UAVs Team and Its Application in Agriculture: A Simulation
Environment
Floriano De Rango, Nunzia Palmieri, Mauro Tropea and Giuseppe Potrino
Dept. Computer Engineering, Modeling, Electronics and System Science, University of Calabria,
Ponte Pietro Bucci, Rende, Italy
Keywords:
UAVs Team, Simulation, Coordination.
Abstract:
The work proposes a simulation environment for UAVs management in the agriculture domain. In these last
years, new technologies can effectively support farmer to face issues and threats such as parasites and sudden
climatic changes that can severely degrade the crop or the quality of the cultivated products. However, to
properly manage a UAVs team, equipped with multiple sensors and actuators, it is necessary to test these
technologies in order to plan specific strategies and coordination techniques able to efficiently support farmers
and achieve the targets. At this purpose, this contribution proposes a simulator suitable for the agriculture
domain, where it is possible to set many parameters of this domain of interest.
1 INTRODUCTION
In this paper, it is supposed to use UAVs (also called
drones) in the agriculture domain in a situation where
it is necessary to monitor plants or products in a land.
Often, farmers need to face many issue such as par-
asites or sudden climate changes that can severely
affect the agricultural products quality. At this pur-
pose, new technologies such as drones equipped with
specific sensors, cameras and fertilizers can support
farmers to face the threats and, in the specific case, to
kill parasites that can destroy plants in the cultivated
field (Zhang and Kovacs, 2012).
Let us suppose to have a cultivated area where a
certain number of plants such as trees are distributed.
In this area, it is possible that some parasites, in any
part of the area, can generate and attack the nearly
plants and to reproduce themselves and finally destroy
the plants in the area. In this case, a UAVs team can
be very useful in overseeing the overall area in order
to localize through cameras the attacked plants or to
see parasites moving among plants.
In order to propose coordination techniques of
drones for precision agriculture domain, it is neces-
sary to have a simulator where it is possible to set the
main parameters of the domain of interest and test the
proposed strategies and protocols for coordinating the
drones in the area.
There is no simulator till now, at the best of our
knowledge, that provide these modules and this tes-
tifies the novelty of the research area although an in-
creasing interest has been shown on the industry and
research field.
The paper is organized in the following way: in
Section 2 the related work about the coordination
techniques of swarm of drones is presented; the pro-
posed simulator and its parameters are presented in
Section 3. Section 4 shows the simulation environ-
ment with results and finally the conclusions are sum-
marized in Section 5.
2 RELATED WORK
Recent works have been proposed regarding the use
of drones or micro-drones in precision agriculture do-
main, showing as the applications of drones for preci-
sion agriculture can be interesting and very useful for
farmers especially if the technology can reduce the
cost of control systems (see (Costa and et.al, 2012);
(Primicerio and et al., 2012); (Zhang and Kovacs,
2012)). Moreover, through the introduction of spec-
trometry or cameras on drones, it is possible to evalu-
ate the height of the crops or soil moisture that can be
useful to understand how the crop is raising such as
shown in (Anthony and et.al, 2014); (Colomina and
Molina, 2014); (Hassan-Esfahani and et.al, 2014).
UAVs applications in agriculture are described in
(Pederi and Cheporniuk, 2015). The authors list some
fields of applications of UAV spraying different pro-
tection chemical like insecticide, fungicides and her-
bicides. Crops spraying UAVs are suited for these
tasks because many fields need of ultra low appli-
cation volume of pesticides per hectare and only on
some specific zones of a field and only at a specific
374
Rango, F., Palmieri, N., Tropea, M. and Potrino, G.
UAVs Team and Its Application in Agriculture: A Simulation Environment.
DOI: 10.5220/0006466303740379
In Proceedings of the 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2017), pages 374-379
ISBN: 978-989-758-265-3
Copyright © 2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reser ved
time. In (Devraj and et.al, 2015) authors propose
a web-based system that facilitate farmers/extension
workers to diagnose insects/pests of major pulse crops
suggesting the suitable treatments.
In (Ye and et al., 2013) four architectures of Preci-
sion Agricultural Management Systems (PAMS) are
proposed. These architectures are: the spatial infor-
mation infrastructure platform, the IoT infrastructure
platform, the agriculture management platform and
the mobile client.
In order to support the study of drones, micro-
drones or UAV applied to precision agriculture, it is
necessary to design a simulator able to support in a
modular and flexible way the different actors playing
in the domain. All these issues are focused at an early
stage in this work, showing as it is possible to model
each single module such as mobility module for par-
asites, mobility module for drones, resistance char-
acteristics of plants, and communication technologies
(De Rango et al., 2008), (Fotino et al., 2007) of drones
useful to implement a coordination strategy among
drones (Palmieri et al., 2017).
