E cient 3D path planning for drone swarm using

improved Sine Cosine Algorithm

Probhat Pachung
South Asian University
Kopal Pandya
University of Manchester
Atulya Nagar
Liverpool Hope University
Jagdish Chand Bansal (  [email protected] )
South Asian University

Research Article

Keywords: Path Planning, Internet of Drones (IoDs), Meta-heuristics, Sine Cosine Algorithm (SCA), Drone
Swarm, Obstacle Avoidance

Posted Date: July 11th, 2023

DOI: https://doi.org/10.21203/rs.3.rs-3142581/v1

License:   This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License

Additional Declarations: No competing interests reported.

 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm

using improved Sine Cosine Algorithm
Probhat Pachung1 , Kopal Pandya2 , Atulya Nagar3
and Jagdish Chand Bansal4*
1 Department of Mathematics, South Asian University, New
Delhi, 110021, New Delhi, India.
2 Department of Mechanical, Aerospace and Civil Engineering,

University of Manchester, Oxford Road, Manchester, M13 9PL,

United Kingdom.
3 Department of Mathematics , Liverpool Hope University, Hope

Park, Liverpool, L16 9JD, United Kingdom.

4* Department of Mathematics, South Asian University, New

Delhi, 110021, New Delhi, India.

*Corresponding author(s). E-mail(s): [email protected];

Contributing authors: [email protected];
[email protected];
[email protected];

Abstract
Path planning for a swarm of drones is primarily concerned with avoiding
collision among the drones and environmental obstacles while determin-
ing the most efficient flight path to the region of interest. This paper
proposes an efficient methodology for drone swarm path planning prob-
lems in 3D environments. An improved population based meta-heuristic
algorithm, Sine Cosine Algorithm (SCA), has been proposed to solve this
problem. As part of the improvements, the population of SCA is initial-
ized using a chaotic map, and a non-linearly decreasing step size is used
to balance the local and global search. In addition, a convergence factor is
employed to increase the convergence rate of the original SCA. The per-
formance of the proposed improved SCA (iSCA) is tested over the drone
swarm path planning problem, and the results are compared with those
of the original SCA, and other state-of-the-art meta-heuristic algorithms.

1
 Springer Nature 2021 LATEX template

2 Efficient 3D path planning for drone swarm using improved SCA

The experimental results show that the drone swarm 3D path planning
problem can be efficiently handled with the proposed improved SCA.

Keywords: Path Planning, Internet of Drones (IoDs), Meta-heuristics, Sine

Cosine Algorithm (SCA), Drone Swarm, Obstacle Avoidance

1 Introduction
In recent years, the field of Unmanned Aerial Vehicles (UAVs) has grown
rapidly, including miniature aircraft, airships, and drones for a wide range of
purposes such as surveillance, military operations, telecommunications, medi-
cal supplies delivery, rescue operations, and monitoring [1–3]. A large number
of UAV systems rely on only one aerial vehicle. Nevertheless, the active coop-
eration of several UAVs is essential in many applications. They can carry out
complicated tasks beyond the ability of a single UAV while being cost effective
and more robust.
A network of drones connected to each other is known as the Internet of
Drones (IoDs) or drones swarm, a layered network control architecture that
primarily coordinates the access of UAVs, controls the airspace, and provides
navigation services between locations known as nodes [4]. Drone swarm can be
used for various applications, such as intelligent transportation systems (ITS),
to enhance vehicle-infrastructure communication. In this application, drone
swarm is an efficient way to improve traffic rules on the ground and provide
ground users with efficient information dissemination. In order to accomplish
such complex tasks, drones must collaborate due to the heterogeneity of their
goals and communication technologies. At the current scenario, drones are
becoming increasingly autonomous as technology advances, and they gain new
capabilities. However, as drones get closer to each other or obstacles in case of
high drone density or challenging missions, they pose new threats.
Obstacles can be static or dynamic. The obstacles that are static are fixed,
such as mountains and buildings, while those that are dynamic include other
drones or air vehicles, birds, etc. Furthermore, controlling drone swarm and
communicating among drones become more complicated tasks. Moreover, if
the drone swarm merge in different directions, the risk of catastrophic collisions
increases. As the likelihood of colliding among drones in a swarm increases,
collision avoidance becomes a more challenging task, and thus, drone swarm
should have a proper method to prevent or avoid collisions.
One of the most important problems for autonomous multi-UAV system
i.e. drone swarm is path planning. Considering the given flight conditions and
flight environment, a collision-free path for drone swarm needs to be planned
based on the given starting and destination points. The planned path should
be cost-effective and comply with relevant constraints. Thus the drone swarm
path planning can be viewed as an optimization problem that involves multiple
constraints [5], and aims to find a feasible minimum path from a given starting
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 3

point to a given destination point based on various optimization criteria and

mission constraints [6, 7]. These constraints include the minimum flight length,
minimum flight time, and state constraints of the drones. Recently, research
on drone swarm path planning has received much attention since it enables
unmanned systems to operate autonomously and intelligently.
Over the past few years, a variety of path planning methods have been
proposed for UAVs and autonomous robots. Graph-based algorithms such as
the Voronoi diagram algorithm [8], A* algorithm [9], Probabilistic road maps
algorithm [10, 11], rapidly-exploring random trees-based algorithm [12, 13],
are some simple path planning methods. Nevertheless, these algorithms rarely
consider UAV kinematic and dynamic constraints, so they cannot be used in
practical applications. In addition. these algorithms are based on cost maps,
which should be created and stored in advance, making it time-consuming to
produce the cost maps. Another type of effective path planning method is
the potential fields-based method. Artificial potential field algorithm [14] and
interfered fluid dynamical system algorithm [15] are two typical examples of
this kind. In order to generate the flyable path for UAVs, such algorithms
need to globally establish the interaction between the attractive and repulsive
fields. Consequently, they are easily trapped in a local minima. Furthermore,
sometimes it is impossible to guarantee a feasible path when the target and
obstacles are too close.
It has been demonstrated that drone swarm path planning problem is an
NP-hard problem, and the complexity of the problem grows with problem size
[16]. In order to solve the NP-hard problems, heuristics algorithms are effective
and easy to implement.
The key challenge in dense swarm and environmental constraints is gener-
ating a collision-free path for drone swarm [17, 18]. In addition, deterministic
approaches for building paths for drone swarm require a large amount of
storage capacity and a long execution time [19].
Hence, to solve such a problem, proper optimization methods are necessary.
Furthermore, optimization criteria may include shortest path length, avoiding
obstacles, shorter time missions, drone constraints (e.g., the amount of energy
required to complete a mission, coverage area, etc.), and so on [20].
In recent years, population-based evolutionary algorithms have benefited
greatly from advancements in swarm intelligence technology [21, 22], and they
maintain a strong ability to find the optimal solution in an efficient and flexible
manner. As a result, researchers are increasingly focusing on UAV path plan-
ning using these methods. A few of the most commonly used algorithms include
Genetic Algorithm (GA) [23], Artificial Bee Colony (ABC) algorithm [24], Ant
Colony Optimization (ACO) [25, 26], Differential Evolution (DE) [27, 28], Par-
ticle Swarm Optimization (PSO) [29], Spider Monkey Optimization (SMO)
[30] etc.
The sine-cosine algorithm is one of the newly introduced swarm intelli-
gence based algorithm, which draws significant attention from the researchers
 Springer Nature 2021 LATEX template

4 Efficient 3D path planning for drone swarm using improved SCA

because of its simplicity and ease of implementation in real-life applications.

