You are on page 1of 41

ENERGY CONSERVATION OF

V-SHAPED SWARMING FIXED WING


DRONES

SEMINAR REPORT

Submitted by

PAUL ROY P J
Reg. No. FIT17ME097

In partial fulfillment for the award of the degree

of

BACHELOR OF TECHNOLOGY
in

MECHANICAL ENGINEERING

FEDERAL INSTITUTE OF SCIENCE AND TECHNOLOGY


(FISAT) ®

NOVEMBER 2020

i
ENERGY CONSERVATION OF
V-SHAPED SWARMING FIXED WING
DRONES

SEMINAR REPORT

Submitted by

PAUL ROY P J
Reg. No. FIT17ME097

In partial fulfillment for the award of the degree

of

BACHELOR OF TECHNOLOGY
in

MECHANICAL ENGINEERING

FEDERAL INSTITUTE OF SCIENCE AND TECHNOLOGY


(FISAT) ®

NOVEMBER 2020

ii
FEDERAL INSTITUTE OF SCIENCE AND TECHNOLOGY
(FISAT) ®
DEPARTMENT OF MECHANICAL ENGINEERING

CERTIFICATE
This is to certify that the Seminar titled

ENERGY CONSERVATION OF
V-SHAPED SWARMING FIXED WING
DRONES
was prepared and presented by

PAUL ROY P J

Reg. No. FIT17ME097

of the Seventh Semester Mechanical Engineering


in partial fulfillment of requirement for the award of
Degree of Bachelor of Technology in Mechanical Engineering under
APJ Abdul Kalam Technological University, during the year 2020.

Martin Antony Dr. Jose Cherian


Seminar Guide Head of the Department
Asst. Professor Professor
Department of Mechanical Engineering Department of Mechanical Engineering

iii
ACKNOWLEDGEMENT

At this pleasing moment of having successfully completed my seminar,


I wish to convey my sincere thanks and gratitude to the management of our
college who provided all the facilities to us.

I hereby express my deep sense of gratitude to my guide Martin


Antony, Asst.Professor Department of Mechanical Engineering, Federal institute
of Science and Technology for the valuable guidance, and suggestions offered
during the course of this seminar, and also in preparing this report.

I would like to extend my thanks to Dr. Jose Cherian, Head of the


Department of Mechanical Engineering, Federal Institute of Science and
Technology.

Along with them I also extend my gratitude towards our principal


Dr.George Issac, Federal Institute of Science and Technology who provided
necessary arrangements for the work to be completed.

I would like to express my thanks to my seminar coordinators, for


their encouragement and guidance for the preparation of this report.

Finally, I take this opportunity for expressing my sincere gratitude to


all my friends for their wholehearted co-operation. Last but not the least I would
like to mentioned here, but who's helping hands have been great support to me,
thank all those people whom I have not mentioned here.

PAUL ROY P J

iv
ABSTRACT

There is currently a growing interest in the area of drag reduction of unmanned


aerial vehicles. Migrating birds take advantage of V-shaped flocking to reduce
the required energy for their flight. Studies have shown that the birds in different
positions in V-shaped flight contend with different drag forces. Lead and tail birds
have to overcome more drag forces than the other birds in V-shaped flock. In this
study, the swarming flight of the fixed-wing drones and a load balancing
mechanism similar to migrating birds during the swarm is investigated. As an
example, the swarm flight of EBee Sensfly wings is analyzed through the
proposed methodology. The aerodynamic drag forces of each individual drone and
the swarm are modeled theoretically. It is shown that drones through the
swarming flight can save up to 70% of their energy and consequently improves
their performance. As swarming drones have different loads and consume a
different level of energy depending on their positions, there is a need to replace
them during the flight in order to enhance their efficiency. To this end, regarding
the number of drones, a replacement algorithm is defined for them so that they
will be able to save more energy during their mission. It is shown that there is
more than 21 percent improvement in flight time and distance of swarming drones
after replacement. This method of replacement and formation can be considered as
one of the effective factors in a drag reduction of swarming aerial vehicles.

v
TABLE OF CONTENTS
Chapter page no
Acknowledgement iv
Abstract v
List of tables vii
List of figures vii
1. Introduction 1
1.1 What is swarming of drones? 1
1.2 What is swarm intelligence? 2
2. Literature Review 3
3. V-formation in birds 5
4. Modeling of drag reduction in V-formation 8
4.1 Drone selected 9
4.2 Assumptions for drag modeling 10
4.3 Drag equation 11
5. Performance enhancements of drones in formation flight 14
6. Sensitivity analysis of swarm parameters 18
7. Load balancing and leader selection protocol 20
7.1 Load balancing 20
7.2 Leader selection protocol in swarming drones 21
7.2.1 STP 22
7.3 Replacement algorithm 23
7.4 Individual drag calculation algorithm 23
8. Control logic 24
8.1 Consensus-based swarm control 25
9. Simulations and results 27
9.1 Even number of drones 27
9.2 Odd number of drones 29
10. Conclusion 32
References 33

