Vehicular Communications 35 (2022) 100469

Contents lists available at ScienceDirect

Vehicular Communications
www.elsevier.com/locate/vehcom

Task assignment algorithms for unmanned aerial vehicle networks:

A comprehensive survey ✩
Sabitri Poudel, Sangman Moh ∗
Department of Computer Engineering, Chosun University, Gwangju 61452, South Korea

a r t i c l e i n f o a b s t r a c t

Article history: Unmanned aerial vehicles (UAVs) have significant prospects in a plethora of public and civic spheres.
Received 22 September 2021 Recently, UAVs have focused primarily on applications where human presence is either impossible
Received in revised form 24 December 2021 or hazardous. A swarm of small UAVs can cooperatively complete operations more proficiently and
Accepted 14 March 2022
economically than a single large UAV. However, many issues must be resolved before stable and reliable
Available online 16 March 2022
multi-UAV networks can be realized. Task assignment in fleets of UAVs is concerned with cooperative
Keywords: decision-making and control. UAVs possess various functional abilities and kinematic constraints while
Unmanned aerial vehicle carrying limited resources onboard. UAVs are nominated to execute multiple sequential tasks supportively
Multi-UAV networks on numerous ground targets. The prime objective of task assignment is to minimalize the task
UAV autonomy accomplishment time and UAV energy consumption. To date, several task assignment algorithms have
Task assignment been designed for UAV networks, and they are comprehensively surveyed in this paper in terms of their
Task coordination main ideas, operational features, advantages, and limitations. These task assignment algorithms are then
Heterogeneity
compared in terms of their significant characteristics and performance factors. To the best of the authors’
knowledge, no survey on task assignment techniques for different UAV missions currently exists in the
literature. We also discuss open issues and challenges and then suggest projections for task assignment
algorithms concerning possible future directions.
© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction munication can be distinguished from other ad-hoc communica-

tion in the following ways: a) dynamic channel for air-to-air (A2A)
The remarkable and prompt growth of unmanned aerial ve- and air-to-ground (A2G) propagations; b) spatial as well as tempo-
hicles (UAVs) in many different applications has recently fasci- ral inconsistencies induced by high UAV mobility; c) shadowing ef-
nated many researchers and operators. The incorporation of UAVs fect due to design structure and rotation of UAV. UAVs of different
is noticeable in such applications, which are either dangerous or sizes and specifications are available. Generally, single-UAV-based
impossible for a human. UAVs can be deployed quickly and eas- applications prefer large UAVs with higher abilities, while small
ily, possess high maneuverability, are cost-efficient, have the self- UAVs are chosen for UAV formations and swarms. A multi-UAV op-
organizing ability, and are flexible and scalable. Hence, UAVs have eration scenario is illustrated in Fig. 1 and is expected to increase
received momentous consideration for civil and military applica- significantly in the future. As shown in the figure, several UAVs are
tions. The applications of UAVs include but are not limited to assigned different tasks and are supposed to execute tasks with
emergency and crisis management [1–3], search and rescue op- proper coordination with other UAVs.
erations in disastrous environments [4,5], monitoring agricultural
fields [6], post-disaster operations [7], remote sensing [8,9], wild- 1.1. Challenges of UAV networks
fire monitoring [10], traffic monitoring [11], freight transportation
[12], and relay networks [13–15]. In the present day, UAVs have
The progress made in microchip technology has helped widen
also shown wide use in pandemic environments to enforce social
the application range of UAVs [19]. Though the UAV network has
distancing [16], spread disinfectants in infected areas [17], carry
been promising, some challenges need to be resolved for effec-
testing kits and other medical supplies [18], and more. UAVs com-
tive UAVs based on reliable networks. Significant research has been
done in the literature that analyzes the challenges and prospects
✩
This work was supported in part by research fund from Chosun University, 2021. for successful UAVs operations. In [20,21], a study focusing on the
* Corresponding author. challenges faced by UAVs during deployment, interference, UAV-
E-mail address: [email protected] (S. Moh). based relaying, regulations, energy consumption, collision avoid-

https://doi.org/10.1016/j.vehcom.2022.100469
2214-2096/© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

on unsolved issues. Real-world UAV-based operations rely on col-

laboration between UAVs to manage the threat of attrition. Cooper-
ation is fundamental for planning efficient missions in profoundly
vulnerable environments. Efficient task allocation algorithms act as
they are designed and account for uncertainties in the dynamic
environment [33]. Task assignment is a combinatorial optimization
process by which a UAV or swarm of UAVs is allocated to accom-
plish many tasks. Fig. 2 shows how a task assignment algorithm
assigns a sequence of tasks within a target region to available UAVs
such that the overall cost of the task is minimized.
The task allocation process for UAVs has become an emergent
issue in the past few years with the widespread use of multi-UAV-
based mission planning systems [34,35]. A task manager is ac-
countable for identifying all tasks and decides the availability and
capability of UAVs to assign tasks to other UAVs proficiently. Sev-
eral task assignment algorithms have been studied in the literature
over several decades. Task assignments can be broadly categorized
into coordinated, distributed, stochastic, deterministic, evolution-
ary, and multi-fusion-based algorithms.
In this survey, we extensively investigated task assignment op-
timization algorithms designed for UAV networks to address their
key features and characteristics. In addition, we provide detailed
explanations of the basic operational principles. In the near fu-
ture, UAVs may require an ever-increasing number of higher-level
planning capabilities to accomplish their missions effectively. Their
missions are expected to be complex, necessitating the use of mul-
tiple heterogeneous UAVs to cooperate efficaciously while achiev-
Fig. 1. Multi-UAV operation scenario. ing the overall mission objective. The main contributions of our
study are discussed in the following subsection.
ance, and security is made. Various challenges regarding spectral
efficiency in multi-UAVs operations are reviewed in [22]. Differ- 1.3. Contribution of the study
ent issues faced in UAV communication, i.e., routing, handover, and
energy efficiency, are deliberated in [23]. Furthermore, some stud- UAVs essentially depend on sensors’ information to catego-
ies have focused on UAVs and UAV-based communications, and rize targets, discard false tasks, and determine comprehensible se-
state-of-the-art works consider UAVs’ particular domain or efficacy quences of decisions and actions to achieve their objectives. The
[24,25]. UAV flights can be understood as several degrees of au- significant contributions of our survey are as follows.
tonomy. Obtaining autonomy for UAVs is a complicated issue, and
the complication level depends on different phases in the decision- • Based on their main ideas, existing surveys on UAVs and their
making process and the necessary level of cooperation between applications are reviewed.
UAVs [26,27]. • The motivation behind the study of task assignment algo-
The maximum cooperation level is essential for task assign- rithms is elaborated on along with design considerations.
ment algorithms [28–30], where UAVs are expected to share in- • Existing task assignment optimizations for UAV networks are
formation, allocate tasks amongst each other, and consign respon- classified and compared based on their main ideas, advan-
sibilities with suitable timing and organization. Task assignment tages, limitations, and possible enhancements. No other earlier
can be described as a graph theory problem and explained using works have surveyed the task assignment algorithms proposed
deterministic search algorithms. Due to their computational com- for the single and multi-UAVs applications.
plexity, traditional deterministic algorithms can handle only small • State-of-the-art task assignment optimizations are extensively
and simple issues. Tasks of a heuristic nature can be solved, but compared and contrasted in terms of their principles, features,
suitable efficiency may not be achieved [31]. Proper communica- advantages, and limitations.
tion and exchange of information among UAV groups can improve • Task assignment optimizations are qualitatively compared with
UAV competence to encounter performance necessities associated various features and characteristics with a comprehensive and
with the speedy and trustworthy implementation of tasks. Appro- comparative discussion.
priate collaboration among UAVs is complex and very challenging • Finally, critical open issues and research challenges faced dur-
to implement. An optimization algorithm that considers task su- ing task assignments in multi-UAV systems are summarized
periority and synchronization, flightworthy trajectories, and other and deliberated for further enhancement and improvement.
constraints of real-world scenarios are highly desired. Furthermore,
UAVs differ in their operating abilities, and the cooperation of mul- 1.4. Organization of the paper
tiple heterogeneous UAVs can complement mission environments
with different types of tasks [32]. The remainder of this paper is organized as follows. Section 2
provides an overview of the state-of-the-art surveys conducted in
1.2. Motivation different fields of UAVs. Section 3 introduces and discusses addi-
tional considerations for UAV task assignment algorithms. Task as-
UAV communication imposes many challenges owing to the signments suggested for UAV networks are classified in to different
outstanding issues present in different constituents of UAV net- categories and extensively reviewed in Section 4. The task assign-
works, as discussed in the previous section. On the other hand, ment algorithms studied in Section 4 are qualitatively compared
these challenges motivate many researchers to focus their studies and discussed in Section 5. In Section 6, open issues, research

2
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Fig. 2. Task assignment diagram.

challenges, and future projections are summarized and discussed. geometric–stochastic models and reviewed extensively. A survey
Finally, concluding remarks are presented in Section 7. The organi- of state-of-the-art medium access control (MAC) protocols pro-
zational structure of the paper is depicted in Fig. 3. posed for UAV networks was presented from the viewpoint of main
features, working ideologies, benefits, and limitations [42]. An ex-
2. Overview of existing surveys on UAVs and our contributions tensive analysis of the latest use of UAVs in different IoT devices
is made, along with challenges faced and design requirements
Some surveys and tutorials for UAVs and UAV-based communi- [43]. Decentralized architectures of multi-layer UAV communica-
cations have existed in the literature of the past several decades. tion with prospects are discussed in [44].
Most of them have focused on the different issues faced in UAV
communication, such as energy and other resource efficiency, se- 3. Considerations for UAV task assignment
curity aspects, charging techniques, and channel modeling. Existing
surveys on various aspects of UAV and UAV-based applications are
Recently, UAVs have been widely deployed in many sectors due
summarized in Table 1 to validate the distinctiveness of our study.
to the unprecedented progress in drone and wireless technologies.
The authors in [36] reviewed the challenges encountered dur-
The proper deployment and operation of UAVs offers reliable and
ing the cooperation of UAVs with the Internet of Things (IoT)
effective solutions to a variety of real-world scenarios. UAVs can
devices. A detailed study of security for critical UAV applica-
be used as base stations (BSs) to collect and deliver on-demand
tions, such as denial of service attacks, man-in-the-middle at-
wireless communications in anticipated regions. In particular, UAVs
tacks, and de-authentication attacks, was presented in [37]. Be-
as flying BSs support and maintain connectivity between wireless
sides this, blockchain, software-defined networks, machine learn-
networks. In comparison to traditional ground BSs, by using UAVs
ing techniques, and edge computing have been studied as emerg-
as aerial BSs, network performance can be improved by adjusting
ing technologies. Different features and provisions anticipated for
their heights, avoiding static and dynamic obstacles, providing line-
networks of UAVs envisioned from the viewpoint of networking
of-sight (LoS) links to ground nodes, etc. [45,46]. Owing to the at-
and communications were reported in [38]. UAV applications are
tributes inherent in UAVs, such as maneuverability, flexibility, and
classified based on their communication necessities, such as search
adaptive altitude, they can excellently complement existing wire-
and rescue, construction, coverage, and delivering goods. Existing
less and ad hoc networks. It can be observed that there has been
UAV applications are categorized as short–long-range, high–low ca-
a tremendous increase in the number of UAV-based applications.
pacity, and real-time vs. delay tolerant. The exclusive features of
According to a TechSci study [47], it is expected that the tentative
UAV networks (such as high maneuverability dispersed UAV nodes
revenue from UAV-based applications will rise from 69 billion dol-
and frequently changing network topology) present challenges dur-
lars in 2018 to 141 billion dollars in 2023. Different aspects affect
ing network design and routing. As a result, the classification of
the task assignment process in UAVs. Hence, the factors influenc-
UAVs, the design of communications and applications, and an ex-
ing the UAV task assignment are briefly studied in the subsections
haustive review of the prevailing routing protocols proposed for
below.
UAVs were presented [39].
A comprehensive study of UAV systems from a cyber-physical
system (CPS) perspective was presented, considering three dif- 3.1. Single- and multi-UAV systems
ferent CPS components [40]. Different channel classification for
UAVs, measurement operations, and statistical channel models Previously, single UAVs were commonly used to achieve mis-
were studied and comprehensively surveyed [41]. It includes the sions [48,49]. Single and large UAVs are used in single UAV-based
following: (1) The UAV channel measuring operations in low- systems, which directly communicate with ground infrastructures.
altitude platforms (LAPs) created on narrow-band or wide-band However, any problems with that one UAV can terminate the en-
channels, low cost or low power channels, and broadly organized tire mission. Utilizing technological advancements and immense
ground infrastructures. (2) The experimental methods for A2G and research, the integration of multiple small UAVs has been widely
A2A propagation channels. (3) Finally, the channel modeling ap- considered recently. Multiple UAV-based systems add reliability,
proaches for UAVs are classified as deterministic, stochastic, and multi-tasking ability, and survivability to wireless communications

3
 S. Poudel and S. Moh
Table 1
Summary of existing surveys.

Reference Year Description Classification Challenges Topic Intelligent IoT Collision Flight UAV Net- Path Channel Application Mobility
of UAVs discussed decision avoidance control working planning modeling
[36] 2019 Presents an ample tutorial ✕ ✓ Smart cities ✓ ✓ ✕ ✕ ✕ ✓ ✕ ✓ ✕
on using the UAVs in
wireless networks

[37] 2021 A survey of security for ✕ ✓ UAV security ✓ ✕ ✕ ✕ ✕ ✕ ✕ ✓ ✕

critical UAV applications

[38] 2016 Surveys UAV-based civil ✕ ✓ UAV for civil ✕ ✕ ✓ ✕ ✓ ✕ ✕ ✓ ✓

applications application

[39] 2020 A survey of routing ✓ ✓ Routing in UAVs ✓ ✕ ✓ ✕ ✓ ✕ ✕ ✓ ✓

protocols for UAVs

[40] 2019 A survey of UAVs from a ✕ ✓ UAV networks ✓ ✕ ✓ ✓ ✕ ✓ ✓ ✓ ✓

cyber-physical system
4

perspective

[41] 2018 Surveys the methods of ✕ ✓ Channel modeling ✕ ✕ ✕ ✕ ✕ ✕ ✓ ✕ ✕

UAV channel modeling in UAVs

[42] 2021 Survey of MAC protocols for ✓ ✓ MAC for UAVs ✓ ✕ ✕ ✕ ✕ ✓ ✕ ✕ ✕

FANETs

[43] 2017 Overview of the ✕ ✕ UAV ✕ ✕ ✕ ✕ ✓ ✕ ✕ ✕ ✓

communication architecture communication
as well as routing protocols architectures
for UAVs is presented

[44] 2016 UAVs and UAVs related ✓ ✓ UAV based IoTs ✕ ✕ ✕ ✕ ✓ ✓ ✕ ✕ ✓

issues are discussed

Our work – Extensively surveys task ✓ ✓ Task Assignment ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

Vehicular Communications 35 (2022) 100469

assignment algorithms for in UAVs
UAVs along with task
assignment considerations
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Fig. 3. Organizational structure of this paper.

[50,51]. A comparison of single- and multi-UAV operations is given Table 2

in Table 2. Comparison of single- and multi-UAV operation.
Furthermore, multi-UAV-based systems can achieve tasks in Parameters Single-UAV operation Multi-UAV operation
much less time and efficiently. In multi-UAV systems, UAVs are as- Failure probability High Low
signed either homogeneous or heterogeneous tasks. In both cases, Survivability Very low High
each UAV requires a continually increasing number of high-level Antenna used Omni-directional Directional
System reconfiguration No Yes
directing and planning competencies to accomplish missions. Most
Complexity Low High
missions are very complex and need effective collaboration to Cost High Low
achieve mission objectives. UAVs mainly rely on information from Coverage Limited High
onboard sensors to effectively identify, classify, and select proper Coordination issues Very low High
Execution time Slow Fast
targets and determine an understandable sequence of choices.

3.2. Classification of UAVs titudes they can fly as high-altitude platforms (HAPs) and LAPs.
HAPs can fly above 17 km, are almost stationary, have high cov-
Depending on the application and mission objectives, an appro- erage, and are durable. Alternatively, LAPs have high mobility, can
priate UAV type must be designated to meet several requirements fly in the range of 10 m to a few kilometers, can fly for a few
and achieve the anticipated quality of service (QoS). UAV selection hours, and are cheaper. UAVs are also categorized based on type,
for a particular application must consider several factors, such as such as fixed-wing and rotary-wing. In comparison to rotary-wing
the capability of the UAV, maximum height it can fly up to, and UAVs, fixed-wing UAVs, such as small aircrafts have more weight,
available onboard devices. UAVs can thus be classified based on al- higher speed, and must move forward to remain airborne. In con-

5
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Fig. 4. Classification of UAVs.

trast, rotary-wing UAVs can hover and stay static over a given area ties while executing robust missions to indeterminate actions is an
[52,53]. An overview of UAV categorization based on types and al- important consideration in the design of task assigning algorithms.
titudes with their functions and capabilities is shown in Fig. 4. Meanwhile, evidence on the environment must be updated and re-
act to the substantial alterations in the mission environment while
3.3. Background study properly rejecting confrontational incorrect information. Neverthe-
less, cooperation and collaboration among agents are most likely to
A task assignment is a combinatorial and open-loop optimiza- achieve better network performance, which is challenging [63–65].
tion process that minimizes a predefined objective function by To boost escalating UAV applications and continue their work-
assigning single or multiple agents to multiple tasks. Task as- ing reliability, task assignment algorithms are compulsory in de-
signment is efficient if all the tasks are completed appropriately, liberating the concerns of UAV communications. The maximum
satisfying the constraints, and is fundamental in many multi-agent- collaboration level is required during task assignment, where UAVs
based applications. Only a few restrictions are optimized in most can promptly share information, tasks and allot tasks to other UAVs
existing research works, ignoring other factors that deteriorate net- with apposite task scheduling and arrangement. This level of coop-
work performance. In multi-agent systems, agents with distinct eration and information sharing complicates the autonomous task
capabilities and types are selected to deal with heterogeneous assignment process. Task assignment algorithms for a swarm of
tasks. While assigning tasks to the agents, the capabilities of agents UAVs have become the subject of much research, and many al-
can be an essential constraint. Another critical factor that deter- gorithms have been suggested for UAV task assignment. Hence, a
mines the efficiency of task assignments is time constraints. Most brief review of all state-of-the-art task assignment algorithms pro-
of the practical applications anticipate real-time task accomplish- posed for multiple UAVs is presented in this paper, which will help
ment. During task execution, risks and uncertainties present in the researchers and engineers to explore this topic further.
environment provide yet other conditions for efficient task assign- The symbols and acronyms used in this paper are listed along-
ment algorithms. side their descriptions in Table 3.
Task assignment has been studied for many other wireless net-
works in the literature. Satisfying decision theory [54] for bal- 4. Design issues of task assignment algorithms
ancing the necessities of the team with requests of individuals,
traveling salesman problem (TSP) [55], dynamic programming [56], Task assignment algorithms for UAVs face significant difficulties,
mixed-integer linear programming (MILP) [57], weapon target as- such as unacceptable computation time for real-time implemen-
signment (WTA) [58] are some optimization algorithms that have tation and degraded performance due to uncertain environmental
been applied to solve task assignment. Complex task assignment aspects. This section addresses the primary design issues of task
can be solved by using multi-objective genetic algorithms (GA) assignment algorithms for different mission planning systems for
[59], ant colony optimization (ACO) [60], reinforcement learning fleets of UAVs.
(RL) [61] and particle swarm optimization (PSO) [62]. However,
due to the inadequacies of constrained resources in agents, deci- 4.1. Multi-UAV system formation
sions are sometimes impacted, resulting in target misinterpreta-
tion. In such circumstances, the capability of agents to configure a Multiple UAVs should be systematically formed for the task as-
prompt solution is highly desired. The ability to integrate ambigui- signment in numerous UAVs. Developing a multi-UAV system is

