Neural Computing and Applications (2022) 34:10133–10155

https://doi.org/10.1007/s00521-022-06998-9 (0123456789().,-volV)(0123456789().
,- volV)

ORIGINAL ARTICLE

Optimal path planning for drones based on swarm intelligence

algorithm
Rashid A. Saeed1 • Mohamed Omri2 • S. Abdel-Khalek3,4 • Elmustafa Sayed Ali5,6 • Maged Faihan Alotaibi7

Received: 12 June 2021 / Accepted: 30 January 2022 / Published online: 29 April 2022
Ó The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2022

Abstract
Recently, Drones and UAV research were becoming one of the interest topics for academia and industry, where it has been
extensively addressed in the literature back the few years. Path planning of drones in an area with complex terrain or
unknown environment and restricted by some obstacles is one of the most problems facing the operation of drones. The
problem of path planning is not only limited to searching for an appropriate path from the starting point to the destination
but also related to how to choose an ideal path among all available paths and provide a mechanism for collision avoidance.
By considering how to construct the best path, several related issues need to be taken into account, that relate to safety,
obstacle avoidance, response speed to overtake obstacles, etc. Swarm optimization algorithms have been used to provide
intelligent modeling for drone path planning and enable to build the best path for each drone. This is done according to the
planning and coordination dimensions among the swarm members. In this paper, we have discussed the features and
characteristics of different swarm optimization algorithms such as ant colony optimization (ACO), fruit fly optimization
algorithm (FOA), artificial bee colony (ABC), and particle swarm optimization (PSO). In addition, the paper provides a
comprehensive summary related to the most important studies on drone path planning algorithms. We focused on analyzing
the impact of the swarm algorithm and its performance in drone path planning. For that, the paper presented one of the
most used algorithms and its models employed to improve the trajectory of drones that rely on swarm intelligence and its
impact on the optimal path cost of drones. The results of performance analysis for the ACO algorithm in a 3D and 2D-
dimensional environment are illustrated and discussed, and then the performance evaluation of the ACO is compared to the
enhanced ACO algorithm. The proposed algorithm achieves fast convergence, accelerating the process of path planning.

Keywords UAV  Swarm intelligence  Ant colony optimization  Path planning  Drones  Artificial bee colony 
Particle swarm optimization

5
Department of Electrical and Electronic Engineering, Red
& S. Abdel-Khalek Sea University, Port Sudan, Sudan
[email protected] 6
Department of Electronics Engineering, College of
1
Department of Computer Engineering, College of Computers Engineering, Sudan University of Science and Technology
and Information Technology, Taif University, (SUST), Khartoum, Sudan
P.O. Box 11099, Taif 21944, Saudi Arabia 7
Department of Physics, Faculty of Science, King Abdulaziz
2
Deanship of Scientific Research, King Abdulaziz University, University, Jeddah 21589, Saudi Arabia
Jeddah, Saudi Arabia
3
Department of Mathematics and Statistics, College of
Science, Taif University, P.O. Box 11099, Taif 21944, Saudi
Arabia
4
Department of Mathematics, Faculty of Science, Sohag
University, Sohag 82524, Egypt

123
10134 Neural Computing and Applications (2022) 34:10133–10155

1 Introduction costs, especially in a complicated environment or in a large

searching space [7].
Drones are used in various applications related to smart Autonomy is enabled in drones that relate to route
city management, in addition to providing many services planning, task scheduling, task assignment, and commu-
such as research and air surveillance, as they have become nications in other considerations. These uses require dif-
widely used in recent [1]. Due to the flexibility of drones to ferent levels of control and the degree of independence to
move freely in an operational environment, the ability to take decisions in managing path setting and planning [8].
provide aerial viewpoint is an important feature of drones Therefore, such applications need intelligent algorithms
that promises cost-efficiency for many military and civilian that are based on making predictions and analyses to create
applications (see Fig. 1). Therefore, the deployment of a and track the paths of their operations [9].
drone to such applications faces difficulties to ensure safety
and reliable drones’ operations area, because of obstacles
and threats [2]. To overcome the mentioned difficulties, the 2 Related works and motivation
path planning problem has become one of the most active
research studies. Path planning methods enable the drone Drones are widely used for risk analysis. It does not require
flight path to be derived so that the drone can move in a any human intervention, but many drones require human
safe and unobstructed area [3]. intervention to supervise and monitor the different drone
Drones usually follow a predetermined path, or they can operations. Such as, some decisions are made by human
also fly independently, and they are under control from a operators to implement targeted drone missions as
ground station [4]. The use of drones in applications that required. Modern drones require being able to make deci-
are related to a complex and harsh environment depending sions about choosing a safe path on their own so that they
on the tasks, whether military or civilian, makes them face can make decisions about the flight path by using path
several challenges. The most important of which is the planning methods [10]. The path planning of drones
possibility of cooperation between them, and the mecha- especially in a 3D environment faces some issues that limit
nism of planning and following the best path to complete the possibility of obtaining a realistic path for the drones.
their tasks [5],6]. Traditional drone path planning methods Therefore, path-planning techniques are used that take into
have weak practical characteristics because they rarely take account all factors such as path length, optimal selection,
into account the limitations of obstacles and heights. The and completeness. By using path planning methods, drones
traditional path planning methods have high computational will be able to discover the optimal and crash-free path, in
addition to reducing energy consumption [11]. Recently,
many studies have been presented to investigate drone
operations and provide different path planning algorithms.
In addition, the emergence of new intelligent trajectory
planning methods based on swarm intelligence optimiza-
tion contributed to developing a smart drone path planning
method.
The study presented in [12], proposed a combination of
two algorithms, particle swarm optimization, and sequen-
tial niche to optimize the UAVs’ multi-paths planning. The
sequential niche method enables to adjustment of the
minimum values of objective function related to the sur-
rounding area to achieve multiple optimizations diversity.
The minimum value can be adjusted by using particle
swarm optimization (PSO). The authors also presented a
new technique to calculate the niche radius, which as are
taken into account were multiple paths planning opti-
mization. The proposed method helps to differentiate
between the drone paths by calculating the distance
between paths. Through the analysis, the authors clarified
that the used method gives results that achieve effective
multiple path planning for drones. In [13] authors review
an efficient method to enable efficient path planning free of
Fig. 1 Drone applications aspects collision in the 3D environment that contains a moving

123
Neural Computing and Applications (2022) 34:10133–10155 10135

obstacle. The proposed method is based on the use of the authors clarified that this method will improve the com-
Glow Worm Swarming Optimization (GSO) algorithm in munication latency.
the process of building a 3D path planning for operating the A study in [18] provides a unique method that helps to
drone in an environment with moving obstacles. The pro- improve the UAV’s path planning to solve the problems of
posed algorithm helps to achieve fast convergence, accel- slow convergence speed and poor search efficiency by
erating the process of path planning. The algorithm using the ACO algorithm. The proposed algorithm is based
improves the accuracy and convergence rate better than on several stages linked to the mechanism of establishing
heuristic optimization algorithms. and correcting the UAV path in addition to the smooth path
In the study presented by [14], the authors propose a planning that the drone follows during flight. Moreover, the
heuristic evolutionary algorithm with three cost functions, ACO algorithm works to avoid collisions with obstacles
to resolve an issue where battery life is short during flight during UAVs operations. The results of the analysis
missions. These functions are related to the path costs of showed that the proposed algorithm is effective and gives
length, security, and smoothness. The proposed algorithm an optimal solution to the path planning based on avoiding
is designed and evaluated in several path processes, like collision, in addition to achieving smoothness in the UAV
exchange, intersection, transition, and agility. The pro- flights. Authors in [19] presented a particle swarm opti-
posed solution enables adjustment and setting weights to mization mechanism that improves the velocity and opti-
convert multi-objective to single-objective optimization. In mization of the automatic trajectory mapping by using a 3D
[15] authors review the concept of using the A*BC algo- path planning to form UAVs. The mechanism initiates the
rithm for path planning to overcome the security problems particle distribution by using the chaos-based logistic map.
during UAVs operations. The study aimed to enhance the So the constant acceleration parameters are set, in addition
A* algorithm processes by using an artificial bee colony to the maximum speed to carry out the process of adapting
algorithm. The combination between ABC and A* algo- the linear variable parameters during the optimization
rithms enables the UAV to reach the desired goal for the process. Through simulations of the proposed method, it
drone flight in record time with the ability to avoid was found that PSO will rapidly and optimally enhance the
obstacles while flying. Through the analysis, the authors trajectory planning.
found that the A*BC algorithm enhances the ability of the Authors in [20] introduce a new approach based on an
drone to carry out the process of planning the best path improved A-star algorithm for real-time stealth drone’s
effectively with the adoption of avoiding obstacles in trajectory planning to enhance the quick breakout which
addition to achieving aviation safety goals. aims to devise a path permeation strategy. The A-star
A study presented in [16] uses three main approaches algorithm addresses the problem of re-planning of drones
known as an artificial potential field (APF), optimal success to stealth in a 3D dynamic environment. The proposed
rate strategy, and chaos strategy to solve the low conver- algorithm helps to enhance the stealth mechanism of the
gence rate and insufficient robustness problems in path drones and detect radar during the breakout process at low
planning methods. PDF enables to accelerate the conver- altitudes. In addition, the authors adopted the proposed
gence of processes, while the optimal success rate strategy algorithm in a real combat environment with various threat
improves the adaptive algorithm weights. The chaos scenarios. Through the analysis, it was found that the
strategy enables to avoid running into the optimum local combination of the path planning algorithm with the pre-
level. The authors improved the combined methods to diction technology enables stealthy UAVs to deal with
dramatically increase the APF success rate in finding threats and completely re-plan the path according to the
suitable planning paths and reducing convergence time. threat scenarios.
Through the results obtained from the analysis, it was In [21] the study demonstrated the possibility of using
found that the improved algorithm enhances the path the gray wolf enhancer algorithm to perform collaborative,
planning process for drones. Authors in [17] proposed a multi-drone path planning, taking into account the potential
method for UAV group modeling and management. The threats from air defenses while flying in restricted areas.
study provided an improved swarm intelligence algorithm The goal of using the proposed algorithm is to try to reduce
to calculate the movement of the whole swarm as well as the cost of potential threats associated with several con-
the movement of members within the swarm during flight. straints during the flight of the drone. The authors worked
Where the study relied on the analysis of previous work in on refining the gray wolf algorithm to optimize the ini-
which drones are used individually. The authors identified tialization process of the population and update individual
two levels of the proposed algorithm based on central positions to avoid suspected threats. In [22] the study
control over the UAV groups, in addition to the modeling presented a method for improving the performance of the
of individual UAV behavior when working as a group PSO algorithm to aid in planning the optimal path of the
interacting with each other, where through the analysis, the UAVs. In this study, the researchers worked on analyzing