vi
LIST OF TABLES

1. The specification of the EBee Sensfly drone. 9


2. Values of swarm parameters. 18

LIST OF FIGURES

1. View of migrating birds in V-flight formation. 7


2. Upwash and Downwash created by trailing vortices 8
of the wing
3. View of EBee Sensefly drone. 9
4. Schematic view of the swarming flight of fixed wing 10
drones with V-shaped formation.
5. Ratio of induced drag for (a)20(b)100 numbers of swarming 14
drone for different values of 2s/b.
6. Total drag of swarming drones for 20 drones for 15
different 2s/b values.
7. Total drag reduction percentage respect to 2s/b=100% 16
for 20 drones for different 2s/b values.
8. Drag for individual drones in swarming flight. 17
9. Sensitivity of the ratio of induced drag to the 18
(a)wingtip spacing(b) wingspan.
10. Sensitivity of the drag of each drone in swarm to the 19
(a)wingtip spacing(b) wingspan.
11. Schematic view of replacement with (a) even number 21
of drones (b) odd number of drones.
12. Snapshots of swarm reformation for nine number of 26
fixed wing drones based on consensus algorithm.
13. Remained energy of individual drones before 27-28
and after replacements for even numbers.

vii
14. View of total (a) covered distance (b) flight time of drones 29
before and after replacements for even numbers.
15. Remained energy of individual drones before 29-30
and after replacements for odd numbers.
16. View of total (a) covered distance (b) flight time of drones 31
before and after replacements for odd numbers.

viii
CHAPTER 1
INTRODUCTION

For a decade, autonomous aerial vehicles have received more attention


than before, which has led to the development of a wide range of drones with
various configurations. In order to perform different types of missions, these
unmanned aerial systems should be able to satisfy given performance
requirements such as high aerodynamic performances, high endurance, hovering
capabilities, agile maneuverability and robustness to withstand potential obstacles.
Today’s drones can cover a wide range of applications that can be categorized into
three different fields based on the type of the Environments that is underwater, air,
and space, Flight zones that is outdoor and indoor, and Missions that is military,
civil and industry. There is now a vastly growing UAS (Unmanned Aerial
Systems) market for many commercial applications, including agriculture,
imagery, real estate, sports photography and shopping/commerce. As the UAS
operations and technology grows in prominence, there exist opportunities for new
roles such as swarming.

1.1 What is swarming of drones?


A swarm is a collection of autonomous agents relying on local sensing and local
implemented behaviours, interacting towards a focused mission goal. Generally
using only one drone for a mission can be risky due to some unpredicted problems
that drone may encounter. To this end, sometimes using multiple drones to
perform a mission is more efficient that is called swarming or formation flight.

1
Swarming flight of drones offers new capabilities such as greater area of
coverage, increased payload carrying capacity, reduced fuel consumption, less
induced drag and a lower pilot/operator effort that is single pilot can control
multiple drones. If one drone of the swarm loses its performance during the flight,
the rest of the drones can carry out the mission. Swarming is a bio-inspired
artificial intelligence based on the behavioural models of swarm flight of birds and
insects in nature.

1.2 What is swarm intelligence?


Swarm intelligence is defined as a complex collective, coordinated, flexible, self-
organized and robust behaviour of a group of individuals which follow a simple
rule. The swarming can be employed by many species of birds. There are many
aerodynamic benefits in formation flights of birds which is more applicable to
larger birds. Also there are other theories that consider the swarming flight as an
anti predator adoption mechanism of birds by small birds. This can be employed
to drones by using various swarming algorithms which should be based on three
basic rules : collision avoidance, swarm centring and velocity matching.

2
CHAPTER 2
LITERATURE REVIEW

A.Mirzaeinia, M.Hassanalian, K.Lee, M.Mirzaeinia (2019)[1]. In this paper,


the swarming flight of the fixed-wing drones and a load balancing mechanism
during the swarm is investigated. As an example, the swarm flight of EBee
Sensefly drones wings are analyzed through the proposed methodology. The
aerodynamic drag forces of each individual drone and swarm are modeled
theoretically. It shows that the drones through the swarming flight can save up to
70% of their energy and consequently improve their performance. As swarming
drones have different loads and consume different levels of energy depending on
their positions, there is a need to replace them during the flight in order to enhance
their efficiency. To this end, regarding the number of drones, a replacement
algorithm is defined for them so that they will be able to save more energy during
their mission.

Amir Mirzaeinia, Mostafa Hassanalian (2019)[2]. Migrating birds take


advantage of V-shaped flocking to reduce the required energy for their flight.
Studies have shown that the birds in different positions in V-shaped flight contend
with different drag forces. Lead and tail birds have to overcome more drag forces
than the other birds in V-shaped flock. Some kinds of flocking birds repositioning
observation have been reported. This paper presents the repositioning
aerodynamics analysis of the V-shaped flocking birds and its energy saving
consequences. This analysis proves that how the Canada geese can fly very far in
a single day through repositioning.

3
John Colombi, David R Jacques, Jacob L. Lambach (2017)[3]. In the seminal
research into simulated swarming, Reynolds developed a methodology that guided
a flock of agents using just three rules: collision avoidance, swarm centering, and
velocity matching. By modifying these rules, an algorithm is created and applied
to unmanned aircraft systems (UAS) so each aircraft in a “swarm” maintains a
precise position relative to the preceding aircraft. Each aircraft experiences a
decrease in induced aerodynamic drag, thus reducing overall fuel consumption,
increasing range and endurance and expanding UAS utility.