6
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Table 3 terdependency of multiple tasks and trajectory optimization deter-

List of notations and abbreviations. mines the simultaneous operations of UAVs. To confirm the proper
Symbol Description dependence among UAVs, the coordination of path plans and task
Nv Number of UAVs assignments is required in real-time. Complexity is the focal aspect
Nt Number of targets in designing cooperative UAV missions [68,69].
X ik, j Decision variable in the range of 0–1 Along with the kinematic constraints of UAVs, factors such as
k Order of UAV
problem size and level of cooperation expected in different types
L ki, j Shortest path length
of mission operations induce complexities. Moreover, UAVs are ex-
f ik, j Waiting time
p i (t ) Survival probability
pected to operate in harsh and critical environments, where real-
S (u t ) Score of targets time service is desired. In such scenarios, UAVs must handle the
rt Remaining targets at time t uncertainties and constraints of dynamic environments. UAVs must
mt Remaining weapons at time t compute and make decisions at the local level to handle these ad-
ut Number of targets to be hit at time t
λ Time discount coefficient
versities, which add to the complexity of UAV networks. Recently,
ci j (xi , pi ) Score function cloud computing and edge computing techniques have been ex-
tu Time required to complete allocated task tensively used to address the computational complexities of UAV
Ck Minimal value at time k communication [70,71].
⊕ Exclusive OR
xk1 Decision variable at time k
k
N Lim Maximum flight path 4.3. Collision-free UAV operation
L kLim Maximum limit of number of tasks
CPS Cyber-physical system
LAPs Low-altitude platforms The existence of multiple UAVs in a network surges the chance
A2G Air-to-ground
of collision. Collision among UAVs can occur when UAVs come
A2A Air-to-air
MAC Medium access control across each other’s way during their operation when they lack au-
LoS Line-of-sight tonomy to handle obstacles and are time-constrained. Hence, one
BS Base station of the furthermost significant necessities in multi-UAV networks is
DE Differential evolution
that UAVs should not clash each other on their flight route and
QoS Quality of service
HAPs High-altitude platforms also their signal should not interfere with each other’s. A num-
MEC Mobile edge computing ber of collision avoidance mechanisms are studied and used in
SO System orchestrator the literature [72]. Mixed-integer based model is studied to ob-
VAIoTS Value added IoTs
tain the collision-free plan for multiple UAVs in [73]. Operating
EAUS Energy-aware UAV selection
DAUS Delay-aware UAV selection
UAVs in different altitudes is considered to avoid collision among
FTUS Fair tradeoff UAV selection UAVs in [74]. Clustering can help to minimize collision to some
IVHM Integrated vehicle health management extent. Another significant aspect in determining the collision-free
RUL Residual useful life UAV operation is the trajectory planning.
BCD Block coordinate descent
EAT Earliest available time
The existence of multiple UAVs in a network surges the chance
G2G Ground-to-ground of collision. Collision among UAVs can occur when UAVs come
TOT Time over target across each other’s way during their operation when they lack the
IoTs Internet of things autonomy to handle obstacles and are time-constrained. Hence,
FTA Fault tree analysis
BSUM Block successive upper-bound minimization
one of the furthermost significant necessities in multi-UAV net-
MILP Mixed integer linear programming works is that UAVs should not clash on their flight route and
TSP Traveling salesman problem their signal should not interfere with each other’s. Some colli-
WTA Weapon target assignment sion avoidance mechanisms are studied and used in the literature
RL Reinforcement learning
GA Genetic algorithm
[72]. Mixed-integer based model is studied to obtain the collision-
PSO Particle swarm optimization free plan for multiple UAVs in [73]. Operating UAVs in different
ACO Ant colony optimization altitudes is considered to avoid collision among UAVs in [74]. Clus-
DE Differential evolution tering can help to minimize collision to some extent. Another sig-
ANN Artificial neural network
nificant aspect in determining the collision-free UAV operation is
the trajectory planning.
required to accomplish a task cooperatively. Recently, the issues
faced during the formation of multi-UAV systems are getting sub- 4.4. Backhaul network
stantial research attention in many studies. The collaborative use
of a multi-UAV system offers many advantages over a single UAV
in large-scale applications, such as accomplishing complex and Robust backhaul is crucial when UAVs are not fully autonomous.
large-scaled tasks, improving the probability of successful task ac- A backhaul network collects and disseminates information for ef-
complishment even during the failure of some UAVs. Moreover, fective UAV cooperation and decision-making processes. Informa-
multi-UAV systems provide diverse competencies in various appli- tion includes the status of flights, sensed data, and flight control.
cation circumstances. However, designing multi-UAV systems and The limitations and requirements of backhaul links for the design
using them cooperatively to achieve the mission objectives brings and deployment of UAVs were studied in [75]. Similarly, in [76],
particular challenges and complexities [66,67]. backhaul and latency-aware UAV positioning with time complexity
were discussed to calculate the optimal height. An aerial back-
4.2. Computational complexity haul scheme was formulated to form a network with multi-hop
backhaul in the sky [77]. All these studies support UAVs to cre-
Having autonomous capacity allows UAVs to execute assign- ate backhaul in a distributed manner. Mm-wave is a recent and
ments with negligible or no human interference. Cooperative UAVs widespread technique for enabling broadband backhaul in UAVs
are expected to outperform the sum of individual UAVs. The in- [78,79].

7
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

4.5. Uniform load balancing the information about the tasks and proficiencies of the UAVs. Af-
ter assigning tasks, they are referred to the trajectory manager,
In multi-UAV operations, the uniform distribution of tasks which plans possible trajectories for each UAV. The results from the
among UAVs is necessary. Differential evolution (DE)-based load trajectory manager are given as a series of waypoints to each UAV.
balancing mechanism has been discussed in [80], where access- UAV controllers compute the definite controls desired to follow the
ing problem is modeled as a generalized assignment problem. waypoints using these path coordinates. The health management
Grounded on the indication of task load within a specific region of issue is focused on using a response mechanism of the perfor-
task execution, task assignment can be done approving even dis- mance model used by the task assignment algorithm.
semination of tasks as far as possible. Another technique for load The developed health-aware task assignment algorithm was
balancing is to assign a group of arbitrarily created non- intersect- validated to be operative through simulation and real aeronautical
ing coordinates. The load balancing technique using multi-criterion experimentations. The preliminary outcomes were very favorable;
decision-making is studied [81]. The optimal user association to nevertheless, many can be achieved in the health management sec-
balance load using hybrid cognitive radio relay [82]. Task assign- tor, accounting for sensor performance and control actuator failure
ment methods can be helpful for load balancing in UAV-assisted modes. Additionally, a robust performance in the face of uncer-
IoT communication [83,84]. tainty was achieved.
Advantages: Improves operational reliabilities and capabilities
4.6. Antenna designs through adaptive task assignment systems and system awareness;
demonstrated effective performance through simulation and exper-
iments.
The design and number of antennas also determine the per-
Disadvantages: Randomly occurring failures provide a level of
formance of UAVs, and hence they must be decided cautiously.
uncertainty; uncertainty present at all planning stages due to in-
UAVs essentially incorporate directional and omnidirectional an-
complete knowledge at the flight plan level, health of sensors
tennas. In short, omnidirectional antennas perform better in en-
during task assignment, and enduring maintenance at the mission
vironments with high mobility but face energy drain and security
planning level.
issues. In contrast, directional antennas have an improved trans-
Possible enhancements: The amount and quality of feedback in-
mission range but require complex protocols and exact antenna
formation can be improved, and a sophisticated stochastic model
orientation algorithms [85,86]. Also, different issues brought about
of health states like energy usage and sensors’ performance can be
by directional communication, such as head-of-line, deafness, and
embedded to deal with uncertainties.
hidden terminals must be addressed. UAVs also use smart and
adaptive antennas.
5.1.2. MILP-TA
The issues of task assignment and optimal formulation to solve
5. Task assignment algorithms combined multi-assignment for a widespread search munition area
were addressed in [91]. MILP can assign infeasible tasks by adding
The task assignment of UAVs is performed to minimize the en- time to UAV paths due to timing constraints. It makes use of the
tire cost by assigning UAVs to accomplish a number of tasks. A discrete approximation of real-world scenarios. Ammunitions are
UAV can be assigned a single task or more. The issues faced by essential to explore, categorize, attack, and confirm the demoli-
task assignment algorithms are computational complexities, task tion of achievable goals. Information on the target area is assumed
coupling, problem size, time constraints, and heterogeneity. Keep- to be communicated between all elements of the UAV swarm. A
ing this in mind, different task assignment algorithms have been formulation based on MILP consists of optimization function, up-
designed for different application-specific UAV operations. We clas- per bound and lower bound on variables and constraints using the
sify task assignment algorithms into four categories: centralized, variables. A UAV is allowed to visit any target only twice to pre-
distributed, bio-inspired, and multi-fusion, as illustrated in Fig. 5. vent looping. UAVs can visit a sink once to find new targets before
The basic operating principles of each algorithm type are briefly reassignment occurs. This helps to avoid inconsistencies in UAVs
discussed along with their advantages, disadvantages, and possible entering and leaving the sink. Existing path planning and lengthen-
improvements in the subsections below. ing algorithms were detailed in [92,93]. Through this formulation,
the UAV flight path varies to ensure that the timing constraints are
5.1. Centralized task assignment algorithms fulfilled and varying task completion times are integrated.
Advantages: Useful in offline task assignment calculations; pro-
Centralized task assignment algorithms require a central plan- vides optimal solution for UAV groups with combined tasks with
ner that gathers information from all UAVs, calculates the optimum timing and task order limitations.
strategy, and passes that information among the UAVs. The central Disadvantages: Makes a discrete representation of real-world
planner can be a ground station receiving information from all the problems, which makes solutions impractical; solution requires
UAVs, calculating optimal plan, and informing all UAVs about the high computation time and makes unreliable for real-time use.
plan. In some cases, one of the UAVs can also act as a planner. In- Possible enhancements: Kinematic constraints of UAVs should
formation sharing is straightforward in centralized task assignment be focused; aspects of complex and dynamic environment such as
schemes, where each UAV communicates with the central planning obstacles, uncertainties, and wind speed must be considered.
agent. Some cooperative algorithms proposed for task assignment
are [87–89]. 5.1.3. Modified Two-Part Wolf Pack Search (MTWPS)
MTWPS, a combinatorial optimization model-based task assign-
5.1.1. Health-Aware Task Assignment (HA-TA) ment using graph and optimization methods, was studied in [94].
A robust decision-making process that improves the group’s An easy computing function is used for large UAVs and target
functioning reliability and competencies of distributed and self- sizes, and to solve the time-sensitive uncertainty; a practical on-
directed UAVs using improved self-awareness systems and adaptive line hierarchical planning algorithm is used. MTWPS includes tra-
mission planning was studied in [90]. The task planner manages ditional offline centralized situations and online capability with
the list of tasks in the task assignment component, which decides time-sensitive uncertainty. A number of UAVs are assigned to clas-
and selects existing UAVs, which can perform tasks depending on sify, reply, and verify tasks on targets sequentially. UAVs have

8
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Fig. 5. Classification of UAV task assignment algorithms.

different flight heights to avoid a collision. The combinatorial opti- Advantages: Deals with online uncertainties such as communi-
mization problem is modeled as follows: cation issues, UAV malfunction, and time-sensitive target problem;
easy computing used to lessen simulation time for large-scale UAV
+3N t N v
N v +3N t
N networks.
min P k=v 1 P k,mod(( j − N v ),3)
{ X ik, j }
Limitations: A UAV’s waiting time depends on the performance
i =1 j =1 (1)
of other UAVs, which makes waiting time complicated for large-
  
× Lki, j + f ik, j X ik, j X ik, j scale networks; the deadlock problem is ignored.
+3N t 
N v Nv
Possible enhancements: Different types of deadlocks require
better handling, which can be focused on in the future; threats
X ik, j = 1, ∀i
and other risks present in the stochastic environment should also
j =1 k =1
s.t. , (2) be focused on.
+3N t 
N v Nv
X ik, j = 1, ∀j 5.1.4. Dynamic Task Assignment (DTA)
i =1
k =1 A mission planning and task assignment framework for the co-
N v+3N t N v+3N t ordination of self-controlled UAVs to improve the robustness and
 k  
L i, j + f ik, j X ik, j X ik, j < L kLim , ∀k (3) scalability of multi-UAV operations was studied in [96]. An outline
i =1 j =1 of the proposed framework is shown in Fig. 6. The main goal of
+3N t N v
N v +3N t the battlefield scenario is to eliminate enemy agents; hence, the
X ik, j ≤ N Lim
k
, ∀k (4) maximum mission time includes search and attacks. For proba-
ble actions to attack enemies, n agents are used, and each agent
i =1 j =1
is given a specific number for identification, i.e., agent_id. Agent
where N v and N t are the numbers of UAVs and targets, respec- information includes position coordinates, energy level, and pay-
tively, and X ik, j is a decision variable ranging from 0 to 1, which load status. However, enemy agents do not have prior knowledge
suggests that the kth UAV travels from the ith point to the jth of the size and location of UAVs. Because of the energy constraints
point to complete task l = mod(( j − N v ), 3). L ki, j and f ik, j are the in UAVs, a cost-effective approach is required to allocate tasks to
length of the shortest path and waiting time of the UAV flying an optimal UAV set. The cost of every attack action is determined
k
from the ith to jth point, respectively. N Lim and L kLim are maximum using the position of the identified target, the position of available
flight path and maximum limit of the number of tasks for kth UAV. agents, available battery, and available payload.
Constraint (2) ensures UAV visits the place once only. Constraints Advantages: Time efficiency, scalability, and robustness are
(3) and (4), on the other hand, control the flying distance and the achieved; achieves the mission goal with minimum cost.
number of tasks, respectively. Discrete wolf pack algorithm-based Limitations: Does not consider the uncertainties and aspects of
task assignment for UAVs is discussed which improves the conver- dynamic environments; preliminary information regarding the size
gence speed [95]. and location of enemy UAVs is unknown.

9
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

based on hierarchical coordinated control. Different heights are

considered for UAVs to avoid collisions. Satisfying decision the-
ory [54], control settings, [114,115], and multi-agent interactions
[116] are used. Two levels are considered for UAV cooperation.
There should be negotiation with UAVs and target assignment vec-
tor at a higher level to allocate UAVs to targets. After that, UAVs
of each team coordinate to identify feasible team time of target
(TOT). Coordination of TOT is resolved through the encapsulation
of important myopic information for optimization.
Advantages: Decomposition of motion planning issues helps
obtain a near-real-time solution; risk is mitigated by increasing
distance between UAV and threats.
Limitations: Tracking error increases due to disturbances; only
static threats are considered, and thus the results obtained are sub-
optimal.
Possible enhancements: Timing of target intersection can be
considered; aspects of the dynamic, complex 3D environment such
Fig. 6. Overview of mission planning and task assignment. as wind speed, risks need consideration.

5.1.8. Energy-Efficient Task Assignment (EETA)

Possible enhancements: Minimizing the combination of all costs
is desired, requiring a solver algorithm; threats and uncertainties To minimize the power consumption of ground devices and
of the complex dynamic environments need consideration. UAVs, assigning tasks and allocating resources were studied for
UAV-assisted mobile edge computing (MEC) systems [117]. Energy
5.1.5. Stochastic Multi-Criteria Acceptability Analysis (SMAA-2) minimization for IoT devices and UAVs through combined opti-
In [97], a group of UAVs was used to complete attacking tasks mization of trajectory, appropriate task assignment, and resource
on ground targets in an uncertain environment. UAVs with differ- assignment. It is quite challenging to solve the non-convex struc-
ent capabilities were allotted tasks to attack before starting the ture imposed by the proposed system using conventional convex
mission. Uncertain information led the criteria values of the task optimization. Hence, block successive upper-bound minimization
assignment process to be random, and the criteria weights were (BSUM) is used.
not known accurately. A novel task assignment process that re- Advantages: Lessens energy depletion, and optimizes task of-
lies on SMAA was presented in some previous works [98–111] to floading, resource sharing, and trajectory without increasing delay;
deal with the abovementioned problem. Target attack tasks were minimizes UAV’s propulsion and computation power.
assigned to multiple UAVs. Limitations: Single UAV is used, which increases the risk of
Advantages: Uncertainties are analyzed; finds satisfactory solu- mission failure.
tion even in severe situations. Possible enhancements: UAVs’ kinematic constraints such as ex-
Disadvantages: Sensor error or target movement leads to un- ternal forces, motions, and physical limitations of the dynamic en-
clear localization; probability of UAVs being destroyed while han- vironment such as threats and uncertainties must be considered
dling attacks remains. during the assignment process.
Possible enhancements: Threats present in the defense system
of targets can be made more apparent; a confidence factor is in- 5.1.9. Energy and Delay Aware Task Assignment (EDATA)
troduced to judge the accuracy of criteria values. A robust system orchestrator (SO) to achieve value-added IoTs
(VAIoTS) attained by incorporating UAVs with IoT payloads was
5.1.6. Filter-Embedded UAV Task Assignment (FEUTA) presented in [118]. The flight of UAVs is organized using a central,
The classical task assignment problem was modified to form co- SO that holds the complete information of UAVs, current positions,
ordinated teams of UAVs to reduce the effect of noise in SA on the energy levels, flight operations, and their onboard IoT devices. SO
solution [112]. The degree of change during the reassignment pro- employs energy-aware UAV selection (EAUS), delay-aware UAV se-
cess was limited appropriately. The measured approach performs lection (DAUS), and fair tradeoff UAV selection (FTUS). EAUS aims
reassignments at the degree of updated knowledge to immediately to reduce the total energy depletion by the UAVs, DAUS seeks to
react to substantial fluctuations observed in the environment. Ad- decrease the UAV time of operation, and FTUS aims to find a fair
ditionally, FEUTA embeds a sophisticated filtering operation during solution between the two conflicting objectives of energy and time.
the task assignment process relying on modified weight coeffi- The power and time required for traveling, sensing, processing, and
cients that depend on environmental variations. data transmission are considered and optimized.
Advantages: Condenses the consequence of noise caused due to Advantages: Energy consumption, operational time, and fair-
high-frequency on the planner; mitigates the effects of noise. ness of the system are optimized jointly; environmental effects on
Limitations: Ignores the aspects such as threats and uncertain- flights and performance are considered.
ties, which makes it unreliable for the real environment; accepts Limitations: GPS localization errors and signal losses exist; in-
noise entirely. crease in execution time for FTUS, which impacts the performance
Possible enhancements: The importance of robust methods pro- of the model.
posed for FEUTA and their connection to noise rejection algorithms Possible enhancements: Cases of accidents and UAV failures
must be investigated. must be considered; the modeling part could be improved con-
cerning the energy consumed by GPS in errors and signal losses.
5.1.7. Coordinated Target Assignment and Intercept (CTAI)
Cooperative control solutions assign UAVs to pass through tar- 5.1.10. Dynamic Programming-based Cooperative Task Assignment
get locations known to have many threats [113]. The complete (DP-CTA)
problem is divided into a target assignment, coordinated UAV, path DP-CTA, which accounts for cooperation among UAVs rather
planning, practicable trajectory design, and asymptotic trajectory than just coordination in adversarial environments, was studied