123
10136 Neural Computing and Applications (2022) 34:10133–10155

the input parameters of the PSO algorithm for drone tra- concept. In addition, the paper also will analyze the per-
jectory planning based on the use of complex and varied formance of the most used swarm intelligent algorithm in
terrain maps. Through the analysis, the researchers found path planning, which is known as ant colony optimization
that the improved algorithm gives an optimum acceleration (ACO).
factor when planning the UAV’s paths. Authors in [23] use
the ferrying UAVs Based on a Generic Algorithm (GA) to
carry out search and rescue operations with high speed and 3 Drones architecture and operations
efficiency. In the given study, the researchers presented a background
new mechanism for the comparison and balance between
the UAV’s energy and the path length obtained from the Intelligent orientation drones enter into many applications,
planning process. The researchers also model path planning especially those that need a quick time response such as for
based on the ability to balance the UAV’s power versus the control and monitoring. Drones need to be characterized by
path length, which determines the operating parameters of the possibilities of free-flying and diversity performing
the UAVs in several factors related to energy consumption tasks and the ability to adopt advanced mechanisms to
in flight. make them have an independent behavior to avoid obsta-
The study presented in [24] provides an adaptive PSO cles and to achieve flight stability according to a method-
algorithm to solve the problem of UAV’s path planning ology that determines the coordinates of the path and
related to inertial weight and learning factors in the process movement. Path identification mechanisms work alongside
of iterative research in a 3D environment. The researchers smart decision-making methods to identify and assess
analyzed both the drone’s self-restriction conditions and conditions surrounding the drones in flight, in addition to
the fitness degree related to the design of the adaptive identifying the critical areas that require immediate inter-
particle swarm improvement algorithm based on the path vention, and coordinating an appropriate response [27].
planning diagram. The results of the simulation analysis Therefore, the intelligent management mechanisms in the
showed that the enhanced algorithm has a high impact to drones enable to evaluate the dynamic and complex
enable the UAVS to perform the optimal path planning emergency data to assess the flight position and then
process. determine the flight path according to the trajectory
In [25] authors present an algorithm that optimizes planning.
swarms of spherical vector-based particles that aids better When designing a drone system, it is important to define
trajectory planning in a potentially threatening environ- decision-making systems and plan flight paths to enable
ment. The algorithm is based on adjusting the cost function autonomy. The decision-making systems are based on
for both optimal and safety constraints in path planning. By putting in place several measures and playing side by side
analyzing the used algorithm, it was found that it is pos- with path planning methods to avoid any obstacles through
sible to obtain an optimal path that achieves less functional the information obtained by sensors to provide information
cost during the search in the UAV flight area by adopting on the actual conditions of the environment during flight
the calculations of the drone position, velocity, and rotation [28]. The system architecture in Fig. 2 shows a drone
angle during rising and fall operations. Authors in [26] system that allows for fully independent decision-making
review path planning for UAVs in a 3D-dimensional and advanced operations with intelligent route planning
complex operation area by using a biogeography-based avoiding any suspected incident.
optimization algorithm (BOA). The study uses BOA to find The architecture consists of different units that work on
an optimal solution to the UAV’s path planning problem. distributing, exchanging, and processing operations data,
The improved path planning by BOA generates the shortest and then evaluating the procedures for accomplishing the
path and gives the lowest process time consumption. flight mission according to the path plan mechanisms [29].
Table 1 summaries the related work studies with a The camera and other sensors located in the drones support
description of the method used and the extracted benefit of the units of path control and decisions making by infor-
the studies. mation obtained during flight, such as the distances from
Recently, artificial intelligence (AI) based on swarm the ground and obstacles. The representation of drone
optimization provides a successful method for solving operations related to image quality limitations and
problems related to path planning. Because of the impor- inspection services are considered as the most important
tance of path planning for drone deployment, this motivates task assessments related to the composition of the drone
us to study the most important AI algorithms that are used camera, which determines its efficiency and the practical
in planning the path of the drone’s applications. The paper and application aspects in which it can be used.
will provide a comprehensive analysis of different path In Fig. 2, the high-level control section contains two
planning algorithms based on the swarm intelligence parts, they are path planning and control formation of the

123
Neural Computing and Applications (2022) 34:10133–10155 10137

Table 1 Summary of swarm intelligence methods for drones path planning approaches
Algorithms approach Feature Advantage Citations

Particle swarm Diversity of multiple optimizations for drones paths Improve the distinguish process for different Cai et al.
optimization and planning issue paths, and plan for efficient multiple paths [12]
sequential niche
method
Glow-worm swarm A method for collision-free UAVs path in a 3D Faster convergence and speeds up the path- Goel
optimization (GSO) environment planning task improves convergence rate et al.
and accuracy [13]
heuristic evolutionary To solve the problem of short battery life in flying Easy conversion from multi-objective Fu et al.
algorithm missions optimization into a single-objective [14]
optimization
A*BC algorithm solve the security problems in UAV flight An effective mechanism to avoid obstacles Depeng
and ensure better path planning et al.
[15]
artificial potential field Overcome the low convergence rate and insufficient Great robustness for the path planning Lin et al.
(APF), robustness in drones path planning process [16]
General swarms Centralize the control of UAVs groups, and model Improves communication latency James
optimization distributed behavior of individual UAVs et al.
[17]
Ant colony optimization Solve the problem of slow convergence speed and High feasibility and effectiveness solutions Li et al.
low searching efficiency for optimal path and collision avoidance [18]
In UAVs trajectory planning
Particle swarm Automatic path planner 3D path planning algorithm Enhance the trajectory planning Shao
optimization for UAV formation et al.
[19]
Improved A- Star Real-time path planning based on the re-planning Efficient threats detection performance in the Zhang
algorithm issue for stealth UAV in a 3D complex dynamic flight environment et al.
environment [86]
Grey wolf optimizer multi-UAV cooperative path planning for air defense Effective in generating paths for multi-UAV Xu et al.
algorithm forces cooperative path planning under threats [21]
Enhanced particle swarm UAV path planning using various complex terrain Improves the acceleration coefficient for Kok
optimization maps with adequate repetitions optimum for UAV path planning et al.
[22]
Genetic algorithm (GA) Energy-balanced path planning for UAVs to measure Minimum planned path length and ferrying Wang
the balance between the planned path length and UAVs’ energy-factor variance et al.
the initial energy [23]
Adaptive particle swarm Path planning problem of fixed-wing UAV in three- Higher global search ability and accuracy Chen
optimization dimensional environment space et al.
[24]
Spherical vector-based UAV path planning in complicated environments Optimal path with minimum searching Phung
particle swarm subjected to multiple threats function cost for UAVs space configuration et al.
optimization [25]
Biogeography-based Trajectory planning for an unmanned aerial vehicle Improved trajectory algorithm generates the Xue
optimization in a three-dimensional complex workspace shortest trajectory and gives lowest the et al.
time consumption [26]

successful operation calculated from the task evaluation complexity of the flight area. The path for the desired
and drone tracks on the reference path. In the path-planning drone’s motion requires more updating to maintain differ-
module, the successive planned path depends on the ent aspects such as maneuverability and obstacle avoidance
intelligence degree of the optimization algorithm used for [28–30]. However, the path-planning problem can be
selecting the best trajectory plan that kept minimum path simplified to create reference paths for drone flights and
cost with obstacles collision avoidance. Path planning speed up the convergence of the optimization process of
processing depends on the environment data collected by the drone system.
the drones themselves and is greatly influenced by the