4
CHAPTER 3
V-FORMATION IN BIRDS

There are different approaches that birds are applying for drag
reduction. Flocking and formation flight are considered as one of the effective
drag reduction approaches. Migrating birds through flocking are able to fly for
long distances. The shape of the formation flight and number of the migration
birds are also important in their performance. “V” formation of migrating birds
like geese is one of the well-known formations to reduce the overall drag of the
flock. In “V” formation flight, the wings of each bird produce vortices due to the
pressure difference of the top and bottom sides of the wings. This wake turbulence
vortex generates an upwash outboard of the wingtips. If a following bird moves in
one of the upwash regions generated by other birds, it will gain free lift so that it
can fly at a lower angle of attack. Basically, the induced drag of the bird is
reduced with decreasing the angle of attack. Considering this drag reduction, the
birds that are flying in a V-formation flight need less thrust and consequently
power for the forward flight. Hence, migrating birds can fly faster and farther and
increase their flight time. Looking at migrating birds indicates that they can gain
optimum positions that will reduce their drag in a formation flight, by optimizing
the spacing between themselves. In formation flight of the migrating birds, not all
birds benefit equally. Generally, the lead bird and the last following birds have to
fly harder into the undisturbed air. The upwash generated by the leader migrating
bird improves the performance of the two following birds, and these two further
improve conditions for the next two birds in the line. Even though the birds
further could have the majority of the benefit compared the other birds, the front
birds do still gain some reduction in induced drag. The presence of the two birds
flanking the leader helps to dissipate the downwash created by the wingtips of the
leader bird and decreases the induced drag of this bird. These two flanking birds

5
also take advantages from a similar reduction in drag if outboard birds flank them
as well. In other words, the birds in the middle of each line of the V-shaped
formation are in the best position. These birds benefit from the upwash of the lead
birds as well as trailing birds. This additional bonus means that birds in the middle
experience less drag force and have a higher level of power than either the lead
bird or the birds at the end of each line. The Canada geese are a large wild goose
species that are native to arctic and temperate regions of North America. In a
study conducted it showed these migrating birds are applying V-shaped formation
for their migration to save more energy and increase their efficiency. In flocking
Canada geese, the front bird of the formation consuming more energy since it is
the first bird to break up the flow of air. As mentioned before, this bird provides
some induced effects on following birds which reduce their drag force.
Consequently, when the leader birds get tired, it drops out the front position and
moves to somewhere at the middle of V-shaped formation, where the bird can take
the advantages of this low drag position. This rotation of position happens many
times in the course of the long journey to warmer climates. In this rotation,
various members of Canada geese have this chance to take the leadership role for
a while. In flocking of Canada geese, each bird generates downwash and upwash
effects. It is assumed that the downwash created by the wing of the bird is
constant along the span. Generally, the generated vortices by the wing of the
Canada geese are undesirable due to the downwash effects and consequently
increasing the induced drag. However, the generated downwashes are
accompanied by upwashes that can be beneficial for the following birds. A goose
that is flying in upwash regions generated by other geese gains free lift so that it
can fly at a lower angle of attack and need less flapping frequency to generate
aerodynamic forces. As the angle of attack is reduced, the induced drag is also
decreased. Therefore, the following birds that are taking advantage of these
upwashes consume less power. Energy savings and drag reduction of the Canada
geese depend on the precision of wing placement in regions of the vortex field
producing maximum lift, and it also varies with the number of migrating birds. In
flocking, switching the lead position is a solution that through evolution,
migrating birds have picked to save more energy during migration. Basically, in
V-formation flight of migrating Canada geese, when the lead goose becomes tired,

6
it will drop out of the lead position and goes further back into one of the lines of
the V. Then another goose on the further back will rapidly move to forward to
take the leading position. Generally, the leader and the last goose in the lines of V-
formation consume more power than the middle ones. Depending on the number
of the flocking geese, there are one or two geese in each line with the highest level
of power. Therefore, the replacements can be performed for these low-level power
birds. Flocking reformation works based on replacing the lead and tail goose with
the two other geese which have the maximum amount of power among all the
flocking Canada geese. In each replacement step, the leader goose functions as a
coordinator. Once a goose with maximum power is notified, the replacement
communication is performed by geese that are involved in replacement scenarios.

Fig 1. View of migrating birds in V-flight formation.

7
CHAPTER 4
MODELING OF DRAG REDUCTION IN V-FORMATION OF
DRONES

The same method of drag reduction can be adopted from swarming birds. A “V”
formation inspired by migrating birds like geese, is one of the well-recognized
formations to reduce the overall work-load. In this flight formation, the wings of
each drone produce vortices because of the pressure differential across the top and
bottom sides of the wings. This wake turbulence vortex creates a downwash
inboard of the wingtips an upwash outboard of the wingtips as shown in fig 2.

Fig 2.Upwash and Downwash created by trailing vortices of the wing

By positioning a trailing drone within the upwash from the previous one the drone
can achieve free lift so that it can fly at a lower angle of attack. Once the angle of
attack is decreased the induced drag is also reduced so that the drones need less
thrust there by reducing the power needed for the flight of the drone.

8
4.1 Drone selected
The drone used for testing this model is EBee Sensefly fig 3, and the specification
is given in table 1.

Fig 3.View of EBee Sensefly drone.

Table 1.The specification of the EBee Sensfly drone.

Specification Value
Wingspan(m) 1.1
Flying weight(kg) 1.1
Cruise speed(m/s) 11-30(15)
Maximum flight time(min) 59

Cruise speed of the drone for this model is assumed to 15m/s, which is constant
throughout the drone’s flight time.