10
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

[119]. As discussed in [120–122], DP-CTA relies on weapon tar- 5.2. Distributed task assignment algorithms
get assignment, where weapons are assigned to targets for opti-
mizing the mission objectives. Two approximation methods of DP Centralized task assignments are sometimes not feasible for
have been developed for large-scale networks. A one-step method UAV swarms due to underlying communication restrictions, robust-
is responsible for generating a cooperative solution quickly, and ness concerns during inconsistencies, and scalability. Computation
a two-step method generates an optimal solution with minimum requirement gets impractical once the number of agents grows,
possible computation time. Using a one-step look-ahead at stage which is also a limiting factor for centralized algorithms. Dis-
and state, maximizes the expression as tributed methods can be helpful in such situations to minimize
these complications. Distributed task assignment algorithms pri-
  marily focus on market-based procedures, i.e., contract network
max S (ut ) + λ J t +1 (rt − ut , mt − |ut |) ,
ut ,|ut |≤mt (5) algorithms and auction algorithms [124–127]. However, decen-
t ∈ {0, . . . , N − 1}, tralized algorithms require proper coordination to improve over-
all performance [128–131]. The fundamental problem is that ac-
where S (ut ) is the score of targets, J t +1 is an estimate of func- complishing harmonization requires UAVs to interchange large
tion J t∗+1 which is defined as optimal expected score in Bellman’s amounts of environmental information, existing states, and forth-
equation, and p i (t ) represents the survival probability. In addition, coming purposes. Communication on this scale is not always pos-
rt , ut , and mt denote the remaining targets, the number of targets, sible and increases UAVs’ visibility to threats. Some recent decen-
and remaining weapons, respectively. N is the number of stages, tralized approaches to overcome the issues of centralized task as-
|ut | is size of ut and λ is time discount factor. In two-step look- signment on each UAV are discussed in the following subsections.
Information sharing is a vital part of decentralized task assignment
ahead at stage t and state (rt , mt ), control ut is used to obtain the
algorithms.
maximum as
  5.2.1. Auction-based Multiple Constraints Task Assignment (ABMC-TA)
max S (ut ) + λ J̃ t +1 (rt − ut , mt − |ut |) ,
ut ,|ut |≤mt (6) An auction-based task assignment algorithm for assigning dy-
t ∈ {0, . . . , N − 1}, namic tasks to UAVs using multi-layer cost calculation was studied
in [132]. The cost computation method divides the computation
where J̃ t +1 is obtained from the one-step method. cost into four layers concerning the four constraint types. UAVs
Advantages: Generates cooperative and optimal solution quick- cooperate in a surveillance task considering four features: sensor
ly; recovers planning of real-world air operations providing im- type (C), field of view (F), lowest distinct pixels (P), and resolu-
provements. tion (R). UAVs that can meet these features are selected when a
Limitations: Computation time grows with the number of tar- new task appears. In addition, the task has a specific time window
gets, and the uncertainties of real complex environment are ig- that specifies the start time for the earliest, latest, and total time
nored. required to accomplish that particular task. An auction-based task
Possible enhancements: Constraints of the real dynamic envi- assignment algorithm is studied in [133], where UAVs are deployed
ronment such as sudden threats, effects of wind, and temperature as traps.
on UAV flights must be focused on. Advantages: Solves multi-constraint task assignment issues ef-
ficiently considering the constraints of sensors, time, danger, and
5.1.11. Prognostics and Health Monitoring (PHM)-based Multi-UAV Task fuel.
Assignment (PHM-based MTA) Limitations: 2D environment with static obstacles considered
Applying integrated vehicle health management (IVHM), based for implementation which cannot justify the constraints of real
on PHM for multiple UAVs, was considered in [123] to achieve dynamic environment; kinematics of UAV such as motion and im-
objectives with negligible spontaneous disruptions. IVHM is the pacts of external forces on its flight path needs consideration.
ability of the system to evaluate current as well as future system Possible enhancements: Complexities imposed by real 3D en-
health state and incorporate that with the structure of existing re- vironment need to be considered to confirm its practical validity;
sources and demands. PHM is the capability to measure different more experimental study and investigation are required to verify
states of health, forecasting forthcoming failures and predicting the the efficiency of the proposed method in real environment.
probable RUL of the system based on measurements. Information
from PHM consists of residual useful life (RUL) approximations. 5.2.2. Robust Decentralized Task Assignment (RDTA)
System-level RUL is designed for decision support to assign a group Decentralized task assignment for a swarm of cooperative UAVs,
of UAVs to weighted tasks. RUL is determined through fault tree which performs well for sparse networks, was studied [134]. A set
analysis (FTA), which is then served by a distribution function from of waypoints is identified for UAVs, and a matrix gives their loca-
the probability distribution function concerning time and proba- tions. A vector provides the waypoint values. A matrix character-
bility of failure in critical scenarios. FTA is a widely used failure izes the capabilities of UAVs; signifies a UAV that can handle tasks
analysis technique where graphs explore potential causes of unde- with waypoint w. Two significant phases, information consensus,
sired states. and planning, are used, as shown in Fig. 7. In the first phase,
Advantages: Applies system-level RUL methodology to accom- UAV communication improves the consistency of the data. How-
plish task assignment including failure probability. ever, due to limitations on communication resulting from noise
Limitations: It may not be feasible to know the probability of and slow convergence, consensus cannot be reached. In this case,
failure in advance; threats and uncertainties of real stochastic en- the second phase, i.e., planning, must be used. RDTA uses a modi-
vironments are ignored. fied planning phase to eliminate potential conflicts [135]. In stage
Possible enhancements: PHM-based TA for extensive haul task 1, UAVs use updated information to generate a set of candidate
reaction of algorithm during failure requires further investigation; plans. Then, the petal algorithm is used, as discussed in [136]. In
consideration of aspects of real environment such as wind, tem- stage 2, UAVs develop the ultimate non-conflicting goal based on
perature, threats is a must. the candidate plan.

11
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Fig. 7. Robust decentralized task assignment algorithm.

Advantages: “No Fly Zones” are considered; a modified candi- the bio-inspired algorithms into swarm intelligence-based algo-
date plan selection algorithm is used to improve performance and rithms, GA-based algorithms, and artificial neural network (ANN)-
increase the robustness to the inconsistencies. based algorithms. On the other hand, evolutionary algorithms are
Disadvantages: Does not consider the heterogeneity of UAVs stochastic methods used to find optimal solutions, and they have
and targets and the constraints of the dynamic environment such been applied in a variety of optimization problems [140–142]. Mo-
as threats and effects of external forces (wind, temperature, etc.) tivated by the nature of biological organisms, individuals learn and
on UAVs performance. adapt to environmental situations for survival. The fitness function
Possible enhancements: Conflicts arising due to low conver- evaluates and decides which individuals are suited for the next
gence need to be eliminated; heterogeneity of UAVs and other generation in each iteration.
environmental aspects such as threats and uncertainties must be
considered. i. Swarm intelligence-based task assignment

5.2.3. Consensus-based Bundle Algorithm (CBBA) Swarm intelligence is the communal behavior of self-organized
A conflict-free task assignment to resolve two real UAV operat- and distributed particles generally witnessed in nature. These al-
ing complications, i.e., collision-free paths and churning behavior gorithms deploy a search-focused approach, where every particle
of UAV flight path, was considered in [137,138]. CBBA was ex- operates independently. Later on, they collaborate with surround-
tended to account for obstacles and mitigate churning to reduce ing neighbors to explore the environment. Two different phases
are involved: exploration and search. While exploring, particles
the algorithm’s sensitivity to noise with the minimum computa-
sense data and then broadcast the sensed data to their neigh-
tional burden. For a given task and agent, CBBA task assignment
bors through different communication channels. The data are then
can be derived as
received by other agents of the swarm. In the search phase, ev-
N  ery agent uses its specific data and the data obtained from its

Nu  t

max c i j ( xi , p i ) xi j (7) neighbors to discover a favorable direction to travel. ACO [60] and
i =1 j =1 PSO [62] are some commonly used swarm intelligence-based algo-
rithms for task assignment in the literature.
where xi j = 1 if UAV i is assigned to task j whereas xi ∈ {0, 1} Nt
is the vector with the jth element being xi j , and ci j (xi , pi ) is the 5.3.1. Earliest available time with PSO (EAT-PSO)
score function of assignment xi and path pi. CBBA based two-layer A task assignment algorithm for UAVs to operate in an in-
task assignment for swarm of UAVs is also studied in [139]. door environment has been studied [143]. A mathematical problem
Advantages: Handles operational complexities of multi-UAVs; model and a heuristic approach to solve the problems faced by
provides reliable, conflict-free UAV task assignments. indoor UAV operation are suggested. For real-time actions, rapid
Limitations: All targets may not be discovered; dynamic threats response to indeterminate actions requires a practicable schedule.
and uncertainties are ignored. EAT is incorporated with PSO to find the optimum plan with a
Possible enhancements: Kinematic and dynamic UAV con- concise calculation time. Indistinguishable UAVs with cameras and
straints on air and ground can be incorporated; UAVs need to be substance handling tools capable of handling the desired tasks in
penalized for massive heading changes. an indoor environment were studied. UAVs can take more than one
task during flight. The framework in [144] was upgraded by fitting
5.3. Bio-inspired task assignment algorithms its solvability. The scheduler allocates tasks to UAVs, and the time
to start a consigned task for every UAV, flight path, hover time,
Bio-inspired algorithms originate from mimicking biological be- waiting time, and recharge schedule are all planned. PSO-based co-
havior based on their analyzing ability to deal with problems. operative task assignment and path planning is presented in [145].
These algorithms leave out the process of building difficult envi- Advantages: Achieves a minimum total make span, forms a
ronmental models, and recommend a strong searching algorithm to seamless UAV operation schedule, and short computation time in
meet the objective firmly. In our study, we have further classified finding a high-quality feasible solution.

12
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Limitations: Allows a single UAV to occupy a position at a par- The central idea of GA is the survival of the fittest using good
ticular time, ignoring uncertain events and obstacles. genetic factor (i.e., efficient assignment) to the succeeding genera-
Possible enhancements: An anti-collision refinement phase can tion. It also produces new search space combining different genes.
be incorporated, a detailed study on the distributed control of A GA-based approach for assigning processing tasks to engines
UAVs can be conducted, and the robustness of the system can be spontaneously is studied in [153]. Some other task assignment al-
assessed. gorithms using GA are studied as well [154–157].

5.3.2. Orchard Picking Algorithms (OPA) 5.3.4. Cooperative Multiple Task Assignment Problem with GA
Influenced by the garnering proficiency of farmers, a task as- (CMTAP-GA)
signment for multiple UAVs was presented in [146]. The nearest The combined issue of task assignment and path planning in
neighbor, which is deliberated with minimum possible neighbor- the form of a graph was presented in [158]. The CMTAP is sug-
ing distance as a pointer, rapidly resolves the optimum arrange- gested for circumstances concerning a group of different types
ment of numerous tasks for collective accomplishment. Every tar- of UAVs. A Dubins car model-based motion planning is used to
get that must be destroyed necessitates various ammunition, and consider each vehicle’s specific constraint’s minimum turn radius.
a collaborative task assignment model for heterogeneous UAVs is Consuming a determinate set for defining the visit angle of a UAV
established accordingly. Three different types of task sets, i.e., re- over a target, the combined problem of task assignment and path
connaissance (D), attack ( A), and evaluation (V ), are considered. optimization was posed as a graph. This novel method results in
Specifies the load task set of UAVs, where subscripts T and U in- suboptimal route assignments. Refining the visitation angle dis-
dicate belonging to the task and UAV, respectively. Depending on cretization results in an upgraded solution. Owing to the computa-
the geometric relation between two triples, i.e., the ammunition tional complications of the subsequent combinatorial optimization
required to destroy the target entirely and the capacity of UAVs to problem, GA for stochastic exploration of the space of solutions
load this ammunition, it is decided whether tasks are fully exe- was proposed. Two different cases of UAVs group configuration
were illustrated: homogeneous, where all UAVs are indistinguish-
cuted. If and denote the sets of targets and UAVs, respectively, and,
able and heterogeneous, with varying competencies of working
where is the number of types of missiles, the total number of type
and kinematic limitations.
missiles for targets to be destroyed is given as
Advantages: Promptly offers feasible task assignment solutions

Nt
q
and real-time implementation for high dimensions; considers task
N qa = Nj, q = 1, 2, .., N a . (8) priority and coordination, time constraints, and trajectory.
j =1 Disadvantages: Performance degrades if the search space is
harsh; it ignores the aspects of real dynamic environments, which
Advantages: Works with short execution time, highly flexible, makes the proposed method impractical in real-time implementa-
strongly robust, and scalable; quickly finds optimal solution; avoids
tion.
difficult cost function based on the tasks, and has high calculation
Possible enhancements: Proper synchronization must be con-
efficiency.
sidered to maintain coordination among UAVs; constraints of the
Limitations: Does not consider the constraints of the actual bat-
dynamic environment such as threats, uncertainties, and external
tlefield; threats and uncertainties that may arise in the dynamic
forces and their impacts on UAV’s motion need to be investigated.
environment are ignored.
Possible enhancements: Adaptive approach of cooperative task
assignment can be introduced to balance the load of UAVs and 5.3.5. Cooperative Multiple Task Assignment Problem (CMTAP)
adapt to real-time battlefield situations. Assigning cooperative UAVs for multiple tasks on multiple tar-
gets was proposed as a combinatorial optimization problem [159].
5.3.3. Cooperative Task Assignment (CTA) The authors used GA to account for distinctive necessities, such as
Cooperative task assignment for UAVs in obstacle-based envi- task preference and harmonization, timing constraints, and flight
ronments was presented in [147]. The upgraded ACO-based opti- restrictions. A matrix exemplification of GA chromosomes simpli-
mization and Hungarian algorithm are collectively used for multi- fies the encoding procedure and the application of genetic opera-
UAV formation. A battlefield environment is considered, where tives. The set of tasks is performed by the teams of the UAVs on
UAVs fly from different positions, search targets in that area, and each target, i.e., classify, attack, verify. Each task requires execution
destroy them. If xi , j ∈ [0, 1] is the decision variable, xi , j = 1 when after a successful target classification from the given observation
the UAV is assigned a task, and xi , j = 0 otherwise. Cooperative task ability. It is assumed that the probability of achieving the assigned
assignment and path planning for multi-UAVs using K-means algo- task if the physical requirements are met sums to one. If not, the
rithm and hybrid optimization algorithms is also studied in [148]. assignment algorithm is conducted again.
Advantages: Efficiently solves considerate task assignment Fig. 8 presents a graph representation of the CMTAP solution
problems of UAVs under multiple circumstances and constraints; with two UAVs, U = {U 1 , U 2 } and two targets T = { T 1 , T 2 } to per-
fast convergence and a highly stable iterative curve are assured. form three tasks N m = 3. The green line indicates the path of the
Limitations: Focuses on flight length only and ignores other first UAV, U 1 , starting from the initial point, visiting the target,
constraints of task assignment; threats and uncertainties of the T 1 , moving ψ = 120◦ , and completing its assignment visiting T 2 .
real environment are ignored, making the solution impractical. However, the purple path shows the routes of U 2 visiting T 2 with
Possible enhancements: Estimation of cost function could be ψ = 0◦ and proceeding towards T 1 , returning to T 2 with ψ = 240◦ ,
adjusted considering the flight mission; in an actual battlefield en- and flying to T 1 with ψ = 240◦ .
vironment, target values vary, and distribution of UAV firepower Advantages: Solves computational complexity for classical com-
must be considered. binatorial optimization methods with higher efficiency and consid-
ers task priority and timing constraints.
ii. GA-based task assignment Disadvantages: Does not consider the dynamic aspects of the
environment such as impacts of threats, uncertainties, and external
Recently, GA has a wide range of applications and it is demon- forces on UAVs flight; heterogeneity of tasks, targets, and UAVs are
strated to be appropriate for task assignment problems [149–152]. ignored.

13
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Fig. 8. Solution of CMTAP.

