Aalborg Universitet

Comparison of path planning algorithms for an unmanned aerial vehicle deployment

under threats

Danancier, Kevin; Ruvio, Delphine; Sung, Inkyung; Nielsen, Peter

Published in:
IFAC-PapersOnLine

DOI (link to publication from Publisher):

10.1016/j.ifacol.2019.11.493

Creative Commons License

CC BY-NC-ND 4.0

Publication date:
2019

Document Version
Publisher's PDF, also known as Version of record

Link to publication from Aalborg University

Citation for published version (APA):

Danancier, K., Ruvio, D., Sung, I., & Nielsen, P. (2019). Comparison of path planning algorithms for an
unmanned aerial vehicle deployment under threats. IFAC-PapersOnLine, 52(13), 1978-1983.
https://doi.org/10.1016/j.ifacol.2019.11.493

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

? Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
? You may not further distribute the material or use it for any profit-making activity or commercial gain
? You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us at vbn@aub.aau.dk providing details, and we will remove access to
the work immediately and investigate your claim.
9th IFAC Conference on Manufacturing Modelling, Management and
9th
9th IFAC
IFAC Conference
Conference on
on Manufacturing
Manufacturing Modelling,
Modelling, Management
Management and
and
Control
Control
9th IFAC Conference on Manufacturing
Control Modelling,
Available Management
online and
at www.sciencedirect.com
Berlin,
9th IFACGermany, August
Conference on 28-30, 2019
Manufacturing Modelling, Management and
Berlin,
Berlin, Germany,
Control Germany, August
August 28-30,
28-30, 2019
2019
Control
Berlin, Germany, August 28-30, 2019
Berlin, Germany, August 28-30, 2019
ScienceDirect
IFAC PapersOnLine 52-13 (2019) 1978–1983
Comparison
Comparison of of Path
Path Planning
Planning Algorithms
Algorithms
Comparison
for an
Comparison
for an of
UnmannedPath Planning
Aerial
of Path Planning
Unmanned Aerial Algorithms
Vehicle
Algorithms
Vehicle
for
for an
an Unmanned
Unmanned Aerial
Aerial Vehicle
Vehicle
forDeployment
an Unmanned
Deployment Under Threats
Aerial
Under Vehicle
Threats
Deployment Under
Deployment Under Threats Threats
Kevin Danancier ∗∗ Delphine Ruvio ∗∗ Inkyung Sung ∗∗
Kevin ∗∗
Kevin Danancier
Danancier ∗ Delphine Delphine Ruvio
Ruvio ∗∗ ∗ Inkyung
∗
Inkyung Sung Sung ∗∗
Kevin Danancier ∗∗ Peter Delphine
Peter Nielsen
NielsenRuvio ∗∗
∗∗ ∗ Inkyung Sung ∗∗
∗∗
Kevin Danancier Peter Delphine NielsenRuvio Inkyung Sung
Peter Nielsen ∗∗ ∗∗
∗ Peter NielsenPolytech
∗ Department
∗
Department of of Industrial
Industrial Engineering,
Engineering, Polytech Polytech Engineering
Engineering School School
∗ Department of Industrial
of Aix-Marseille Engineering,
University, France Engineering School
∗ Department of Industrial
of
of Aix-Marseille
Aix-Marseille Engineering,
University,
University,Polytech France
France Engineering School
Department of Industrial Engineering,
(kevin.danancier@etu.univ-amu.fr, Polytech Engineering School
delphine.ruvio@etu.univ-amu.fr)
of ∗∗Aix-Marseille University,
(kevin.danancier@etu.univ-amu.fr,
(kevin.danancier@etu.univ-amu.fr, France
delphine.ruvio@etu.univ-amu.fr)
delphine.ruvio@etu.univ-amu.fr)
of ∗∗Aix-Marseille
Operations University,
Research France
Group,
(kevin.danancier@etu.univ-amu.fr,
∗∗ Operations Research
Operations delphine.ruvio@etu.univ-amu.fr)
Research Group,
Group,
(kevin.danancier@etu.univ-amu.fr,
Department delphine.ruvio@etu.univ-amu.fr)
Department of
Department of Materials
of Materials
Materials
∗∗ and
and Production,
∗∗ Operations
and
Operations
ResearchAalborg
Production,
Production,
Research
Group,University,
Aalborg
Aalborg University, Denmark
Group,University, Denmark
Denmark
Department of (inkyung
Materials
(inkyung sung@mp.aau.dk,
and Production,
sung@mp.aau.dk, peter@mp.aau.dk)
Aalborg
peter@mp.aau.dk) University, Denmark
(inkyung sung@mp.aau.dk,
Department of Materials and Production, Aalborg University, Denmarkpeter@mp.aau.dk)
(inkyung sung@mp.aau.dk, peter@mp.aau.dk)
(inkyung sung@mp.aau.dk, peter@mp.aau.dk)
Abstract:
Abstract: Following
Following the
the massive interests
massive interests
interests in in unmanned
in unmanned
unmanned aerial aerial vehicles
vehicles (UAVs),
aerial vehicles (UAVs), various
various
Abstract:
optimization Following
algorithms the
have massive
been proposed for a path planning problem that (UAVs),
allow the various
units to
Abstract:
optimization
optimization Following
algorithms
algorithms the
have
have massive
been
been interests
proposed
proposed for
for in
a
a unmanned
path
path planning
planning aerial
problem
problemvehicles
that
that (UAVs),
allow
allow the
the various
units
units to
to
Abstract:
navigate in aFollowing
region the
filled massive
with threats interests
such as in
a radarunmanned
detection aerial
in airvehicles
defence (UAVs),
systems. various
Among
optimization
navigate
navigate in
in aa algorithms
region
region filledhave
filled with
with been proposed
threats
threats such
such as foraaaradar
as path detection
radar planning problem
detection in
in air that allow
air defence
defence the units
systems.
systems. Among
Among to
optimization
the algorithms,
algorithms, algorithms
we address
address have been
Dijkstra’s proposed
algorithm for aand path planningalgorithm
a heuristic
heuristic problem that
for the allow
the theplanning
path units to
navigate
the
the in a region
algorithms, we
we filled with
address threatsalgorithm
Dijkstra’s
Dijkstra’s such as aand
algorithm radar
and a
a detectionalgorithm
heuristic in air defence
algorithm for
for thesystems.
path
path Among
planning
planning
navigate
of a UAV. in The
a region filled with
algorithms are threats
compared suchunderas a radar
various detection in air defence
configurations of a systems.
region to Among
navigate
thea
of
of aalgorithms,
UAV.
UAV. The we address Dijkstra’s
algorithms are compared algorithm
underand
under variousa heuristic algorithm
configurations of for the path
aa region to planning
navigate
the
with
of
witha respectThe
algorithms,
UAV.
respect Theto we
to
algorithms
the address
optimality
algorithms
the optimality
are and
are
compared
Dijkstra’s thealgorithm
compared
and the computational
under
computational
various
and
various
configurations
a heuristic
complexity algorithm
configurations
complexity of theof
of the of
for region
the path
algorithms.
a region
algorithms.
toCopyright
to
navigate
planning
navigate
Copyright
with
of
 a
c 2019 respect
UAV.
2019 Theto the optimality
algorithms are and the
compared computational
under various complexity
configurations
IFACto the optimality and the computational complexity of the algorithms. Copyright of the ofalgorithms.
a region toCopyright
navigate
with