9
4.2 Assumptions for drag modeling
For modeling of drag various assumptions where made like the drone is flying in
laminar regime of flow. Flow type is taken as incompressible so that the air
density is constant during the analysis. Since the drones are generally using
electric motor and batteries for propulsion, the density of air is not much affected
compared to the higher speed aircrafts. A horseshoe model is followed in which
the bound vortex and trailing vortex is considered as constant which means that
lift is same throughout the span of the wing. In this modeling, it is assumed that
the induced drag is the main portion of the total drag generated, and the total
induced drag of the swarming fixed wing drones depends on the scatters along y-
axis and not on the depth distribution along x-axis. Assumptions made in the
modeling has reduced the complexity of the analysis.

Fig 4. Schematic view of the swarming flight of fixed wing drones with V-shaped
formation.

From fig 4, ‘b’ denotes the wingspan, ‘a’ denotes the length of bound vortex, ‘V’
is the cruise speed of the drone, ‘i’ and ‘j’ are the positions of drones, ‘s’ denotes
the wingtip spacing, ‘δxij’ denotes the vertical distance between two drones and
‘δyij’ denotes the horizontal distance between two drones.

10
4.3 Drag equation
Downward velocity (downwash)by Biot-Savarts law is given by:-

Γ
𝑤𝑑 = 2𝑏 (1)

Generated lift per unit span is given by Kutta-Joukowski theorem:-

𝐿́ = 𝜌Γ𝑉 (2)

To generate same lift as a wing with a wingspan b, the bound vortex should have a
length of 2a, corresponding to the distance between the tip vortices. The ratio of
a to b is assumed to be π/8.
Therefore the total generated lift by the wing of drone is given by:-

𝐿 = 2𝑎𝜌Γ𝑉 (3)

Since the fixed wing drones are flying at low speed, in this model it is assumed
that induced drag is the main portion of the total generated drag.
The induced drag for the wing of the drone is calculated by:-

𝐷𝐼 = 2𝑎𝜌Γ𝑤𝑑 (4)

Substituting wd from equation (1) in equation (4),

Γ Π
𝐷𝐼 = 2𝑎𝜌Γ 2𝑏 = 8 𝜌Γ 2 (5)

The induced drag can be defined with substituting equation (3) in equation (5) as:-

𝜋 4𝐿 2 2𝐿2 2(𝑚𝑔)2
𝐷𝐼 = 8 𝜌 [𝜋𝜌𝑏𝑉] = 𝜋𝜌𝑏2𝑉 2 = 𝜋𝜌𝑏2𝑉 2 (6)

11
Where ‘m’ and ‘g’ are the mass and acceleration due to gravity of the drone. For a
fixed wing drones with similar characteristics such as wingspan, weight and cruise
speed, the drag force will be the same.
The vertical velocity component or the upwash at a distance δx, δy away from the
center of the wing is calculated by applying the Biot-Savarts law as:-

Γ 1 𝛿𝑦+𝑎 𝛿𝑦−𝑎 Γ 𝑎
𝑤(𝛿𝑥, 𝛿𝑦) = − 4𝜋 𝛿𝑥 { − } − 2𝜋 𝛿𝑦 2−𝑎2 {1 −
√𝛿𝑥 2 +(𝛿𝑦+𝑎)2 √𝛿𝑥 2 +(𝛿𝑦−𝑎)2

𝛿𝑥
} (7)
√𝛿𝑥 2 +(𝛿𝑦+𝑎)2

Generally, the total induced drag of swarming of fixed wing drones is equal to
sum of the self-induced drags of each drone and as many mutual drag as there are
permutations of their wings in pair.
The total induced drag for n swarming fixed wing drone is given by:-

𝐷𝐼 (𝑛) = ∑𝑛𝑖=1 ∑𝑛𝑗=1 𝐷𝐼𝑖𝑗 (8)

In equation (8), DI11, DI22…..and DInn are assumed to be equal and can be replaced
by n×DI11. Assume an unstaggered flight formation of fixed wing drones, the
mutually induced drag terms can be written as:-

𝐷𝐼𝑖𝑗 = 𝐷𝐼𝑗𝑖 (9)

Mutually induced drag is given by:-

𝐷𝐼12 = 2𝑎𝜌Γ𝑤(𝛿𝑥, 𝛿𝑦) (10)

Taking the integral mean value of the upwash over the span a of drone 1 gives:-

𝐷𝐼12 = 2𝑎𝜌Γ𝑤
̅ (11)

12
The average upwash velocity induced by the drone at a point (δx, δy) is:-

1 𝛿𝑦+𝑎
𝑤
̅(𝛿𝑥, 𝛿𝑦) = 2𝑎 ∫𝛿𝑦−𝑎 𝑤(𝛿𝑥, 𝜂) 𝑑𝜂 (12)

Munks stagger theorem is applied to simplify the calculations. The theorem states
that “a collection of lifting surfaces (drones) may be translated in the streamwise
direction (x axis) without affecting the total induced drag of the system (swarm)
as long as the circulation of every wing (or lift) is unchanged”.
Substituting equation (12) in equation (11) and simplifying the equations we get:-

2𝐷𝐼11 2𝑎 2
𝐷𝐼𝑖𝑗 = log [1 − (|𝑖−𝑗|(𝑏+𝑠)) ] (13)
𝜋2

𝛿𝑦𝑖𝑗 = |𝑖 − 𝑗|(𝑏 + 𝑠) (14)

Total induced drag = self- induced + mutual induced drags, therefore equation
(8) can be modified as:-