Possible enhancements: Moving, sudden targets, and actual en- Limitations: Does not cover practical mission scenarios such as
vironmental scenarios must be considered; computational effi- weather constraints, radar, terrain barriers during simulation.
ciency needs to be realized to achieve a real-time solution. Possible enhancements: Distributed model with more stochastic
features can be considered to cover practical mission scenarios; a
5.3.6. Modified GA-based Cooperative Task Assignment (MGA-CTA) comparison of a centralized and distributed algorithm for stochas-
Task assignment for heterogeneous multiple UAVs, concerned tic task assignment needs to be made.
with cooperative decision-making and control, was studied in
[160]. In comparison to previous GA-based algorithms, genetic fac- 5.3.8. Battle Antennae Search with GA (BAS-GA)
tors of chromosomes were distinct with regard to the tasks to be The BAS algorithm-based multi-UAV task assignment using an
accomplished on targets in this work. In addition, a mirror demon- improved GA was discussed in [162]. First, the target sequence is
stration of UAVs for expressing inadequate resources of UAVs was
searched using the BAS algorithm, and then double-crossing oper-
used. Information on UAV resources is maintained and updated in
atives are used to increase the multiplicity of the target arrange-
a timely manner to process distinct types of genes. A task assign-
ment to find the optimal solution. Finally, dynamic adjustment
ment that assigns multiple UAVs to several tasks on various targets
of mutation probability is considered to improve the local search
is a possible solution.
ability to avoid local optima. Homogeneous UAVs are assigned to
Advantages: Considers and addresses the heterogeneity of
accomplish tasks within a target area. Three main types of tasks
UAVs, targets, and environments; mirror representation to deal
are considered, i.e., Task = {Reconnaissance, Attack, Verify}.
with the limited resources.
Advantages: Diversity of searching capacity improves conver-
Limitations: Only one kind of resource, i.e., weapon, is con-
gence time; local search ability is increased; has good convergence
sidered; stationary targets considered in simulation, weapons are
from small to large missions.
used in one way only.
Limitations: Does not consider the kinematics of UAV which
Possible enhancements: Different operational and kinematic
impacts on performance and constraints of the realistic dynamic
constraints of UAV must be considered; aspects of real dynamic en-
environment such as threats and uncertainties.
vironments such as threats, uncertainties, and other external forces
Possible enhancements: Rapid task adjustment using central-
and their impacts need to be analyzed.
ized and distributed methods can be experimented on; integra-
tion of path plan into task assignment for broad task planning for
5.3.7. Cooperative Multiple Task Assignment Problem with Stochastic
multi-UAVs can be done.
Velocities and Time Windows (CMTAPSVTW)
A unique improved GA is anticipated to acquire a competent so-
lution for realistic mission environments [161]. UAVs with superior 5.3.9. Opposition-based GA Using Double-Chromosomes Encoding and
abilities and some limitations are considered for task consignment. Multiple Mutation Operators (OGA-DEMMO)
The model of Dubins car is accepted to create UAV flight paths Multi-UAV-based reconnaissance for heterogeneous targets us-
along with two extra flights to coordinate the paths and create ac- ing GA to optimize the task sequence of UAVs was considered in
tual paths. [163]. Targets are classified as line targets, point targets, and area
Advantages: Considers kinematic constraints (motion, exter- targets, considering the features of the target geometry and view of
nal forces, propulsion), resource constraints (fuel and onboard the sensor’s field. To solve the computational issues, OGA-DEMMO
weapons), and time constraints (time windows and sequence of was developed to enhance the variety of the population to improve
tasks). the global exploration ability. OGA-DEMMO aims to find the best

14
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Fig. 9. Q learning-based FTA.

task assignment to UAVs, maximizing overall performance effec- Disadvantages: Considers only stationary targets; complexities
tiveness. and aspects of the real battlefield environment are ignored.
Advantages: Minimizes the execution time and total UAV con- Possible enhancements: Moving and sudden targets and com-
sumption and improves the optimality of the algorithm and con- plexities of real battlefield environmental scenarios must be con-
vergence energy. sidered.
Limitations: No constraints are considered for reconnaissance;
if the sensor’s field of view covers the ground target, the task on 5.4. Multi-fusion-based task assignments
that target is ended; constant UAV velocity is considered.
Possible enhancements: Constraints of natural and dynamic en- To meet the requirements of UAV applications, the fusion of
vironments such as threats, uncertainties, and other external forces two or more task assignment algorithms is practiced in the lit-
can be considered. erature. GA, ACO, and PSO-based multi-UAV task assignments for
team orienteering problems are studied [174]. In this manner, the
iii. Artificial Neural Network (ANN)-based Task Assignment limitations imposed by specific algorithms can be complemented
by the advantages of other algorithms, and hence, the best solu-
Coordination and cooperation always remain as a challenge in tion can be achieved. A few studies that utilized a combination of
multi-robot and multi-agent systems. The ultimate goal of the two or more algorithms are presented below.
agents is to find an optimal policy to maximize the cumulative
reward instead of the local optimal solution in real time. In re- 5.4.1. Robust Filter Embedded Task Assignment (RFETA)
cent years, ANN has been developed as a promising approach to
A combination of robust and FETA algorithms for uncertain
solve the issues of dynamic environment by continuously interact-
and dynamic environments was presented in [175]. One of the
ing with the environment. DRL [164,165], DNN [166], SOM [167],
approaches is used to design task plans robust to uncertainties,
and DQN [168] are some commonly used ANN algorithms. Existing
reduce sensitivity to errors in situational awareness, and work
ANN-based task assignment algorithms were studied in the litera-
ture [169–172]. for long durations. After updating these results, the following ap-
proach plans again, resulting in the best plan with the current
5.3.10. Fast Task Assignment (FTA) information. However, this may lead to churning if updates are
A Q-learning-based task assignment for heterogeneous UAVs frequent. RFETA uses proactive and reactive methods for handling
in uncertain environments through neural network (NN) estima- uncertainties and enhances the worst-case scenario expressed as
tion and experience replay with priority to solve RL problems was
studied in [173]. T = { T 1, T 2, ..., T N } represents tasks in order max (C k − λk .σk ) T xk1 − βkT (xk1 ⊕ xk−1 ), (9)
xk1 ∈ X k
of priority, and {att( T k), def ( T k), ele( T k)} is attack capability, de-
fense capability, and electronic jamming capability, respectively, to where C k is the minimal value at time k, ⊕ is exclusive OR, xk1
complete task T k. Environmental uncertainty is denoted as envk = is the decision variable at time k1, and λ k . σk represents element-
{ wk, rk}, where wk represents wind speed, and rk represents rain- by-element multiplication.
j j j
fall. In addition, UAVs are denoted as U = {U 1 , U 12 , ..., U m }, and Advantages: Presents robust results in uncertain environments;
{att(U j ), def ( T k), ele( T k)} is the attack capability, defense capabil- sensitive to errors; minimizes the effects of noise; yields fast con-
ity, and electronic jamming capability of a j-type UAV. Task as- vergence, and reduces churning.
signment can be formulated by repetitively choosing UAV type in Limitations: It does not perform better if lots of information
correspondence with the maximum Q-value using the greedy strat- must be considered while planning.
egy as depicted in Fig. 9. Possible enhancements: Effects of intermittent measurements
Advantages: Computationally fast and efficient, highly adaptive, on the performance of RFETA can be studied; adaptive formula-
and handles uncertainty; deals with different types of tasks. tions of tuning parameter λ for optimization can be studied.

15
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

5.4.2. Energy-Efficient Multi-UAV Assisted Multi- Access Edge of the algorithms, many unforeseen issues may arise while imple-
Computing (EMU-MEC) menting the algorithms in real systems.
Energy efficient multi-UAV task assignment using MEC that of-
fers computing services to resource-constrained UAVs has been 6. Comparison of task assignment algorithms for UAVs
studied [176]. Task assignment, joint device association, and com-
puting resource assignment problems were formulated for the
minimization of energy depletion considering the power budget, In this section, existing task assignment algorithms studied for
existing resources for computing at UAVs, and deadline restraints. UAVs and UAV networks are qualitatively compared in terms of
The problem is first decomposed into three sub-problems and their main ideas, algorithms used for the task assignment process,
solved using the iterative block coordinate descent (BCD) algo- performance metrics considered, and qualitative and operational
rithm. A single UAV-aided MEC was studied in [177,178] where features. Table 4 summarizes the main ideas of all 27 algorithms
UAV roams around the target area to aid as a server. Performance discussed in the previous section along with the metrics consid-
and stability of the single UAV based system is constrained. Thus, ered for their improvement. It can be observed that the most
constraints of energy, computing power, task completion are im- popular algorithms used for task assignment are GA, RHTA, MILP,
proved by controlling association of devices and resource assign- and DP. The primary focus of the task assignment algorithms is to
ment variables in multi-UAV system in the proposed method. minimize the convergence time and complexities while maintain-
Advantages: Total energy consumption of sensor devices and ing cooperation between UAVs.
UAVs is minimized. In Table 5, task assignment algorithms are compared in terms
Limitations: Inter-cell inferences from mobile devices of other of qualitative features and key characteristics, such as task inter-
UAVs are experienced. dependence, types of target, task precedence, types of UAVs con-
Possible enhancements: UAV trajectory, communication, and sidered, types of task assignment algorithms, level of operational
computation resource assignment can be investigated. complexities of the algorithms, performance evaluation tools used,
and environmental scenario considered. Only a few of the algo-
5.4.3. Multiple Time-Window-based Dubins Traveling Salesmen rithms consider real environmental scenarios that are full of uncer-
Problem (MTWDTSP) tainties and adversities. An ideal task assignment algorithm must
Task assignment for multiple UAVs with altered capacities using consider the dynamics and complexities of real-world scenarios. In
a multi-objective symbiotic organism search algorithm to enhance some missions, task priorities are significant, and hence, they must
UAV task arrangement was studied in [179]. In addition, a task be considered during the task assignment process. In addition,
model based on the time window and task assignment based on from the tabular comparison, we can observe that heterogeneous
MTWDTSP was developed for various targets. Double-chain and UAVs have been used in recent studies. Most task assignment al-
some criteria are recognized for task assignment in logical and gorithms are centralized and evolutionary. However, a combination
environmental constraints. A group of targets requires an investi- of centralized and distributed task assignment algorithms has not
gation using heterogeneous UAVs subjected to different restraints. been considered in the literature. The fusion of multiple algorithms
The Dubins model [180,181] is introduced for UAVs with the fol- can be observed in some task assignment processes. Most algo-
lowing suppositions: (i) constant UAV velocity, (ii) reconnaissance rithms are aimed at minimizing complexities. Moreover, popular
job at certain heights, (iii) UAV flies at different heights without simulation tools include Monte Carlo, MATLAB, and mathematical
collision, and (iv) limited flight time. The dimensions of the UAVs formulations.
are abridged from 3D to 2D. Similarly, Table 6 summarizes different performance metrics,
Advantages: Improved convergence speed and efficiency; pro- such as energy consumption, latency, efficiency of algorithms,
vides optimized task sequence. adaptability scalability, and load balancing. Most task assignment
Limitations: Complexities increases with the number of UAVs algorithms focus on increasing the lifetime of the network by
and tasks; not all tasks are executed. reducing energy consumption. Delay is another major consider-
Possible enhancements: Constraints of complex dynamic envi- ation in task assignment algorithms. We can observe that most
ronments such as the effect of speed of the wind, flight height, task assignment algorithms have managed to achieve low latency
and temperature can be considered for better and realistic results. and high efficiency. Balancing the load among UAVs has not been
The advantages, limitations, and possible enhancements of each considered much in state-of-the-art algorithms. However, load bal-
task assignment algorithms discussed in this section are listed in ancing also plays a significant role in large-scale mission planning
Table 4. This will help readers to understand better through com- systems. Moreover, performance determinants, such as adaptability
parison. and scalability, have not been considered much in previous stud-
ies.
5.4.4. Lesson learned By comparison, for practical applications, stochastic evolution-
A good task assignment algorithm should be adaptive that can ary algorithms can perform better as they can learn from experi-
acquires the best performance in any circumstances. Thus, creating ence and adapt to the aspects of complex dynamic environments
an adaptive system, which is capable of adapting not only to the in real time. Moreover, decentralized algorithms assuring coordi-
objectives of the assignment but also to the limitations of the en- nated UAV operation can provide alternative solutions to enhance
vironment, is highly anticipated. Otherwise, massive degradation the performance of multi-UAV operations. In addition, robust and
in the performance of the target system can be observed with highly scalable algorithms are desirable for large missions. See Ta-
the increasing amount of limitations imposed by the environment ble 7.
and UAVs. The task assignment algorithms, reviewed in this sec-
tion, were evaluated through computer simulations by the original
authors and many of them were also validated through theoret- 7. Open issues, research challenges, and future directions
ical analysis. In the future, one can consider implementing these
algorithms in real systems and evaluate the performance accurate- Researchers in both academic and industrial fields are striving
ness of the algorithms in actual environment, which is necessary to utilize the full potential of UAVs. Regardless of the promising
when handling the intricate environmental issues. Though com- roles of UAVs, there are also a number of design challenges to be
puter based simulations gives preliminary results for the validation addressed. In fact, every UAV application has its own challenges

16
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Table 4
Comparison of task assignment algorithms in terms of their advantages, limitations, and possible enhancements.

Protocol Advantages Limitations Possible enhancements

HA-TA [90] Improves operational reliabilities and Randomly occurring failures provide a The amount and quality of feedback
capabilities through adaptive task level of uncertainty; uncertainty present information can be improved, and a
assignment systems and system at all planning stages due to incomplete sophisticated stochastic model of health
awareness; demonstrated effective knowledge at the flight plan level, states like energy usage and sensors’
performance through simulation and health of sensors during task performance can be embedded to deal with
experiments. assignment, and enduring maintenance uncertainties.
at the mission planning level.
MILP-TA [91] Useful in offline task assignment Makes a discrete representation of Kinematic constraints of UAVs should be
calculations; provides optimal solution for real-world problems, which makes focused; aspects of complex and dynamic
UAV groups with combined tasks with solutions impractical; solution requires environment such as obstacles, uncertainties,
timing and task order limitations. high computation time and makes and wind speed must be considered.
unreliable for real-time use.
MTWPS [94] Deals with online uncertainties such as A UAV’s waiting time depends on the Different types of deadlocks require better
communication issues, UAV malfunction, performance of other UAVs, which handling, which can be focused on in the
and time-sensitive target problem; easy makes waiting time complicated for future; threats and other risks present in the
computing used to lessen simulation time large-scale networks; the deadlock stochastic environment should also be
for large-scale UAV networks. problem is ignored. focused on.
DTA [96] Time efficiency, scalability, and robustness Does not consider the uncertainties and Minimizing the combination of all costs is
are achieved; achieves the mission goal aspects of dynamic environments; desired, requiring a solver algorithm; threats
with minimum cost. preliminary information regarding the and uncertainties of the complex dynamic
size and location of enemy UAVs is environments need consideration.
unknown.
SMAA-2 [97] Uncertainties are analyzed; finds Sensor error or target movement leads Threats present in the defense system of
satisfactory solution even in severe to unclear localization; probability of targets can be made more apparent; a
situations. UAVs being destroyed while handling confidence factor is introduced to judge the
attacks remains. accuracy of criteria values.
FEUTA [112] Condenses the consequence of noise Ignores the aspects such as threats and The importance of robust methods proposed
caused due to high-frequency on the uncertainties, which makes it unreliable for FEUTA and their connection to noise
planner; mitigates the effects of noise. for the real environment; accepts noise rejection algorithms must be investigated.
entirely.
CTAI [113] Decomposition of motion planning issues Tracking error increases due to Timing of target intersection can be
helps obtain a near-real-time solution; disturbances; only static threats are considered; aspects of the dynamic, complex
risk is mitigated by increasing distance considered, and thus the results 3D environment such as wind speed, risks
between UAV and threats. obtained are suboptimal. need consideration.
EETA [117] Lessens energy depletion, and optimizes Single UAV is used, which increases the UAVs’ kinematic constraints such as external
task offloading, resource sharing, and risk of mission failure. forces, motions, and physical limitations of
trajectory without increasing delay; the dynamic environment such as threats
minimizes UAV’s propulsion and and uncertainties must be considered during
computation power. the assignment process.
EDATA [118] Energy consumption, operational time, GPS localization errors and signal losses Cases of accidents and UAV failures must be
and fairness of the system are optimized exist; increase in execution time for considered; the modeling part could be
jointly; environmental effects on flights FTUS, which impacts the performance of improved concerning the energy consumed
and performance are considered. the model. by GPS in errors and signal losses.
DP-CTA [119] Generates cooperative and optimal Computation time grows with the Constraints of the real dynamic environment
solution quickly; recovers planning of number of targets, and the uncertainties such as sudden threats, effects of wind, and
real-world air operations providing of real complex environment are temperature on UAV flights must be focused
improvements. ignored. on.
PHM-based MTA [123] Applies system-level RUL methodology to It may not be feasible to know the PHM-based TA for extensive haul task
accomplish task assignment including probability of failure in advance; threats reaction of algorithm during failure requires
failure probability. and uncertainties of real stochastic further investigation; consideration of
environments are ignored. aspects of real environment such as wind,
temperature, threats is a must.
ABMC-TA [132] Solves multi-constraint task assignment 2D environment with static obstacles Complexities imposed by real 3D
issues efficiently considering the considered for implementation which environment need to be considered to
constraints of sensors, time, danger, and cannot justify the constraints of real confirm its practical validity; more
fuel. dynamic environment; kinematics of experimental study and investigation are
UAV such as motion and impacts of required to verify the efficiency of the
external forces on its flight path needs proposed method in real environment.
consideration.
RDTA [134] “No Fly Zones” are considered; a modified Does not consider the heterogeneity of Conflicts arising due to low convergence
candidate plan selection algorithm is used UAVs and targets and the constraints of need to be eliminated; heterogeneity of
to improve performance and increase the the dynamic environment such as UAVs and other environmental aspects such
robustness to the inconsistencies. threats and effects of external forces as threats and uncertainties must be
(wind, temperature, etc.) on UAVs considered.
performance.
CBBA [137,138] Handles operational complexities of All targets may not be discovered; Kinematic and dynamic UAV constraints on
multi-UAVs; provides reliable, conflict-free dynamic threats and uncertainties are air and ground can be incorporated; UAVs
UAV task assignments. ignored. need to be penalized for massive heading
changes.
(continued on next page)

17
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Table 4 (continued)