c
c respect
2019 IFAC
IFAC
©
with
c 2019,
respect
2019 IFAC
to the optimality and the computational complexity of
IFAC(International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
the algorithms. Copyright
c 2019 IFAC
Keywords:
Keywords: Optimization and Control, Control, Scheduling, Operations Operations Research
Keywords: Optimization
Optimization and and Control, Scheduling,
Scheduling, Operations Research Research
Keywords: Optimization and Control, Scheduling, Operations Research
Keywords: Optimization and Control, Scheduling, Operations Research
1. INTRODUCTION
1. INTRODUCTION First, Dijkstra
First, Dijkstra algorithm,
algorithm, aa well-known
well-known exact exact approach
approach for for
1. INTRODUCTION aFirst, Dijkstra
shortest path algorithm,
problem aiswell-known
discussed exact
with approach
an emphasis for
1. INTRODUCTION aaFirst, Dijkstra
shortest
shortest path
path algorithm,
problem
problem ais
iswell-known
discussed
discussed exact
with
with approach
an
an emphasis
emphasis for
Unlike other other automated 1. INTRODUCTION
automated vehicles,
vehicles, whichwhich require
require aa physical
physical on onFirst, Dijkstra
how to to path algorithm,
generate a
a waypoint
waypoint well-knownnetwork exact approach
foran a region
region for
to
Unlike a
on shortest
how
how to generate
generateproblema
a is
waypoint discussed
network
network with for
for a
a emphasis
region to
to
Unlike
line or other
makers automated
to move, vehicles,
e.g. an which
automated require
guided a physical
vehicle, a shortest
navigate and path
a problem
cost for a is
flightdiscussed
between with
two an emphasis
waypoints. A
Unlike
line or other automated
makers to move, vehicles,
e.g. an which require
automated guided a physical
vehicle, on how
navigate
navigate to
and
and generate
a cost
acoordinate
cost for a
for waypoint
a flight
a flight network
between
between two
two for a region
waypoints.
waypoints. to
A
A
line
anUnlike or makers
other
unmanned to move,
automated
aerial e.g. an
vehicles,
vehicle automated
(UAV) whichhas guided
require
less a vehicle,
physical
restrictions on how
waypoint to is generate
a a waypoint
that a network
UAV may for
use a toregion
change to
an line or makers
unmanned to move,
aerial e.g.
vehicle an automated
(UAV) has guided
less vehicle,
restrictions navigate
waypoint
waypoint and
is
is a
a a cost for
coordinate
coordinate a flight
that
that a
abetween
UAV
UAV may
maytwo waypoints.
use
use to
to change
change A
an
on unmanned
lineitsoroperational
makers to aerial
move, vehicle
e.g. an(UAV)
environment hasof less
automated
because restrictions
guided
its vehicle,
flexibility in its navigate
its and a cost such
flight conditions
conditions for a as flight betweenand
a direction
direction two awaypoints.
speed. A
onan unmanned
its operational aerial vehicle
environment (UAV)
because has of less
its restrictions
flexibility in waypoint
flight is a coordinate
such as thata a UAV may
and use
a to
speed. change
on
an its operational
unmanned
movement. aerial
Providing environment
vehicle
an aerial because
(UAV)
viewpointhasofisless
its flexibility
alsorestrictions
an impor- in its flight conditions
waypoint is a coordinate such as thata direction
a UAV may and use a speed.to change
on its operational
movement. Providing environment
an aerial because ofis its flexibility in One its flight
of the conditions
the drawbacks suchof as a direction
Dijkstra’s and a speed.
algorithm in the path
movement.
on itsfeature
tant Providing
operational aerial viewpoint
anmaking
environment
of aa UAV,
UAV, viewpoint
because ofisits
its application
application
also
also an
an impor-
impor-
flexibility
promising in One its flight
One of
of the conditions
drawbacks
drawbacks suchof as
of a direction
Dijkstra’s
Dijkstra’s and a speed.
algorithm
algorithm in
ina the
the path
path
movement.
tant
tant feature
feature Providing
of
of a UAV, an aerial
making
making viewpoint
its
its is
application also an impor-
promising
promising planning
One of theis that it
drawbacks takesof a long
Dijkstra’s time to derive
algorithm in thesolution
path
movement.
and Providing
cost-efficient. A an aerial
military viewpoint
and civil is also an
engineering impor-are planning
planning
One of theis
is that
that it
it
drawbacks takes
takesof a
a long
long
Dijkstra’s time
time to
to derive
derive
algorithm in a
a thesolution
solution
path
andtant cost-efficient.
and feature of a UAV, A makingand
A military its application
civil promising
engineering are as the scale of aa it
network (e.g. the numbers of nodes
nodes and
tant cost-efficient.
examples feature
of of a UAV,
application military
making
areas, and civil
its application
where engineering
deploying promising
a are asplanning
theinscale
the is that
of takes a(e.g.
network longthe time to derive
numbers of a solution and
and cost-efficient.
examples
examples of
of A military
application
application areas,
areas, and civil
where
where engineering
deploying
deploying aa fleet
fleet
fleet
of
are
of
of
as
planning
arcs
as
arcs thein
scale
the
scale
the
of
is that a it
network
network)
of a
takes
network
network)
a(e.g.
longthe
increases,
(e.g.
increases, the
numbers
time
whichto is
numbers
which
derive
is
of nodes