S.Mirjalili initially proposed this algorithm to solve optimization problems [31].
Besides, a variety of meta-heuristic algorithms have been used to study the
UAV path planning problems [32–34]. However, SCA has not been proposed
for path planning for drone swarm in a 3D environment. This is due to its
shortcomings of slow convergence and falling into local optimality when solving
complex problems. Therefore, this paper proposes an improved sine cosine
algorithm (iSCA) to address this issue. The main contributions of this paper
are as follows.
• In iSCA, the chaos based initialization of the population for better unifor-
mity is used.
• It uses non-linearly decreasing step size to balance between local and global
search process of SCA.
• The convergence factor is employed for faster convergence of SCA.
• The proposed iSCA is tested over drone swarm path planning problem and
compared with other state-of-the-art algorithms.
• Applied the iSCA for tackling the 3D path planning problem for the drone
swarm.
The remainder of this paper is arranged as follows: Section 2 describes
the mathematical model for drone swarm path planning problem. The path
planning algorithm based on the proposed iSCA is presented in Section 3.
Section 4 discusses the simulation results with a detailed comparison among
the algorithms. Finally, the conclusion of this work is summarized in Section 5.

2 Mathematical model for drone swarm path

planning problem
When planning the paths for drone swarm, it is important to consider some
factors such as terrain area, the cost associated with each path, and drone’s
safety. The mission environment may contain obstacles such as mountains,
radars, buildings, or other threats. In addition, the drone swarm consists of a
large number of drones. Hence, objective functions must incorporate all these
environmental factors as well as reflect their effects on performance. Drone
swarm path planning problem is formulated as an optimization problem and
then solved using the proposed iSCA. In the following sections, environmental
constraints and objective functions are discussed.

2.1 Representations of flying space for drone swarm

In drone swarm path planning, the goal is to find an optimal and feasible path
for the drones from their starting position to their target position under com-
plex environmental constraints. Throughout this study, we refer to (x, y, z) as
the three-dimensional coordinates of waypoints of the path. The flying spaces
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 5

for drone swarm is expressed as follows [35].

S = {(x, y, z)|xlb ≤ x ≤ xub , ylb ≤ y ≤ yub , zlb ≤ z ≤ zub } (1)

Where xlb , ylb , and zlb are the lower limits of the flying space while xub , yub ,
and zub are the upper bounds.

2.2 Obstacle model

Nowadays, it is possible to obtain accurate and up-to-date terrain maps and
obstacles position using various sensing technologies such as infrared, LiDAR,
GPS, etc. In this paper, it is assumed that the spatial boundaries and the
locations of the obstacles are well known in advance. We model the obstacles
as given in [36]. If (xk1 , yk1 , zk1 ) are the coordinates of the k1th circular obstacle
in a 3D environment with radius Rk1 then the k1th obstacle can be represented
as follows [36].
Ok1 = (xk1 , yk1 , zk1 , Rk1 ) (2)
Where the coordinates (xk1 , yk1 , zk1 ) are calculated as follows:

xk1 = Rk1 cos (θ) sin (φ) + xc (3)

yk1 = Rk1 sin (θ) sin (φ) + yc (4)
zk1 = Rk1 cos (φ) + zc (5)

Where (xc , yc , zc ) are the coordinates of center of the k1th obstacle and θ ∈
[0, 2π] , φ ∈ [0, π/2].

2.3 Objective function modeling

In path planning, the objective function includes determining the length of
the path, considering environmental constraints, and avoiding collisions with
obstacles and other drones in the swarm. Our objective function aims to
minimize the overall path length while avoiding the obstacles. The objective
function can therefore be expressed as follows [35].

F = Fpl + Foc + Fmc (6)

Where Fpl is the cost associated with path length, Foc is the cost of drones
collision with obstacles and Fmc is the collision cost among drones. The goal
is to minimize the objective function F . The next subsection describes the
mathematical formulations of Fpl , Foc , and Fmc .

2.3.1 Cost associated with path length

The expected flight path of a mission is a shorter one because shorter paths
consume less fuel and are less likely to incur unforeseen threats. In order to
 Springer Nature 2021 LATEX template

6 Efficient 3D path planning for drone swarm using improved SCA

evaluate the cost associated with path length, we use the following path length
ratio (PLR) [37].
PD−1 p
j=1 (xj+1 − xj )2 + (yj+1 − yj )2 + (zj+1 − zj )2
Fpl = p . (7)
(xD − x1 )2 + (yD − y1 )2 + (zD − z1 )2

Where D denotes the total number of waypoints in the path, (xj , yj , zj )

are the coordinates of the j th waypoint, (x1 , y1 , z1 ) and (xD , yD , zD ) are the
coordinates of the start and end waypoints of the path, respectively.
Here, the denominator represents the length of the shortest path between
the start and the end waypoint while the numerator represents the length of
the flight path. So, Fpl is always ≥ 1 and a smaller value of Fpl corresponds
to a flight path with shorter length.

2.3.2 Obstacle cost

In order to fly a drone safely, the planned path must avoid all obstacles. Even
one point in the solution that passes through an obstacle may incur high
cost. If pathi is the planned path for dronei and (xj , yj , zj ), j = 1, 2, . . . , D
are the coordinates of the j th waypoint in the pathi then every waypoint
(j = 1, 2, . . . , D) of the pathi should be checked against all obstacles to see if
they fall into them. In order to do so, the distance between the waypoint’s and
the center of obstacles is taken into account. It is assumed that the waypoint
does not fall into the obstacle if the distance between the waypoint and the
center of the obstacle is greater than the radius of the obstacle. In this case,
a negligible cost is given to the objective function as the obstacle cost. In
contrast, when the distance between them is shorter than the radius of the
obstacle, then the high cost is assigned as the penalty.
Thus, the cost for the obstacle avoidance is defined as follows [35].

D X
k  
X α × dist(j, k1 )
Foc = exp − (8)
j=1 k1 =1
R k1

Where α ∈ [0, 1] is a control parameter, k is the total number of obstacles, Rk1

is the radius of k1th obstacle and dist(j, k1 ) represents the distance between j th
waypoint of the pathi and center of the k1th obstacle and is defined as follows:
p
dist(j, k1 ) = (x(j) − x0 (k1 ))2 + (y(j) − y0 (k1 ))2 + (z(j) − z0 (k1 ))2 (9)

Where (x(j), y(j), z(j)) represents the coordinates of the j th waypoint and
(x0 (k1 ), y0 (k1 ), z0 (k1 )) are the coordinates of the center of the k1th obstacle.