𝐷𝐼 (𝑛) = 𝑛 × 𝐷𝐼11 + 2 ∑𝑛−1 𝑛


𝑖=1 ∑𝑗=𝑖+1 𝐷𝐼𝑖𝑗 (15)

Substituting equation (13) in equation (15):-

4𝐷𝐼11 2𝑎 2
𝐷𝐼 (𝑛) = 𝑛 × 𝐷𝐼11 + ∑𝑛−1 𝑛
𝑖=1 ∑𝑗=𝑖+1 log [1 − ( ) ] (16)
𝜋2 |𝑖−𝑗|(𝑏+𝑠)

The total drag of each individual drone (drone-k) considering just one side of the
V-shaped formation is given by:-

2𝐷𝐼11 2𝑎 2
𝐷𝐼𝑘 = 𝐷𝐼11 + ∑𝑛𝑗=1 log [1 − ( ) ] (17)
𝜋2 |𝑘−𝑗|(𝑏+𝑠)

13
CHAPTER 5
PERFORMANCE ENHANCEMENTS OF DRONES IN
FORMATION FLIGHT

To study the effects of different parameters in swarming flight of


the proposed drone, analysis based on the proposed drag modeling was conducted.
The main parameters affecting the individual drag of each drone in swarming
flight are:- wing span, wingtip spacing, mass of drone, length of the bound vortex,
cruising velocity and number of drones in formation.
A ratio of induced drag in formation flight to solo flight of fixed wing drones is
defined to understand how formation flight affects performance of drones, which
is given by:-

𝐷𝐼11 (𝑛) 4 2𝑎 2
𝐷. 𝑅 = = 1 + 𝑛𝜋2 ∑𝑛−1 𝑛
𝑖=1 ∑𝑗=𝑖+1 log [1 − (|𝑖−𝑗|(𝑏+𝑠)) ] (18)
𝑛𝐷𝐼11

Where D.R is calculated for ratios of the distance between the wings to the semi-
wingspan, 2s/b=100, 50, 10, 0, -10 and -24% as shown in Fig 5.

Fig 5.Ratio of induced drag for (a)20(b)100 numbers of swarming drone for
different values of 2s/b.

14
In Fig 5(a), for different values of the s/b, the ratio of the induced drag is changed.
A ratio of −24% has the maximum drag reduction which is equal to s =2a −b. In
this special case when the wing-tip spacing is such that the tip vortices of the two
adjacent drones overlap s =2a −b. Fig 5(a) indicates that for this 2s/b ratio, the
maximum drag reduction can be achieved.
In Fig 5(b), the effect of the drones’ numbers in the ratio of the induced drag is
shown. It is visible that if the numbers of the drones increased more than 50, there
will be a slight change in the ratio of the induced drag. For 2s/b =−24% a drag
reduction of almost 80% can be seen in swarming flight of the considered drone.

The total drag of the swarming versus the number of the drones is calculated for
different values of the 2s/b as shown in Fig 6.

Fig 6.Total drag of swarming drones for 20 drones for different 2s/b values.

In Fig 6, the summation of the drag of swarm is shown. As indicated in Fig 6, with
decreasing the 2s/b the total drag of the swarming is reducing considerably.

The percentages of the total drag reduction respect to change of the s are shown in
Fig 7. For calculating the total drag reduction percentage, the following equation is
applied.

𝐷(2𝑠⁄𝑏) −𝐷(2𝑠⁄𝑏=100)
𝑡𝑜𝑡𝑎𝑙 𝑑𝑟𝑎𝑔 𝑟𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 = (19)
𝐷(2𝑠⁄𝑏=100)

15
Fig 7.Total drag reduction percentage respect to 2s/b=100% for 20 drones for
different 2s/b values.

Significant amount of drag reduction can be achieved by modifying the value of s.


For instance, for 2s/b =−0.24 a drag reduction of almost 70% can be gained
compared to 2s/b =1.The results in Fig.6 &7 demonstrate that the amount of drag
reduction can be increased if the distance s is decreased. By spacing drones apart
properly, they can achieve optimum positions that reduce the drag of every drone
in the formation. However, not all drones benefit equally in the swarm.
In this study to avoid the collision of the drones during the swarming in real
applications, the 2s/b is considered 10%. Therefore, in the rest of this study s
=0.055m. Applying the values of the considered fixed-wing drone in the proposed
modeling, the drag of the individual drones can be calculated through equation
(17).

16
In Fig 8, the drag of each drone in swarming flight is shown.

Fig 8. Drag for individual drones in swarming flight.

The obtained results in Fig 8 show that with increasing the number of swarming
drones, the drag force of each individual drone decreases. Also, depending on the
odd or even number of the drones always there are one or two drones at the
middle that have the lowest value of the drag. For example, if there are 2k +1
drones that are performing swarming, the drone k +1 has the lowest drag and if
there are 2k drones the drones k and k +1 both have the lowest amount of drag. In
these two cases, the rest of the drones have higher values of the drag which
increases symmetrically from the middle ones. It can be concluded from Fig 8 that
the leader and the tail drones are consuming more power than the rest of the
drones during the swarming flight.