123
10138 Neural Computing and Applications (2022) 34:10133–10155

Fig. 2 Drone general system-

based path planning architecture
(Hoang, 2019)

After a successful path planning process, the high-level to avoid collision with potential obstacles during flight.
control unit is responsible to configure the planned path Otherwise, the drones must have intelligent decisions to
with the mission allocation and formation system for avoid collisions. The path planning concept is related to
controlling the flight process and determining the reference motion and trajectory planning, in addition to navigation.
path to carry out the process of evaluating the path that is In motion planning, drones consider the constraints such as
followed by the drones. Accordingly, the dynamic move- flight path, and ensure that calculations of path length and
ment of the drones is controlled by what is known as the minimum angle of rotation are taken into account when
low-level control section, which is related to the drone performing path optimization [32]. The trajectory mapping
controllers. serves as the motion planning, attaching the route planning
that contains the speed, time, and kinematics of the drones.
Both motion and trajectory planning relies on navigation to
4 Drones path planning and environments help avoid collision and localization and allow the drone to
be monitored during its journey [33].
Drone Path planning is an important issue encountered in Path planning for the drone is done according to the
operating a drone to move from an initial point to a target requirements of the application environment structure,
[31]. The drone path must be determined in such a way as whether it is 2-D (two-dimensional) or 3-D (three-

123
Neural Computing and Applications (2022) 34:10133–10155 10139

dimensional), as the path planning based on a 3-D envi- according to a prior geographic map, but cannot give
ronment can find obstacles and objects, which are used in optimal route planning and the possibility of effective
monitoring the drone and following its navigation, espe- drone control when there are unexpected obstacles during
cially in the complex work environment [34]. According to the flight [39]. The local route is planned by analyzing the
the state and nature of the environment in which the drones information obtained from the surrounding operating
are applied and deployed, a path-planning algorithm is environment to calculate an optimal path. Local environ-
chosen that aims to reduce the computational cost and time ment information helps to estimate drone flight states when
to plan the optimal path [35]. The path generated by this planning a path to avoid a potential collision with object
algorithm should be optimized as it consumes minimal movements in a dynamic environment [40]. Therefore,
energy, takes less time, and reduces the impact of the local path planning algorithms must acquire a degree of
drone’s collision with impeders, and between drones intelligence to adapt to the dynamic of the application
themselves [36]. environment and take advantage of environmental infor-
Path planning methods provide effective strategies that mation such as size, shape, and site structure.
analyze the data collected by the drone according to Traditional path planning such as Dijkstra algorithm
resource constraints such as energy, time, or distance enables the finding of the shortest path global graphs. The
budgets [37]. To calculate the optimal path (v), the maxi- Bellman–Ford algorithm can also be used to find the path
mum gain is achieved in the measurements according to the depending on the flexible iteration process concerning the
following equation. choice of initial estimates and updates [41]. The shortest
infoðmeasureðvÞÞ path can be extracted by estimating the shortest walk from
v ¼ arg max ; costðvÞ  b ð1Þ a given initial point to the next point subject to the con-
costðvÞ
straint that the walk contains at most walk links and goes
where the measure (v) gives the finite set of measurements through the initial point only once [41],42]. These algo-
along trajectory v in the 3-D space above the environment. rithms enable to obtain the shortest path from the digraph
Info represents the information objective quantifying the and build a path from the origin to the target in the original
utility of the acquired measurements. costðvÞ provides the probability map.
corresponding cost, which cannot exceed a predefined Besides the shortest path selection, an efficient drone’s
budget that is denoted by b [38]. operation requires that the planned path must be optimized
When carrying out the drone’s path planning process, depending on the application [43]. The drones are usually
some performance indicators must be taken into account, controlled from the ground base station and make the
such as the minimum cost of work, the shortest flight time, drones follow flight path waypoints which are corre-
the shortest flight route, and verification of the environment sponded to a path point in the search map (see Fig. 4). We
information used to calculate the optimal path [32]. Based assumed that flight path X is represented as a list of n
on these requirements, path planning methods can be waypoints, and the map coordinates are Cij ¼ ðxij ; yij ; zij Þ,
divided into two types, global and local path planning (see the Euclidean distance between two points can be found as
Fig. 3). Global path planning works to find the path

Fig. 3 Global and local path planning approaches

123
10140 Neural Computing and Applications (2022) 34:10133–10155

Fig. 4 Optimize trajectory

planning based on waypoints
allocation time

 
vector Cij Cijþ1 . The cost of path length plc can be found the swarm update the search state according to the detected
by the following equation [44]. channel information when they sensed a value higher than
the threshold, and otherwise, agents continue to explore
X
n1  
plc ¼ vectorCij Cijþ1  ð2Þ behavior until they reach their target.
j¼1 The swarm intelligence approaches enable to find opti-
mal direction to move by collaborating with the neighbors
Moreover, collision avoidance must also be considered in swarm environment. The collaboration with surrounding
in path planning methods to ensure safe drone operations agents in swarm enables to detect of search data values
by guiding the drone through possible threats that may be concerning the threshold and hence, agents who detect the
caused by obstacles in the operating environment [25, 44]. data values become a member of the search team to
The optimal path planning can be calculated based on autonomously emerging the different other cooperating
considering to set all threats and measures the diameter of agents’ teams [47]. Many intelligent swarm algorithms
the drones and the danger distance to the collision zone. based on swarm optimization have been studied exten-
The danger distance can be measured from the radio sively for search and path planning. In the following sec-
between the ten meters in a static environment with a good tion, the most important algorithms that are used in the
GPS signal and the hundred meters for moving objects in drone’s path planning are theoretically and mathematically
an environment with a weak GPS signal for positioning. presented and evaluated [48].
A. Particle swarm optimization (PSO)
5 Intelligent path planning algorithms Particle Swarm Optimization (PSO) is one of the swarm
intelligence approaches, acts as an optimization scheme,
Due to the complex drone flight procedures related to tra- and is based on social behavior. PSO performs population-
jectory planning and safe flight operations, the drones must based stochastic optimization processes with randomiza-
be able to smartly fly free in every operation environment tion according to interactions between population members
and avoid any probable threats or collation to obstacles. [49]. PSO is an effective technology that uses the concept
One of the most recently used techniques for search plan- of swarm intelligence, which can address complex prob-
ning is swarmed intelligence, which describes the behavior lems with high quality and fast calculation time. However,
of self-organized and decentralized particles acting and PSO can provide weak performance when considering
cooperating in their nature. Algorithms based on swarm optimization, which appears in its’ speed, calculation cost,
intelligence are enabled to provide a smart search strategy and global convergence [50]. Despite the aforementioned
by the concept of swarm agents cooperation [45]. The disadvantages, it offers better advantages over traditional
agents explore the work averment by collaborating with the algorithms for path planning. It provides a fast and easy
surrounding neighbors. Swarm agents are following search mechanism to improve path planning due to its
strategies of exploration behavior and smart search plan- acceptable convergence [50, 51]. Table 2 shows the com-
ning. The exploration behavior is based on collection data parison of the general features between the PSO and tra-
in the environment and compares the data with a predefined ditional algorithms.
threshold, while agents sense the neighbor’s data in the The strategy of PSO depends on initiating a population of
environment by using different kinds of communication random solutions, where each point in random solutions is
schemes according to the swarm kind [46]. The agents in treated as a particle that can move from one region to another

123
Neural Computing and Applications (2022) 34:10133–10155 10141

Table 2 Comparison between

Feature Traditional PCO
PCO and traditional algorithms
Search space Single point path search Potential solution population
Initialization By the user Automatically by PCO
Transition point Deterministic Probabilistic
Applicability Apply to a specific problem Apply to a variety of problems
Output Local optimization Global optimization
Advantages Convergence index Global search, high speed, parallelism
Disadvantages Computational cost No general convergence index

using a specific mechanism [52]. The algorithm evaluates Fig. 5) [54]. The performance of PSO is significantly
each random solution and extracts its cost to obtain the best affected by the particle’s speed, which can be expressed as a
local and global points in each waypoint associated with the speed for each particle by Eq. (3).
solutions of the population [53]. Through the evaluation, the    
solution points are arranged so that the local best point is SGþ1 G G G
i;k ¼ wSi;k þ Q1 rp pbest;k  xi;k þ Q2 rg gbest;k  xG
i;k

considered the best location for the current generation, and ð3Þ
the global best point expresses the absolute best location (see
where S represents the particle speed, Q1 and Q2 are the