2.3.3 Cost of drone member collision

When planning the paths for drone swarm, collision avoidance must be con-
sidered. All drones should maintain a reasonable distance from one another.
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 7

The probability of collision among drones increases as the drone swarm den-
sity increases space. It is therefore extremely important to ensure that drones
are not too close to each other when drone swarm paths are generated. If pathi
is the planned path for dronei and patho is the planned path for any other
drone then every waypoint’s of pathi must be checked with every waypoint’s
of other paths (patho ). To do so, it is necessary to consider a safety distance
(sd) between paths.
The cost associated with collision among drones can be written as follows
[35].
D X D̃  
X α × dist(j, j1 )
Fmc = exp − (10)
j=1 j =1
sd
1

Where D and D̃ are the number of waypoints in the pathi and patho , respec-
tively. α ∈ [0, 1] is a control parameter, sd is the inter-drone distance, and
dist(j, j1 ) represents the distance between pathi and patho and is defined as
follows: q
dist(j, j1 ) = (xj − xj1 )2 + (yj − yj1 )2 + (zj − zj1 )2 (11)
Where (xj , yj , zj ) and (xj1 , yj1 , zj1 ) are the waypoints of pathi and patho ,
respectively.
In the above model, the cost of member collisions among drones is mostly
driven by the distance between pathi and other paths (patho ). As the safety
distance (sd) should be maintained, the cost will increase when the distance
between paths is ≤ sd, and it decreases rapidly as the distance between the
paths increases.

3 Path planning algorithm

3.1 Sine Cosine Algorithm (SCA)
The SCA is a new population-based meta-heuristic algorithm developed by
S. Mirjalili [31], which utilizes a set of candidate solutions for performing the
search. This is a method whereby guided randomness is created through the use
of sine and cosine trigonometric functions. In SCA, the global solution is called
a destination point and the solution vectors are called candidate solutions.
For drone swarm path planning problem, let each feasible path represent
the feasible candidate solution of the population in SCA. Assuming that the
number of waypoints for each candidate solution is D, then for the three-
dimensional path planning problem, the waypoints of the pathi (ith candidate
solution) can be expressed as follows:

Xi = (xi,1 , . . . , xi,D )T (12)

Where xi,j , j ∈ 1, . . . , D represents the j th waypoint of the ith candidate

solution and are denoted as follows:

xi,1 = (xxi,1 , xyi,1 , xzi,1 ) (13)

 Springer Nature 2021 LATEX template

8 Efficient 3D path planning for drone swarm using improved SCA

... (14)
xi,D = (xxi,D , xyi,D , xzi,D ) (15)

(xxi,j , xyi,j , xzi,j ), j ∈ 1, . . . , D represents the coordinates of j th waypoint of the

ith candidate solution in a three-dimensional space.
Thus if N denote the total number of candidate solutions then the
population (swarm) can be represented as

P = (X1 , X2 , . . . , XN )T (16)

For SCA with N number of candidate solutions, there is one destination point
(global best solution), which can be written as

Gbest = (gbest,1 , . . . , gbest,D ) (17)

Now, in tth iteration the position of each candidate solution is updated based
on the following formula [31]:
(
(t)
(t+1) Xi,j + r1 × sin(r2 )× | r3 × Gtbest − Xi,j
t
|, if r4 < 0.5
Xi,j = (t) t t
(18)
Xi,j + r1 × cos(r2 )× | r3 × Gbest − Xi,j |, Otherwise

Where r2 , r3 , and r4 are random numbers in the ranges (0, 2π), (0, 1), and
(0, 1), respectively. Here, the parameter r4 is known as the switching param-
eter, because it is used to choose the search paths using sine or cosine
function.
The parameter r1 is known as the control parameter, which decreases lin-
early from a number β to 0. It is responsible to manage the exploration and
exploitation during the search by changing its value. r1 > 1 indicates the explo-
ration of the search space, while r1 < 1 indicates exploitation. r1 is defined as
follows:  
t
r1 = β × 1 − (19)
M axiteration
Where t and M AXiteration are the current iteration number and the maximum
number of iterations, respectively .
The pseudo-code of SCA is shown in Algorithm 1.

3.2 Improved Sine Cosine Algorithm (iSCA)

Despite the fact that the original SCA has enough exploration capability to
diversify the search space, it often gets stuck in local optima and undergoes
premature convergence when tackling complex problems [38]. Drone swarm
path planning problem is a complex problem. It needs an efficient algorithm
which balances the exploration and exploitation, efficiently. Therefore, it is
essential to balance exploration and exploitation in SCA when performing the
search operations to find the optimal path.
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 9

Algorithm 1 Pseudo-code of SCA

Initialize the N number of candidate solutions randomly.
Evaluate all the solutions.
Identify the best candidate solution.
while t < M axiteration do
Generate r1 using (19).
for i = 1 : N do
for j = 1 : D do
Get the values of r2 , r3 and r4 .
Update the position of each candidate solution using (18).
end for
end for
Evaluate the updated population.
Identify the best candidate solution.
end while

In order to prevent trapping in local optima and to search more accurately

and rapidly for global optima, the present study proposes improvements in the
SCA as follows:
• Chaos based initialization of candidate solutions.
• Better position update strategy by introducing non-linearly decreasing step
size.
• Incorporation of the convergence factor in the search mechanism to speed
up the convergence rate.

3.2.1 Chaos based population initialization

The population initialization in any evolutionary algorithm plays a very impor-
tant role in the convergence speed and quality of the final solution. In general,
random initialization is the most commonly used method of generating initial
population in the absence of any information about the solution. The SCA uses
uniformly distributed random solutions to initialize the population of candi-
date solutions. According to [39], when the distribution is more uniform, the
population maintains rich diversity, which increases the chance of faster conver-
gence and better solution quality. Hence, chaos based initialization contributes
in maintaining better diversity among the potential drone swarm paths. Logis-
tic maps have the advantage of a more uniform distribution when compared
with random distribution over 10,000 times [39].
In this work, in order to enrich the diversity of the initial population, the
logistic map, which is one of the simplest and the most widely used chaotic
map, is used [40].

yj+1 = µ × yj × (1 − yj ), j = 0, 1, 2, . . . (20)
 Springer Nature 2021 LATEX template

10 Efficient 3D path planning for drone swarm using improved SCA

Where yj is the j th chaotic variable. µ is the bifurcation coefficient. A chaotic

state occurs if µ ∈ [3.57, 4]. When µ = 4, y0 ∈ (0, 1), the system produces a uni-
form chaotic signal, which will be employed for initialization of the candidate
solutions.
Steps to implement logistic map based initialization are as follows:
1. First, set y0 ∈ (0, 1) and generate D (population dimension) chaotic
variables using following Equation (21).

yj+1 = µ × yj × (1 − yj ), j = 0, 1, 2, . . . , D (21)

Where yj denotes the j th variable.

2. Repeat step 1 for i = 1, 2, . . . , N (population size), and generate the initial
chaotic variables for each candidate solution i.
3. Initialize the candidate solutions as follows:

xi,j = xmin,j +yi,j ×(xmax,j −xmin,j ), i = 1, 2, . . . , N, j = 1, 2, . . . , D (22)

Where xmax,j and xmin,j are the upper and lower bounds of the j th variable,
respectively.
4. Finally, ith candidate solution using logistic map is

Xi = (xi,1 , xi,2 , . . . , xi,D ); ∀i = 1, 2, . . . , N (23)

Pseudo-code of chaos based population initialization is presented in Algorithm

2.