17
CHAPTER 6
SENSITIVITY ANALYSIS OF SWARM PARAMETERS

Sensitivity analysis is carried out to study the effects of each swarm


parameter. The sensitivity of the ratio of induced drag and individual drag of
swarming drones to the independent parameters, such as wingspan (b) and wingtip
spacing (s) is investigated. The reference values for b and s are equal to 1.1 m and
0.055 m, respectively. Sensitivities of ±5%, ±10%, ±15% are considered in this
analysis. Table 2 shows the value of parameters.
Table 2.Values of swarm parameters.
Parameters Wingspan (b,m) Wingtip spacing(s,m)

0.935, 0.99, 1.045, 1.1, 0.055


Wingspan
1.155, 1.21, 1.265

0.04675, 0.0495,
Wingtip spacing 1.1
0.05225, 0.055, 0.05775,
0.0605, 0.06325
In Fig 9(a) and 9(b), the sensitivities of the ratio of induced drag (formation/solo
flight) to the wingtip spacing and wingspan are shown, respectively.

Fig 9. Sensitivity of the ratio of induced drag to the (a) Wingtip spacing
(b) Wingspan.

18
The results in Fig 9(a) indicate that with increasing the wingtip spacing from the
reference value, the ratio of induced drag is increased. Fig 9(b) demonstrates that
with increasing the wingspan of the drones from the reference value, the ratio of
induced drag is decreased. Fig 9(a) and (b) also show the sensitivity of the ratio of
the induced drag to the number of the drones.
The sensitivities of the drag of each drone in a swarm with 10 drones to the
Wingtip spacing and Wingspan, are shown in Fig 10 (a) and (b), respectively.

Fig 10.Sensitivity of the drag of each drone in swarm to the (a) Wingtip spacing
(b) Wingspan.

It can be seen that the drag of each drone in swarm increases with increasing the
wingtip spacing and decreasing the wingspan.

19
CHAPTER 7
LOAD BALANCING AND LEADER SELECTION PROTOCOL

7.1 Load balancing


Load balancing is a critical part of any distributed systems. There are situations
that some parts of the system are overloaded while the other parts of the
distributed system might not be that much loaded. Optimum resource usage is the
main goal of load balancing. The same phenomenon can be seen in the swarming
flight of migrating birds. Even though all the birds can take advantage of
swarming flight, the bird in the lead position has to work the hardest. Generally, in
V-formation flight, when the lead bird becomes tired, it will drop out of the lead
position and goes further back into one of the lines of the V. Then another bird on
the further back will rapidly move to forward to take the leading position.
Generally, swarming drones have different loads. Assuming all the drones in the
swarm have the same applications and same processing load, they still consume
different energy. The reason behind this is that the different drones, as indicated in
Fig 8 have to overcome different drag forces as they are flying together.
Therefore, they need to spend different energy in order to fly together. Those
drones which are spending more energy during swarming flight may lose their
energy functionality and performance consequently. However, the ones that
consume less energy due to trailing aerodynamic effects of the front drones can
have more endurance and save more energy. In more complicated scenarios in
swarm drones, they may have different functionalities which make them spend
more energy to fly and perform their mission. Load balancing in swarming drones
can be carried out in different ways. The very first solution can be reconfiguring
the drones’ positions.

20
7.2 Leader selection protocol in swarming drones
In swarming drones’ scenarios, load balancing needs some communications
between drones to find out how energy consumption is happening in different
drones and decide how to replace. The very first solution is to have a centralized
manager inspiring from the queen honey bees. The central manager or the leader
drone is needed to gather all the drones’ information and find out how to perform
position reconfiguration or functions they are doing. Here in this work, it is
assumed that all swarming drones are able to communicate with each other. The
second question here is who is the swarm leader? The very preliminary solution is
to pre-configure the leader drone in initialization step just before they start to fly.
Then the leader broadcasts a special signal code to everyone. This broadcasting
signal helps all swarming drones know about the leader, and they forward their
information to the leader. Leader calculates the required parameters and transmits
back to the drones to issue their position replacement commands. Fig 11,Shows
different scenarios for replacement of leader and tail drone in a swarm.

Fig 11. Schematic view of replacement with (a) even number of drones (b) odd
number of drones.
There are 2k +1drones, the leader drone can be replaced with drone k +1 and the
drone 2k +1 can be replaced with drone k +2 in that line and if there are 2k drones,
the leader will be substituted with drone k and the last one, drone 2k will be
replaced with drone k +1. The proposed method is easy to implement and perform
because everything is pre-designed and pre-configured into the drones, and there
is not much intelligence in drones. The problem with this method is that the pre-
configure leader is a single point of failure and if it dies and there will be no ways
to continue the mission.

21
The second method which is more intelligent than the pre-configuration scenarios.
This method is inspired by Spanning Tree Protocol (STP) in switched networks.

7.2.1 STP
In switched network when a node tries to communicate with other nodes it
broadcast an address through the switches but it forms a loop between each
switches and it will never establish a connection. This is where STP comes to play
STP selects the first possible connection and eliminates the rest. If the first
connection is interrupted next possible connection is enabled. So that a single
point failure never occurs in a system.