Fig. 5 Particle swarm

optimization with the best
fitness

123
10142 Neural Computing and Applications (2022) 34:10133–10155

acceleration coefficients, and rg , rp are random numbers ACO algorithms based on the updating pheromones impact
between (0 and 1). W denoted for inertial weight and x for concerning the ACOs function optimization.
current location, where i and k represent the particle In general, the differences between the algorithms in
numbers from the population, and waypoint, respectively. Fig. 6 lie in the effect of the number of ants and the number
The term gbest represents the global best point, and pbest is of rounds calculated with different quantities of ants within
for local best point. The position of d (dimensional com- a limited number of iterations. The figure shows the dif-
ponent) of the ith the particle can be calculated as follows ference between the three algorithms, which determine the
accuracy of each of them according to how the ants and
xGþ1 G Gþ1
i;k ¼ xi;k þ Si;k ð4Þ
their number interact with the calculated solutions. The
In Eqs. (3) and (4), the parameters must be set to enable optimization can be deduced and the correct data obtained
updates to the local and global best points until the ter- from the optimization mechanism is related.
mination condition is met and calculates the cost of each Generally, the three algorithms work similarly by fol-
new solution. When using this algorithm, it is possible to lowing the same structure; however, the main difference is
reach the termination state at the maximum limit to cal- the workings of pheromones. The updating and maintain-
culate solutions and calculate a cost that is closer to the ing of pheromones in each iteration will result in different
optimal solution [56]. About planning the path and solutions according to the governed next search probabil-
movement in multi-dimensional directions, it finds that the ities by the number of pheromones and its distance
traditional algorithms face some limitations such as long- according to Eq. 5. The process of each iteration updating
time computations, and are likely to converge to a local depends on the q. The pheromone values are multiplied by
minimum. However, the PSO enables the finding of a this factor. The factor multiplication ensures to obtain of
globally optimal solution and is less likely to converge to a new paths for the best solution.
local minimum. C. Artificial bee colony (ABC)
B. Ant colony optimization (ACO) Swarm intelligence based on bee behavior, is a new
Ant colony optimization (ACO) is another type of algo- algorithm that enables optimization of the search mecha-
rithm that relies on swarm intelligence that solves the path nism depending on the nature of the bee agent’s interaction
optimization problem [57]. This algorithm relies on simu- in the colony [62]. The ABC algorithm is based on three
lating the behavior of real ants in their colony to manage artificial bee colony; they are as follows: worker bees,
the processes of creating and arranging paths. According to spectator bees, and scout bees [63]. The path is planned
the study of the ant’s interaction with each other, the from the bee colony site to the food site by finding the best
algorithm uses the same interaction behavior that the ants solution to follow the path. The path selection is according
adopt to determine paths between the starting points to the to the tasks assigned to the bees during the algorithm’s
destination by depositing a chemical substance known as process. The scout bee searches to find new food sources
the pheromone [58]. [64]. Worker bees locate food and store the location of
ACO enables to delivery of fast and efficient solutions their memories, while the spectator bees will receive
with distributed computation and collective interaction of a information on the source and location of food from the
population of agents. Through previous studies, ACO has worker bee. According to the mentioned bee’s operations,
provided solutions to several optimization problems in the ABC algorithm processes can be described as shown in
backward applications such as vehicle paths, path planning Fig. 7. The initial process occurred by the scout bees and
for moving robots, as well as optimal path planning for controls the parameters of xi vector of the population of
drone systems [59]. There are several types of ants’ food source [65]. Each xi can be calculated as follows.
behavior-based algorithms, such as ant system algorithm, xi ¼ li þ rand(0; 1Þðui  li Þ ð5Þ
ant colony algorithm for the iteration best, in addition to
ant colony system for the global best [60]. The difference where li and ui represent the lower and upper xi bound
between these ant colony algorithms is due to the pher- parameters, respectively. The spectator bees set the new
omone updating procedures. For example, in the ant colony food source, which is increased to have more nectar around
algorithm for the best iteration, the ant that gets the best the neighborhood of the source food population xi . The
result in each iteration of the pheromone updating proce- newly founded neighborhood of the source is evaluated and
dure is used. While in an ant colony of the best globally, can be calculated by the following equation [66].
the pheromone updating procedure may be left to the ant at vi ¼ xi þ Uðxi  xj Þ ð6Þ
the best performance during previous total iterations [61].
Figure 6 shows the performance comparison of different where U is a random number chosen between (-a, a), i and
j represent the current and next points (city), respectively.

123
Neural Computing and Applications (2022) 34:10133–10155 10143

Fig. 6 General impact of

pheromone updating in the
ACO algorithms performance

xj represents the randomly selected food source. The new compared with PSO, which can achieve acceptable con-
food source fitness is measured depending on the objective vergence speed and search for better solutions for drone
function value of the solution fi ðxi Þ. The spectator bees path planning. Although the DE and PSO address the
choose the food source depending on the probability values optimization problems and give an acceptable performance
calculations based on the fitness value as in Eq. (9). in accuracy, robustness, and convergence time, ABC gives
8 the highest performance when applied for multiple drone
< 1
; if fi ðxi Þ  0 path planning.
fiti ðxi Þ ¼ 1 þ f i ðx i Þ ð7Þ
:
1 þ absðfi ðxi ÞÞ; if fi ðxi Þ  0 D. Firefly algorithm (FA)
fiti ðxi Þ Firefly algorithm is a new mechanism for improving swarm
pi PSN ð8Þ intelligence, as it relies on research methods based on the
i¼1 fiti ðxi Þ
social behavior of fireflies [69]. The Firefly algorithm
The ABC algorithm is characterized by ease of imple- simulates the social behavior of flying fireflies, which are,
mentation and high flexibility since it requires fewer con- communicating, and searching to find mates by using
trol parameters compared with other search techniques. bioluminescence with varied flashing patterns. Fireflies’
The ABC algorithm has been used in many applications algorithm improves some fireflies’ flashing properties, so
such as engineering design, image processing, and network the improvement process depends on fireflies being
management, in addition to contributing to solving attracted to each other, regardless of their gender, the firefly
numerical optimization problems [15]. Moreover, it can brightness, and the objective function based on the firefly
provide an efficient unmanned aerial vehicle operation to brightness [70, 71].
avoid obstacles. The algorithm relies on an indication that flashing fire-
For drone path planning applications, ABC helps drones flies of lower brightness are stimulated and move toward
to safely fly toward the target point by improving obstacles the brightest. So that the attractiveness is proportional to
avoidance mechanisms. ABC algorithm enables to re-for- the brightness, which decreases relatively with the increase
mulate paths to avoid obstacles in flight, as it improves the in the distance between the fireflies [72]. The flowchart in
effectiveness and efficiency of drone path planning [84]. Fig. 9 shows the process of fireflies and position updating
The performance of ABC is shown in Fig. 8, where the according to the attraction toward the brightest firefly. The
best, median, and average cost values of the artificial bee attractiveness of the fireflies can be calculated by Eq. (11).
colony (ABC) are being compared to the particle swarm
b ¼ bo exp ðcdijm Þ; m1 ð9Þ
optimization (PSO) and Differential Evolution (DE) algo-
rithms. The Figure shows the performance of the ABC where bo represents the attractiveness of fireflies when the
against the PSO and DE in the drone path planning opti- distance dij = 0, and c is the coefficient of light absorption.
mization applications [96]. The ABC achieves small costs dij is the distance between two fireflies i and j.