3 UAVS SIMULATOR
We consider a team of UAVs/drones, that performs a
mission of monitoring a land for destroying parasites
in the plants. Once the parasites are detected by a
UAV, it tries to destroy them through a certain quan-
tity of pesticide that the UAVs carry. However, the
pesticide resources of each drones are limited in quan-
tity. Once a drone ( called leader) notices parasites in
a region, it has to form a coalition based on its current
resource level such as pesticide and energy resources
and it broadcasts some information (i.e, its location,
type and number of its capabilities) to the other UAVs.
The other drones, that have the required resources,
will evaluate their availability to reach the leader or
not. It should be noticed that the coalitions formed are
temporary by nature; once the detected parasites are
destroyed, the coalition members can perform other
tasks.
Drones movement is modeled through the use of
a local map that is stored on-board. This map collects
info about already visited fields, if there are plants in
the near field (cell) and if some plants are infected or
not. Each of these info last for a time on the basis of
the time set by the simulator. This temporal depen-
dence of plants and insects reflect the real situation
where an insect can attach again the plant or a plant
can be planted again on the field.
The map is implemented by a cells square matrix and
each cell represents a portion of field. The drone is
assumed to be in the middle of the cell. The info
related to the cell can be the presence or not of the
plant, the presence of parasites and consequently if
the plant can be infected or not. Moreover, the info
in the cell can also express if a plant is under care or
if the cell has not been explored yet. Drone updates
its map after any movement removing the old info on
the map and replacing them with more updated info.
This local map helps the drone about the next field
(cell) to visit. It selects one of the unexplored cells
performs the field selection. The cell selection prob-
ability is uniformly distributed among all unexplored
cells. However, after visiting some cells, the selec-
tion probability can be distributed again among the
remaining cells.
When all cells in the local map cannot be visited be-
cause they have been already visited from a few time
or because these cells are occupied by a health plant,
the cell selection probability can be changed.
Each drone can exchange its map with all neigh-
bor drones (drones within the drones transmission
range). This possibility can assure that a drone can
know the situation of neighbor cells also without go-
ing to explore them and this speed up the convergence
of the overall field exploration.
Moreover, we consider many plants disseminated on
the field that can attract the parasites. It is assumed
that each parasite does not communicate with other
parasites to perform a strategy but it moves on the ba-
sis of its local scope. It is oriented towards a single
direction and it is limited by its local scope that is as-
sumed in our case to be its three cells in front. During
the movement the parasite can go in one of the adja-
cent and visible cells.
4 SIMULATION ENVIRONMENT
4.1 Simulator Indicators in the GUI
Each drone is equipped with a battery providing the
energy to fly and move in the interested area. It is
also equipped with a pesticide tank to spray on the
parasite to kill them and a wireless module to com-
municate with neighbor drones. The communication
range of each UAV is assumed to be limited. How-
ever, a UAV can reach another UAV by a sequence
of communication links. The fuel tank is considered
in milliliters (ml) whereas the battery power in Watts
(W). All these parameters are selected before assess-
ing the simulation through a GUI in the front-end of
the simulator.
The battery power can be dissipated in three ways:
drone movements, pesticide spraying and communi-
UAVs Team and Its Application in Agriculture: A Simulation Environment
375
Figure 1: Example of UAVs level.
cation, which is due to radiation, signal processing as
well as other circuitry. It is assumed that to eliminate
a parasite is necessary a fixed amount of pesticide.
It is also assumed that drones can move at a con-
stant speed from a cell to another one or they can stop
to spray the pesticide on the tree attached by parasites.
This assumption could be removed extending the mo-
bility model and without affecting the validity of the
overall proposed simulator model.
Moreover, when the fuel tank is below a minimum
level the drone needs to come back at its base station
or when the drone terminated its pesticide. In order to
provide a friendly graphical interface, all drones are
represented with some indicators that can change the
color on the basis of the quantity of remaining pesti-
cide or the battery level such as shown in Fig.1.
In the Fig.1 it is shown the circle colored in the most
internal part to indicate the residual pesticide quantity
of each drone i (S
i
) using colors listed in Table 1. The
residual battery levels of each drone i (P
i
), instead, is
represented with colors indicated in Table 2.
Indicators are also related to the tree health in order
to represent the resistance level to the parasites attack
(Table 3). More specifically, when a tree has been
attached to parasites, its health decreases on the ba-
sis of number of parasites attaching the three and the
time of the parasites remain into the three. We refers
to H
i
as the current health state of each tree and H as
the initial state.
Table 1: Indicator of pesticide remaining quantity.
S
i
>
1
2
S (tank capacity)
1
4
S < S
i
1
2
S (tank capacity)
0 < S
i
1
4
S (tank capacity)
Table 2: Indicator of battery remaining quantity.