Algorithm 2 Chaos based population initialization

Input: Population size (N ), population dimension (D), bifurcation coeffi-
cient (µ) & y0 = rand.
for i = 1 : N do
for j = 1 : D do
Generate the chaotic variables using (21).
end for
end for
for i = 1 : N do
for j = 1 : D do
Generate the initial candidate solution using (22).
end for
end for
Output: N initial solutions with D number of variables.
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 11

3.2.2 Improved position updating mechanism

In SCA, the control parameter r1 controls exploration in the early iteration
and exploitation in the later iteration. This transition parameter can be further
modified in order to balance the exploration and exploitation of the search
process. The control parameter r1 in SCA is a linear function that decreases
linearly from β to 0. Because of its linearity, sometimes it creates abrupt
changes in jumping from one iteration to the next. In some cases, abrupt
changes may result in the skipping of good solutions; thus, valuable information
about the quality of search areas might be lost. In this study, a modified
formula for r1 so that it decreases exponentially from γ to 0 is proposed to
avoid all these issues.

t2
 
r1new = γ × exp − (24)
(β × M AXiteration )2
Where t and M AXiteration are current and maximum number of iterations,
respectively. γ is a user-defined parameter.

3.2.3 Convergence factor

Further, a convergence factor CF is employed in the search mechanism.
This convergence factor CF helps iSCA to converge rapidly while balancing
exploration and exploitation. CF is defined as follows:
 
t
CF = β × 1 − (25)
M AXiteration

In Equation (25), the convergence factor (CF) is inversely proportional

to the number of iterations. Its small value corresponds to less dependency
over the current position, and its higher value plays more role of the cur-
rent position in deciding the new position. Initially, when CF is large, the
search process is significantly guided by the current position, and in the later
iteration, when CF is small, it plays less role, and the new position is more
depends upon the global best solution.

As mentioned above, the non-linearly decreasing step size (Equation 24)

helps in balancing exploration and exploitation of the search process very well,
while the convergence factor (Equation 25) helps in fast convergence. Thus
the following proposed search Equation (26) is used in iSCA which merges
both the techniques, cancels the absolute value term, to obtain better perfor-
mance in drone swarm path planning in terms of solution quality, accuracy,
and convergence speed.
(
(t)

(t+1) CF × Xi,j + r1new × sin(r2 ) × r3 × Gtbest − Xi,j
t
, if r4 < 0.5
Xi,j = (t) t t


CF × Xi,j + r1new × cos(r2 ) × r3 × Gbest − Xi,j , Otherwise
(26)
 Springer Nature 2021 LATEX template

12 Efficient 3D path planning for drone swarm using improved SCA

Symbols have their usual meaning. The pseudo-code of improved SCA (iSCA)
for drone swarm path planning is shown in Algorithm 3.

Algorithm 3 The improved SCA (iSCA) for drone swarm path planning
Environment construction:
Set the boundary for drones flying area.
Set obtacles positions.
Set the starting and destination position for each drone.
for d = 1 : total number of drones do
Initialize the N number of candidate solutions using chaos based
initialization (Equation (20)).
Evaluate all the solutions.
Identify the best candidate solution.
Main loop:
while t < M axiteration do
Calculate r1new using (24) .
Calculate CF using (25).
for i = 1 : N do
for j = 1 : D do
Generate r2 , r3 and r4 .
Update the position of each candidate solution using (26).
end for
end for
Evaluate the updated population.
Identify the best candidate solution.
end while
Output: The optimal path for drone d.
end for

4 Simulation results & discussions

In this section, simulation results and comparisons are presented in order to
show the performance of the proposed iSCA over drone swarm path planning
problem. The experiments are carried out in a MATLAB environment on a
server with a 3.70 GHz CPU, 64 GB of RAM, and a 64-bit operating system.

4.1 Parameter settings

Parameter settings play an important role in the performance of an algorithm
as appropriate parameters may lead to better results of the algorithm. In the
simulation environment, five drones are assumed to fly, simultaneously from
their starting position to their destination position. Table 1 shows the posi-
tions of the starting and destination positions of each drone in a 3D space
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 13

of size 16000 × 16000 × 16000. Four static obstacles are placed in the search
space, whose positions are listed in Table 2. The location of each obstacle and
the current positions of drones with their destinations (goals) are presented in
2D and 3D views in Figures 1 and 2, respectively. The parameters correspond-
ing to all the considered algorithms are presented in Table 3. The parameters
in Table 4 are common to all the algorithms. In Table 4, sd stands for safety
distance for collision avoidance among the drones, D stands for the total num-
ber of waypoints, N represents the population size, and µ is the bifurcation
coefficient.

Table 1 Drones starting and destination positions

Drones Starting position Destination position
Drone1 (200,800,350) (16000,4000,350)
Drone2 (200,3800,350) (16000,7000,350)
Drone3 (200,6800,350) (16000,10000,350)
Drone4 (200,9800,350) (16000,13000,350)
Drone5 (200,12800,350) (16000,16000,350)

Table 2 Obstacles position

Obstacles Positions Radius
Obstacle1 (5000,10000,0) 1800
Obstacle2 (10000,2000,0) 1200
Obstacle3 (10000,8000,0) 1100
Obstacle4 (5000,2000,0) 1500

Table 3 Parameters setting for all algorithms

Algorithms Parameters
SCA same as in [31]
RCN β = 2, γ = 1
CL β = 2, µ = 4
iSCA β = 2, γ = 1
PSO wmax = 0.8, wmin = 0.4, c1 = 1.47, c2 = 1.47
IPSO wmax = 0.9, wmin = 0.4, c1 = 1.47, c2 = 1.47, ϵmax = 0.9, ϵmin = 0.05, Vmax = 0.3
ABC same as in [24]

4.2 Results and comparisons

In this section, to analyze the performance of the proposed iSCA, some per-
formance indicators such as solution quality, convergence speed, failure and
success rate, and drone swarm formation running time are considered. The
results obtained from the considered algorithms are recorded and are analyzed
in the respective subsections.
 Springer Nature 2021 LATEX template

14 Efficient 3D path planning for drone swarm using improved SCA

Fig. 1 Two-dimensional view of the obstacles, starting and destination points of the drone
swarm.

Fig. 2 Three-dimensional view of the obstacles, starting and destination points of the drone
swarm.

4.2.1 Comparison with original SCA

In this subsection, the performance of the proposed iSCA is compared with
the original SCA. Since iSCA is proposed with two important inclusions, that
is, convergence factor and non-linearly decreasing step-size, so to examine the
significance of each modifications, we additionally considered both factors,
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 15

Table 4 Common parameters

Population size (N) = 300
Maximum iteration = 150
Safety distance (sd) = 80m
Number of way points (D) = 20
µ=4

Fig. 3 Framework of RCN

Fig. 4 Framework of CL

independently for the comparison. Following two variants of SCA are consid-
ered using each modification independently for the comparison.

• RCN: This algorithm is formed by including uniformly distributed pop-

ulation initialization along with non-linearly decreasing step size and
convergence factor in the original SCA as shown in Figure 3.

• CL: This algorithm initializes the population using a chaos map and linearly
decreasing step size in SCA as shown in Figure 4.