not
of
not
aa solution
nodes
a
and
desired
and
desired
and cost-efficient.
UAVs could A military
dramatically reduce andthecivil
cost engineering
and dangers
dangers are
of characteristic
arcs
as theinscale
the of network)
a network increases,
(e.g. thewhich
numbers is path
not
of aplanning,
nodes desired
and
examples
UAVs
UAVs of
could
could application
dramatically
dramatically areas,
reduce
reduce where
the
the deploying
cost
cost and
and a fleet
dangers of arcs in the of a
network) solution method
increases, for
which the is not a desired
examples
the of
operations application
related, while areas,
while where
maximising deploying
the a fleet
performance of characteristic
characteristic
arcs in the of a solution
of atime.
network) solution method
methodwhich
increases, for the
for the is path
path
not aplanning,
planning,
desired
theUAVs
the could
operations
operations dramatically
related,
related, while reduce the
maximising
maximising cost and
the dangers
performance of especially in real
ofUAVsthe could dramatically
operations. reduce the costthe andperformance
dangers of especially characteristic in of atime.
in real solution method for the path planning,
of thethe
of operations
operations. related, while maximising the performance especially characteristic real
of atime.
solution method for the path planning,
thethe operations.
operations related, while maximising the performance Motivated especially inbyreal
especially in the
real
time.
fact, we
time. we present aa heuristicheuristic algorithmalgorithm
of the
However, operations.
deploying UAVs UAVs to to such
such applications
applications brings brings aa with Motivated
Motivated by the fact,
fact, we present
by thegeneration present a The
heuristic algorithm
of the
However, operations.
However, that deploying
deploying UAVsdifficult
to suchtoapplications brings a with a
Motivated waypoint
by thegeneration
fact, we present scheme. a The
heuristicalgorithm algorithm first
challenge it is often guarantee the safety with aa waypoint
Motivated waypoint
by the generation
fact, we scheme.
scheme.
present a The
heuristicalgorithm
algorithm algorithm first
first
However,
challenge
challenge deploying
that
that it
it is
is UAVs
often
often to
difficult
difficultsuch to
to applications
guarantee
guarantee the
thebrings
safety
safety a draws a straight
with aaawaypoint line from
generation an origin to a destination and
and
However,
and
challenge
deploying
reliability
that
reliability it
of
UAVs
of isUAVs
UAVs
often
to such
during
difficult
during the
to
the
applications
deploymentthe
guarantee
deployment
brings
because
safety
because
a draws
draws
with
adjusts a straight
straight
waypoint
the line
line
line
by
from
from an
generation
inserting anscheme.
origin
scheme.
new toThe
to
originwaypoints
The
algorithm first
aa destination
destination
algorithm
to the
and
and
first
line,
and
an reliability
challenge
operationthatarea of UAVs
it is for
often
a during
difficult
UAV the deployment
to guarantee
deployment often the because
safety adjusts
involves draws
adjusts a straight
the
the line
line line
by
by from
inserting
inserting an origin
new
new to a
waypoints
waypoints destination
to
to the
the and
line,
line,
an and
an reliability
operation
operation of
area
area UAVs
for
for a
a during
UAV
UAV the deployment
deployment
deployment often
often because
involves
involves draws
when a
the straight
line is line from
overlapped an origin
with an to a destination
obstacle or and
threats.
and
obstaclesreliability
and area of
threatsUAVs such during the deployment
as aadeployment
high-rise building
building because
and an an Since adjusts
when
when the
thethe line
line
line is by
is by inserting
overlapped
overlapped new
with
with waypoints
an
an obstacle
obstacle toor
or the the line,
threats.
threats.
an operation
obstacles
obstacles and threats
and area
threats for a
such
suchUAV as high-rise
as adeployment
high-rise buildingoften involves
and
and an Since adjusts the the line
scheme is inserting
simple, the new waypoints
algorithm could tofind a line,
path
an operation
enemy’s air-defence for a UAV
system. often involves when the
Since the scheme
line is is overlapped
simple, thewith
the an obstacle
algorithm could find or threats.
aa path
obstaclesair-defence
enemy’s
enemy’s and threats such as a high-rise building and an quickly.
system. when the theAscheme
line is is
performance simple,
overlappedcomparison algorithm
with thecould
anofobstacle two findor path
threats.
algorithms
obstaclesair-defence
and threats system.
such as a high-rise building and an quickly. Since theA
quickly.
Since the A scheme is simple,
performance
performance
scheme is simple,
the algorithm
comparison
comparison
the of
of
algorithm the
the could
two
two
could
find
algorithms
algorithms
find
a path
a path
Toenemy’s
addressair-defence system.
the challenge,
challenge, is then carried out under various scenarios.
To enemy’s
To address
addressair-defence
the
the system.aaa path
challenge,
path planning
path
planning problem
planning
problem has
problem
has is
has
quickly.
is then
quickly.
A performance
then carried
carried
A out
performance undercomparison
out under various
comparison
of the two algorithms
various scenarios.