In the drone scenario, it seems using the available power as node ID is more
reliable since the node with more power has more chance to survive more than the
others. After advertising the node IDs, everyone which is receiving higher ID
from others stops advertising more and accept the other one as a leader. In this
method, leader selection is performed adaptively based on the drones’ energies.
Swarming reformation works based on replacing the lead and tail agent with the
two other agents which have the maximum amount of energy among all the
swarming agents. A specific protocol is needed to find the agent with maximum
energy. This algorithm is inspired by the root selection protocol, which has been
used as an initial step in a network spanning protocol. Iterative minimum energy
agent elimination is used to find the max energy agent to be replaced with lead or
tail agent as they are spending more energy than the others. To find the max
energy agent, they start advertising their energy to all the others, once an agent
received an agent with higher level of energy then it will finish more advertising.
It is also assumed that direct communication is possible between all the agents. In
each replacement step, the leader agent functions as a coordinator. Once an agent
with maximum energy is found, the replacement command will be issued to
agents that are involved in replacement scenarios. As it is discussed before, four
agents are involved in the replacement process, which are head/tail and two agents
with maximum power.

22
The following algorithms demonstrate the replacement and load balancing
mechanisms:-

7.3 Replacement mechanism


1: Function Leader_and_tail_Swapping(a, s, b, N, v, init_energy)
d_11 =(2 *mˆ2 *9.8ˆ2)/(pi *1.225 *bˆ2 *vˆ2);
(note:- induced drag of one drone)
2: D_total =Drag_Calculation_function(a, s, b, N, d_11);
(note:- Vector contains drag of all drones)
3: While(min(Remind energy) >0.01 *initial energy)
4: Distance =0.5 *Remind Energy of the leader/D_total (1)
5: Remind Energy =Remind Energy – Distance *D_total
(note:-Vector contains remind energy of all drones)
6: Swap(leader, drone with maximum energy)
7: Swap(tail, drone with second maximum max energy)
8: End Function

7.4 Individual drag calculation algorithm


1: Function D_total =Drag_Calculation_function(a, s, b, N, d_11)
2: For i =1:N
3: For j =1:N
4: IF i ~=j
5: Denum =abs(i −j) *(b +s);
Drag_val =Drag_val *(1 −(2 *a/Denum).ˆ2);
6: END
7: D_total(i) =log(Drag_val) *(2 *d_11)/(piˆ2) +d_11;
8: End
9: End
10: End Function

23
CHAPTER 8
CONTROL LOGIC

Generally swarming drones use rules to be in the


formation. The complexity of swarming increases as the number of the rules
increase. The main rules followed in a swarming flight are:- collision avoidance,
velocity matching, swarm centering. So here is an example of the rules followed
in a case study in swarming of aerosonde drone to find its efficiency during
swarming:-
1. Velocity Matching:- The output of the rule is designed to accelerate or
decelerate each aircraft to an aerodynamically optimum airspeed.
2. Swarm Centering:- Each aircraft uses the swarm’s current speed and
desired direction to evaluate an aim point in the future, and then adjusts
speed and heading to arrive at that point at the right time. As the aircraft
nears the desired position, the forward look-ahead distance decreases.
3. Collision Avoidance:- A buffer radius is established. If any aircraft are
within a circular region defined by buffer radius, current velocities are
compared to see if separation distance is increasing or decreasing. If
decreasing, an immediate evasive turn away and acceleration/deceleration
is commanded.
4. Formation position changes:- Each aircraft determines if it should maintain
its current formation position or move to a different position. Formation
position changes are normally initiated after turns in order to reform the
proper formation in a minimum amount of time.

24
8.1 Consensus-based swarm control
Swarming control of drones can be achieved by the consensus algorithm.
Generally, a consensus algorithm is to find an agreement of certain quantities
through communications in a multi-agent system. For the position of the drone-i
in Cartesian coordinates denoted by qi=[xi, yi, zi]T, where the superscript T is a
transpose operator, the swarming is achievable by the following consensus
algorithm:-

𝑞𝑖̇ (𝑡) = − ∑𝑗∈𝑁𝑖 𝑎𝑖𝑗 (𝑞𝑖 (𝑡) − 𝑞𝑗 (𝑡)) (20)

where aij is some nonnegative weight, and Ni is a set of neighbors for ith drone. In
the case of swarm control problems, equation 20 can be modified by:

𝑞𝑖̇ (𝑡) = − ∑𝑗∈𝑁𝑖 𝑎𝑖𝑗 (𝑞𝑖 (𝑡) − 𝑞𝑗 (𝑡) − ∆𝑖𝑗 ) (21)

where ∆ij=[∆x,ij, ∆y,ij, ∆z,ij]T with some constant ∆x,ij, ∆y,ij, and ∆z,ij for predefined
formation.
The reachability of swarm control indicates qi→qj+∆ij, ∀(i, j)as t→∞, regardless of
initial conditions. For the planar swarming (i.e., the altitude of all drones is
identical to a certain value, which is determined by a leader), ∆z,ij=0, ∀(i, j).

25
Fig 12.Snapshots of swarm reformation for nine number of fixed wing drones
based on consensus algorithm.
Whenever the swarming reformation event occurs to balance the load through the
proposed leader selection protocol, the consensus-based swarm control is carried
out in order to reshape the formation. Fig 12 (a)-(f) show snapshots of swarm
reformation. Initially, each drone is numbered in order from the right to the left as
indicated in 12 (a). Then the proposed leader selection protocol computes the new
order for each drone, followed by reshaping the formation based on consensus
algorithm Fig12 (b)-(f). After a certain distance of the flight, this process takes
place repeatedly for the purpose of load balancing.