The performances over drone swarm path planning problem of SCA, iSCA,
RCN, and CL are discussed and comparative study has been carried out
based on the performance indicators given below:

Comparison based on solution quality: The solution quality can only

be measured through objective function value (fitness value (FV)). Firstly, we
have recorded the fitness values of all the considered algorithms over all the
iterations in a single run. Figures 5-8 show the graphical representation of fit-
ness values over iterations of all the algorithms considered in this subsection.
From these results, it is clear that the FV of the proposed iSCA outperforms
SCA, RCN, and CL. This comparison justifies that both the modifications
along with chaos based initialization is necessary to achieve this superior per-
formance of iSCA. In other words, all three modifications in SCA are jointly
responsible for better performance of iSCA.
Since randomness is present in all the considered algorithms so it is not
enough to take a decision through a single run. Thus in order to do fair com-
parison among algorithms, we use the Monte-Carlo simulations with 40 runs
to each algorithms and analysed the results. Figures 9-12 show the planned
paths for drones by SCA, iSCA, RCN, and CL, respectively in 2D views.
 Springer Nature 2021 LATEX template

16 Efficient 3D path planning for drone swarm using improved SCA

SCA iSCA
1.7 2
DRONE1
DRONE2 DRONE1
1.6
DRONE3 DRONE2
DRONE4
1.8
DRONE3
1.5 DRONE5 DRONE4
Fitness value

Fitness value
DRONE5
1.6
1.4

1.3
1.4

1.2
1.2
1.1

1 1
0 50 100 150 0 50 100 150
Iteration Iteration

Fig. 5 Fitness value over iteration for Fig. 6 Fitness value over iteration for
drone swarm path formation by SCA drone swarm path formation by iSCA

Table 5 Average fitness value of the formation

Algorithm Drone1 Drone2 Drone3 Drone4 Drone5 FAFV Improved %
SCA 1.150535239 1.045553413 1.111839756 1.141272442 1.063580563 1.102556283 NA
iSCA 1.125486909 1.038086312 1.097011195 1.111504654 1.026271119 1.079672038 2.076
RCN 1.133059711 1.037975431 1.103182588 1.119762864 1.028719366 1.084539992 1.634
CL 1.150075141 1.048581918 1.118792639 1.139484139 1.061501466 1.103687061 -0.103

While Figures 13-16 represent the 3D views of the planned path for drones for
considered algorithms. From Figures 9-12 and 13-16, it can be seen that the
planned path for each drone by each algorithm is obtained without collision
with obstacles and among drones. Thus it is guaranteed that all the algorithms
SCA, iSCA, RCN, and CL can generate a feasible path for each drone. Table
5 shows the average fitness value (AFV) of the formation during Monte-Carlo
simulations. The best results are highlighted with bold face. Figures 17-20
show the iteration-wise average fitness value (AFV). It can be observed from
Table 5 that from the single drone perspective, iSCA has outperformed all
the algorithms SCA, RCN, and CL for drones 1, 3, 4, and 5 but in the case of
drone 2, iSCA no longer outperforms RCN. However, it is reasonable that the
AFV of the formation is more important and fair than the individual drone’s
performance. Because the AFV of the formation indicates the overall solution
quality of an algorithm, as well as the safety and cost-effectiveness of drone
operations in the flying environment. Moreover, AFV for a single drone can
easily be affected by different environmental constraints, so it is not good
enough to evaluate the performance of an algorithm from a single drone’s
perspective in case of drone swarm optimal path planning. The AFV of the
formation in Table 5 shows that iSCA has the least AFV, RCN performs
second, and CL worst. As compared to original SCA, the AFV of iSCA, RCN,
and CL has decreased to 2.076%, 1.634%, and -0.103%, respectively. Note
that CL is not able to outperform SCA while RCN outperforms SCA inde-
pendently, but RCN and CL which forms iSCA outperforms SCA as shown in
Table 5. Overall, iSCA has a better solution quality than SCA, RCN, and CL.

Comparison based on convergence speed: The convergence speed

is an important performance indicator in analyzing the effectiveness of the
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 17

RCN CL
1.7

1.8 DRONE1
1.6
DRONE2
1.7 DRONE3
DRONE1
DRONE2
1.5 DRONE4
1.6
Fitness value

Fitness value
DRONE5
DRONE3
DRONE4 1.4
1.5 DRONE5

1.4 1.3
1.3
1.2
1.2
1.1
1.1

1 1
0 50 100 150 0 50 100 150
Iteration Iteration

Fig. 7 Fitness value over iteration for Fig. 8 Fitness value over iteration for
drone swarm path formation by RCN drone swarm path formation by CL

Fig. 9 2D view of the planned path for Fig. 10 2D view of the planned path for
drone swarm by SCA drone swarm by iSCA

algorithm. Figures 5-8 and Figures 17-20 are the convergence curves of SCA,
iSCA, RCN, and CL in single and multiple runs, respectively. To show the
evolution process of the algorithms over iterations, we assume that the x−axis
represents the iteration number and y−axis represents the average fitness
value in these Figures. In addition, we also adopt the minimum number of
iterations that requires to reach the optimal solution as another indicator
to evaluate the convergence speed. In this process, firstly, we set the crite-
ria for the feasible solution as |AF Vt−20 − AF Vt | < 0.001. In other words,
|AF Vt−20 − AF Vt | < 0.001 represents the difference in AFV obtained in 20
consecutive iterations. Here, t represents the current iteration number and
AF V represents the average fitness value that is obtained in 40 runs. Table 6
represents the average minimum iterations (AMIs) of the formation for each
of the considered algorithms. It can be seen from Table 6 that SCA has the
least AMI with a value of 34.2, while iSCA has the largest AMI with a value
of 62.8. On the other hand, CL and RCN are 2nd and 3rd in the list with val-
ues 38 and 38.6, respectively. Though SCA obtained the highest convergence
speed than iSCA, RCN, and CL, but it can be observed that the solution
obtained by SCA with this fast convergence speed is sub-optimal and inferior
 Springer Nature 2021 LATEX template

18 Efficient 3D path planning for drone swarm using improved SCA

Fig. 11 2D view of the planned path for Fig. 12 2D view of the planned path for
drone swarm by RCN drone swarm by CL

Fig. 13 3D view of the planned path for Fig. 14 3D view of the planned path for
drone swarm by SCA drone swarm by iSCA

Table 6 Average minimum iterations of the formation

Algorithm DRONE1 DRONE2 DRONE3 DRONE4 DRONE5 FAMI
SCA 25 38 37 31 40 34.2
iSCA 38 78 111 37 50 62.8
RCN 22 28 61 41 41 38.6
CL 29 45 57 38 21 38

than iSCA, RCN, and CL. This means that SCA can easily get trapped to
local optima. Since the non-linearly decreasing step-size along with conver-
gence factor are present in iSCA, these factors help iSCA to avoid stagnation
and provide the searching ability for more iterations to obtain global optima.