scenarios.
of the two algorithms
been actively
To address
been actively studied,
thestudied,
challenge, which
which derives
a path
derives a flight
planning
a flightproblem
path for
path forhasaaa Specifically,
is then carried
Specifically, we out undertest
generate various scenarios.
scenarios varying aa distribu-distribu-
been
UAV actively
To address
from aathe studied,
challenge,
point which
to another
another derives
a path a flight
withplanning
respect path
problem
to navigating forhas is then carried
Specifically, we
weandout undertest
generate
generate various
test scenarios.
scenarios
scenarios varying
varying a distribu-
been actively studied, which derives a flight path for a tion of threats targets on a region. A threat is
UAV
UAV
been from
from a
actively point
point
studied,to
to another
which with
with respect
respect
deriveset aal., flightto
to navigating
navigating
path for a tion Specifically,
tion of threats we generate
and targets test
on scenarios
a region. Avarying
threat a
is aaa danger
danger
distribu-
aaaUAV
UAV
region
region
regionfrom
from
of ainterest
of interest
point
of ainterest
point to
safely
to another
safely
safely
another
(Alotaibi
with
(Alotaibi
(Alotaibi
with
respect
et al.,
respect
2018).
to
2018).
et al., 2018).
toenergy
Various
navigating
Various
Various
navigating a UAV
aation
UAV
of threats
Specifically,
should
of threats
should
we and
and
avoid
targets
generate
avoid during
targets
during
onitascenarios
test
onit
region.Unlike
flight.
aflight.
Avarying
region.Unlike
threat
A threat an
an
is
a
obstacle,
is
danger
distribu-
a dangeraaa
obstacle,
operational constraints
a region of interest such
safely as the
(Alotaibi maximum
et al., 2018). level
Various UAV UAV
tion of
can should
threats
penetrateavoid
and during
targets on ita flight.
region. Unlike
A threatan obstacle,
is a danger
operational
operational
ofa region
a UAV
UAV
constraints
constraints
of interest
and
such
such
safely
safety
as
as the
(Alotaibi
distance
maximum
thefrom
maximum
et al.,
an 2018).
energy
energy
object
level
levela UAV
Various
(e.g. a UAV
UAV can
can should
penetrate
penetrateavoidaaa during
threat but
threat
threat
but
but
it increases
it
it
increases
it flight. Unlikethe
increases
the
the
unit’s
an unit’s
obstacle,
unit’s
treat
treat
treata
ofoperational
of a
a UAV constraints
and
and safety
safety such
distance
distanceas the maximum
from
from an
an energy
object
object (e.g.
(e.g.levela
a a UAV
level. A should
UAV avoid
also during
needs to it flight.
conduct Unlike
tasks, an
i.e. obstacle,
to survey a
operational
building or constraints
another such as
UAV) are the maximum
often imposed energy into level
the UAV
level.
level. can
A
A penetrate
UAV
UAV also
also a threat
needs
needs to
to but it
conduct
conduct increases
tasks,
tasks, the
i.e.
i.e. unit’s
to
to treat
survey
survey
of
buildinga
building UAV or and safety
another UAV)distanceare from
often an object
imposed (e.g.
into thea UAV
targets. can penetrate
The path a threat
planning but it
algorithmsincreases
are the
then unit’s
applied treat
to
of a UAVor and
problem. another
safetyUAV) distanceare often
from an imposed
objectinto (e.g.the level. A The
a targets.
targets.
level. A The
UAVpath
UAV
alsoplanning
path
also
needs toalgorithms
planning
needs
conduct tasks,
toalgorithms
conductavoidingare
are then
tasks,
i.e. applied
then
i.e.
to survey
applied
to survey to
to
building or another UAV) are often imposed into the generate
problem.
problem. targets. flight
The paths
path to visit
planning visit targets
algorithms are then threats.
applied Theto
building or another UAV) are often imposed into the generate generate
targets. flight
flight
The paths
paths
path to
to
planning visit targets
targets
algorithms avoiding
avoiding
are then threats.
threats.
applied The
Theto
problem. the importance of the path planning in a UAV algorithms
Following generate are paths
flight then evaluated
evaluated
to visit in terms
targets terms of the
avoiding thethreats.
threat level level
The
problem.
Following
Following the
the importance
importance of
of the
the path
path planning
planning in
in a
a UAV
UAV algorithms
algorithms
generate are
are
flight then
then
paths evaluated
to visit in
in
targets terms of
of
avoiding the threat
threat
threats. level
The
deployment, various
Following thevarious approaches
importance based
of thebased
path onon exact and
planning and heuris-
UAV of
in aheuris- of paths derived
algorithms
paths derived
are then andevaluated
and time to
time to in compute
terms of
compute the
the paths.
thepaths.
threatBased Based
level
deployment,
deployment,
Following thevarious approaches
approaches
importance thebased
ofproposed on
path(Zhao exact
exact
planning and
in aheuris-
UAV on of
on paths
algorithms
the derived
are then
evaluation, and
we time
evaluated
finally to characterize
compute
in terms of the thepaths.
the threat
favourableBased
level
tic
tic approaches
deployment,
approaches have
various
have been
approaches
been based(Zhao
proposed on exact et
et al.,
and
al., 2018).
heuris-
2018). of paths
the derived
evaluation, and
we time
finally to compute
characterize the paths.
the favourableBased
tic approaches
deployment, have
various been