Protocol Advantages Limitations Possible enhancements

EAT-PSO [143] Achieves a minimum total make span, Allows a single UAV to occupy a An anti-collision refinement phase can be
forms a seamless UAV operation schedule, position at a particular time, ignoring incorporated, a detailed study on the
and short computation time in finding a uncertain events and obstacles. distributed control of UAVs can be
high-quality feasible solution. conducted, and the robustness of the system
can be assessed.
OPA [146] Works with short execution time, highly Does not consider the constraints of the Adaptive approach of cooperative task
flexible, strongly robust, and scalable; actual battlefield; threats and assignment can be introduced to balance the
quickly finds optimal solution; avoids uncertainties that may arise in the load of UAVs and adapt to real-time
difficult cost function based on the tasks, dynamic environment are ignored. battlefield situations.
and has high calculation efficiency.
CTA [147] Efficiently solves considerate task Focuses on flight length only and Estimation of cost function could be adjusted
assignment problems of UAVs under ignores other constraints of task considering the flight mission; in an actual
multiple circumstances and constraints; assignment; threats and uncertainties of battlefield environment, target values vary,
fast convergence and a highly stable the real environment are ignored, and distribution of UAV firepower must be
iterative curve are assured. making the solution impractical. considered.
CTMAP-GA [158] Promptly offers feasible task assignment Performance degrades if the search Proper synchronization must be considered
solutions and real-time implementation space is harsh; it ignores the aspects of to maintain coordination among UAVs;
for high dimensions; considers task real dynamic environments, which constraints of the dynamic environment such
priority and coordination, time makes the proposed method impractical as threats, uncertainties, and external forces
constraints, and trajectory. in real-time implementation. and their impacts on UAV’s motion need to
be investigated.
CMTAP [159] Solves computational complexity for Does not consider the dynamic aspects Moving, sudden targets, and actual
classical combinatorial optimization of the environment such as impacts of environmental scenarios must be considered;
methods with higher efficiency and threats, uncertainties, and external computational efficiency needs to be realized
considers task priority and timing forces on UAVs flight; heterogeneity of to achieve a real-time solution.
constraints. tasks, targets, and UAVs are ignored.
MGA-CTA [160] Considers and addresses the heterogeneity Only one kind of resource, i.e., weapon, Different operational and kinematic
of UAVs, targets, and environments; is considered; stationary targets constraints of UAV must be considered;
mirror representation to deal with the considered in simulation, weapons are aspects of real dynamic environments such
limited resources. used in one way only. as threats, uncertainties, and other external
forces and their impacts need to be
analyzed.
CMTAPSVTW[161] Considers kinematic constraints (motion, Does not cover practical mission Distributed model with more stochastic
external forces, propulsion), resource scenarios such as weather constraints, features can be considered to cover practical
constraints (fuel and onboard weapons), radar, terrain barriers during simulation. mission scenarios; a comparison of a
and time constraints (time windows and centralized and distributed algorithm for
sequence of tasks). stochastic task assignment needs to be made.
BAS-GA [162] Diversity of searching capacity improves Does not consider the kinematics of Rapid task adjustment using centralized and
convergence time; local search ability is UAV which impacts on performance and distributed methods can be experimented
increased; has good convergence from constraints of the realistic dynamic on; integration of path plan into task
small to large missions. environment such as threats and assignment for broad task planning for
uncertainties. multi-UAVs can be done.
OGA-DEMMO [163] Minimizes the execution time and total No constraints are considered for Constraints of natural and dynamic
UAV consumption and improves the reconnaissance; if the sensor’s field of environments such as threats, uncertainties,
optimality of the algorithm and view covers the ground target, the task and other external forces can be considered.
convergence energy. on that target is ended; constant UAV
velocity is considered.
FTA [173] Computationally fast and efficient, highly Considers only stationary targets; Moving and sudden targets and complexities
adaptive, and handles uncertainty; deals complexities and aspects of the real of real battlefield environmental scenarios
with different types of tasks. battlefield environment are ignored. must be considered.
RFETA [175] Presents robust results in uncertain It does not perform better if lots of Effects of intermittent measurements on the
environments; sensitive to errors; information must be considered while performance of RFETA can be studied;
minimizes the effects of noise; yields fast planning. adaptive formulations of tuning parameter λ
convergence, and reduces churning. for optimization can be studied.
EMU-MEC [176] Total energy consumption of sensor Inter-cell inferences from mobile UAV trajectory, communication, and
devices and UAVs is minimized. devices of other UAVs are experienced computation resource assignment can be
investigated.
MTWDTSP [179] Improved convergence speed and Complexities increases with the number Constraints of complex dynamic
efficiency; provides optimized task of UAVs and tasks; not all tasks are environments such as the effect of speed of
sequence. executed. the wind, flight height, and temperature can
be considered for better and realistic results.

and opportunities. From the above detailed review of the literature, 7.1. UAV regulations
we present the following research gaps along with future efforts
that could be pursued for the practical deployment of UAVs. In this Regulatory concerns are a key limiting factor in the deployment
section, the major characteristics and research gaps faced during of UAV operations. Regardless of the widespread UAVs applications,
the design of task assignment schemes for a set of UAVs are briefly there are numerous issues concerning confidentiality, security, col-
addressed. Furthermore, future research directions for these issues lision avoidance, safety, and protection of data. In this respect, UAV
are suggested. guidelines are uninterruptedly being established to govern the op-

18
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Table 5
Comparison of task assignment algorithms in terms of main ideas and design approaches.

Protocol Main idea Design approach

HA-TA [90] Step towards allowing robust decision making for disseminated autonomous RHTA algorithm
UAVs.
MILP-TA [91] Optimal formulation for solving the coupled multiple-assignment problem. MILP algorithm
MTWPS [94] Task assignment with constraint and time-sensitive uncertainties. Meta-heuristic optimization called MTPS
DTA [96] A dynamic task assignment using enemy information for combinatorial Hungarian algorithm
optimization.
SMAA-2 [97] Offers suitable task assignment under severely uncertain circumstances. Stochastic multi-criteria acceptability analysis
method
FEUTA [112] Performs reassignment at the rate that information is updated, which permits Noise rejection algorithm and linear integer
instantaneous response to any noteworthy fluctuations in the environment. program
CTAI [113] Provides end-to-end explanation to cooperative control problem in the Satisfying decisions theory
occurrence of dynamic threats.
EETA [117] Joint task assignment and resource assignment in UAV-aided MEC. Block successive upper-bound minimization
algorithm
EDATA [118] EAUS, DAUS, and FTUS are proposed. Linear integer problem
DP-CTA [119] Two models of DP estimation methods are designed i.e. one-step look-ahead Dynamic programming based
and two-step look-ahead
PHM-based MTA [123] Information from a PHM method is applied so as to support decision making Receding horizon task assignment algorithm
through IVHM structure.
ABMC-TA [132] Auction-based technique that deals with assigning tasks of multi-UAV system Auction algorithm and multi-layer cost computation
under the constrictions of UAV capability, time window, cost of fuel and
dangers during the travel.
RDTA [134] Confirms the resultant plans calculated by team are conflict-free and feasible RDTA
CBBA [137,138] Modified CBBA to handle obstacles and mitigate churning behavior Consensus-based bundle algorithm
EAT-PSO [143] A mathematical model that solves the problem and heuristic-based Earliest available time (EAT) and PSO algorithm
methodology.
OPA [146] Resolves the ideal order of numerous tasks for cooperative accomplishment. Nearest neighbor method
CTA [147] Cooperative multi-task assignment model. Hungarian algorithm
CTMAP-GA [158] Allows unique requirements of the scenario to be considered, such as task GA
precedence and coordination, timing constraints, and trajectory limitations
CMTAP [159] Integrated task assignment and path optimization refining the visitation angle Dubins car and GA
MGA-CTA [160] Concerned with cooperative decision making and control Modified GA with multi-type genes
CMTAPSVTW[161] Cooperative multi-task assignment problem with stochastic velocities and Meta-heuristic based on a modified GA
time windows for heterogeneity
BAS-GA [162] Uses two cross-operator methods to preserve the globality of the algorithm Improved GA based on the beetle antenna search
and perform a deep search of the solution space algorithm
OGA-DEMMO [163] Presents a multi-UAV reconnaissance task assignment model for Opposition-based GA
heterogeneous targets and an effective genetic algorithm to optimize UAV task
sequence
FTA [173] Neural network approximation and prioritized replay to offload online MDP and Q-learning-based
computation to offline learning procedure
RFETA [175] Provides substitute approach that associates robust scheduling with Robust filter embedded task assignment
procedures established to eradicate churning
EMU-MEC [176] A joint device association, task assignment, and computing resource Mixed-integer non-linear programming and block
assignment problem considering the energy budget, available computing coordinate descent
power, and task completion deadline constraints
MTWDTSP [179] Problem of task consignment is articulated as a manifold time window-based Modified multi-objective symbiotic organism search
Dubins traveling salesman problem algorithm

eration of UAVs while bearing in mind various factors, such as UAV nels. Communication between nodes is significantly affected when
type, range, height, and speed of UAVs [182,183]. Generally, five the topology of the network fluctuates due to failures and mobil-
different criteria are frequently considered when developing UAV ity. Delayed or incorrect information misleads the communication
regulations. Despite their usefulness in each area, UAVs are banned process. As a result, the network may experience delays, which
in different places; hence, appropriate airspace protocols and reg- are intolerable in many UAV applications. FANETs require exact
istration of drones are necessary for appropriate UAV operation. and timely network information to determine the best and real-
Specific areas of flying should be recognized for UAV practitioners, time communication. Moreover, adaptive and online communica-
where they can operate UAVs without any obstacles following the tion protocols must be explored and designed for dynamic network
regulations. Apps that indicate restricted zones must be developed topologies [184,185].
to ensure safe UAV operation. Moreover, combative counter-drone
measures should be established to circumvent the utilization of 7.3. Localization
UAVs for malicious intent or in an unintentional security breach.
Localization plays a crucial role in the safe operation of UAVs
7.2. Dynamic network topology beyond LoS applications. Because of the high and unpredictable
mobility of UAVs, accurate localization of UAVs within a short time
The foundation of a peer-to-peer wireless network is a dy- is problematic. The position of UAVs in the network affects the
namic network topology that determines the relationships among formation of appropriate communication between UAVs. A widely
neighboring nodes to be maintained within the network. Due to adopted solution is to use GPS [186,187] to obtain the position
the high maneuverability of UAVs and constraints of the dynamic coordinates of any wireless node. However, the accuracy of UAV
environment, the topology of UAV networks remains intermittent location is limited due to its 3D mobility and highly dynamic
most of the time. High UAV mobility causes rapid changes in the and obstacle-based environment. Furthermore, due to the probable
network topology and fluctuations in radio communication chan- large deployment of UAVs, the use of GPS may add unnecessary

19
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Table 6
Comparison of task assignment algorithms in terms of key features and characteristics.

Protocol Task interdependency Target Priority Environment

HA-TA [90] No Heterogeneous Yes Complex and uncertain
MILP-TA [91] Yes Heterogeneous Yes -
MTWPS [94] No Mobile No Complex and dynamic
DTA [96] No – No Battlefield
SMAA-2 [97] No Mobile No Severely uncertain
FEUTA [112] – Uncertain No Dynamic
CTAI [113] Yes Homogeneous Yes Dynamic and threat-based
EETA [117] No Mobile No –
EDATA [118] No IoT devices Yes Unknown and uncertain
DP-CTA [119] Yes Heterogeneous No Adversarial
PHM-based MTA [123] No Stationary Yes Critical
ABMC-TA [132] No Heterogeneous No Dynamic
RDTA [134] No – No Uncertain
CBBA [137,138] No Stationary and mobile No Obstacle- and noise-based
EAT-PSO [143] No Heterogeneous No Indoor
OPA [146] No Stationary No Complex
CTA [147] No Mobile No Obstacle-based
CTMAP-GA [158] Yes Heterogeneous Yes –
CMTAP [159] No Heterogeneous and stationary No –
MGA-CTA [160] Yes Heterogeneous No –
CMTAPSVTW[161] Yes Stationary Yes Uncertain and dynamic
BAS-GA [162] Yes Stationary No –
OGA-DEMMO [163] No Heterogeneous No Real-world mission environments
FTA [173] No Stationary Yes Complex and uncertain
RFETA [175] No Heterogeneous No Uncertain and dynamic
EMU-MEC [176] Yes IoT No Infrastructure-less
MTWDTSP [179] No Heterogeneous Yes Obstacle- and threat-based

Note: “–” in the table indicates that the information is not addressed in the corresponding literature.

Table 6
Comparison of task assignment algorithms in terms of key features and characteristics (continued).

Protocol Year UAV type Task assignment type Operational complexity Performance evaluation tool
HA-TA [90] 2008 Draganfly VTi Pro R/C Centralized Low MIT’s Real-time Indoor Autonomous Vehicle
helicopters Test Environment
MILP-TA [91] 2003 WASM Centralized High Numerical formulations
MTWPS [94] 2017 Heterogeneous Centralized Low MATLAB
DTA [96] 2019 – Centralized Low SITL and Gazebo
SMAA-2 [97] 2015 Homogeneous Stochastic and centralized Low Monte Carlo
FEUTA [112] 2004 Homogeneous Centralized and stochastic High Numerical formulations
CTAI [113] 2002 Homogeneous Coordinated and stochastic High MATLAB and Simulink
EETA [117] 2021 Single UAV Centralized and stochastic Low Numerical formulation
EDATA [118] 2019 Rotary UAVs Centralized and stochastic – Python and Gurobi optimization tools
DP-CTA [119] 2005 Homogeneous Stochastic and centralized Lower Numerical formulations
PHM-based MTA [123] 2014 Homogeneous Decentralized – MATLAB and CPLEX
ABMC-TA [132] 2016 Homogeneous Centralized and stochastic – Experimented
RDTA [134] 2006 Heterogeneous Decentralized Low Monte Carlo
CBBA [137,138] 2009 Heterogeneous Decentralized Low Monte Carlo
EAT-PSO [143] 2019 Multi-copters Bio-inspired High Numerical formulations
OPA [146] 2021 Heterogeneous Bio-inspired Low Numerical formulation
CTA [147] 2017 Homogeneous Bio-inspired – MATLAB
CTMAP-GA [158] 2005 – Bio-inspired Low Monte Carlo
CMTAP [159] 2010 Heterogeneous Bio-inspired Low Monte Carlo
MGA-CTA [160] 2012 Heterogeneous Bio-inspired Low Monte Carlo
CMTAPSVTW[161] 2018 Heterogeneous Bio-inspired Low MATLAB
BAS-GA [162] 2020 Homogeneous Bio-inspired Low -
OGA-DEMMO [163] 2018 Heterogeneous Bio-inspired High MATLAB
FTA [173] 2019 Heterogeneous Bio-inspired – Python 3.7.0, Tensorflow
RFETA [175] 2007 Homogeneous Multi-fusion – Monte Carlo
EMU-MEC [176] 2021 Edge server- equipped UAVs Multi-fusion Low Numerical formulations
MTWDTSP [179] 2019 Heterogeneous Multi-fusion Low Visual C++ and MATLAB

Note: “–” in the table indicates that the information is not addressed in the corresponding literature.

costs undermining its use prospective. Different localization meth- through which location information is shared must be reliable and
ods have been studied for proficient UAV location. Network-based robust in the network design [189].
positioning, which is based on exchanging packets and height-
based positioning, can be used to enhance position information. 7.4. UAV speed
Recently, localization techniques based on fuzzy logic that relies on
weighted centroid and vision-based localizations [188] have been The mobility of UAVs is one of the major reasons for their rapid
greatly favored for precise localization. Moreover, wireless links adoption in many different applications. Owing to the high mobil-

20
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

Table 7
Qualitative comparison of task assignment algorithms in terms of performance.

Protocol Energy consumption Latency Efficiency Adaptability Scalability Load balancing

HA-TA [90] Low Low High High – No
MILP-TA [91] – High High No No No
MTWPS [94] No Low High No No No
DTA [96] Low Low High High High -
SMAA-2 [97] No No Improved No No No
FEUTA [112] – Low Improved No No No
CTAI [113] – Low High No No No
EETA [117] Low Satisfactory High No No No
EDATA [118] Low Low High No No No
DP-CTA [119] No Low Improved High – No
PHM-based MTA [123] Low Low High No No No
ABMC-TA [132] Low Low High – – No
RDTA [134] – Low High High Low No
CBBA [137,138] No Low High – – No
EAT-PSO [143] Recharging Low High No Yes No
available
OPA [146] No Low High High High Yes
CTA [147] No Low Improved No No No
CTMAP-GA [158] Low Low High High High No
CMTAP [159] Low Low Improved High High No
MGA-CTA [160] Low Low Improved No – No
CMTAPSVTW[161] Low Low High High High No
BAS-GA [162] Low Low High No Yes No
OGA-DEMMO [163] Low Low High No No No
FTA [173] No Low High High No No
RFETA [175] No Low High High High No
EMU-MEC [176] Low Low High – – No
MTWDTSP [179] Low Low High No No No

Note: “–” in the table indicates that the information is not addressed in the corresponding literature.

ity of UAVs, frequently changing and unpredictable topologies have for UAV-to-UAV and UAV-to-ground communications requires am-
been observed. The speed of UAVs varies with their size; small- ple simulations and measurements. In [193,41], a summary of the
sized UAVs usually move at 15 m/s approximately, and large UAVs prevailing A2G channel modeling works were presented. Ref. [193]
can have a remarkable speed of 150 m/s. The types of UAVs must provides both simulation and measurement-based results for path
be carefully selected considering the desired speed for their tar- loss, delay spread, and fading in A2G communications. Thus, accu-
geted applications. The trade-off between the speed of the UAV rate channel characterization is crucial for performance optimiza-
and its turning agility must also be prioritized. Different emergent tion and the design of efficient UAV communication. Measurement
technologies, such as designing dynamic and optimized trajectories operation while modeling the UAV-UAV and UAV-GBS channels
for UAVs in 3D for better accuracy and speed, have sustained the with various velocities and moving directions of UAVs must be
durability of small UAV nodes. Consequently, restrictions on speed done in the presence of regularly and irregularly shaped infras-
must be considered via field experiments. The optimized and adap- tructures.
tive speed of UAVs benefits highly enhanced UAV communication
[190]. 7.7. Synchronization

7.5. UAV types It is assumed that for every application sphere of UAV time
synchronization, proper time-stamping is required not only to be
Depending on the goals and application of UAV operation, one updated but also for coordination among other UAVs and devices.
must select an appropriate UAV type and meet numerous require- Synchronization plays a crucial role in UAV applications in disas-
ments imposed by the anticipated QoS and environmental con- ter or emergency environments, where if one of the UAVs misses
straints. In fact, for the proper utilization of UAVs, numerous fac- information at a particular time, the network will be partitioned,
tors, such as UAV capabilities and onboard sensor devices, must which degrades the network performance [194].
be considered [191,192]. UAVs are categorized based on their size,
weight, speed, coverage range, and other operational capabilities. 7.8. UAV antennas
Inappropriate UAV selection may deteriorate complete network
performance. Owing to the ability to move in any direction and at differ-
ent speeds, a new antenna design for UAV networks is essential
7.6. Channel modeling to achieve higher data rates. One of the alternatives for improving
the data rate during UAV-to-UAV and UAV-to-BS transmission is to
Wireless media amongst transmitters and receivers affect the install a UAV tracking antenna. For UAVs cruising at high speeds,
signal propagation. The characteristics of A2G channels are consid- it is desirable to have small, aerodynamic antennas that limit
erably different from those of ground-to-ground (G2G) channels. drag but can still yield sufficient bandwidth and coverage. More-
A2G is highly susceptible to blockage, whereas A2A communica- over, directional antennas are widespread due the restricted en-
tion has dominant LoS communications. To design and implement ergy and space limitations of particularly small sized UAVs. More-
optimal UAV applications, an accurate A2G channel model is a over, tilted-beam circularly-polarized antenna can be used to save
perquisite. In addition, A2G channels are highly dependent on alti- space. Performance is expected to be improved in terms of return
tude, type, angle of elevation, and type of propagation environment losses, axial ratio, and radiation pattern, which can be achieved by
of UAV employment. Therefore, finding a standard channel model means of such antennas. Additional propose for scheming and us-