Comparison based on success and failure rates: Besides solution

quality and convergence speed, success and failure rate is also another impor-
tant factor in analysing the performance of the considered algorithms. This
indicator measures the reliability of the algorithm. In this work, Monte-Carlo
simulations are carried out with 40 runs, and failure number (FN) is defined as
the number of times the final fitness value (F Vt=150 ) is greater than or equal
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 19

Fig. 15 3D view of the planned path for Fig. 16 3D view of the planned path for
drone swarm by RCN drone swarm by CL

SCA iSCA
1.18 1.14

1.16
Average fitness value 1.12
Average fitness value

1.14
1.1
1.12
DRONE1 DRONE1
1.1 1.08
DRONE2 DRONE2
DRONE3 DRONE3
1.08 DRONE4 DRONE4
DRONE5 1.06 DRONE5
1.06
1.04
1.04

1.02 1.02
0 50 100 150 0 50 100 150
Iteration Iteration

Fig. 17 Average fitness value over iter- Fig. 18 Average fitness value over iter-
ation on 40 runs for drone swarm path ation on 40 runs for drone swarm path
formation by SCA formation by iSCA

RCN CL
1.16 1.16

1.14 1.14
Average fitness value

Average fitness value

1.12
1.12
1.1 DRONE1
1.1 DRONE2
DRONE1
1.08 DRONE2
DRONE3
DRONE3 DRONE4
1.08 DRONE5
1.06 DRONE4
DRONE5

1.04 1.06

1.02 1.04
0 50 100 150 0 50 100 150
Iteration Iteration

Fig. 19 Average fitness value over iter- Fig. 20 Average fitness value over iter-
ation on 40 runs for drone swarm path ation on 40 runs for drone swarm path
formation by RCN formation by CL

to 1.13. This means that the path generated by the algorithm is too long,
and/or the drone collides with an obstacle, and/or the collision happens among
drones. On the other hand, a zero failure rate indicates that F Vt=150 < 1.13.
The average failure number (AFN) of the formation and the total path fail-
ure rate (FR) during the Monte-Carlo simulations on 40 runs are presented
in Table 7. Figure 34 is the graphical representation of the formation failure
 Springer Nature 2021 LATEX template

20 Efficient 3D path planning for drone swarm using improved SCA

Table 7 Failure rate of the formation

Algorithm DRONE1 DRONE2 DRONE3 DRONE4 DRONE5 AFN FR % improvement
SCA 40 0 4 25 0 13.8 34% NA
iSCA 9 0 1 1 0 2.2 5.5% 84.057
RCN 16 0 3 5 0 4.8 12% 65.217
CL 40 0 9 27 0 15.2 38% -10.144

Table 8 Formation running time by different algorithms

Algorithm Formation running time (seconds) % improvement

SCA 1.482247 NA
iSCA 1.163345 21.5147
RCN 1.163345 21.5147
CL 1.467294 1.0088

number (FN) of considered algorithms. From Table 7, one can see that the pro-
posed iSCA has the least failure rate with an improved percentage of 84.057
as compared to the original SCA, while 65.217% improvement is obtained by
RCN. On the other hand, there is no improvement in CL. This shows that
the proposed iSCA can achieve the highest success rate or reliability in the
path generation and guarantees a significant improvement in solution quality
as compared to SCA, RCN, and CL.

Comparison based on running time: Running time is also an essential

factor in evaluating the performance of an algorithm as it plays a significant
role in path planning and collision avoidance among a drone swarm. In this
work, the running time is the time required to complete the task from starting
position to destination position of the drone swarm. The results are presented
in Table 8 and Figure 35. From Table 8, one can see that the increased percent-
age of the iSCA, RCN, and CL as compared to the original SCA are 21.5147,
21.5147, and 1.0088, respectively. These results demonstrate the ability of the
proposed iSCA to generate the optimal paths for drone swarm in a faster way.

4.2.2 Comparison with other algorithms

In this subsection, we have compared the proposed iSCA with other meta-
heuristics such as PSO [29], ABC [24], and one of the recently proposed
improved PSO (IPSO) [35], which increases the solution quality and conver-
gence speed of basic PSO. The population size and the maximum number of
iterations are set to 300 and 150, respectively. Table 3 shows the parameter
settings for the considered algorithms. For a fair comparison among the algo-
rithms, we have performed the Monte-Carlo simulations with 40 runs. The
detailed comparison among the algorithms based on the performance indica-
tors: solution quality, convergence speed, success and failure rate, and drone
swarm formation running time are described as follows.
Figures 21-23 and Figures 24-26 are the 2D and 3D views of the planned
paths generated by PSO, IPSO, and ABC, respectively. The planned paths for
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 21

Fig. 21 2D view of the planned path for Fig. 22 2D view of the planned path for
drone swarm by PSO drone swarm by IPSO

Fig. 23 2D view of the planned path for Fig. 24 3D view of the planned path for
drone swarm by ABC drone swarm by PSO

Table 9 Average fitness value of the formation

Algorithm Drone1 Drone2 Drone3 Drone4 Drone 5 FAFV
iSCA 1.125486909 1.038086312 1.097011195 1.111504654 1.026271119 1.079672038
PSO 1.274060312 1.031918232 1.150279379 1.137454152 1.039364774 1.12661537
% improved in iSCA 11.6613 -0.5977 4.6309 2.2813 1.2598 4.1667
IPSO 1.167843329 1.030065184 1.152098958 1.116741579 1.039040203 1.10115785
%improved in iSCA 3.6274 -0.7786 4.7815 0.4689 1.229 1.9512
ABC 1.298236247 1.163505086 1.15079026 1.466059323 1.103041915 1.236326566
%improved in iSCA 13.3064 10.7794 3.5704 35.4554 6.9599 12.671

the proposed iSCA are shown in Figure 10 and Figure 14 in 2D and 3D views,
respectively. One can see from these figures that iSCA and IPSO can generate
a feasible path for each drone, while PSO and ABC are not able to generate a
feasible path. Moreover, ABC can not even maintain the safety distance among
the drones as the collision happened among themselves also. This is because of
the extensive exploration capability of ABC and unable to maintain the proper
balance between exploration and exploitation. While due to the non-linearly
decreasing step size and convergence factor in iSCA, it can easily generate a
feasible path for the drones while maintaining the safety distance among the
 Springer Nature 2021 LATEX template

22 Efficient 3D path planning for drone swarm using improved SCA

Fig. 25 3D view of the planned path for Fig. 26 3D view of the planned path for
drone swarm by IPSO drone swarm by ABC

Table 10 Failure rate of the formation

Algorithm DRONE1 DRONE2 DRONE3 DRONE4 DRONE5 AFN FR % improved in iSCA
iSCA 9 0 1 1 0 2.2 5.5% NA
PSO 40 0 35 21 0 19.2 48% 60.416
IPSO 40 0 40 0 0 16 40% 86.25
ABC 40 27 25 40 8 30 55% 92.142

drones. This comparison exhibits the advantages of the proposed method as

compared to other considered algorithms.
In Tables 9 and 10, the average fitness value (AFV) of the formation and
path failure rate (FR) of the considered algorithms are given, respectively. It is
clear from Table 9 that iSCA obtains the least AFV with a reduced percentage
of 4.1667, 1.9512, and 12.671 than PSO, IPSO, and ABC, respectively. This
proves the ability of the proposed method to obtain the best solution than
PSO, IPSO, and ABC. When the reliability of the algorithms is taken into
consideration, Table 10 shows that iSCA has the highest reliability than that
of PSO, IPSO, and ABC. In Table 10, one can see that the proposed iSCA
has the least failure rate while ABC has the highest failure rate. The improved
percentage in iSCA as compared to PSO, IPSO, and ABC are 60.416, 86.250,
and 92.142, respectively. All these results demonstrate the effectiveness of the
proposed iSCA both in terms of reliability as well as solution quality.
When the convergence speed is taken into account among the considered
algorithms, Table 11 shows the average minimum iteration (AMI) of the for-
mation that requires to satisfy the feasible criterion. The definition for feasible
criterion is the same as defined above. Figures 27-29 represent the FV of PSO,
IPSO, and ABC respectively, while Figures 30-32 represent the AFV of PSO,
IPSO, and ABC respectively. The AMI of the formation is graphically pre-
sented in Figure 33. As it can be seen from Table 11, ABC has the least AMI,
while IPSO is 2nd in the list. This shows that ABC has the highest conver-
gence speed as compared to other algorithms. On the other hand, ABC has the
largest function value. Therefore, though PSO, IPSO, and ABC have smaller
AMI than iSCA, their function values are no longer outperformed iSCA. This
is a clear indication of the fact that these algorithms suffers from premature
convergence while iSCA continues to improve through iterations.
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 23