approaches proposed
based (Zhao
on exact et al.,
and 2018).
heuris- on
of the evaluation,
paths
conditions derived
for each we
and finally
time
algorithm. to characterize
compute the the
paths.favourableBased
Among them, we we present two algorithms, Dijkstra’s algo- conditions
on the evaluation,
tic approaches
Among
Among them,
them,
tic approaches wehave
have
beentwo
present
present
been
proposed
two (ZhaoDijkstra’s
algorithms,
algorithms,
proposed
et al., 2018).
(ZhaoDijkstra’s
et al., 2018).
algo-
algo- conditions
on the evaluation,for each we
for each finally characterize the favourable
algorithm.
algorithm.
we finally characterize the favourable
rithm
Among
rithm and
and a heuristic
them,
a heuristic
we algorithm
present two
algorithm we designed.
designed.
algorithms,
we Dijkstra’s algo- conditions for each algorithm.
rithm
Amongand a heuristic
them, we present algorithm we designed.
two algorithms, Dijkstra’s algo- conditions for each algorithm.
rithm and a heuristic algorithm we designed.
rithm
2405-8963and© a2019,
heuristic algorithm we
IFAC (International designed.
Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright
Copyright © 2019 IFAC 2013
Peer review©
Copyright 2019
2019 IFAC
© under 2013
responsibility of International Federation of Automatic
IFAC 2013Control.
Copyright © 2019 IFAC
10.1016/j.ifacol.2019.11.493 2013
Copyright © 2019 IFAC 2013
2019 IFAC MIM
Berlin, Germany, August 28-30, 2019 Kevin Danancier et al. / IFAC PapersOnLine 52-13 (2019) 1978–1983 1979
2. RELATED WORK
The path planning for UAVs can be categorized into
two classes depending on the mission type of a UAV
deployment.
First, the path planning problem can be found for a mis-
sion, which requires a placement of a UAV on a position,
e.g. delivering a material to a site. In this case, a path
planning problem is often defined as to find a sequence
of waypoints to avoid threats or obstacles (Bertuccelli
et al., 2009; Alotaibi et al., 2018). Fig. 1 illustrates a path
that allows a UAV to visit multiple points while avoiding
collisions with objects.
Fig. 1. Illustration of a path
Please refer to Radmanesh et al. (2018); Zhao et al. (2018)
for comprehensive reviews on the exact and heuristic
approaches for the planning algorithms. The path planning
can also be embedded into a scheduling system for a fleet
of UAVs. Coupled with a decision that assigns tasks to
UAVs, the path planning problem is solved to find a path
between tasks (Coutinho et al., 2018).
The path planning problem has also been studied to
cover a region, termed as the coverage path planning
(CPP). For a region of interest, the CPP consists of 1)
a decomposition of the region into a set of sub-polygons,
2) finding a covering direction for each sub-polygon, and 3)
determining a sequence of the sub-polygons to form a final
covering path (Jiao et al., 2010; Barrientos et al., 2011).
The concept of the approach is described in Fig. 2.
Fig. 2. General steps in the CPP solving
In this paper, we focus on the first type of the path
planning problem. Specifically, we address a situation
where a UAV needs to visit multiple target points following
a given order. A region to navigate is filled with threats,
which can damage a UAV. A UAV can fly over or through
a threat, if necessary, unlike physical obstacles such as
a building, but it increases the treat level of the unit.
The objective of the path planning problem is set as to
minimise the threat level of a UAV during a flight visiting
multiple positions.
2014
To find a path for the problem, we first discuss Dijkstra’s
algorithm, an exact algorithm for a shortest path problem.
Given a network and a function to compute a cost between
two nodes, Dijkstra’s algorithm derives an optimal path
to a destination which minimises the cost of the path.
However, it takes a long time as the size of a problem,
e.g. the number of nodes in a network, increases. Moreover,
the performance of the algorithm is critically depending on
the network used for the path planning, which brings an
additional decision-making problem for the path planning.
A grid network (Zhang et al., 2014) or a set of waypoints
generated by heuristic rules (Alotaibi et al., 2018) are
generally used to construct the network.
To overcome the challenges, we develop a simple heuristic
algorithm expected to produce a path with an acceptable
performance level within a short time. The performance of
the algorithm is compared to that of Dijkstra’s algorithm
under various scenarios. Finally, favourable conditions for
each algorithm are investigated.
Note that a region of interest in this paper is represented
as a two-dimensional space. Although, actual operational
environment of a UAV is a three-dimensional space, the
path planning in two-dimensional environments has been
extensively studied due to its simplicity with an assump-
tion that a UAV flies by maintaining its altitude or manual
adjustment (Zhao et al., 2018). For the path planning in
three-dimensional environment, please refer to Yang et al.
(2014); Ahmad et al. (2017).
3. PATH PLANNING ALGORITHMS
Let us first define a path for a UAV as a sequence of
waypoints. Recall that a waypoint is a coordinate that a
UAV may use to change its flight direction. Threats, which