26
CHAPTER 9
SIMULATIONS AND RESULTS

Considered drones almost have one hour endurance and if they fly
with 15m/s, they would be able to fly a distance of 54km. If it is assumed that an
individual drone generate 1N drag force during their flight, so that the energy
resource of 54KJ will be need for the entire flight. In this model it is assumed that
the leader and the last drones will change their positions during the flight once
they reach 50% of their initial energy. Two simulations where conducted
considering the above factors to cover both odd and even numbers of drones in
each swarm line. The simulations are ends once the drone with minimum power
reaches 1% of its initial energy. In the first scenario ten drones are considered as a
swarming line (Fig 13) and nine drones are considered as the second scenario
(Fig15).
9.1 Even number of drones

Fig 13. Remained energy of individual drones before and after replacements for
even numbers.

27
Fig 13. Remained energy of individual drones before and after replacements for
even numbers.

Fig 13 shows the remained energy of each agent and their level of energy before
and after replacements. Fig 13 (a), (c), (e), and (g) are showing the drones
reformations before replacement and Fig 13(b), (d), (f), and (h) are indicating the
drones reformations after replacement. Three replacements happen in this
scenario. The simulations are ended once the drone with minimum power reaches
1% of its initial energy. All the drones have the same initial energy. Fig 13 (a) and
(b) are demonstrating the drones formations before and after the first replacement
step. For example, D1 and D6 are replaced with D5 and D10, respectively, in the
first replacement step. Figs 13 (c)-(h) are showing this information for other
replacement steps in three consecutive rows. D5 and D7 are replaced with D4 and
D6, respectively, in the second step and D3 and D8 are also repositioned with D4
and D7 in the third step, respectively. In the fourth replacement step, there is no
replacement because the first and last agents have the highest energy in the
swarm. As it is illustrated in Fig 13, swarm reformation energy is always
symmetric if the swam consist of even number of drones.

28
Fig 14. View of total (a) covered distance (b) flight time of drones before and after
replacements for even numbers.
Fig 14(a) and (b) shows the total swarm flight displacement and time for ten
drones. The dashed red line is showing the case that no replacement is happening.
Simulation results in Fig 14 show that more than 21% enhancement in flight time
and displacement can be achieved for the swarming drones with replacement.
Swarming flight time and displacement without replacement are calculated based
on the first agent flight capability. In the non-replacement scenario, once the
energy level of the leader agent is drained then it is considered as the swarm’s
achieved flight time and displacement.

9.2 Odd number of drones

Fig 15. Remained energy of individual drones before and after replacements for
odd numbers.

29
Fig 15. Remained energy of individual drones before and after replacements for
odd numbers.

Fig 15 shows the swarm reformation scenarios for nine drones. Fig15 (a)-(h)are
showing swarm drones reformations before and after replacement in four steps.
Fig 15 (a), (c), (e) and (g) and Fig15 (b), (d), (f), and (h) are indicating energy
level before and after replacement, respectively. In the first replacement step of
this scenario, D1 and D9 are replaced with D5 and D4, respectively. D5 with D6
and D4 with D3 are replaced in the second step and in the third step, D6 with D7
and D2 with D3 are relocated. Finally, in the last step, D7 is not replaced because
it still has the highest energy but D2 is replaced with D6. As shown in Fig 15,
swarm reformation energy is always asymmetric if the swam consist of an odd
number of drones.

30
Fig 16. View of total (a) covered distance (b) flight time of drones before and after
replacements for odd numbers.

Fig 16 (a) and 16 (b) show the total swarm flight displacement and time for nine
drones. The dashed red line is showing the scenario without replacement. The
results indicate that for this case there is more than 21% enhancement in flight
time and displacement of the drones.

31
CHAPTER 10
CONCLUSION

• The assumptions made in this modeling have reduced the complexity of


the analysis of this study. More advanced computational fluid dynamics
theories can be performed in future works, to extend the analysis.
• Load balancing was carried out on swarming drones to increase their
efficiency. Inspiration by migrating birds and their leader replacement was
the main idea of reconfiguring the drones’ positions.
• In this proposed concept, the drones which have spent a higher level of
energy during the swarming flight are replaced with drones that consumed
less energy
• It was shown that lead and tail drones in V-formation flight are the agents
that are consuming the highest level of the energy.
• A specific protocol was defined to find the agents with the highest
remaining energy to replace them with lead and tail agents. This protocol
was inspired by root node selection in spanning tree protocol in computer
networks.
• The extensive simulations and results showed that swarms with
replacement could have more than 21% enhancement in their flight range
and endurance compared with swarms without reformation.

32
REFERENCES

[1] Mirzaeinia, A., Hassanalian, M., Lee, K., & Mirzaeinia, M. (2019).
“Energy conservation of V-shaped swarming fixed-wing drones through
position reconfiguration”. Aerospace Science and Technology, 105398.
doi:10.1016/j.ast.2019.105398
[2] Mirzaeinia, A., & Hassanalian, M. (2019). “Energy Conservation of V-
Shaped Flocking Canada Geese through Leader and Tail Switching”.
AIAA Propulsion and Energy 2019 Forum. doi:10.2514/6.2019-4152
[3] Colombi, J., Jacques, D. R., & Lambach, J. L. (2017). “Integrating UAS
swarming with formation drag reduction” 2017 Annual IEEE International
Systems Conference (SysCon). doi:10.1109/syscon.2017.7934764
[4] Kshatriya, M., & Blake, R. W. (1992). “Theoretical model of the optimum
flock size of birds flying in formation”. Journal of Theoretical Biology,
157(2), 135–174. doi:10.1016/s0022-5193(05)80618-6

33

You might also like