, based on which, the topological design
should pass from the classic grid implementation to a
more sophisticated diagram.

2.1 Management Zones

The field management zones are smaller parts of
the field which show certain variability from each
other and each of them needs different management
treatment and tactics. The criterion to create the
management zones is not only one and is not always
fixed. The categorisation depends on several factors
which the agronomists decide that are capable to do so,
for example humidity, organic matter or agile. In figure
2, four management zones are depicted, each of which
have individual characteristics.

2.3 Topology efficiency

The aforesaid approach is applied for the
segregation of field in management areas, in order to
avoid the node placement in grid placement. An
appropriate number of sensors which will completely
cover the informative needs from the aspect of data are
placed so that a complete picture of the total field is
drawn.
The following methodology for the optimal
placement of sensors is used:
1. Scanning of all extent with the system Veris
3100.
2. Explicit segregation between zones.
3. Registration of points in the field which can
cause problem in the communication between the
nodes.

Figure 2: Management Zones

The size of the zones depends on the ability of the
agronomist to differentiate them according to
measurements. So there is no rule to define the size of

527

4. Placement of nodes in places in which

communication is achieved and will give central
measurements for each area, and also the routing
protocol used for sensors, can reach convergence
according to some rules which will state afterwards.
In an existing field in which cotton is cultivated, in
the region of Karditsa (Central Greece), with an extent
of 50 acres, the aforesaid methodology is used in
management zones that have been drawn by Veris,
according to the work done by A. Markinos et al.[8].
Using Veris to scan in a depth of 90cm, the 50 acres
field is segmented in five distinguishable zones.
Simulations are done using the two following
network topologies as shown in figure 3.

A grid topology of 56 sensors with 32m vertical

and horizontal distance each.

A proposed methodology of 26 sensors with

medium distance of 34,8m among sensors.

proposed topology is that for specific nodes more

power consumption [10] is demanded comparing to the
grid topology. All sensors we use are greatly
responsible for their zone through their measurements
and they function at higher rates comparing to those
used in a grid topology.

2.4 Technical Settings

Because a variety for ad hoc routing algorithms
for wireless sensor network have been developed and
each algorithm has its own scenario, it is hard to
compare all these algorithms. The solution is given by
Rmase which deals with the challenge to compare
different routing algorithms. Rmase application [9],
written in Matlab code, is the base application we used
to run all our simulations. Rmase consists of a network
topology model, an application model and a
performance model. Rmase has been used to develop
new routing algorithms, to analyze performance tradeoffs. Rmase is implemented as application in Prowler.
The simulation was done in Prowler, which is an event
based simulator, a framework based to TinyOS /
NesC.
The routing algorithm we used is Logical Grid
Routing (LGR), which builds a spanning-tree, the root
of which is node (0, 0) who is the ultimate destination
of all packets in the network. Every node in the path
from the source to the destination is connected to a set
of potential parents.

Figure 3: Grid Topology & Optimal nodes

placement
The coverage distance for each node depends on
the node and RF technology. The nodes were not put at
the maximum coverage distance for more power
efficiency. The medium distance of 35m results from
distances of nodes, in which there is a high possibility
of communication between them (the Signal to
Noise ratio was above a certain threshold). Clearly
there is a trade-off between the number of nodes used
(lower cost in euros) and the network robustness.
The reduction in the number of sensors that is
reached in the cotton cultivation, reaches 56%. This
ratio occurs when the absolutely necessary nodes are
used to cover the management zones. In a more robust
solution, some extra nodes are used which are placed
in key communication positions in the field.
A key point in the design of the network is to take
under consideration factors such as scattering, the
absorption and weakening of the signal that depend
each time on the type of the cultivation, the height of
leafage as also hillocks and pieces of machinery in the
field, and lastly the sources of electric noise such as
high voltage cables. Other factors are still the
technology of sensors used and real climatic conditions
that prevail in the region. One drawback of the

3. Simulation Results
Rmase provides a set of performance metrics for
comparing
different
routing
algorithms,
including latency, throughput, success/loss rate and
energy consumption/efficiency. The simulation time is
calibrated to 100 seconds. Some definitions of the
metrics are:
Latency: Time to send a message from source to
destination. For any destination, if n packets have
arrived, latency for that destination is given by
n di
=1 n , where di is the latency of the ith packet.
Network latency is then averaged by the number of
destinations as shown in figure 4.
Throughput: Number of messages per second
received at destination. The throughput of the network
is the sum of the throughputs from all destinations as
shown in figure 5.
Success Rate: The total number of packets received
at all the destinations vs. the total number of packets
sent from all the sources as shown in figure 6.

528

Loss Rate: Number of lost packets vs. the total

expected number of packets for that destination, as
shown in figure 7.
Energy consumption: Sum of used energy of all the
nodes, where the used energy of a node is the sum of
the energy used for communication, including
transmitting (Pt), receiving (Pr) and idling (Pi), as
shown in figure 8.

throughput is increasing. Maximum measured value is

above 3 messages per second.

Figure 6: Network Success Rate

In veris topology success rate for a long portion of
simulation time is increasing. This means that more
packets are likely to be delivered to the destination.
Approximately after 60 seconds in veris topology the
success rate is near 80%. On the contrary, in grid
topology, we observe that not only success rate is on
average constant, but it has converged approximately
to 40%. The maximum difference in success rate
measurements in both topologies reaches about 50%.