123
10144 Neural Computing and Applications (2022) 34:10133–10155

Fig. 7 Artificial bee colony with

allocated onlookers

Equation (11) shows that it is a nonlinear equation, which xki ðt þ 1Þ ¼ xki ðtÞ þ bo expðcdij2 Þððxkj ðtÞ  xki ðtÞÞÞ þ ak ki
may lead to rich characteristics in the context of algorith-
ð11Þ
mic behaviors [73]. Since the degree of attractiveness is
depending on the distance between the fireflies, the where a is a scaling factor controlling the step sizes of the
movement of one firefly toward the brighter is affected by random walks that can be expressed by ki which is a ran-
this distance, which can be calculated by Eq. (10) [73], dom number vector that can be generated by uniform dis-
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi tribution with time ‘k’ According to the equations, for any
u D  2
uX
dij ¼ t xki  xkj ð10Þ two fireflies xi and xj, if the brightness of xj is greater than
k¼1 xi, then xi will be attracted by xj, and the position of firefly
xi will be updated [73]. That means each firefly is attracted
where dij is the distance between xi and xj, and D is the to other fireflies brighter. This algorithm enables the opti-
dimension xki represents the current value of ith firefly at mization method to be easier and leads to slow
kth dimension, and the same for the next value xkj . The convergence.
attraction of the fireflies toward each other can be calcu- Firefly algorithm provides a solution to several different
lated by the following equation. improvement problems through its unique technology

123
Neural Computing and Applications (2022) 34:10133–10155 10145

Fig. 8 Performance comparison between ABC, DE, and PSO for drones path planning

based on the nature of the firefly interacting with each where a is denoted for step size usually equal to 1. levy(b)
other. The FA algorithm can be integrated into the fuzzy is representing the parameter of levy flights to provide
controller to optimize different drone parameters such as random walk, where random steps can be calculated by the
obstacle distance, and mutual distance between the motion
levy distributed equation equal to t1b ð0\b\2Þ. The
of the drone. The use of the FA algorithm enables opti-
cuckoo bird algorithm achieves a higher global conver-
mization of the drone fuzzy controller parameters.
gence and efficiency when using Levi’s trips compared to
E. Cuckoo Search Algorithm (CSA) the algorithms that use standard Gaussian operations [83].
Cuckoo Search Algorithm (CSA) is an algorithm inspired A few numbers of studies have been conducted to use the
by the behavior of cuckoo birds. This algorithm theory is cuckoo search algorithm for drone trajectory planning since
based on basic rules that cuckoo birds take to place only using it within a longer period reduces the chances of obtaining
one egg in each stage, hence cuckoos randomly select the an optimal solution. The cuckoo search algorithm enables to
nest to place the eggs. In addition, that, cuckoos carried analysis and study of the problem of path planning for drones
forward only the high-quality eggs and nests [74]. This and works much better by combining it with the genetic algo-
algorithm works according to the behavior of the cuckoo rithm [77]. Hybrid cuckoo search and genetic algorithms will
when searching for a nest so that it represents a possible provide a fast and high success rate performance spectrum
solution egg, and it updates the solution by updating the when the control and threat points are many.
nest position continuously [75].
The cuckoo search mechanism relies on Levi’s flight to
simulate the pattern of birds in nature as illustrated in 6 Optimal path planning-based ACO
Fig. 10. The host bird can choose whether to throw the egg algorithm
away or abandon the nest and build a new nest completely
by using probability factor pa [ [0,1]. The cuckoo search Path planning is a key part of a drone’s assignment planning
algorithm uses the parameter pa to control local search and system. It is aiming at generating optimal or appropriate opti-
global exploration [8177]. Due to the concept of the mal flight paths, to safely finish the task when breaking through
cuckoo bird’s behavior, the new solution x (t ? 1) of n the enemy threats area. Based on the different task require-
cuckoos can be found for global exploration by Eq. (12), ments and battlefield environment, several different planning
methods have been put forward, such as optimized control,
xn ðt þ 1Þ ¼ xn ðtÞ þ a levyðbÞ ð12Þ Singular perturbation method, and artificial potential field
method [78]. In this paper, we presented some famous intelli-
gent optimization technology-based swarm planning methods.

123
10146 Neural Computing and Applications (2022) 34:10133–10155

Fig. 9 The firefly algorithm

ACO is one of the most efficient swarm intelligent methods that environment of the area of the drone flight in addition to its
enable to optimization of the path planning for the drone’s impact on changes and other weather factors. In this study,
operations. In this section, we evaluate the performance of we consider using an ant colony optimization algorithm to
ACO for drone path planning and extract its impact in different plot the flight path of the drone, while ignoring the various
scenarios. weather factors and the ability of the drone to maneuver.
The algorithm was analyzed based on the path planning
6.1 ACO performance test correction problem [79].
According to the ant colony behaviors, the ACO algo-
In general, the process of path planning for large flight rithm enables a selection of the best path depending on the
distances is linked to several restrictions depending on the probability of edge selection by following the procedure of

123
Neural Computing and Applications (2022) 34:10133–10155 10147

Fig. 10 Cuckoo search algorithm

flowchart as shown in Fig. 11. We consider an ant that on the use of theories related to the nature of ants they are,
would like to choose the path in a given network node, ant pheromone updating, and evaporation [67]. The pher-
will start to move from one node to another and deposit omones are updated according to Eq. (14) as follows.
pheromone [80]. The first ant chooses randomly the loca- Xm
tion of the direct path, and updates the current path by sij ¼ ð1  qÞsij þ Dskij ð14Þ
depositing the pheromone, and hence the evaporated path is k¼1

considered the worst choice, and the path with a high where (1 - q) represents the pheromone evaporation rate
pheromone concentration is the best one [81]. By consid- and q is a pheromone persistence coefficient and Dskij is the
ering k to denote the ant, start and next node represented by
number of deposited pheromones. Equation 6 ensures
i and j, respectively. Ant will choose to go to the next node
continuously best path updates, which promotes strong
based on the probability given to be Eq. (13).
paths and neglects weak paths [83].
ðskij Þa ðgkij Þb The ACO algorithm is analyzed based on planning the
pkij ¼ P ; if j 2 Nik ; otherwise best flight path from the starting point to the endpoint of the
R k ðsk Þa ðgk Þb ð13Þ
j Ni ij ij
drone flight, so that the drone path is corrected during the
pkij ¼ 0 flight, taking into account the selection of the shortest flight
path. The algorithm of the ACO will also be analyzed after
where pkij is the probability of edge selection, ðskij Þa and
improving so that the drone can avoid obstacles during the
ðgkij Þb represent the pheromone level, and desirability of flight by measuring a new parameter into account of the
edge, respectively. The parameters a and b are used to traditional ACO by assuming that the drone can measure
control the impact of ðskij Þa and ðgkij Þb . If consider setting the distance from the obstacles during flight and adjust the
a = 0, the probability will depend on heuristic values. By flight path quickly and accurately at the lowest cost. We
considering to set b = 0, the probability will have calcu- compare the ACO and the proposed enhancement in terms
lated depending on the pheromone value [82]. of optimal path costs. The simulation will take two analysis
Due to the ant moving in the path according to the scenarios by considering the evaluation of the ACO algo-
concentration of pheromone value, the algorithm will work rithm in 2D and 3D flight environments. The simulation

123
10148 Neural Computing and Applications (2022) 34:10133–10155

trajectory is determined through iterative searches for ant

swarm optimization. The path plan is adopted from a
starting point to an endpoint in a three-dimensional
scheme. Also, the two-dimensional path planning of the
drone using the ACO algorithm was analyzed in the same
way as in the first scenario, and then the proposed
enhancement apothems are evaluated compared to the
traditional ACO after improving the path planning update
mechanism depending on the drone’s distance from the
potential obstacles.

6.2 Parameter settings

In the analysis, the MATLAB program version R2020a

was used to simulate the planning of the three (3D) and
(2D) dimensional trajectory of drones and measure the
performance of the ACO algorithms, in addition to measure
its effectiveness in selecting the best path at the lowest
cost. Table 3 shows the simulation parameters for the ACO
path planning settings.

6.3 Analysis results

This section shows the results of performance analysis for

the ACO algorithm in a 3D and 2D-dimensional environ-
ment, and then presents the performance evaluation of the
Fig. 11 The proposed ACO for drone’s path planning ACO compared to the improved ACO algorithm. The
results are divided into scenarios according to the 2D and
will take place according to the Algorithm 1 procedures. 3D environments as follows.
In the first scenario, the ACO algorithm is analyzed on a From the simulation of the 3D path planning environ-
3D drone path-planning model and the cost of the optimum ment, as shown in Fig. 12, we note that the use of the ACO

Algorithm 1: Simulation procedure and sequences

// Simulation procedure and sequences
// for the proposed ACO algorithm
1. Initiate the ACO algorithm setting
2. Assume the maximum number of ants used
3. Explore a path planning scenario
4. Consider iterations for the ACO.
5. Set the edge pheromone for paths
6. Ants are allowed accord to the evaporation and update process
7. Consider the distance parameter in calculations.
8. Drones measure the distances from each obstacle
9. Set according to the critical, safe, and crash distances.
10. Calculate the path covered by the ants
11. Make conditions for the best update path.
12. Return ACO path planned

123
Neural Computing and Applications (2022) 34:10133–10155 10149

Table 3 Simulation parameters simulation environment selects to go for shorter paths to