damage a UAV, and targets, which will be surveyed by a
UAV, are distributed on a region of interest. The sequence
of the targets to survey is given.
With the setting, the path planning problem in this paper
is defined as to find a flight path that minimises the
exposure of a UAV to threats, while surveying targets
following a given order. The assumptions made in this
paper are listed as follows:
• Threats are the only factors of risk;
• A threat is modelled as a stationary circle area;
• The locations of threats and targets are known a
priori ;
• The threat level to UAV is proportional to its flight
time over threats and distance to threats;
• A UAV flies at a same speed during its flight.
To solve the problem, we implement Dijkstra’s and heuris-
tic algorithms.
3.1 Dijkstra’s Algorithm
Given σ, a sequence of targets to survey, we first form a set
of pairs of targets specifying an origin and a destination
of a sub-path. For an origin and a destination of a sub-
path, Dijkstra’s algorithm is applied to find a path between
them. For example, for sequence b − t1 − t2 − b, where b
is a base for a UAV, pairs of waypoints, (b, t1 ), (t1 , t2 )
and (t2 , b), are generated. Then, for each pair, Dijkstra’s
2019 IFAC MIM
1980 Germany, August 28-30, 2019
Berlin, Kevin Danancier et al. / IFAC PapersOnLine 52-13 (2019) 1978–1983
algorithm is used to find a sub-path between the waypoints
in the pair. A complete path that visits all targets is finally
created by connecting the sub-paths for each pair following
σ.
A square grid-based waypoint network as illustrated in
Fig. 3 is constructed for Dijkstra’s algorithm. Such a
network is easy to get and allows Dijkstra’s algorithm to
generate almost all possible paths on a region, when the
density of nodes in the network is high enough.
Fig. 3. Network structure for Dijkstra’s algorithm
Here, the density of the network is determined by a
parameter termed as precision. When the precision level
is high, we have a dense network with many waypoints,
whereas we get a sparse network, when the precision level
is low. Fig. 4 illustrates the function of the precision
parameter. Obviously, the performance of a path derived
by Dijkstra’s algorithm is significantly affected by the
precision level.
Fig. 4. Network with different precision parameters (a)
high precision (b) low precision
A cost to travel from a waypoint to another is calculated
considering a distance between them and the treat level
of being at the destination waypoint. Following the work
of Zhang et al. (2014), the threat level at a position (x, y)
by a threat located at (xt , yt ), can be calculated using an
equation as follows:
1 d2
ft (x, y) = √ e− 2σ , (1)
2πσ
where d is the Euclidean distance between (x, y) and
(xt , yt ). The total threat level at point (x, y) is then
computed as:
F (x, y) = 1 − [1 − ft (x, y)],
t∈T
where T is a set of threats in a region. Finally, the
travelling cost from node i to node j in a network, cij ,
is calculated as follows:
cij = dij + w × Fj ,
where dij is Euclidean distance between node i and j, w is
the parameter for the weight of threats termed as danger,
and Fj is the total threat level at node j. When w is zero,
the problem is a common shortest path problem.
Based on the network structure and the cost function,
Dijkstra’s algorithm with a priority queue (Goldberg and
Tarjan, 1996) is implemented to generate the minimum
cost path between two targets.
3.2 A Heuristic Algorithm – Waypoint Generation
The heuristic algorithm termed as waypoint generation
first draws a straight line between two positions and
adjusts the line when the line is overlapping with a threat
represented as a circle. The size of the circle is determined
considering the threat level. In case of overlapping, in
order to adjust the line, and thus avoid the threat, new
waypoints generated are inserted into the line.
Let us first explain how to detect whether a line between
point A and B is overlapped with a threat. Given a
threat represented as a circle with radius r, if the distance
between A and the centre of the threat or the distance
between B and the centre are less than or equals to r, line
AB is overlapped with the threat. We also need to check if
the minimum distance d between the centre of the threat
and AB is less than or equals to radius r. A situation,
where the last condition is met, is illustrated in Fig. 5.
Fig. 5. Overlap between a path and a threat
When all conditions are not satisfied, a UAV can fly
directly from A to B. Otherwise, additional waypoints to
avoid threats are inserted into the line. The way to adjust a
straight line between two targets with the minimum travel
distance increase is illustrated in Fig. 6. Based on the
concept, the waypoint generation algorithm is designed,
as set out in Algorithm 1.
(2) Fig. 6. Straight line update to avoid a threat
3.3 Path Smooting
As the final step, the paths derived by both Dijkstra’s
(3) algorithm and the waypoint generation algorithm are
smoothed using a function described in Algorithm 2.
This function is essential for the path planning algorithms,
which can bring a 5% reduction in the travel distance. An
example of the path smoothing is presented in Fig. 7.
2015
2019 IFAC MIM