Figure 4: Network Latency

Latency in veris case is much less than latency in
grid topology. We have calculated the average network
latency throughout the simulation for the two cases
which is 0.0842sec for the grid and 0.038 for the veris.
In veris latency is decreased by approximately 55%.

Figure 7: Network Loss Rate

In our optimized veris topology loss rate is much
less than that in grid topology. We also observe that in
veris topology the graph of loss rate is decreasing
throughout simulation time. This does not happen in
grid topology graph which tends to be stable, but it is
not decreasing. Moreover after 60 seconds of
simulation time in veris topology the loss rate seems to
have converged around 0.2 value. On the contrary in
grid topology this value is about 0.6. So in average the
loss rate (after 50-60 seconds) has the maximum
difference and it is decreased by 26%.

Figure 5: Network Throughput

We observe that in veris topology throughput is
lower than that in grid, because in grid topology the
total packets sent to the destination (implying that we
consider 100% delivery rate) are much more than the
packets sent by sources in the veris topology. (About
1180 packets sent in grid and 396 packets in veris
topology, which is 30% of the grid packets.). However
we see that in veris case the graph of throughput tend
to be increasing (convex). This does not happen in grid
topology in which the graph is not increasing, but tends
to converge around 5. In veris topology for a large
portion of simulation time (at about 60 seconds)

529

prove that local measurements taken by sensors are

more concise and should be used instead all others.
Another future step as far as the routing protocol is
concerned to built the spanning tree is to produce a
procedure that will decide if the node coordinates
given by the user are sufficient for the LGR protocol to
form a spanning tree and converge. This will be an
automatic way to decide more precisely about the exact
number of sensors and their placement in the grid.

6. References
Figure 8: Network Energy Consumption

[1] A. Baggio, 2005. Wireless sensor networks in precision

agriculture, Delft University of Technology The
Netherlands,
available
at:
http://www.sics.se/realwsn05/papers/baggio05wireless.pdf

The
difference
in
energy
consumption,
energy_grid(i) energy_veris(i) for i = 1..100 is
greater than zero and keeps increasing. That shows us
that if we continue simulation for more than 100
seconds energy consumption in veris keeps decreasing
compared to grid. That observation is very important in
a sensor network.

[2] J. Burrell, T. Brooke, R. Beckwith 2004. Vineyard

Computing:
Sensor
Networks
in
Agricultural
Production,Published by the IEEE CS and IEEE ComSoc
1536-1268/04
[3] Virginia Tech. 2002. Precision Farming:
Comprehensive Approach, Publication Number: 442-500

4. Conclusions

[4] Kikiras, P.C., Drakoulis D.2003. The European Approach

to Augmented Satellite Based Positioning Systems and their
Application in Precision Farming, Proceedings of the
International Sumposium held at Volos, Greece, 7-9
November 2003

Precision Farming and WSN applications combine

an exciting new area of research that will greatly
improve quality in agricultural production, water
management and will have dramatic reduction in cost
needed. Furthermore, the ease of deployment and
system maintenance opens the way for the adoption of
WSN systems in precision farming. Using the
proposed methodology, in finding the optimal sensor
topology, we contrive to lower implementation cost
and thus make WSN a more appealing solution for all
kinds of fields and cultivations.

[5] Vellidis, G. 2005. A Real-Time Smart Sensor Array for

Scheduling Irrigation in Cotton, NESPAL National
Environmentally Sound Production Agriculture Lab.
[6] Kahn, K. et al. 2002. Ad Hoc Sensor Networks A New
Frontier for Computing Applications, Intel Corporate
Technology Group
[7] Zhang, W. et al. 2003. Integrated Wireless
Sensor/Actuator Networks in an Agricultural Application,
Carnegie Mellon University
[8] Gemptos, T., Foundas, S., Markinos, A., Blackmore.
2003. Precision Agriculture:Applications Perspectives in
Greece and in South Europe. Proceedings of 3rd Conference
on Agriculture Engineering. May, 29-31, 2003. p. 230-237

5. Discussion Future Work

A future aim is the implementation of the proposed
solution to a real field and the mixture of different
measurement technologies (data loggers and Veris).
Mica or Mica-like motes will be used. Another target is
to use sensors that measure locally electrical
conductivity so as to obtain the maps needed. There is
a close collaboration with the Department of
Agriculture Crop Production and Rural Environment of
University of Thessaly to make a complete comparison
of humidity and temperature measurements taken from
data loggers, which take ground measurements and use
wires to send all data in a central unit. As a next step
we aim to verify the measurement system with data
loggers used by the agronomists using Veris and
electrical conductivity measurements and possibly to

[9]Zhang Ying et al. Routing Modeling Application

Simulation Environment, Palo Alto Research Center
(PARC),www2.parc.com/isl/groups/era/nest/Rmase/default.h
tml
[10] K. Ferentinos, T. Tsiligridis, K. Arvanitis 2005. Energy
Optimization of Wireless Sensor Networks for
Environmental Measurements, Computational Intelligence
for Measurement and Applications, 2005, CIMSA. 2005
IEEE International Conderence, p.250-255

530