P
i
>
1
2
P (Initial battery capacity)
1
4
S < P
i
1
2
P (Initial battery capacity)
0 < P
i
1
4
P (Initial battery capacity)
Table 3: Tree health indicators.
H
i
>
1
2
H
1
4
H < H
i
1
2
H
0 < H
i
1
4
H
H
i
= 0 (Three completely damaged)
4.2 Simulator Front-end
In Fig.2 it is shown an example of the graphical inter-
face of the designed simulator. Fig. 3 shows the GUI
to set all simulation parameters. More specifically, the
parameters adopted in the simulation are listed in the
following:
Pesticide tank: it indicates the capacity of the pes-
ticide tank (ml);
Maximum battery level: it indicates the maximum
battery level when the battery is charged;
Minimum battery level: it indicates the minimum
battery level to be preserved to reach the base sta-
tion;
Percentage of trees in the field: it represents the
tree percentage to distribute on the considered
field;
Tree lifetime: it is the initial health state of all
trees; it is represented by a number that can be
decreased if a parasite attaches the tree;
Base station number: it represents the number of
base station from which drones can leave or can
come back to recharge the batteries;
Energy spent in unit of pesticide spray: it indi-
cates how much energy is dissipated in spray 1 ml
of pesticide;
Movement energy: it represents the energy dissi-
pated to move by 1 m in the considered area;
Tree damage due to parasites (for each time unit):
it indicates the damage attributed to the tree by the
parasite for each time unit that parasite attaches
the plant; it is important because reduces the tree
health state and it can reduce the number of trees
on the field;
Minimum pesticide amount to kill a parasite: it in-
dicates the minimum amount of pesticide (in ml)
to kill a single parasite;
Communication link bandwidth : it is the band-
width related to the communication link; it is im-
portant because it affects the transmission delay of
all info that a drone sends to other drones (for ex-
ample the local map distribution among neighbor
drones);
SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications
376
Simulation number: it is the number of indepen-
dent simulations before elaborating statistics; it
is important to present simulations results in the
confidence interval of 95 % ; this parameter can
be selected at the beginning of the simulation and
it can change on the basis of the simulation pa-
rameters adopted;
Communication distance: it is the transmission
range among drones expressed in meter;
Pesticide spraying delay : it represents the time
necessary to perform 1 ml spray of pesticide;
Recharge time: it indicates the minimum amount
of time to recharge the battery and refill the pesti-
cide tank at the base station;
Parasite movement delay : it is the time necessary
to the parasite to move by 1m;
Map update time: it is the periodic time that
drones exchange their local map with other neigh-
bor drones; it is important because it can affect the
convergence time of the link state routing and it
can affect also the topological info of the drones
network
Map lifetime: it is the time till MAP informa-
tion is considered valid by drone; it assured that
info needs to be updated otherwise they will be
removed because obsoleted; this assures also that
a drone can visit again a cell if it does not receive
any more info of other drones about the cell sta-
tus; this assured that if some parasites move from
one cell to another cell already visited, they can
be detected again and killed;
TTL for help request: it represents the maximum
number of hops that a help request can propagate
in the network; it is useful also to avoid loop or
help request diffusion without time limit;
Time-out help request: it represents the waiting
time of a drone requesting a help by other drones;
after this time a drone can recall the best drones
among all drones answering to help request on the
basis of selection criteria.
In the front-end interface it is possible to select
the data diffusion strategy. In particular, map info can
be disseminated in a restricted flooding with different
maximum-hop. For example a local map distribution
inside the local range of each drone can be considered
a particular case of a restricted flooding with number
of hops equal to 1. If the number of hop increases, the
restricted flooding can become the classical flooding
where all drones are involved in the communication
and can receive the local map of each drone updating
in a fast way the overall network topology.
On the other hand, it is possible in the GUI to se-
lect also a link-state routing as map distribution pro-
tocol in order to build a map and drones topology in
the overall network with the possibility to compute
the minimum cost paths from each drone towards all
drones in the network. The shortest paths are consid-
ering changing the classical minimum-hop count met-
ric with the residual pesticide of drones. This means
that the modified link-state routing allows to select
drones with higher amount of pesticide guaranteeing
that the infected area can be cured after the arriving
of the involved drones.
Figure 2: Graphical Interface.
4.3 Simulation Results
In the following some simulation results related to
consumed energy, killed parasites and pesticide con-
sumption are shown. It is worth pointing out that the
hand designed simulator is a discrete event simulator.
In Fig. 4 it is shown the consumed energy for
flooding with restriction to 1 hop until 5 hops and
link-state routing. LS and restricted flooding to 3,
4 hops perform better than restricted flooding to 1
hop. Thus is due to the map exchange that is not ex-
ploited when the flooding is too restrictive. On the
other hand, extending the flooding to more hops, the
advantage of the map distribution allow more drones
to reduce their movement in already explored areas.