21
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

ing smart antennas is to reduce the UAV communication energy Security communication is necessary for many UAV applications.
by minimizing the transmission power [85,86]. Backscatter anten- Security issues are a foremost concern in UAV communications
nas and wireless power transfer techniques could also be adopted because wireless media are used. Highly secure and safe commu-
for longer-term network life. The integration of several adaptive nication is of utmost importance for military applications, where
antennas also contributes in the communication of highly mobile data secrecy is the principal focus. Any alterations in information
UAVs. A compact, sabre-like antenna that is capable of switching by a fraudulent node may cause severe degradation of UAV appli-
between two radiation patterns can be considered for better cov- cations. Furthermore, invaders may cause network partitioning and
erage in UAV communications [195,196]. decrease network efficiency. Consequently, it is very important to
develop robust and secure UAV communications that can meet the
7.9. Network lifetime diverse needs of application reliability with minimum untrusted
node involvement [206]. The security mechanisms of UAV com-
Power consumption is always a primary concern in UAV com- munications are highly affected by the presence of faulty nodes
munications. Most often, UAVs are battery-powered devices, and and erroneous communication channels, which must be controlled
UAVs use their energy for thrust, data gathering, processing, and with effective approaches. Different cryptographic measures, such
dissemination. The proficiency of UAV networks and network life- as data encryption and hashing techniques, can be used to prevent
time are greatly determined by the energy levels of the UAVs. network attacks [207]. Block chain technology can be utilized to
Hence, ensuring energy-efficient UAV mission operations is of pri- ensure the safety and security of UAVs based on collected data.
mary importance. Different energy harvesting techniques, such as
wireless charging and solar energy can help increase the network 7.13. Fault-tolerant system
lifetime of UAVs. An autonomous concept of battery swapping was
introduced and enhanced in [197–200]. In the swapping process, In contrast to manned aerial networks, UAVs allow a wider
UAVs are connected to an external power supply for smooth func- range of flight operating points and are highly vulnerable to fail-
tioning and to prevent data loss due to dead UAV nodes [201,202]. ures. To augment safe and reliable UAV communications, fault-
Additionally, exhausted batteries are sometimes replaced by abun- tolerant methods must be considered when designing control sys-
dantly charged batteries to continue the current mission. The clus- tems for multiple UAVs. Different strategies of fault-tolerant con-
tering of nodes also helps to minimize energy consumption by trol for UAV communications are receiving significant attention
limiting the number of transmissions [203]. from researchers because of the growing awareness of hazards re-
sulting from the failure of different components and the necessity
7.10. Coordinated communication
of trustworthy and secure systems in emergent scenarios. A net-
work of UAVs must share critical data, and faults in a particular
The number of UAVs employed in any application may vary
UAV can break down the entire network. Keeping this in mind,
from one to numerous. Proper coordination and control among
robust fault-tolerant mechanisms should be adopted to overcome
UAVs is necessary for better network performance. To this end,
the different failures observed in UAV networks [208,209]. Inter-
several efforts have been made to design error-free mechanisms
UAV support must be restructured in case the network partitions
to guarantee cooperation among UAVs. However, most of them
due to the failure of one or more UAVs.
are still immature and are subject to several challenges. The dis-
tributed approach divides the workload between all nodes, making
7.14. Integration with cloud computing
peer-to-peer communication suitable. In the centralized approach,
a central entity is responsible for designing and controlling all
Cloud services have become the next frontier in advancing the
other nodes. Collision avoidance and proper sharing of tasks be-
workflows of UAVs. Cloud services have facilitated UAVs by allow-
tween the employed UAVs are significant for efficient cooperation
ing the use of resources at their leisure. The limitations of UAV
and collaboration. A bandwidth-efficient multi-robot coordination
networks can be incorporated and supported using cloud comput-
algorithm tackling the problem of UAV swarm formation and main-
ing techniques, as discussed in [210]. To accomplish higher data
tenance was studied in [204]. Network time synchronization and
throughput in UAV-based applications, aerial caching could be a
proper time-stamping for UAV networks require intense investiga-
promising solution. A cloud-assisted approach can also be useful
tion for proper coordination among UAVs.
for deriving optimal flight and data acquisition. Some research ef-
7.11. Network density forts have been made to provide techniques for linking UAVs to the
cloud.
A UAV network consists of flying nodes that rely on battery
power and limited onboard resources; therefore, communication 7.15. Path planning
protocols that consume the minimum possible energy with new
inbuilt intelligent technologies must be focused on. Based on the Path/trajectory planning is an important and challenging issue
types of UAV and objectives of the applications, the density of in the design of UAV-based communication systems. In fact, task
UAVs can vary from small to extremely dense. In dense UAV assignment and path planning are highly interrelated aspects of
networks, suitable clustering methods must be considered to aid any UAV mission. The UAV path must be optimized with respect
communication. Designing adaptive density-based communication to key performance metrics, such as spectral efficiency, through-
protocols can be another credible solution [205,27]. In addition, put, energy, and delay. Although many studies have focused on
adaptive and energy-efficient clustering approaches improve the path planning algorithms for UAVs, UAV networks still experience
performance of dense UAV networks. The assessment of wind and issues of target location and identification due to high UAV mo-
weather conditions affects drone robustness against harsh environ- bility. Thus, it is desirable to rapidly develop dynamic trajectory
mental conditions to enhance network lifetime. planning approaches for UAVs to increase the likelihood of end-
to-end connections while maintaining adequate intact target area
7.12. Security and privacy coverage. Various path planning techniques proposed for UAV path
planning were analyzed in [211]. There are still several issues
UAVs are exposed to many possible security risks attributed to in the path-planning process, in which path optimization consid-
the susceptibility of protocols and resource-constrained UAV nodes. ers the mobility patterns, obstacles, delay, and energy. Moreover,

22
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

the algorithm can be equipped with deep learning approaches to requirements. UAV mobility at low speed reduces the coverage
help UAVs generate an environment map in real time. Genetic and significantly, which results in network delays. Regarding the emer-
chaotic operators can also be introduced to improve the conver- gency application of UAVs, mobility models with minimal latency
gence rate and hence the efficiency. 3D real-time path planning are highly desired. However, continuous connectivity with the net-
for different obstacle densities and moving obstacle environments work is very challenging, which may result in degraded network
in the presence of uncertainty was studied in [212–214]. performance in such scenarios. Thus, new mobility models for a
specific environment and applications that can cope with the chal-
7.16. Integration with IoT lenges of mobility must be focused on. Recently, different mobil-
ity models have been designed for UAVs [219], and they can use
The efficient operation of UAVs requires addressing several key variable speeds during different communication phases. Dynamic
problems [215]. For instance, the number of UAVs that are neces- UAV speed control using machine learning techniques to adapt the
sary for full coverage of a given geographical area must be con- UAV speed according to network requirements is another possible
sidered. UAVs are self-efficient for sensing and decision making, solution. Based on recent advances in deep reinforcement learn-
and this has made UAV networks more demanding. The integration ing algorithms, green mobility management has been proposed for
of UAVs with IoT adds huge dimensions and remarkable results. UAV-assisted IoT.
However, there are many additional challenges [216]. Designing
scalable UAV networks is in high demand as the application areas 7.20. Task offloading
to be covered, as well as the number of tasks to be accomplished,
are exponentially increasing. In many mission planning systems, numerous tasks must be
conducted simultaneously. Multi-UAV systems are in high demand
7.17. Quality of service for such applications. Tasks must be divided among the employed
UAVs such that all tasks receive proper attention in a timely man-
QoS support over UAV communications remains as a challeng- ner. Owing to time-sensitive and computation-intensive UAV appli-
ing issue when previously built-up arrangements of the network cations and high mobility scenarios, cost-efficient task offloading
are not accessible due to the changes in speed, location, architec- remains a challenging issue. Many intelligent task offloading algo-
ture, or separation between UAV nodes. Different types of QoS pro- rithms have been presented in the literature for the proper utiliza-
visions for UAV-aided IoT networks were recommended in [217]. tion of UAV networks [220,221] that aim to minimize the overall
New resource innovation and planning procedures are necessary energy consumption for the accomplishment of predefined tasks.
to dynamically change the requirements of QoS of UAV networks.
Therefore, well-organized approaches for the efficient handover 7.21. Impact of 5G, B5G, and mm-wave communication
mechanism, proficient methods for determining and solving UAV
battery related issues, user-facilitation, and network assortment, Providing universal connectivity to miscellaneous IoT devices is
capability, and spectral effectiveness needs an additional explo- a prime challenge for 5G and B5G. 5G communications were de-
ration. Furthermore, designing of the protocols supporting multiple ployed globally from 2020, and many more competences are in the
classes of traffic and preemption allowance, identification of posi- progression for standardization, such as mass connectivity, guaran-
tion and prioritization of the packets are some unexplored areas. teed low latency, and ultra-reliability. Concerning future require-
Timely update of the control information guarantees précised in- ments and demands, researchers are now interested in beyond
formation but adds consumption of energy. Competent routing, 5G/6G communication, which is expected to provide worldwide
flight self-sufficiency, and policy-making metrics can confirm the coverage, improved energy/spectral/cost efficiency, better intelli-
essential QoS of the targeted UAV applications. gent security, and more benefits. The challenges, issues, and con-
cerns for these networks are immense, and integrating them with
7.18. Scalability highly mobile aerial nodes adds more complications and techni-
cal problems. Mm-wave communication with multi-gigahertz/ter-
The large-scale ever-growing use of UAVs in a wide range of ahertz bandwidth availability enhances the proficiency and data
applications is anticipated in the next few decades. Thus, UAV net- transmission rate in comparison to microwave communications.
works as well as task assignment algorithms designed for a swarm Recently, UAVs have been expected to serve ample applications,
of UAVs must be highly scalable. Small-scaled UAV networks are and the use of mm-wave bands can be very promising for han-
the best choice for commercial applications because of their easy dling the higher demands for data rate to achieve greater through-
positioning, low maintenance, low acquisition costs, and high mo- put. However, mm-wave suffers from propagation-related inade-
bility. The number of UAVs employed, urgency of the UAV applica- quacies, which can be compensated by the flexibility of UAVs.
tions, and range to be covered are highly interrelated. Large-scaled Many other deficiencies present in mm-wave signals must be re-
UAVs are required when the mission area is large and delay is solved while preserving its benefits. 5G, B5G, and mm-wave-based
strictly undesirable. However, the complexity of task assignment UAV communications were discussed in [222]. Further investiga-
algorithms increases with increasing number of UAVs and other tion of mm-wave and 6G communications is extremely desired for
IoT nodes. Designing scalable and adaptive communication proto- reliable, flexible, robust, and spectrally efficient UAV communica-
cols is a principal challenge for both light and dense UAV networks tions [223,224].
[218].
7.22. Intelligence-based algorithms
7.19. Mobility management
Machine learning techniques play a promising role in UAV-
Networks of UAVs face many design problems related to net- based applications as they can deal with uncertain environmental
work mobility. Frequent topology changes affect the communi- constraints and can be implemented in real time. Machine learn-
cation and cooperation among UAVs. Different mobility models ing improves UAV network performance by learning from the sur-
have been studied to manage the movement of UAVs, but they rounding environment and past experience. Machine learning can
are insufficient to completely solve the issues faced in UAV com- possibly be leveraged to design and optimize UAV-based wire-
munication. Mobility models must be selected based on network less communication systems [225–228]. Using RL methods, UAVs

23
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

can automatically modify their locations, directions, and motion to Acknowledgements

serve ground nodes. UAVs can adapt to the constraints of dynamic
environments by independently optimizing their paths. Further- The authors thank the editor and anonymous referees for their
more, neural networks and data analysis models can be leveraged comments, which helped improve the quality of this paper.
to envisage the behavior of ground nodes and efficiently arrange
and control UAVs. The mobility of users and load sharing can be References
achieved for the best distribution and route planning of UAVs.
[1] M.Y. Arafat, S. Moh, Localization and clustering based on swarm intelligence
in UAV networks for emergency communications, IEEE Int. Things J. 6 (2019)
7.23. Lesson learned 8958–8976, https://doi.org/10.1109/JIOT.2019.2925567.
[2] S. Poudel, S. Moh, Energy-efficient and fast MAC protocol in UAV-aided wire-
less sensor networks for time-critical applications, Sensors 21 (2020), https://
In this subsection, key lessons learned while surveying on the doi.org/10.3390/s20092635.
[3] S. Poudel, S. Moh, Hybrid path planning for efficient data collection in UAV-
task assignment algorithms are provided. Multiple UAVs can be
aided WSNs for emergency applications, Sensors 21 (2021), https://doi.org/10.
used in many applications to enhance the area of coverage, to 3390/s21082839.
minimize the cost of operation and maintenance, to do real-time [4] S. Hayat, E. Yanmaz, T.X. Brown, C. Bettstetter, Multi-objective UAV path plan-
operation, and to meet the network requirements. The task as- ning for search and rescue, in: Proc. - IEEE Int. Conf. Robot. Autom., 2017,
pp. 5569–5574.
signment is the significant aspect of multi-UAV based applications.
[5] S. Waharte, N. Trigoni, Supporting search and rescue operations with UAVs,
Task allocation in UAVs needs several considerations such as UAV in: Proc. - EST 2010 - 2010 Int. Conf. Emerg. Secur. Technol. ROBOSEC
capability, time for task execution, cost of fuel, and risks imposed 2010 - Robot. Secur. LAB-RS 2010 - Learn. Adapt. Behav. Robot. Syst., 2010,
by the dynamic environment. An efficient task assignment algo- pp. 142–147.
[6] C. Chen, D. Shi, S. Cui, Y. Kang, Cloud-based UAV monitoring and management
rithm can assure cooperative and coordinated UAV operation to
framework, in: Proc. - 2018 3rd Int. Conf. Control. Robot. Cybern., CRC 2018,
meet the application requirements. The task assignment is mutu- 2018, pp. 61–66.
ally participated, discussed, and contended using multiple UAVs, [7] M.Y. Arafat, S. Moh, Location-aided delay tolerant routing protocol in UAV net-
or each UAV selects task independently in accordance to the ob- works for post-disaster operation, IEEE Access 6 (2018) 59891–59906, https://
doi.org/10.1109/ACCESS.2018.2875739.
servation of environmental constraints. Distributed algorithms em-
[8] N. Yin, R. Liu, B. Zeng, N. Liu, A review: UAV-based remote sensing, IOP
phasizes on achieving efficiency whereas they do not take global Conf. Ser., Mater. Sci. Eng. 490 (2019), https://doi.org/10.1088/1757-899X/490/
considerations into account, which may not result in accurate task 6/062014.
assignment. Coordinated algorithms, on the other hand, have a [9] G. Yang, J. Liu, C. Zhao, Z. Li, Y. Huang, H. Yu, B. Xu, X. Yang, D. Zhu, X. Zhang,
R. Zhang, H. Feng, X. Zhao, Z. Li, H. Li, H. Yang, Unmanned aerial vehicle
central controller with global information and assign tasks that
remote sensing for field-based crop phenotyping: current status and perspec-
benefit the overall system. Proper connectivity between UAVs is tives, Front. Plant Sci. 8 (2017), https://doi.org/10.3389/fpls.2017.01111.
always required while designing the multi-UAV systems. [10] M.Y. Arafat, S. Moh, Bio-inspired approaches for energy-efficient localization
and clustering in UAV networks for monitoring wildfires in remote areas, IEEE
Access 9 (2021) 18649–18669, https://doi.org/10.1109/ACCESS.2021.3053605.
8. Conclusion [11] M. Elloumi, R. Dhaou, B. Escrig, H. Idoudi, L.A. Saidane, Monitoring road traffic
with a UAV-based system, in: IEEE Wirel. Commun. Netw. Conf., WCNC 2018,
April, 2018, pp. 1–6.
In this article, we have presented a detailed review and discus- [12] R. Kellermann, T. Biehle, L. Fischer, Drones for parcel and passenger trans-
sion of the recent research and development of UAVs in industry portation: a literature review, Transp. Res. Interdiscip. Perspect. 4 (2020)
as well as academia. UAV fleets are the primary choice for many 100088, https://doi.org/10.1016/j.trip.2019.100088.
[13] H. Lu, X. Wei, H. Qian, M. Chen, A cost-efficient elastic UAV relay network
mission planning systems because of their effectiveness. We have construction method with guaranteed QoS, Ad Hoc Netw. 107 (2020), https://
reviewed the prevailing studies on UAV networks and different as- doi.org/10.1016/j.adhoc.2020.102219.
pects of UAV communications. Task assignment algorithms play a [14] W. Chen, S. Zhao, Q. Shi, Improve stability in UAV relay networks by jointly
crucial role in multi-UAV operations for proper cooperation and optimizing communication, in: Trajectory and Power, 2018 IEEE Int. Conf.
Commun. Syst., ICCS 2018, 2018, pp. 180–185.
coordination and need further research. We have extensively sur- [15] I. Rubin, R. Zhang, Placement of UAVs as communication relays aiding mobile
veyed the state-of-the-art task assignment and optimization algo- ad hoc wireless networks, in: Proc. - IEEE Mil. Commun. Conf. MILCOM, 2007,
rithms for multi-UAV based mission planning systems. We have pp. 1–7.
[16] A. Kumar, K. Sharma, H. Singh, S.G. Naugriya, S.S. Gill, R. Buyya, A drone-based
also reviewed existing algorithms with their advantages, disadvan-
networked system and methods for combating coronavirus disease (COVID-
tages, and possible enhancements in the near future. Then, all the 19) pandemic, Future Gener. Comput. Syst. 115 (2021) 1–19, https://doi.org/
task assignment algorithms for UAV networks were classified and 10.1016/j.future.2020.08.046.
compared in terms of their main principles, operational character- [17] A. Kumar, K. Sharma, H. Singh, P. Srikanth, R. Krishnamurthi, A. Nayyar,
Drone-based social distancing, sanitization, inspection, monitoring, and con-
istics, and performance metrics.
trol room for COVID-19, Artif. Intell. Mach. Learn. COVID-19 (2021), https://
This review has exposed the blossoming attention of re- doi.org/10.1007/978-3-030-60188-1_8.
searchers and engineers towards the harnessing and utilization of [18] J. Euchi, Do drones have a realistic place in a pandemic fight for deliver-
the full potential of UAVs. A number of deep RL algorithms have ing medical supplies in healthcare systems problems?, Chin. J. Aeronaut. 34
(2021) 182–190, https://doi.org/10.1016/j.cja.2020.06.006.
been suggested in the literature for the mitigation of collision, tra-
[19] M. Azmat, S. Kummer, Potential applications of unmanned ground and aerial
jectory optimization, etc. Nevertheless, there is significant scope vehicles to mitigate challenges of transport and logistics-related critical suc-
for designing task assignment algorithms for multi-UAV commu- cess factors in the humanitarian supply chain, Asian J. Sustain. Soc. Responsib.
nications. We believe that this review will assist researchers and 5 (2020), https://doi.org/10.1186/s41180-020-0033-7.
[20] A. Fotouhi, H. Qiang, M. Ding, M. Hassan, L.G. Giordano, A. Garcia-Rodriguez, J.
engineers in conducting further research on task assignment and
Yuan, Survey on UAV cellular communications: practical aspects, standardiza-
optimization algorithms. tion advancements, regulation, and security challenges, IEEE Commun. Surv.
Tutor. 21 (2019) 3417–3442, https://doi.org/10.1109/COMST.2019.2906228.
[21] H. Shakhatreh, A.H. Sawalmeh, A. Al-Fuqaha, Z. Dou, E. Almaita, I. Khalil, N.S.
Declaration of competing interest Othman, A. Khreishah, M. Guizani, Unmanned Aerial Vehicles (UAVs): a sur-
vey on civil applications and key research challenges, IEEE Access 7 (2019)
48572–48634, https://doi.org/10.1109/ACCESS.2019.2909530.
The authors declare that they have no known competing finan-
[22] S. Hayat, E. Yanmaz, R. Muzaffar, Survey on unmanned aerial vehicle networks
cial interests or personal relationships that could have appeared to for civil applications: a communications viewpoint, IEEE Commun. Surv. Tutor.
influence the work reported in this paper. 18 (2016) 2624–2661, https://doi.org/10.1109/COMST.2016.2560343.