Table 11 Average minimum iterations of the formation

Algorithm DRONE1 DRONE2 DRONE3 DRONE4 DRONE5 FAMI
iSCA 38 78 111 37 50 62.8
PSO 48 77 30 44 68 53.4
IPSO 45 42 52 57 32 45.6
ABC 21 21 21 21 21 21

PSO IPSO
2 2.2
DRONE1 DRONE1
DRONE2 DRONE2
DRONE3 2 DRONE3
1.8 DRONE4 DRONE4
DRONE5
DRONE5
1.8
Fitness value

Fitness value
1.6
1.6
1.4
1.4

1.2
1.2

1 1
0 50 100 150 0 50 100 150
Iteration Iteration

Fig. 27 Fitness value over iteration for Fig. 28 Fitness value over iteration for
drone swarm path formation by PSO drone swarm path formation by IPSO

PSO
ABC 1.3
1.8

1.7 1.25 DRONE1

Average fitness value

DRONE2
DRONE1
DRONE3
1.6 DRONE2
1.2 DRONE4
DRONE3
Fintess value

DRONE5
1.5 DRONE4
DRONE5
1.15
1.4

1.3 1.1

1.2
1.05
1.1
1
1 0 50 100 150
0 50 100 150 Iteration
Iteration
Fig. 30 Average fitness value over iter-
Fig. 29 Fitness value over iteration for
ation on 40 runs for drone swarm path
drone swarm path formation by ABC
formation by PSO

The formation running time for the considered algorithms is presented in

Table 12 and Figure 36. It can be easily seen from Table 12 that the increased
percentage of iSCA as compared to PSO, IPSO, and ABC are 1.8762, 20.2827,
and 20.2827, respectively. Thus the proposed iSCA can generate feasible paths
for the drone swarm more accurately and faster, as compared to PSO, IPSO,
and ABC.

5 Conclusion and future direction

Drone swarm path planning problem in a 3D environment has been dealt
with using an improved variant of the Sine Cosine Algorithm. A case study
with 4 obstacles, 5 drones, and 16000 × 16000 × 16000 size flying space has
 Springer Nature 2021 LATEX template

24 Efficient 3D path planning for drone swarm using improved SCA

IPSO ABC
1.2 1.5

1.18 1.45 DRONE1

DRONE2

Average fitness value

Average fitness value

1.16
1.4 DRONE3
DRONE4
1.14
1.35 DRONE5

1.12
1.3
1.1 DRONE1
DRONE2 1.25
1.08 DRONE3
DRONE4
1.2
1.06 DRONE5

1.04 1.15

1.02 1.1
0 50 100 150 0 50 100 150
Iteration Iteration

Fig. 31 Average fitness value over iter- Fig. 32 Average fitness value over iter-
ation on 40 runs for drone swarm path ation on 40 runs for drone swarm path
formation by IPSO formation by ABC

Table 12 Formation running time of different algorithms

Algorithm Formation running time(seconds) % improved in iSCA

iSCA 1.163345 NA
PSO 1.18559 1.8762
IPSO 1.45934 20.2827
ABC 1.45934 20.2827

70 140
Formation average minimum iteration

60 120
Formation failure number

50 100

40 80

30 60

20 40

10 20

0 0
SCA RCN CL ABC PSO IPSO ISCA SCA RCN CL ABC PSO IPSO ISCA
Different algorithms Different algorithms

Fig. 33 Formation average minimum Fig. 34 Formation failure number by

iterations by different algorithms different algorithms

been used to check the performance of the proposed improved SCA (iSCA).
The results are compared with the original SCA and other state-of-the-art
meta-heuristic algorithms. The comparison results show that the proposed
iSCA can generate the optimal paths for the drones more accurately with high
convergence speed as compared to other considered algorithms. The same
study with the dynamic obstacle environment is the future research agenda.

Acknowledgements: Author Jagdish Chand Bansal Acknowledges the

funding from Liverpool Hope University UK.

Author Contributions: Probhat Pachung: Conceptualization, Method-

ology, Writing original draft. Kopal Pandya, Atulya Nagar: Review, Jagdish
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 25

1.5 1.5
Formation running time

Formation running time

1 1

0.5 0.5

0 0
SCA RCN CL ISCA ABC PSO IPSO ISCA
Different algorithms Different algorithms

Fig. 35 Formation running time by dif- Fig. 36 Formation running time by dif-
ferent algorithms ferent algorithms

Chand Bansal: Supervision, Review.

Funding: Not Applicable

Code or Data Availability: Not Applicable.

Declarations
Conflicts of Interest: The authors declare that they have no known com-
peting financial interests or personal relationships that could have appeared
to influence the work reported in this paper.

Ethics Approval: This article does not contain any studies with human par-
ticipants or animals performed by any of the authors.

Informed Consent: Informed consent was obtained from all individual

participants included in the study.

References
[1] Mozaffari, M., Saad, W., Bennis, M. and Debbah, M., 2015, December.
Drone small cells in the clouds: Design, deployment and performance anal-
ysis. In 2015 IEEE global communications conference (GLOBECOM) (pp.
1-6). IEEE.

[2] Valavanis, K.P. and Vachtsevanos, G.J. eds., 2015. Handbook of

unmanned aerial vehicles (Vol. 1). Dordrecht: Springer Netherlands.

[3] Al-Hourani, A., Kandeepan, S. and Lardner, S., 2014. Optimal LAP alti-
tude for maximum coverage. IEEE Wireless Communications Letters,
3(6), pp.569-572.
 Springer Nature 2021 LATEX template

26 Efficient 3D path planning for drone swarm using improved SCA

[4] Gharibi, M., Boutaba, R. and Waslander, S.L., 2016. Internet of drones.
IEEE Access, 4, pp.1148-1162.

[5] Huo, L., Zhu, J., Wu, G. and Li, Z., 2020. A novel simulated annealing
based strategy for balanced UAV task assignment and path planning.
Sensors, 20(17), p.4769.

[6] Ma, Y., Hu, M. and Yan, X., 2018. Multi-objective path planning for
unmanned surface vehicle with currents effects. ISA transactions, 75,
pp.137-156.