In Fig. 5 it is shown the number of killed parasites
for restricted flooding and LS. Also in this case it is
perceived a dependence flooding by the number of
hops. For this performance metric the best results are
achieved for restricted flooding with 2-3 hops. LS is a
good compromise between number of killed parasites
and consumed energy. Concerning the last graphic in
Fig. 6, the used pesticide is shown for restricted flood-
ing and LS. In this case, LS is able to perform better
than flooding because the distributed protocol is able
to better consider the residual pesticide among drones
involving in the spraying action drone with more pes-
ticide capacity.
UAVs Team and Its Application in Agriculture: A Simulation Environment
377
Figure 3: GUI to set the simulation parameters.
5 CONCLUSIONS
In this paper a novel simulator working in the agri-
culture domain is proposed. It allows to instantiate
Figure 4: Consumed energy vs topology update time for
flooding and link-state routing with pesticide quantity as
metric.
Figure 5: Killed parasites vs topology update time for flood-
ing and link-state routing with pesticide quantity as metric.
the agriculture field, drones with related communica-
tion module and sensor/actuators, parasites with their
local movement and scope and plants with some re-
sistance characteristics. Moreover, two coordination
strategy have been proposed and studied: the first
one is based on a local field map distribution in or-
der to let neighbor drones about the presence of par-
asites or already visited pieces of land; the second
one is based on the extension of a link-state routing
protocol to the agriculture application domain to dis-
tribute all topology info among all drones. It is shown
as the distributed link-state distribution is more suit-
able since it allows to reduce the energy consumption
and to increase te number of killed parasites. More-
Figure 6: Used pesticide vs topology update time for flood-
ing and link-state routing with pesticide quantity as metric.
SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications
378
over, the distribution of pieces of already visited maps
among all drones allow to speed up the algorithm con-
vergence reducing the number of cells visited more
times. The link-state routing has been modified to ac-
count more metrics such as remaining pesticide quan-
tity and already visited cells. Future work includes the
introduction of others parameters, coordination tech-
niques, and many constraints related to the domain of
interest.
REFERENCES
Anthony, D. and et.al (2014). On crop height estimation
with uavs. In IEEE/RSJ Int. Conference on Intelligent
Robots and System.
Colomina, J. and Molina, P. (2014). Unmanned aerial sys-
tems for photogrammetry and remote sensing: a re-
view. Journal of Photogrammetry and Remote Sens-
ing (ISPRS), 92:79–97.
Costa, F. and et.al (2012). The use of unmanned aerial vehi-
cles and wireless sensor network in agriculture appli-
cations. In IEEE Int. Geoscience and Remote Sensing
Symposium.
De Rango, F., Lonetti, P., and Marano, S. (2008). MEA-
DSR: A Multipath Energy-aware Routing Protocol for
Wireless Ad Hoc Networks, pages 215–225. Springer
US.
Devraj, R. and et.al (2015). Expert systems for manage-
ment of insect-pests in pulse crop. In , in 2nd Int.
Conference on Computing for Sustainable Global De-
velopment (INDIACom), pages 1144–1150.
Fotino, M., Gozzi, A., Cano, J.-C., Calafate, C., De Rango,
F., Manzoni, P., and Marano, S. (2007). Evaluating
Energy Consumption of Proactive and Reactive Rout-
ing Protocols in a MANET, pages 119–130. Springer
US.
Hassan-Esfahani, L. and et.al (2014). Topsoil moisture
estimation for precision agriculture using unmanned
aerial vehicle multispectral imagery. In IEEE Int.
Geoscience and Remote Sensing Symposium.
Palmieri, N., Yang, X.-S., De Rango, F., and Marano, S.
(2017). Comparison of bio-inspired algorithms ap-
plied to the coordination of mobile robots consider-
ing the energy consumption. Neural Computing and
Applications, pages 1–24.
Pederi, Y. and Cheporniuk, H. (2015). Unmanned aerial ve-
hicles and new technological methods of monitoring
and crop protection in precision agriculture. In Proc.
of 3rd IEEE Int. Conference Actual Problems of Un-
manned Aerial Vehicles Developments (APUAVD).
Primicerio, J. and et al. (2012). A flexible unmanned aerial
vehicle for precision agriculture. Precision Agricul-
ture, 13:517–523.
Ye, J. and et al. (2013). A precision agriculture management
system based on internet of things and webgis. In 21st
Int. Conference on Geoinformatics, pages 1–5.
Zhang, C. and Kovacs, J. (2012). The application of small
unmanned aerial systems for precision agriculture: a
review. Precision Agriculture.
UAVs Team and Its Application in Agriculture: A Simulation Environment
379