24
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

[23] L. Gupta, R. Jain, G. Vaszkun, Survey of important issues in UAV communi- [47] T. Research, Global drones market by type (vtol/rotary, fixed wing, etc.), by
cation networks, IEEE Commun. Surv. Tutor. 18 (2016) 1123–1152, https:// segment (consumer, commercial & military), by application (aerial photogra-
doi.org/10.1109/COMST.2015.2495297. phy, agriculture, industrial inspection, etc.), by payload, by region, competi-
[24] D.C. Tsouros, A. Triantafyllou, S. Bibi, P.G. Sarigannidis, Data acquisition and tion forecast & opportunities, https://www.techsciresearch.com/report/global-
analysis methods in UAV-based applications for precision agriculture, in: drones-market/1345.html, 2013. (Accessed 15 November 2019), online.
Proc. - 15th Annu. Int. Conf. Distrib. Comput. Sens. Syst., DCOSS 2019, 2019, [48] M. Elloumi, R. Dhaou, B. Escrig, H. Idoudi, L.A. Saidane, Monitoring road traffic
pp. 377–384. with a UAV-based system, in: IEEE Wirel. Commun. Netw. Conf., WCNC 2018,
[25] D.C. Tsouros, S. Bibi, P.G. Sarigiannidis, A review on UAV-based applica- April, 2018, pp. 1–6.
tions for precision agriculture, Information 10 (2019), https://doi.org/10.3390/ [49] H. Liu, X. Li, M. Fan, G. Wu, W. Pedrycz, P.N. Suganthan, An autonomous path
info10110349. planning method for unmanned aerial vehicle based on a tangent intersec-
[26] X. Li, Y.D. Zhang, Multi-source cooperative communications using multiple tion and target guidance strategy, IEEE Trans. Intell. Transp. Syst. (2020) 1–13,
small relay UAVs, in: 2010 IEEE Globecom Work, GC’10, 2010, pp. 1805–1810. https://doi.org/10.1109/tits.2020.3030444.
[27] S. Yin, Y. Zhao, L. Li, UAV-assisted cooperative communications with power- [50] H. Liu, X. Li, M. Fan, G. Wu, W. Pedrycz, P.N. Suganthan, An autonomous path
splitting SWIPT, in: 2018 IEEE Int. Conf. Commun. Syst., ICCS 2018, 2018, planning method for unmanned aerial vehicle based on a tangent intersec-
pp. 162–167. tion and target guidance strategy, IEEE Trans. Intell. Transp. Syst. (2020) 1–13,
[28] J.W. Curtis, R. Murphey, Simultaneous area search and task assignment for a https://doi.org/10.1109/tits.2020.3030444.
team of cooperative agents, in: AIAA Guid. Navig. Control Conf. Exhib., 2003. [51] C. Rasche, C. Stern, L. Kleinjohann, B. Kleinjohann, A distributed multi-UAV
[29] Y. Jin, A.A. Minai, M.M. Polycarpou, Cooperative real-time search and task as- path planning approach for 3D environments, in: ICARA 2011 - Proc. 5th Int.
signment in UAV teams, in: 42nd IEEE International Conference on Decision Conf. Autom. Robot. Appl., 2011, pp. 7–12.
and Control (IEEE Cat. No. 03CH37475), 2003, pp. 7–12. [52] Y. Zeng, R. Zhang, T.J. Lim, Wireless communications with unmanned aerial
[30] J. Wu, X. Su, UAV task assignment algorithm based on splitting and reorgani- vehicles: opportunities and challenges, IEEE Commun. Mag. 54 (2016) 36–42,
zation, in: Proc. - 2019 IEEE Int. Conf. Ubiquitous Comput. Commun. Data Sci. https://doi.org/10.1109/MCOM.2016.7470933.
Comput. Intell. Smart Comput. Netw. Serv., IUCC/DSCI/SmartCNS 2019, 2019, [53] A. Al-Hourani, S. Kandeepan, A. Jamalipour, Modeling air-to-ground path loss
pp. 35–42. for low altitude platforms in urban environments, in: 2014 IEEE Glob. Com-
[31] F. Semiz, Task assignment and scheduling in UAV mission planning, The- mun. Conf., GLOBECOM 2014, 2014, pp. 2898–2904.
sis [Online] Middle East Technical University, School of Natural and Ap- [54] W.C. Stirling, M.A. Goodrich, Conditional preferences for social systems, Proc.
plied Sciences, 2015, https://open.metu.edu.tr/bitstream/handle/11511/25002/ IEEE Int. Conf. Syst. Man Cybern. 2 (2001) 995–1000, https://doi.org/10.1109/
index.pdf. icsmc.2001.973048.
[32] F. Jiang, F. Pourpanah, Q. Hao, Design, implementation, and evaluation of a [55] K. Savla, E. Frazzoli, F. Bullo, Traveling salesperson problems for the Dubins
neural-network-based quadcopter UAV system, in: IEEE Transactions on In- vehicle, IEEE Trans. Autom. Control 53 (2008) 1378–1391, https://doi.org/10.
dustrial Electronics, 2020. 1109/TAC.2008.925814.
[33] P. Wu, J. Xie, J. Chen, Safe path planning for unmanned aerial vehicle under [56] L. Wang, J. Liu, J. Hu, Q. Zhuge, E.H.M. Sha, Optimal assignment for tree-
location uncertainty, in: IEEE Int. Conf. Control Autom. ICCA, 2020-October, structure task graph on heterogeneous multicore systems considering time
2020, pp. 342–347. constraint, in: Proc. - IEEE 6th Int. Symp. Embed. Multicore SoCs, MCSoC 2012,
[34] J.J. Wang, Y.F. Zhang, L. Geng, J.Y.H. Fuh, S.H. Teo, Mission planning for het- 2012, pp. 121–127.
erogeneous tasks with heterogeneous UAVs, in: 2014 13th Int. Conf. Control [57] S. An, H.J. Kim, Simultaneous task assignment and path planning using mixed-
Autom. Robot. Vision, ICARCV 2014, 2014, pp. 1484–1489. integer linear programming and potential field method, in: Int. Conf. Control.
[35] F. Semiz, Task assignment and scheduling in UAV mission planning, The- Autom. Syst., 2013, pp. 1845–1848.
sis [Online], Middle East Technical University, School of Natural and Ap- [58] Y. Lu, D.Z. Chen, A new exact algorithm for the weapon-target assignment
plied Sciences, 2015, https://open.metu.edu.tr/bitstream/handle/11511/25002/ problem, Omega (United Kingdom) 98 (2021) 102138, https://doi.org/10.1016/
index.pdf. j.omega.2019.102138.
[36] S.H. Alsamhi, O. Ma, M.S. Ansari, F.A. Almalki, Survey on collaborative smart [59] Y. Qiu, H. Jiang, Q. Li, X. Dong, Z. Ren, Application of an adapted genetic algo-
drones and internet of things for improving smartness of smart cities, rithm on task allocation problem of multiple UAVs, in: 2018 IEEE CSAA Guid.
IEEE Access 7 (2019) 128125–128152, https://doi.org/10.1109/ACCESS.2019. Navig. Control Conf., CGNCC 2018, 2018.
2934998. [60] H. Falaghi, M.R. Haghifam, ACO based algorithm for distributed generation
[37] V. Hassija, V. Chamola, A. Agrawal, A. Goyal, N.C. Luong, D. Niyato, F.R. Yu, sources allocation and sizing in distribution systems, in: 2007 IEEE Lausanne
M. Guizani, Fast, reliable, and secure drone communication: a comprehensive POWERTECH, Proc., 2007, pp. 555–560.
survey, IEEE Commun. Surv. Tutor. (2021) 1, https://doi.org/10.1109/comst. [61] D. Ben Noureddine, A. Gharbi, S. Ben Ahmed, Multi-agent deep reinforcement
2021.3097916. learning for task allocation in dynamic environment, in: ICSOFT 2017 - Proc.
[38] S. Hayat, E. Yanmaz, R. Muzaffar, Survey on unmanned aerial vehicle networks 12th Int. Conf. Softw. Technol., 2017, pp. 17–26.
for civil applications: a communications viewpoint, IEEE Commun. Surv. Tutor. [62] T. Issac, S. Silas, E.B. Rajsingh, Investigations on PSO based task assignment al-
18 (2016) 2624–2661, https://doi.org/10.1109/COMST.2016.2560343. gorithms for heterogeneous wireless sensor network, in: 2nd Int. Conf. Signal
[39] D. Shumeye Lakew, U. Sa’Ad, N.N. Dao, W. Na, S. Cho, Routing in flying ad Process. Commun. ICSPC 2019 - Proc., 2019, pp. 89–93.
hoc networks: a comprehensive survey, IEEE Commun. Surv. Tutor. 22 (2020) [63] L. Lei, G. Shen, L. Zhang, Z. Li, Toward intelligent cooperation of UAV swarms:
1071–1120, https://doi.org/10.1109/COMST.2020.2982452. when machine learning meets digital twin, IEEE Netw. 35 (2021) 386–392,
[40] H. Wang, H. Zhao, J. Zhang, D. Ma, J. Li, J. Wei, Survey on unmanned aerial https://doi.org/10.1109/MNET.011.2000388.
vehicle networks: a cyber physical system perspective, IEEE Commun. Surv. [64] S. Yin, Y. Zhao, L. Li, UAV-assisted cooperative communications with power-
Tutor. 22 (2020) 1027–1070, https://doi.org/10.1109/COMST.2019.2962207. splitting SWIPT, in: 2018 IEEE Int. Conf. Commun., Syst. ICCS 2018, 2018,
[41] A.A. Khuwaja, Y. Chen, N. Zhao, M.S. Alouini, P. Dobbins, A survey of chan- pp. 162–167.
nel modeling for UAV communications, IEEE Commun. Surv. Tutor. 20 (2018) [65] Y. Li, R. Zhang, J. Zhang, L. Yang, Cooperative jamming via spectrum sharing
2804–2821, https://doi.org/10.1109/COMST.2018.2856587. for secure UAV communications, IEEE Wirel. Commun. Lett. 9 (2020) 326–330,
[42] M.Y. Arafat, S. Poudel, S. Moh, Medium access control protocols for flying ad https://doi.org/10.1109/LWC.2019.2953725.
hoc networks: a review, IEEE Sens. J. 21 (2021) 4097–4121, https://doi.org/10. [66] T. Huang, Y. Wang, X. Cao, D. Xu, Multi-UAV mission planning method,
1109/JSEN.2020.3034600. in: 2020 3rd International Conference on Unmanned Systems, ICUS, 2020,
[43] M.A. Khan, A. Safi, I.M. Qureshi, I.U. Khan, Flying ad-hoc networks (FANETs): pp. 325–330.
a review of communication architectures, and routing protocols, in: 2017 1st [67] J. Liu, Z. Shi, Y. Zhang, A new method of UAVs multi-target task assignment,
Int. Conf. Latest Trends Electr. Eng. Comput. Technol., INTELLECT 2017, 2018- in: DEStech Trans. Eng. Technol. Res., 2018, pp. 388–394.
January, 2018, pp. 1–9. [68] A. Bahabry, H. Ghazzai, G. Vesonder, Y. Massoud, Space-time low complexity
[44] N. Hossein Motlagh, T. Taleb, O. Arouk, Low-altitude unmanned aerial algorithms for scheduling a fleet of UAVs in smart cities using dimensionality
vehicles-based internet of things services: comprehensive survey and future reduction approaches, in: SysCon 2019 - 13th Annu. IEEE Int. Syst. Conf. Proc.,
perspectives, IEEE Int. Things J. 3 (2016) 899–922, https://doi.org/10.1109/ 2019.
JIOT.2016.2612119. [69] U. Challita, W. Saad, C. Bettstetter, Interference management for cellular-
[45] R. Shakeri, M.A. Al-Garadi, A. Badawy, A. Mohamed, T. Khattab, A.K. Al-Ali, connected UAVs: a deep reinforcement learning approach, IEEE Trans. Wirel.
K.A. Harras, M. Guizani, Design challenges of multi-UAV systems in cyber- Commun. 18 (2019) 2125–2140, https://doi.org/10.1109/TWC.2019.2900035.
physical applications: a comprehensive survey and future directions, IEEE [70] F. Zhou, R.Q. Hu, Z. Li, Y. Wang, Mobile edge computing in unmanned aerial
Commun. Surv. Tutor. 21 (2019) 3340–3385, https://doi.org/10.1109/COMST. vehicle networks, IEEE Wirel. Commun. 27 (2020) 140–146, https://doi.org/
2019.2924143. 10.1109/MWC.001.1800594.
[46] F. Lagum, I. Bor-Yaliniz, H. Yanikomeroglu, Strategic densification with UAV- [71] S.Y.M. Mahmoud, N. Mohamed, Toward a cloud platform for UAV resources
BSS in cellular networks, IEEE Wirel. Commun. Lett. 7 (2018) 384–387, https:// and services, in: Proc. - IEEE 4th Symp. Netw. Cloud Comput. Appl., NCCA
doi.org/10.1109/LWC.2017.2779483. 2015, 2015, pp. 23–30.