Parameters Values
visit all targets using the same obstacles distribution. The
time calculation for the two algorithms that were taken to
Flight area 80 9 40 km reach the endpoint is out of this paper’s scope and can be
Ant numbers 20 considered in future work.
Iterations number 100 As shown in Fig. 16, the performance of improved ACO
Coefficient a 1 is impacted by the distance measured from obstacles which
Coefficient b 4 show that, when drones become closer to the obstacles, the
Coefficient Q 8 convergence curve for the optimal path shows the worst
Evaporating rate q 0.75 cost case, while the safe distance enables to obtain a good
Safe obstacle distance 6m convergence curve. The proposed safe distance shows that
Critical obstacle distance 1m less optimal cost can be achieved than the other two
Crush obstacle distance 0.5 m schemes. In addition, the optimal cost in the safe distance
can be reached very fast compared with the critical and
crash distances where the optimal path cost was reached
after 20 and 13 iterations, respectively. By comparing, the
algorithm in planning the three-dimensional trajectory of performance of the ACO and improvement in path updat-
the drone flight directly affects the implementation effi- ing based on distances state, the proposed mechanism
ciency and the quality of the path. As shown in the illus- shows it outperformed the ACO as shown in Fig. 17.
trated Fig. 13, it found that the algorithm establishes a The proposed algorithm shows less iteration number
compromise between three-dimensional environmental than the ACO, where it needs only three iterations before it
complexity and computational efficiency. Simulations goes into a steady-state, while ACO requires about 20
show that the ACO algorithm would produce a good iterations to reach the steady-state [84]. One can notice also
solution in an acceptable time for best path planning with that, ACO went through various steps and values called
acceptable cost. stairs effects before reaching its steady states. This hap-
For the 2D path planning environment, Figs. 14 and 15 pened from 9 iterations until 20 iterations where it went
show the simulation of drone Path planning by using ACO through 5 steps to reach finally the steady-state. In general,
and proposed improvement, respectively. By considering the results show that the improved ACO algorithm can find
the obstacles distance calculations with the ACO path optimal paths at higher convergence speeds. The ACO
updating procedure, we notice that the parameter distance algorithm may not make the path meet the mission con-
between the drone and obstacle will serve to determine the straints even though it can get the path from starting node
quality of the path planning. From the illustration of the to destination, but by considering the distance obstacles
results, it can be shown that our proposed method in a criteria in the ACO path updating process, the proposed

Fig. 12 3D simulation for drone

path planning using ACO
algorithm

123
10150 Neural Computing and Applications (2022) 34:10133–10155

Fig. 13 Convergence curves of 3D optimal paths according to the iterative process

Fig. 14 2D simulation for drone path planning using ACO algorithm

methods will show that the convergence speed and global 7 Conclusions
optimality are improved significantly [85].
Drones represent an important role in the emergence of a
wide range of surveillance and research applications.
However, the application of the drone in different

123
Neural Computing and Applications (2022) 34:10133–10155 10151

Fig. 15 2D simulation for drone path planning using improved path distance algorithm

Fig. 16 Performance of obstacle distance measurement impact on the ACO algorithm

environments represents a major challenge in planning comprehensive analysis of several algorithms that rely on
tasks and a safe path to obtain data efficiently, especially in swarm intelligence for drone path planning, which works to
complex environments. This paper provided a solve many problems related to the possibility of learning

123
10152 Neural Computing and Applications (2022) 34:10133–10155

Fig. 17 Convergence curves of optimal paths for ACO and proposed improvement

and focus on the limitations of the working environment In the future study, machine-learning techniques will be
for drones. The performance of some algorithms used in used to optimize the process of analyzing the relative
planning the drone’s trajectory has also been analyzed and positions of the detected objects and obstacles to determine
simulated on various forms of the environment, such as the the optimal collision-free paths.
two-dimensional and the three-dimensional environment.
In this paper, the performance of the ACO algorithm to
Funding This paper was funded by the Deanship of Scientific
solve the drone’s path planning problem in a two-dimen-
Research (DSR), King Abdulaziz University, Jeddah, under Grant No.
sional and three-dimensional environment space was ana- (D-78-305-1442). The authors, therefore, gratefully acknowledge
lyzed assuming the presence of obstacles on the flight area. DSR technical and financial support.
The ACO was also optimized by adopting measurements of
Data availability The data used to support the findings of this study
the distance between the drone and obstacles during flight
are available from the corresponding author upon request.
to update the route planning during flight and then per-
forming a simulation to compare the ACO algorithm and
the proposed method in a two-dimensional environment Declarations
where it was found that the improvement gives high effi-
ciency in optimizing the path planning with acceptance Conflicts of interest The authors declare that they have no conflicts of
interest to report regarding the present study.
convergence and cost.
Our future study is based on the fact that the presented
study relied on choosing an appropriate path at the lowest References
cost on the distance between the drones and the obstacles
so that the information is taken by measuring the distance 1. Mohamed N, Al-Jaroodi J, Jawhar I, Idries A, Mohammed F
through sensors during flight as a value that determines the (2020) Unmanned aerial vehicles applications in future smart
cities. Technol Forecast Soc Change 153:119293
dangerous and permitted flight levels according to the
2. Wu Y, Wu S, Hu X (2021) Multi-constrained cooperative path
distance of fixed obstacles with the path-determination planning of multiple drones for persistent surveillance in urban
mechanism based on the swarm algorithm. However, this environments. Complex Intell Syst. https://doi.org/10.1007/
method cannot be efficient if the drones face moving or s40747-021-00300-5
sudden obstacles, and therefore there is a need for a smart 3. Alsolami F, Alqurashi FA, Hasan MK, Saeed RA, Abdel-Khalek
S, BenIshak A (2021) Development of self-synchronized drones’
mechanism to maneuver while flying to the destination network using cluster-based swarm intelligence approach. IEEE
through an obstacle-free path.

123
Neural Computing and Applications (2022) 34:10133–10155 10153

Access 9:48010–48022. https://doi.org/10.1109/ACCESS.2021. 21. Xu C, Xu M, Yin C (2020) Optimized multi-UAV cooperative