Berlin, Germany, August 28-30, 2019 Kevin Danancier et al. / IFAC PapersOnLine 52-13 (2019) 1978–1983 1981

Algorithm 1 Waypoint Generation The first configuration, Random, represents a random dis-
tribution of threats and targets. The second configuration,
1: function getWaypoints(σ)
Barrier, describes a situation where targets are protected
2: Clear Saf eP ath, Clear Q
by threats as a barrier. The last configuration, Zone, rep-
3: Add the first point in σ to Saf eP ath
resents a situation where threats and targets are relatively
4: for each pair (o,d) in the σ do
collocated in a zone.
5: o ← o
6: while there are threats overlapping with a line Given the threat/target configurations, we also vary the
between o and d do numbers of threats and targets and the parameters for
7: Among all threats overlapping with the line, Dijkstra’s algorithm, i.e. the network density (precision)
find threat t with the shorted distance to the line and a weight of a threat (danger), as listed in Table 1.
8: Generate waypoint wp on the boundary of
threat t, which leads the minimum distance increase Table 1. Parameter setting for the experiments
in Saf eP ath Parameter Levels
9: Add wp to Saf eP ath The number of threats 40 – 20 – 5
10: o ← wp The number of targets 40 – 20 – 5
11: end while Precision (Dijkstra’s) High – Medium – Low
12: Add d to Saf eP ath Danger (Dijkstra’s) High – Low
13: end for
14: return Saf eP ath We generate 10 problem instances for each set of the
15: end function parameter setting and threat/target configuration. Thus,
we conduct 1,620 experiments for Dijkstra’s algorithm
(54 parameter settings x 10 instances x 3 threat/target
Algorithm 2 Path Smoothing (P ath) configurations) and 270 experiments for the waypoint gen-
1: for each triple (o,m,d) in P ath do eration algorithm (9 parameter settings x 10 instances
2: if m is a waypoint and o and d are not a base of x 3 threat/target configurations) in total. The experi-
a UAV and there is no threat then ments have been conducted on a PC with Intel Core i7-
3: Remove m from P ath 8700K@4.3GHz equipped with 16 GB of RAM.
4: end if
5: end for 4.2 Random Configuration
6: return P ath
Fig. 9 summarises the quality of the paths, i.e. the travel
time spent in threats, provided by the algorithms.