[7] YongBo, C., YueSong, M., JianQiao, Y., XiaoLong, S. and Nuo, X., 2017.
Three-dimensional unmanned aerial vehicle path planning using modified
wolf pack search algorithm. Neurocomputing, 266, pp.445-457.

[8] Pehlivanoglu, Y.V., 2012. A new vibrational genetic algorithm enhanced

with a Voronoi diagram for path planning of autonomous UAV. Aerospace
Science and Technology, 16(1), pp.47-55.

[9] Bayili, S. and Polat, F., 2011. Limited-Damage A*: A path search algo-
rithm that considers damage as a feasibility criterion. Knowledge-Based
Systems, 24(4), pp.501-512.

[10] Baumann, M., Leonard, S., Croft, E.A. and Little, J.J., 2010. Path
planning for improved visibility using a probabilistic road map. IEEE
Transactions on Robotics, 26(1), pp.195-200.

[11] Brubaker, M.A., Geiger, A. and Urtasun, R., 2015. Map-based proba-
bilistic visual self-localization. IEEE transactions on pattern analysis and
machine intelligence, 38(4), pp.652-665.

[12] Kothari, M. and Postlethwaite, I., 2013. A probabilistically robust path

planning algorithm for UAVs using rapidly-exploring random trees.
Journal of Intelligent & Robotic Systems, 71(2), pp.231-253.

[13] Moon, C.B. and Chung, W., 2014. Kinodynamic planner dual-tree RRT
(DT-RRT) for two-wheeled mobile robots using the rapidly explor-
ing random tree. IEEE Transactions on industrial electronics, 62(2),
pp.1080-1090.

[14] Chen, Y., Yu, J., Su, X. and Luo, G., 2015. Path planning for multi-UAV
formation. Journal of Intelligent & Robotic Systems, 77(1), pp.229-246.

[15] Chen, Y.B., Luo, G.C., Mei, Y.S., Yu, J.Q. and Su, X.L., 2016. UAV path
planning using artificial potential field method updated by optimal control
theory. International Journal of Systems Science, 47(6), pp.1407-1420.
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 27

[16] Zhang, X. and Duan, H., 2015. An improved constrained differential evolu-
tion algorithm for unmanned aerial vehicle global route planning. Applied
Soft Computing, 26, pp.270-284.

[17] Besada-Portas, E., de la Torre, L., Jesus, M. and de Andrés-Toro, B., 2010.
Evolutionary trajectory planner for multiple UAVs in realistic scenarios.
IEEE Transactions on Robotics, 26(4), pp.619-634.

[18] Zheng, C., Li, L., Xu, F., Sun, F. and Ding, M., 2005. Evolutionary route
planner for unmanned air vehicles. IEEE Transactions on robotics, 21(4),
pp.609-620.

[19] Ma, Y., Zamirian, M., Yang, Y., Xu, Y. and Zhang, J., 2013. Path
planning for mobile objects in four-dimension based on particle swarm
optimization method with penalty function. Mathematical Problems in
Engineering, 2013.

[20] Ma, Y., Hu, M. and Yan, X., 2018. Multi-objective path planning for
unmanned surface vehicle with currents effects. ISA transactions, 75,
pp.137-156.

[21] Ma, H., Shen, S., Yu, M., Yang, Z., Fei, M. and Zhou, H., 2019.
Multi-population techniques in nature inspired optimization algorithms:
A comprehensive survey. Swarm and evolutionary computation, 44,
pp.365-387.

[22] Zhao, Y., Zheng, Z. and Liu, Y., 2018. Survey on computational-
intelligence-based UAV path planning. Knowledge-Based Systems, 158,
pp.54-64.

[23] Roberge, V., Tarbouchi, M. and Labonté, G., 2012. Comparison of parallel
genetic algorithm and particle swarm optimization for real-time UAV path
planning. IEEE Transactions on industrial informatics, 9(1), pp.132-141.

[24] Karaboga, Dervis, and Bahriye Basturk. ”A powerful and efficient algo-
rithm for numerical function optimization: artificial bee colony (ABC)
algorithm.” Journal of global optimization 39.3 (2007): 459-471.

[25] Dorigo, M., Birattari, M. and Stutzle, T., 2006. Ant colony optimization.
IEEE computational intelligence magazine, 1(4), pp.28-39.

[26] Konatowski, S. and Pawlowski, P., 2018, February. Ant colony opti-
mization algorithm for UAV path planning. In 2018 14th International
Conference on Advanced Trends in Radio electronics, Telecommunications
and Computer Engineering (TCSET) (pp. 177-182). IEEE.
 Springer Nature 2021 LATEX template

28 Efficient 3D path planning for drone swarm using improved SCA

[27] Price, K.V., 1996, June. Differential evolution: a fast and simple numerical
optimizer. In Proceedings of North American fuzzy information processing
(pp. 524-527). IEEE.

[28] Pan, J.S., Liu, N. and Chu, S.C., 2020. A hybrid differential evolution
algorithm and its application in unmanned combat aerial vehicle path
planning. IEEE Access, 8, pp.17691-17712.

[29] Kennedy, James, and Russell Eberhart. ”Particle swarm optimization.”

Proceedings of ICNN’95-international conference on neural networks. Vol.
4. IEEE, 1995.

[30] Bansal, Jagdish Chand, et al. ”Spider monkey optimization algorithm for
numerical optimization.” Memetic computing 6.1 (2014): 31-47.

[31] Mirjalili, S., 2016. SCA: a sine cosine algorithm for solving optimization
problems. Knowledge-based systems, 96, pp.120-133.

[32] Duan, H. and Qiao, P., 2014. Pigeon-inspired optimization: a new swarm
intelligence optimizer for air robot path planning. International journal
of intelligent computing and cybernetics.

[33] Wang, G., Guo, L., Duan, H., Liu, L. and Wang, H., 2012. A modified
firefly algorithm for UCAV path planning. International Journal of Hybrid
Information Technology, 5(3), pp.123-144.

[34] Zhu, W. and Duan, H., 2014. Chaotic predator–prey biogeography-based

optimization approach for UCAV path planning. Aerospace science and
technology, 32(1), pp.153-161.

[35] Ahmed, G., Sheltami, T., Mahmoud, A. and Yasar, A., 2020. IoD
swarms collision avoidance via improved particle swarm optimization.
Transportation Research Part A: Policy and Practice, 142, pp.260-278.

[36] Weisstein, E.W., 2002. Hemisphere. https://mathworld.wolfram.com/.

[37] Yang, P., Tang, K., Lozano, J.A. and Cao, X., 2015. Path planning for
single unmanned aerial vehicle by separately evolving waypoints. IEEE
Transactions on Robotics, 31(5), pp.1130-1146.

[38] Gupta, S., 2022. Enhanced sine cosine algorithm with crossover: A com-
parative study and empirical analysis. Expert Systems with Applications,
198, p.116856.

[39] Shao, S., Peng, Y., He, C. and Du, Y., 2020. Efficient path plan-
ning for UAV formation via comprehensively improved particle swarm
optimization. ISA transactions, 97, pp.415-430.
 Springer Nature 2021 LATEX template

Efficient 3D path planning for drone swarm using improved SCA 29

[40] Tian, D. and Shi, Z., 2018. MPSO: Modified particle swarm optimization
and its applications. Swarm and evolutionary computation, 41, pp.49-68.