3064905 path planning under the complex confrontation environment.
4. Kangunde V, Jamisola RS Jr, Theophilus EK (2021) A review on Elsevier, Amsterdam
drones controlled in real-time. Int J Dyn Control. https://doi.org/ 22. Kok KY, Rajendran P (2020) Enhanced particle swarm opti-
10.1007/s40435-020-00737-5 (Epub ahead of print) mization for path planning of unmanned aerial vehicles. ECTI
5. Alladi T, Chamola V, Sahu N, Guizani M (2020) Applications of Trans Comput Inf Technol 14(1):67–78
blockchain in unmanned aerial vehicles: a review. Veh Commun 23. Wang L et al (2020) An energy-balanced path planning algorithm
23:100249 for multiple ferrying UAVs based on GA. Int J Aerosp Eng
6. Giordan D, Adams MS, Aicardi I et al (2020) The use of 2020:15 (Article ID 3516149)
unmanned aerial vehicles (UAVs) for engineering geology 24. Chen J, Zhao H, Wang L (2021) Three-dimensional path planning
applications. Bull Eng Geol Environ 79:3437–3481. https://doi. of UAV based on adaptive particle swarm optimization algo-
org/10.1007/s10064-020-01766-2 rithm. In: ISSMAS 2021. Journal of physics: conference series,
7. Mukherjee A, Saeed RA, Dutta S, Naskar MK (2017) Fault vol 1846
tracking framework for software-defined networking (SDN). In: 25. Phung MD, Ha QP (2021) Safety-enhanced UAV path planning
Singhal C, De S (eds) Resource allocation in next-generation with spherical vector based particle swarm optimization. arXiv:
broadband wireless access networks. IGI Global, Hershey, 2104.10033v1 [cs.NE] 13 April 2021
pp 247–272. https://doi.org/10.4018/978-1-5225-2023-8.ch011 26. Xue Z, Liu X (2021) Trajectory planning of unmanned aerial
8. Zhang T, Li Q, Zhang C et al (2017) Current trends in the vehicle based on the improved biogeography-based optimization
development of intelligent unmanned autonomous systems. Front algorithm. Adv Mech Eng 13(3):1–15
Inf Technol Electron Eng 18:68–85. https://doi.org/10.1631/ 27. Hodge VJ, Hawkins R, Alexander R (2021) Deep reinforcement
FITEE.1601650 learning for drone navigation using sensor data. Neural Comput
9. Azar AT, Koubaa A, Ali Mohamed N, Ibrahim HA, Ibrahim ZF, Appl 33:2015–2033. https://doi.org/10.1007/s00521-020-05097-x
Kazim M, Ammar A, Benjdira B, Khamis AM, Hameed IA, 28. de Assis RA, Pazim R, Malavazi MC, Petry PPC, de Assis LM,
Casalino G (2021) Drone deep reinforcement learning: a review. Venturino E (2020) A mathematical model to describe the herd
Electronics 10(9):999. https://doi.org/10.3390/ behaviour considering group defence. Localiz: Appl Math Non-
electronics10090999 linear Sci 5(1):11–24
10. Besada JA, Bergesio L, Campaña I, Vaquero-Melchor D, López- 29. Mosa AA, Abdalla AH, Saeed RA (2012) Evaluation of MAN-
Araquistain J, Bernardos AM, Casar JR (2018) Drone mission EMO route optimization schemes. J Netw Comput Appl
definition and implementation for automated infrastructure 35(5):1454–1472. https://doi.org/10.1016/j.jnca.2012.02.001
inspection using airborne sensors. Sensors (Basel) 18(4):1170. 30. Popović M, Vidal-Calleja T, Hitz G et al (2020) An informative
https://doi.org/10.3390/s18041170 path planning framework for UAV-based terrain monitoring.
11. Yosuf RH, Mokhtar RA, Saeed RA, Alhumyani H, Abdel-Khalek Auton Robot 44:889–911. https://doi.org/10.1007/s10514-020-
S (2022) Scheduling algorithm for grid computing using shortest 09903-2
job first with time quantum. Intell Autom Soft Comput 31. Alsaqour R, Abdelhaq M, Saeed RA, Al-Hubaishi M, Alsaqour
31(1):581–590. https://doi.org/10.32604/iasc.2022.019928 O, Uddin M, Alahdal T (2014) Effect of mobility parameters on
12. Cai Q, Long T, Wang Z, Wen Y, Kou J (2016) Multiple paths position information inaccuracy of position-based MANET
planning for UAVs using particle swarm optimization with routing protocol. Int J Wirel Mob Comput (Inderscience)
sequential niche technique. In: 28th Chinese control and decision 7(1):68–77. https://doi.org/10.1504/IJWMC.2014.058886
conference (CCDC) 32. Madridano Á, Al-Kaff A, Martı́n D, de la Escalera AA (2020) 3D
13. Goel U, Varshney S, Jain A, Maheshwari S, Shukla A (2018) trajectory planning method for UAVs swarm in building emer-
Three-dimensional path planning for UAVs in dynamic envi- gencies. Sensors (Basel) 20(3):642. https://doi.org/10.3390/
ronment using glow-worm swarm optimization. In: International s20030642
conference on robotics and smart manufacturing (RoSMa2018) 33. Alsharif S, Saeed RA, Albagory Y (2022) An efficient HAPS
science direct. Procedia Computer Science, vol 133, pp 230–239 cross-layer design to mitigate COVID-19 consequences. Intell
14. Fu Z et al (2018) A heuristic evolutionary algorithm of UAV path Autom Soft Comput 31(1):43–59. https://doi.org/10.32604/iasc.
planning. Wirel Commun Mob Comput 2018:11 (Article ID 2022.019493
2851964) 34. Majeed A, Lee S (2018) A fast global flight path planning
15. Depeng X, Ziying L, Mengshan L (2019) Research and simula- algorithm based on space circumscription and sparse visibility
tion of UAV security strategy based on A*BC algorithm. Comput graph for unmanned aerial vehicle. Electronics 7(12):375. https://
Sci Eng 9(1):1–5 doi.org/10.3390/electronics7120375
16. Lin N, Tang J, Li X, Zhao L (2019) A novel improved bat 35. Saeed RA, Mokhtar RA, Chebil J, Abdallah AH (2012) TVBDs
algorithm in UAV path planning. Comput Mater Contin CMC coexistence by leverage sensing and geo-location database. In:
61(1):323–344 2012 international conference on computer and communication
17. James S, Raheb R, Hudak A (2020) UAV swarm path planning. engineering (ICCCE), pp 33–39. https://doi.org/10.1109/ICCCE.
In: 2020 integrated communications navigation and surveillance 2012.6271147
conference (ICNS), pp 2G3-1–2G3-12 36. Nurelmadina N, Hasan MK, Memon I, Saeed RA, ZainolAriffin
18. Li B, Qi X, Yu B, Liu L (2020) Trajectory planning for UAV KA, Ali ES, Mokhtar RA, Islam S, Hossain E, Hassan MA (2021)
based on improved ACO algorithm. IEEE Access 8:2995–3006. A systematic review on cognitive radio in low power wide area
https://doi.org/10.1109/ACCESS.2019.2962340 network for industrial IoT applications. Sustainability 13(1):338.
19. Shao S, Peng Y, He C, Du Y (2020) Efficient path planning for https://doi.org/10.3390/su13010338
UAV formation via comprehensively improved particle swarm 37. Wang Y, Zhang G, Shi Z, Wang Q, Su J, Qiao H (2020) Finite-
optimization. ISA Trans 97:415–430 time active disturbance rejection control for marine diesel engine.
20. İlhan E, Kıymaz İO (2020) A generalization of truncated Appl Math Nonlinear Sci 5(1):35–46. https://doi.org/10.2478/
M-fractional derivative and applications to fractional differential amns.2020.1
equations. Appl Math Nonlinear Sci 5(1):171–188. https://doi. 38. Rupa C, Midhunchak D, Hasan MK, Alhumyani H, Saeed RA
org/10.2478/amns.2020.1.00016 (2021) Industry 5.0: Ethereum blockchain technology based