Fig. 7. Example of the path smoothing

4. EXPERIMENTAL RESULTS

4.1 Problem Instance Generation
The performance of the path planning algorithms is inves-
tigated under various operational scenarios with respect
to the travel time in threats (optimality) and the compu-
tation time to derive a path (complexity).
We first define three different configurations for treats and
targets as described in Fig. 8.
Fig. 8. Configurations of a threat/target distribution
Fig. 9. The travel time in threats (sec) in the random
configuration
First, it is observed that Dijkstra’s algorithm with a
low danger parameter (a low weight on a threat) works
poorly. Intuitively, with a low danger parameter, Dijkstra’s
algorithm tends to generate a path penetrating threats to
minimise a total travel distance.
On the other hand, the waypoint generation algorithm
shows a comparable performance to Dijkstra’s algorithm.
The waypoint generation algorithm also outperforms Di-
jkstra’s algorithm in terms of the computation time to get
a path as presented in Fig. 10. The waypoint generation
algorithm generates a path, on average, within 0.05 sec.
Fig. 10 also shows that the computation time of the algo-
rithms is a function of the number of targets, whereas the
computation time of the algorithms are relatively insensi-
tive to the number of threats. Obviously, the computation

2016
2019 IFAC MIM
1982 Germany, August 28-30, 2019
Berlin, Kevin Danancier et al. / IFAC PapersOnLine 52-13 (2019) 1978–1983
Fig. 10. Computational time (sec) of the path planning
algorithms in the random configuration
time of Dijkstra’s algorithm increases as the precision of a
network for the algorithm increases.
4.3 Barrier Configuration
Fig. 11 shows the performance comparison between the
path planning algorithms in terms of the travel time in
threats. Unlike the results in the random configuration,
the performance of the waypoint generation algorithm
becomes poor in the barrier threat configuration. When
threats are overlapped, the waypoint generation algorithm
often fails to find a path that avoids the overlapped threats
because a scheme to avoid a single treat is implemented in
the algorithm. Dijkstra’s algorithm with a high precision
network provides the best results.
Fig. 11. The travel time in threats (sec) in the barrier
configuration
Fig. 12 presents the analysis of the computation time of
the path planning algorithms. From the figure, we can
see that the computation time of Dijkstra’s algorithm
is proportional to the number of threats. The waypoint
generation algorithm still finds a path within a short time.
Fig. 12. Computational time (sec) of the path planning
algorithms in the barrier configuration
2017
4.4 Zone Configuration
Fig. 13 summarises the threat level of paths derived by the
two path planning algorithms. We observe that the per-
formance gap between the algorithms becomes significant
as the number of threats increases. When the number of
threats is high, the waypoint generation algorithm works
poorly.
Fig. 13. The travel time in threats (sec) in the zone
configuration
Fig. 14 shows the computational time of the algorithms.
As shown in the figure, Dijkstra’s algorithm with a low
precision network can generate a path faster than the way-
point generation algorithm, when the numbers of targets
and threats are high. Considering the results presented in
Fig. 13, we conclude that Dijkstra’s algorithm outperforms
the waypoint generation algorithm under the zone threat
configuration.
Fig. 14. Computational time (sec) of the path planning
algorithms in the barrier configuration
5. SUMMARY
Following the active interest in path planning algorithms
in a UAV development, this study analyses two types of
path planning algorithms in terms of their ability to avoid
threats and computation time to compute a path. We first
present Dijkstra’s algorithm, which is a common exact
approach to solve the path planning problem. However,
there are many design issues to apply the algorithm, i.e.
how to design a network and the cost function of the
algorithm. As observed, the performance of the algorithm
is quite depending on the design. We also present the
waypoint generation algorithm that finds a path in a
simple manner. In the experiments, under the random
configuration, the algorithm provides comparable paths to
those of Dijkstra’s algorithm with respect to the path’s
threat level, while reducing the time to compute the paths
dramatically. Under the barrier and zone configurations,
Dijkstra’s algorithm outperforms the waypoint generation
algorithm.
2019 IFAC MIM

Berlin, Germany, August 28-30, 2019 Kevin Danancier et al. / IFAC PapersOnLine 52-13 (2019) 1978–1983 1983

REFERENCES
Ahmad, Z., Ullah, F., Tran, C., and Lee, S. (2017).
Efficient energy flight path planning algorithm using 3-
d visibility roadmap for small unmanned aerial vehicle.
International Journal of Aerospace Engineering, 2017.
Alotaibi, K.A., Rosenberger, J.M., Mattingly, S.P.,
Punugu, R.K., and Visoldilokpun, S. (2018). Unmanned
aerial vehicle routing in the presence of threats. Com-
puters & Industrial Engineering, 115, 190–205.
Barrientos, A., Colorado, J., Cerro, J.d., Martinez, A.,
Rossi, C., Sanz, D., and Valente, J. (2011). Aerial
remote sensing in agriculture: A practical approach to
area coverage and path planning for fleets of mini aerial
robots. Journal of Field Robotics, 28(5), 667–689.
Bertuccelli, L., Choi, H.L., Cho, P., and How, J. (2009).
Real-time multi-uav task assignment in dynamic and
uncertain environments. In AIAA guidance, navigation,
and control conference, 5776.
Coutinho, W.P., Battarra, M., and Fliege, J. (2018).
The unmanned aerial vehicle routing and trajectory
optimisation problem, a taxonomic review. Computers
& Industrial Engineering.
Goldberg, A.V. and Tarjan, R.E. (1996). Expected per-
formance of dijkstra’s shortest path algorithm. NEC
Research Institute Report.
Jiao, Y.S., Wang, X.M., Chen, H., and Li, Y. (2010).
Research on the coverage path planning of uavs for poly-
gon areas. In Industrial Electronics and Applications
(ICIEA), 2010 the 5th IEEE Conference on, 1467–1472.
IEEE.
Radmanesh, M., Kumar, M., Guentert, P.H., and Sarim,
M. (2018). Overview of path-planning and obstacle
avoidance algorithms for uavs: a comparative study.
Unmanned Systems, 6(02), 95–118.
Yang, L., Qi, J., Xiao, J., and Yong, X. (2014). A literature
review of uav 3d path planning. In Proceeding of the 11th
World Congress on Intelligent Control and Automation,
2376–2381. IEEE.
Zhang, B., Liu, W., Mao, Z., Liu, J., and Shen, L. (2014).
Cooperative and geometric learning algorithm (CGLA)
for path planning of uavs with limited information.
Automatica, 50(3), 809–820.
Zhao, Y., Zheng, Z., and Liu, Y. (2018). Survey
on computational-intelligence-based uav path planning.
Knowledge-Based Systems, 158, 54–64.

2018