123
10154 Neural Computing and Applications (2022) 34:10133–10155

DApp smart contract. Math Biosci Eng 18(5):7010–7027. https:// J Bionic Eng 6:161–173. https://doi.org/10.1016/S1672-
doi.org/10.3934/mbe.2021349 6529(08)60113-4
39. Wen L et al (2020) Collision-free trajectory planning for auton- 56. Ahmed ZE, Hasan MK, Saeed RA, Hassan R, Islam S, Mokhtar
omous surface vehicle. arXiv:2005.09857v1 RA, Khan S, Akhtaruzzaman M (2020) Optimizing energy con-
40. Yang L, Qi J, Song D, Xiao J, Han J, Xia Y (2016) Survey of sumption for cloud internet of things. Front Phys 8:358. https://
robot 3D path planning algorithms. J Control Sci Eng 2016:22. doi.org/10.3389/fphy.2020.00358
https://doi.org/10.1155/2016/7426913 (Article ID 7426913) 57. Perez-Carabaza S, Besada-Portas E, Lopez-Orozco JA, de la Cruz
41. Eltahir AA, Saeed RA, Alawi MA (2013) An enhanced hybrid JM (2018) Ant colony optimization for multi-UAV minimum
wireless mesh protocol (E-HWMP) protocol for multihop time search in uncertain domains. Appl Soft Comput 62:789–806
vehicular communications. In: 2013 international conference on 58. Tamura Y, Sakiyama T, Arizono I (2021) Ant colony optimiza-
computing, electrical and electronic engineering (ICCEEE), tion using common social information and self-memory. Com-
pp 1–8. https://doi.org/10.1109/ICCEEE.2013.6633899 plexity 2021:7. https://doi.org/10.1155/2021/6610670 (Article ID
42. Alarcón V et al (2020) Procedures for the integration of drones 6610670)
into the airspace based on U-space services. Aerospace 7:128. 59. Hu B, Sun Z, Hong H et al (2020) UAV-aided networks with
https://doi.org/10.3390/aerospace7090128 optimization allocation via artificial bee colony with intellective
43. Schranz M, Umlauft M, Sende M, Elmenreich W (2020) Swarm search. J Wirel Commun Netw 2020:40. https://doi.org/10.1186/
robotic behaviors and current applications. Front Robot AI 7:36 s13638-020-1659-y
44. Li T, Yang W (2020) Solution to chance constrained program- 60. Chiu CL, Zhang J, Li M et al (2020) A study of environmental
ming problem in swap trailer transport organisation based on disclosures practices in Chinese energy industry. AJSSR 5:9.
improved simulated annealing algorithm. Appl Math Nonlinear https://doi.org/10.1186/s41180-020-00036-1
Sci 5(1):47–54. https://doi.org/10.2478/amns.2020.1.00005 61. Annepu V, Rajesh A (2020) Implementation of an efficient arti-
45. Alfeo AL, Cimino MGCA, De Francesco N, Lazzeri A, Lega M, ficial bee colony algorithm for node localization in unmanned
Vaglini G (2018) Swarm coordination of mini-UAVs for target aerial vehicle assisted wireless sensor networks. Wirel Pers
search usingimperfectsensors. Intell Decis Technol Commun 114:2663–2680. https://doi.org/10.1007/s11277-020-
12(2):149–162. https://doi.org/10.3233/IDT-170317 07496-8
46. Alawi MA, Saeed RA, Hassan AA, Alsaqour RA (2014) Sim- 62. Li B, Gong LG, Yang WL (2014) An improved artificial bee
plified gateway selection scheme for multi-hop relay vehicular ad colony algorithm based on balance-evolution strategy for
hoc network. Int J Commun Syst 27(12):3855–3873. https://doi. unmanned combat aerial vehicle path planning. Sci World J
org/10.1002/dac.2581 2014:10. https://doi.org/10.1155/2014/232704 (Article ID
47. del Valle Y, Venayagamoorthy GK, Mohagheghi S, Hernandez J, 232704)
Harley RG (2008) Particle swarm optimization: basic concepts, 63. Han YQ, Li JQ, Liu Z, Liu C, Tian J (2020) Metaheuristic
variants and applications in power systems. IEEE Trans Evol algorithm for solving the multi-objective vehicle routing problem
Comput 12(2):171–195. https://doi.org/10.1109/TEVC.2007. with time window and drones. Int J Adv Rob Syst. https://doi.org/
896686 10.1177/1729881420920031
48. Janga Reddy M, Nagesh Kumar D (2020) Evolutionary algo- 64. Mao W, Lan HY, Li HR (2016) A new modified artificial bee
rithms, swarm intelligence methods, and their applications in colony algorithm with exponential function adaptive steps.
water resources engineering: a state-of-the-art review. H2Open J Comput Intell Neurosci 2016:13. https://doi.org/10.1155/2016/
3(1):135–188. https://doi.org/10.2166/h2oj.2020.128 9820294 (Article ID 9820294)
49. El-Shorbagy MA, Hassanien AE (2018) Particle swarm opti- 65. Pang B, Song Y, Zhang C, Wang H, Yang R (2018) A modified
mization from theory to applications. Int J Rough Sets Data Anal artificial bee colony algorithm based on the self-learning mech-
(IJRSDA) 5(2):1–24. https://doi.org/10.4018/IJRSDA. anism. Algorithms 11(6):78. https://doi.org/10.3390/a11060078
2018040101 (Accessed (May 04, 2021)) 66. Sheoran S, Mittal N, Gelbukh A (2020) Artificial bee colony
50. Fan S-KS, Jen C-H (2019) An enhanced partial search to particle algorithm in data flow testing for optimal test suite generation. Int
swarm optimization for unconstrained optimization. Mathematics J Syst Assur Eng Manag 11:340–349. https://doi.org/10.1007/
7:357. https://doi.org/10.3390/math7040357 s13198-019-00862-1
51. Ghorpade SN, Zennaro M, Chaudhari BS, Saeed RA, Alhumyani 67. Feng J, Meng M, Liu S, Zhang X, Yuan J, Zhang Z (2020)
H, Abdel-Khalek S (2021) Enhanced differential crossover and Prediction of Chinese automobile growing trend considering
quantum particle swarm optimization for IoT applications. IEEE vehicle adaptability based on Cui–Lawson model. Appl Math
Access. https://doi.org/10.1109/ACCESS.2021.3093113 Nonlinear Sci 5(2):367–376. https://doi.org/10.2478/amns.2020.
52. Martı́nez-Ledesma M, Jaramillo Montoya F (2020) Performance 2.00054
evaluation of the particle swarm optimization algorithm to 68. Patel B, Patle B (2020) Analysis of firefly-fuzzy hybrid algorithm
unambiguously estimate plasma parameters from incoherent for navigation of quad-rotor unmanned aerial vehicle. Inventions
scatter radar signals. Earth Planets Space 72:172. https://doi.org/ 5:48. https://doi.org/10.3390/inventions5030048
10.1186/s40623-020-01297-w 69. Fu Q et al (2015) A novel firefly algorithm based on improved
53. Ali ES, Hasan MK, Hassan R, Saeed RA, Hassan MB, Islam S, learning mechanism. In: International conference on logistics
Nafi NS, Bevinakoppa S (2021) Machine learning technologies engineering, management and computer science (LEMCS 2015)
for secure vehicular communication in internet of vehicles: recent 70. Henrio J, Deligne T, Nakashima T et al (2019) Route planning for
advances and applications. J Secur Commun Netw (SCN). https:// multiple surveillance autonomous drones using a discrete firefly
doi.org/10.1155/2021/8868355 algorithm and a Bayesian optimization method. Artif Life
54. Lu YX, Wang JS, Guo SS (2019) Solving path planning problem Robotics 24:100–105. https://doi.org/10.1007/s10015-018-0454-
based on particle swarm optimization algorithm with improved x
inertia weights. IAENG Int J Comput Sci 46:4 71. Ahmed KEB, Saeed RA, Mokhtar RA (2017) Real time CAM-
55. Duan H, Zhang X, Wu J et al (2009) Max–min adaptive ant SHIFT tracking algorithm using TMS320DM6437 EVM. In:
colony optimization approach to multi-UAVs coordinated tra- 2017 international conference on communication, control, com-
jectory replanning in dynamic and uncertain environments. puting and electronics engineering (ICCCCEE), pp 1–6. https://
doi.org/10.1109/ICCCCEE.2017.7866698

123
Neural Computing and Applications (2022) 34:10133–10155 10155

72. Yu X, Xu F (2020) Random inverse packet information and its 80. Wang G, Guo L, Duan H, Wang H, Liu L, Shao M (2012) A
acquisition. Appl Math Nonlinear Sci 5(2):357–366. https://doi. hybrid metaheuristic DE/CS algorithm for UCAV three-dimen-
org/10.2478/amns.2020.2.00042 sion path planning. Sci World J 2012:11. https://doi.org/10.1100/
73. Cheng L et al (2020) A staged adaptive firefly algorithm for UAV 2012/583973 (Article ID 583973)
charging planning in wireless sensor networks. Comput Commun 81. Cuevas E, Reyna-Orta A (2014) A cuckoo search algorithm for
161:132–141 multimodal optimization. Sci World J 2014:20. https://doi.org/10.
74. Lu Y, Dong C, Wang Q (2017) Control allocation for aircraft 1155/2014/497514 (Article ID 497514)
with input constraints based on improved cuckoo search algo- 82. Pan JS, Song PC, Chu SC, Peng YJ (2020) Improved compact
rithm. Defence Technol 13(1):1–5 cuckoo search algorithm applied to location of drone logistics
75. Hu K-C, Tsai C-W, Chiang M-C, Yang C-S (2015) A multiple hub. Mathematics 8:333. https://doi.org/10.3390/math8030333
pheromone table based ant colony optimization for clustering. 83. Senan S, Hashim AA, Saeed RA, Hameed SA, Zeki AM, Daoud
Math Probl Eng 2015:11. https://doi.org/10.1155/2015/158632 JI (2012) Route optimization scenario of a new scheme based on
(Article ID 158632) nested mobile network. In: 2012 international conference on
76. Tao Y et al (2021) A mobile service robot global path planning computer and communication engineering (ICCCE), pp 717–721.
method based on ant colony optimization and fuzzy control. Appl https://doi.org/10.1109/ICCCE.2012.6271310
Sci 11:3605. https://doi.org/10.3390/app11083605 84. Iglesias A et al (2018) Cuckoo search algorithm with Lévy flights
77. Ahmed MZ, Hashim AHA, Khalifa OO, Saeed RA, Alsaqour RA, for global-support parametric surface approximation in reverse
Alkali AH (2021) Connectivity framework for rendezvous and engineering. Symmetry 10:58. https://doi.org/10.3390/
mobile producer nodes using NDN interest flooding. In: 2021 sym10030058
international congress of advanced technology and engineering 85. Zhao Y (2020) Analysis of trade effect in post-Tpp era: based on
(ICOTEN), pp 1–5.https://doi.org/10.1109/ICOTEN52080.2021. gravity model and Gtap model. Appl Math Nonlinear Sci
9493555 5(1):61–70. https://doi.org/10.2478/amns.2020.1.00007
78. Chen Y, Li N, Zhong X, Xie W (2019) Joint trajectory and 86. Zhang Z, Wu J, Dai J, He C (2020) A novel real-time penetration
scheduling optimization for the mobile UAV aerial base station: a path planning algorithm for stealth UAV in 3D complex dynamic
fairness version. Appl Sci 9(15):3101. https://doi.org/10.3390/ environment. IEEE Access 8:122757–122771. https://doi.org/10.
app9153101 1109/ACCESS.2020.3007496
79. Pan JS, Liu JL, Hsiung SC (2019) Chaotic cuckoo search algo-
rithm for solving unmanned combat aerial vehicle path planning Publisher’s Note Springer Nature remains neutral with regard to
problems. In: Proceedings of the 2019 11th international con- jurisdictional claims in published maps and institutional affiliations.
ference on machine learning and computing (ICMLC ‘19),
pp 224–230. Association for Computing Machinery, New York,
NY, USA

123