25
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

[72] P. Sánchez, R. Casado, A. Bermúdez, Real-time collision-free navigation of [100] R. Lahdelma, P. Salminen, Pseudo-criteria versus linear utility function in
multiple UAVs based on bounding boxes, Electronics 9 (2020) 1–19, https:// stochastic multi-criteria acceptability analysis, Eur. J. Oper. Res. 141 (2002)
doi.org/10.3390/electronics9101632. 454–469, https://doi.org/10.1016/S0377-2217(01)00276-4.
[73] Y. Liu, X. Zhang, Y. Zhang, X. Guan, Collision free 4D path planning for multi- [101] R. Lahdelma, P. Salminen, J. Hokkanen, Locating a waste treatment facility
ple UAVs based on spatial refined voting mechanism and PSO approach, Chin. by using stochastic multicriteria acceptability analysis with ordinal criteria,
J. Aeronaut. 32 (2019) 1504–1519, https://doi.org/10.1016/j.cja.2019.03.026. Eur. J. Oper. Res. 142 (2002) 345–356, https://doi.org/10.1016/S0377-2217(01)
[74] N. Ahmadian, G.J. Lim, M. Torabbeigi, S.J. Kim, Collision-free multi-UAV flight 00303-4.
scheduling for power network damage assessment, in: 2019 Int. Conf. Un- [102] R. Lahdelma, K. Miettinen, P. Salminen, Reference point approach for multi-
manned Aircr. Syst., ICUAS 2019, 2019, pp. 794–798. ple decision makers, Eur. J. Oper. Res. 164 (2005) 785–791, https://doi.org/10.
[75] E. Kalantari, M.Z. Shakir, H. Yanikomeroglu, A. Yongacoglu, Backhaul-aware ro- 1016/j.ejor.2004.01.030.
bust 3D drone placement in 5G+ wireless networks, in: 2017 IEEE Int. Conf. [103] T. Tervonen, R. Lahdelma, J.A. Dias, J. Figueira, P. Salminen, SMAA-TRI: a Pa-
Commun. Work., ICC Work, 2017, 2017, pp. 109–114. rameter Stability Analysis Method for ELECTRE TRI, 2007.
[76] S. Kumar, S. Suman, S. De, Backhaul and delay-aware placement of UAV- [104] R. Lahdelma, P. Salminen, Prospect theory and stochastic multicriteria accept-
enabled base station, in: INFOCOM 2018 - IEEE Conf. Comput. Commun. ability analysis (SMAA), Omega 37 (2009) 961–971, https://doi.org/10.1016/j.
Work., 2018, pp. 634–639. omega.2008.09.001.
[77] U. Challita, W. Saad, Network formation in the sky: unmanned aerial vehi- [105] A.S. Kangas, J. Kangas, R. Lahdelma, P. Salminen, Using SMAA-2 method with
cles for multi-hop wireless backhauling, in: 2017 IEEE Glob. Commun. Conf., dependent uncertainties for strategic forest planning, Forest Policy Econ. 9
GLOBECOM 2017 - Proc. 2018-January, 2017, pp. 1–6. (2006) 113–125, https://doi.org/10.1016/j.forpol.2005.03.012.
[78] M. Gapeyenko, V. Petrov, D. Moltchanov, S. Andreev, N. Himayat, Y. Kouch- [106] J. Hokkanen, R. Lahdelma, P. Salminen, Multicriteria decision support in a
eryavy, Flexible and reliable UAV-assisted backhaul operation in 5G mm-wave technology competition for cleaning polluted soil in Helsinki, J. Environ.
cellular networks, IEEE J. Sel. Areas Commun. 36 (2018) 2486–2496, https:// Manag. 60 (2000) 339–348, https://doi.org/10.1006/jema.2000.0389.
doi.org/10.1109/JSAC.2018.2874145. [107] A. Menou, A. Benallou, R. Lahdelma, P. Salminen, Decision support for cen-
[79] B. Galkin, J. Kibiłda, L.A. DaSilva, Backhaul for low-altitude UAVs in urban en- tralizing cargo at a Moroccan airport hub using stochastic multicriteria ac-
vironments, in: IEEE Int. Conf. Commun. 2018-May, 2018, pp. 1–6. ceptability analysis, Eur. J. Oper. Res. 204 (2010) 621–629, https://doi.org/10.
[80] L. Yang, H. Yao, J. Wang, C. Jiang, A. Benslimane, Y. Liu, Multi-UAV-enabled 1016/j.ejor.2009.11.021.
load-balance mobile-edge computing for IoT networks, IEEE Int. Things J. 7 [108] M.M. Rahman, J.V. Paatero, R. Lahdelma, Evaluation of choices for sustainable
(2020) 6898–6908, https://doi.org/10.1109/JIOT.2020.2971645. rural electrification in developing countries: a multicriteria approach, Energy
[81] S. Nathaniel, S.H.S. Ariffin, A. Farzamnia, A.J. Adegboyega, Multi-criteria load Policy 59 (2013) 589–599, https://doi.org/10.1016/j.enpol.2013.04.017.
balancing decision algorithm for LTE network, in: 2014 4th Int. Conf. Eng. [109] A. Pesola, A. Serkkola, R. Lahdelma, P. Salminen, Multicriteria evaluation of
Technol. Technopreneuship, ICE2T 2014, 2014-August, 2015, pp. 57–62. alternatives for remote monitoring systems of municipal buildings, Energy
[82] S. Mishra, A. Trivedi, Relay selection with channel allocation for cognitive ra- Build. 72 (2014) 229–237, https://doi.org/10.1016/j.enbuild.2013.10.037.
dio relay channels in CRN, in: IFIP Int. Conf. Wirel. Opt. Commun. Networks,
[110] T. Tervonen, R. Lahdelma, Implementing stochastic multicriteria acceptability
WOCN., 2014, pp. 14–17.
analysis, Eur. J. Oper. Res. 178 (2007) 500–513, https://doi.org/10.1016/j.ejor.
[83] B.S. Roh, M.H. Han, J.H. Ham, K. Il Kim, Q-LBR: Q-learning based load balanc-
2005.12.037.
ing routing for UAV-assisted VANET, Sensors (Switzerland) 20 (2020) 1–17,
[111] T. Tervonen, JSMAA: open source software for SMAA computations, Int. J. Syst.
https://doi.org/10.3390/s20195685.
Sci. 45 (2014) 69–81, https://doi.org/10.1080/00207721.2012.659706.
[84] S. Jung, K. Il Kim, B. Roh, J.H. Ham, Load balancing algorithm for multiple
[112] M. Alighanbari, L.F. Bertuccelli, J.P. How, Filter-embedded UAV task assignment
UAVs relayed tactical ad hoc networks, in: Proc. - Int. Comput. Softw. Appl.
algorithms for dynamic environments, in: Collect. Tech. Pap. - AIAA Guid.
Conf., vol. 1, 2019, pp. 944–945.
Navig. Control Conf., vol. 4, 2004, pp. 2560–2574.
[85] E. Yanmaz, R. Kuschnig, C. Bettstetter, Achieving air-ground communications
[113] R.W. Beard, T.W. McLain, M. Goodrich, Coordinated target assignment and
in 802.11 networks with three-dimensional aerial mobility, in: Proc. - IEEE
intercept for unmanned air vehicles, in: Proceedings 2002 IEEE Interna-
INFOCOM, 2013, pp. 120–124.
tional Conference on Robotics and Automation (Cat. No. 02CH37292), 2002,
[86] Z. Huang, C.C. Shen, A comparison study of omnidirectional and directional
pp. 2581–2586.
MAC protocols for ad hoc networks, in: Conf. Rec. / IEEE Glob. Telecommun.
[114] M.A. Goodrich, W.C. Stirling, R.L. Frost, A theory of satisficing decisions and
Conf., vol. 1, 2002, pp. 57–61.
control, IEEE Trans. Syst. Man Cybern., Part A, Syst. Hum. 28 (1998) 763–779,
[87] C. Schumacher, P.R. Chandler, S.J. Rasmussen, Task allocation for wide area
https://doi.org/10.1109/3468.725348.
search munitions via iterative network flow, in: AIAA Guid. Navig. Control
[115] M.A. Goodrich, W.C. Stirling, R.L. Frost, Model predictive satisficing fuzzy logic
Conf. Exhib., 2002, pp. 1917–1922.
control, IEEE Trans. Fuzzy Syst. 7 (1999) 319–332, https://doi.org/10.1109/91.
[88] Y. Jin, A. Minai, M. Polycarpou, Cooperative real-time search and task alloca-
771087.
tion in UAV teams, in: Proceedings of the IEEE Conference on Decision and
[116] W.C. Stirling, M.A. Goodrich, Satisficing games, Inf. Sci. (NY) 114 (1999)
Control, Maui, Hawaii, Dec. 2003, pp. 7–12.
255–280, https://doi.org/10.1016/S0020-0255(98)10048-8.
[89] E. King, Y. Kuwata, M. Alighanbari, J. How, Coordination and control exper-
iments on a multi-vehicle testbed, in: Proceedings of the 2004 American [117] Y.K. Tun, Y.M. Park, N.H. Tran, W. Saad, S.R. Pandey, C.S. Hong, Energy-efficient
Control Conference, 2004. resource management in UAV-assisted mobile edge computing, IEEE Commun.
[90] B.Y.B. Bethke, M. Valenti, J.P. How, UAV Task Assignment, 2008, pp. 39–44. Lett. 25 (2020) 249–253, https://doi.org/10.1109/lcomm.2020.3026033.
[91] C. Schumacher, M. Pachter, P. Chandler, L.S. Pachter, UAV task assignment with [118] N.H. Motlagh, M. Bagaa, T. Taleb, Energy and delay aware task assign-
timing constraints, in: AIAA Guid. Navig. Control Conf. Exhib., 2003. ment mechanism for UAV-based IoT platform, IEEE Int. Things J. 6 (2019)
[92] C. Schumacher, P.R. Chandler, S.J. Rasmussen, D. Walker, Task allocation for 6523–6536, https://doi.org/10.1109/JIOT.2019.2907873.
wide area search munitions with variable path length, in: Proc. Am. Control [119] M. Alighanbari, J.P. How, Cooperative task assignment of unmanned aerial ve-
Conf., vol. 4, 2003, pp. 3472–3477. hicles in adversarial environments, in: Proc. Am. Control Conf., vol. 7, 2005,
[93] C. Schumacher, P.R. Chandler, S.J. Rasmussen, D. Walker, Path elongation for pp. 4661–4666.
UAV task assignment, in: AIAA Guid. Navig. Control Conf. Exhib., 2003. [120] T. Chang, D. Kong, N. Hao, K. Xu, G. Yang, Solving the dynamic weapon tar-
[94] Y. Chen, D. Yang, J. Yu, Multi-UAV task assignment with parameter and time- get assignment problem by an improved artificial bee colony algorithm with
sensitive uncertainties using modified two-part Wolf pack search algorithm, heuristic factor initialization, Appl. Soft Comput. J. 70 (2018) 845–863, https://
IEEE Trans. Aerosp. Electron. Syst. 54 (2018) 2853–2872, https://doi.org/10. doi.org/10.1016/j.asoc.2018.06.014.
1109/TAES.2018.2831138. [121] R.A. Murphey, Target-Based Weapon Target Assignment Problems, 2000,
[95] Y. Lu, Y. Ma, J. Wang, L. Han, Task assignment of UAV swarm based on pp. 39–53.
wolf pack algorithm, Appl. Sci. 10 (2020) 1–17, https://doi.org/10.3390/ [122] R.K. Ahuja, A. Kumar, K.C. Jha, J.B. Orlin, Exact and heuristic algorithms for the
app10238335. weapon-target assignment problem, Oper. Res. 55 (2007) 1136–1146, https://
[96] J. Autenrieb, N. Strawa, H.S. Shin, J.H. Hong, A mission planning and task allo- doi.org/10.1287/opre.1070.0440.
cation framework for multi-UAV swarm coordination, in: Int. Work. Res. Educ. [123] I.P. De Medeiros, L.R. Rodrigues, S.A. Embraer, R. Santos, PHM-based multi-
Dev. Unmanned Aer. Syst., RED-UAS 2019, 2019, pp. 297–304. UAV task assignment, in: IEEE International Systems Conference Proceedings,
[97] X. Hu, J. Cheng, H. Luo, Task assignment for multi-UAV under severe uncer- 2014, pp. 2–9.
tainty by using stochastic multicriteria acceptability analysis, Math. Probl. Eng. [124] J. Bellingham, M. Tillerson, A. Richards, J.P. How, Multi-Task Allocation and
2015 (2015), https://doi.org/10.1155/2015/249825. Path Planning for Cooperating UAVs, 2003, pp. 23–41.
[98] R. Lahdelma, J. Hokkanen, P. Salminen, SMAA - stochastic multiobjective ac- [125] P.R. Chandler, M. Pachter, S.R. Rasmussen, C. Schumacher, Distributed control
ceptability analysis, Eur. J. Oper. Res. 106 (1998) 137–143, https://doi.org/10. for multiple UAVs with strongly coupled tasks, in: AIAA Guid. Navig. Control
1016/S0377-2217(97)00163-X. Conf. Exhib., 2003, pp. 1–9.
[99] R. Lahdelma, P. Salminen, SMAA-2: stochastic multicriteria acceptability anal- [126] C. Chen, W. Bao, T. Men, W. Zhou, D. Liu, L. Ma, A damage-tolerant task as-
ysis for group decision making, Oper. Res. 49 (2001) 444–454, https://doi.org/ signment algorithm for UAV swarm in confrontational environments, Int. J.
10.1287/opre.49.3.444.11220. Aerosp. Eng. 2020 (2020), https://doi.org/10.1155/2020/8878136.

26
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

[127] H. Choi, Y. Kim, H. Kim, Genetic algorithm based decentralized task assign- [151] S. Yan, X. Zhou, L. Wang, H. Wang, GA-based automated task assignment on
ment for multiple unmanned aerial vehicles in dynamic environments, Int. network processors, in: Proc. Int. Conf. Parallel Distrib. Syst. - ICPADS, vol. 1,
J. Aeronaut. Space Sci. 12 (2011) 163–174, https://doi.org/10.5139/IJASS.2011. 2005, pp. 112–118.
12.2.163. [152] C.B.P. Del Notario, R. Baert, M. D’Hondt, Multi-objective genetic algorithm for
[128] M. Alighanbari, Task assignment algorithms for teams of UAVs in dynamic task assignment on heterogeneous nodes, Int. J. Digit. Multimed. Broadcast.
environments, Thesis [Online], Massachusetts Institute of Technology, Dept. 2012 (2012), https://doi.org/10.1155/2012/716780.
of Aeronautics and Astronautics, and Massachusetts Institute of Technology, [153] S. Yan, X. Zhou, L. Wang, H. Wang, GA-based automated task assignment on
Operations Research Center, 2004, http://hdl.handle.net/1721.1/17754. network processors, in: Proc. Int. Conf. Parallel Distrib. Syst. - ICPADS, vol. 1,
[129] H. Choi, Y. Kim, H. Kim, Genetic algorithm based decentralized task assign- 2005, pp. 112–118.
ment for multiple unmanned aerial vehicles in dynamic environments, Int. [154] Y. Eun, H. Bang, Cooperative task assignment and path planning of multiple
J. Aeronaut. Space Sci. 12 (2011) 163–174, https://doi.org/10.5139/IJASS.2011. UAVs using genetic algorithm, in: Collect. Tech. Pap. - 2007 AIAA InfoTech
12.2.163. Aerosp. Conf., vol. 3, 2007, pp. 2680–2697.
[130] F. Ye, J. Chen, Y. Tian, T. Jiang, Cooperative task assignment of a heterogeneous [155] L. Cheng, L. Zhong, S. Tian, J. Xing, Task assignment algorithm for road patrol
multi-UAV system using an adaptive genetic algorithm, Electronics 9 (2020), by multiple UAVs with multiple bases and rechargeable endurance, IEEE Ac-
https://doi.org/10.3390/electronics9040687. cess 7 (2019) 144381–144397, https://doi.org/10.1109/ACCESS.2019.2944881.
[131] X. Guo, M. Ji, D. Wen, X. Zhang, S. Tian, Task assignment and path plan- [156] G. Xu, T. Long, Z. Wang, L. Liu, Target-bundled genetic algorithm for multi-
ning for distributed multiple unmanned aerial vehicles in the “last mile”, unmanned aerial vehicle cooperative task assignment considering precedence
Xitong Gongcheng Lilun Yu Shijian/Syst. Eng. Theory Pract. 41 (2021) 946–961, constraints, Proc. Inst. Mech. Eng., G J. Aerosp. Eng. 234 (2020) 760–773,
https://doi.org/10.12011/SETP2020-0031. https://doi.org/10.1177/0954410019883106.
[132] Q. Cheng, D. Yin, J. Yang, L. Shen, An auction-based multiple constraints task [157] F. Ye, J. Chen, Y. Tian, T. Jiang, Cooperative task assignment of a heterogeneous
allocation algorithm for multi-UAV system, in: Proc. - 2016 Int. Conf. Cybern. multi-UAV system using an adaptive genetic algorithm, Electronics 9 (2020),
Robot. Control., CRC 2016, 2017, pp. 1–5. https://doi.org/10.3390/electronics9040687.
[133] M.V. Dileep, B. Yu, S. Kim, H. Oh, Task assignment for deploying unmanned [158] T. Shima, S.J. Rasmussen, A.G. Sparks, K.M. Passino, Multiple task assignments
aircraft as decoys, Int. J. Control. Autom. Syst. 18 (2020) 3204–3217, https:// for cooperating uninhabited aerial vehicles using genetic algorithms, Comput.
doi.org/10.1007/s12555-019-1073-6. Oper. Res. 33 (2006) 3252–3269, https://doi.org/10.1016/j.cor.2005.02.039.
[134] M. Alighanbari, J.P. How, Robust decentralized task assignment for cooperative [159] E. Edison, T. Shima, Integrated task assignment and path optimization for co-
UAVs, in: Collect. Tech. Pap. - AIAA Guid. Navig. Control Conf., vol. 5, 2006, operating uninhabited aerial vehicles using genetic algorithms, Comput. Oper.
pp. 3232–3247. Res. 38 (2011) 340–356, https://doi.org/10.1016/j.cor.2010.06.001.
[135] M. Alighanbari, J.P. How, Decentralized task assignment for unmanned aerial [160] Q. Deng, J. Yu, N. Wang, Cooperative task assignment of multiple hetero-
vehicles, in: Proc. 44th IEEE Conf. Decis. Control. Eur. Control Conf., CDC-ECC geneous unmanned aerial vehicles using a modified genetic algorithm with
’05, 2005, pp. 5668–5673. multi-type genes, Chin. J. Aeronaut. 26 (2013) 1238–1250, https://doi.org/10.
[136] J. Bellingham, M. Tillerson, A. Richards, J.P. How, Multi-Task Allocation and 1016/j.cja.2013.07.009.
[161] Z. Jia, J. Yu, X. Ai, X. Xu, D. Yang, Cooperative multiple task assignment
Path Planning for Cooperating UAVs, 2003, pp. 23–41.
problem with stochastic velocities and time windows for heterogeneous un-
[137] L.F. Bertuccelli, H.L. Choi, P. Cho, J.P. How, Real-time multi-UAV task assign-
manned aerial vehicles using a genetic algorithm, Aerosp. Sci. Technol. 76
ment in dynamic and uncertain environments, in: AIAA Guid. Navig. Control
(2018) 112–125, https://doi.org/10.1016/j.ast.2018.01.025.
Conf. Exhib., 2009.
[162] Z. Wang, B. Wang, Y. Wei, P. Liu, L. Zhang, Cooperative multi-task assignment
[138] J. Chen, X. Qing, F. Ye, et al., Consensus-based bundle algorithm with local
of multiple UAVs with improved genetic algorithm based on beetle antennae
replanning for heterogeneous multi-UAV system in the time-sensitive and dy-
search, in: Chinese Control Conf., CCC 2020, July, 2020, pp. 1605–1610.
namic environment, J. Supercomput. (2021), https://doi.org/10.1007/s11227-
[163] Z. Wang, L. Liu, T. Long, Y. Wen, Multi-UAV reconnaissance task allocation
021-03940-z.
for heterogeneous targets using an opposition-based genetic algorithm with
[139] X. Fu, P. Feng, B. Li, X. Gao, A two-layer task assignment algorithm for UAV
double-chromosome encoding, Chin. J. Aeronaut. 31 (2018) 339–350, https://
swarm based on feature weight clustering, Int. J. Aerosp. Eng. 2019 (2019),
doi.org/10.1016/j.cja.2017.09.005.
https://doi.org/10.1155/2019/3504248.
[164] H. Xiang, M. Peng, Y. Sun, S. Yan, Mode selection and resource allocation
[140] J. Balicki, Numerical experiments on Pareto-optimal task assignment repre-
in sliced fog radio access networks: a reinforcement learning approach, IEEE
sentations by tabu-based evolutionary algorithm, WSEAS Trans. Inf. Sci. Appl.
Trans. Veh. Technol. 69 (2020), https://doi.org/10.1109/TVT.2020.2972999.
5 (2008) 695–705, http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.
[165] Y. Wei, L. Pan, S. Liu, L. Wu, X. Meng, DRL-scheduling: an intelligent QoS-
551.2698&rep=rep1&type=pdf.
aware job scheduling framework for applications in clouds, IEEE Access 6
[141] J. Balicki, Immune systems in multi-criterion evolutionary algorithm for task
(2018) 55112–55125, https://doi.org/10.1109/ACCESS.2018.2872674.
assignments in distributed computer system, in: Lect. Notes Comput. Sci. (In-
[166] W. Zhu, T.Y. Liang, C.K. Shieh, Hopfield neural network based task mapping
cluding Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 3528,
method, Comput. Commun. 22 (1999) 1068–1079, https://doi.org/10.1016/
2005, pp. 51–56.
S0140-3664(99)00086-9.
[142] J. Li, J.Q. Zhang, Q. Kang, C.J. Jiang, Team of Bayesian optimization algo- [167] A. Zhu, S.X. Yang, A neural network approach to dynamic task assignment of
rithms to solve task assignment problems in heterogeneous computing sys- multirobots, IEEE Trans. Neural Netw. 17 (2006) 1278–1286, https://doi.org/
tems, in: Conf. Proc. - IEEE Int. Conf. Syst. Man Cybern., 2014-January, 2014, 10.1109/TNN.2006.875994.
pp. 127–132. [168] L. Hou, L. Lei, K. Zheng, X. Wang, A Q-learning-based proactive caching strat-
[143] Y. Khosiawan, Y. Park, I. Moon, J.M. Nilakantan, I. Nielsen, Task scheduling egy for non-safety related services in vehicular networks, IEEE Int. Things J. 6
system for UAV operations in indoor environment, Neural Comput. Appl. 31 (2019) 4512–4520, https://doi.org/10.1109/JIOT.2018.2883762.
(2019) 5431–5459, https://doi.org/10.1007/s00521-018-3373-9. [169] S. Sheng, P. Chen, Z. Chen, L. Wu, Y. Yao, Deep reinforcement learning-based
[144] Y. Khosiawan, I. Nielsen, A system of UAV application in indoor environ- task scheduling in iot edge computing, Sensors 21 (2021) 1–19, https://doi.
ment, Prod. Manuf. Res. 4 (2016) 2–22, https://doi.org/10.1080/21693277. org/10.3390/s21051666.
2016.1195304. [170] J. Wang, J. Wang, Task Assignment Based on a Dual Neural Network, Springer
[145] X. Du, Q. Guo, H. Li, Y. Zhang, Multi-UAVs cooperative task assignment and International Publishing, 2018.
path planning scheme, J. Phys. Conf. Ser. 1856 (2021), https://doi.org/10.1088/ [171] M. Melnik, D. Nasonov, Workflow scheduling using neural networks and re-
1742-6596/1856/1/012016. inforcement learning, Proc. Comput. Sci. 156 (2019) 29–36, https://doi.org/10.
[146] W. Liu, X. Zheng, H. Garg, Multi-UAV cooperative task assignment based on 1016/j.procs.2019.08.126.
orchard picking algorithm, Int. J. Comput. Intell. Syst. 14 (2021) 1461–1467, [172] J. Su, J. Wang, S. Liu, N. Zhang, C. Li, A method for efficient task assign-
https://doi.org/10.2991/ijcis.d.210423.003. ment based on the satisfaction degree of knowledge, Complexity 2020 (2020),
[147] J. Shi, Y. Wang, J. Tian, Research on cooperative task assignment of UAV for- https://doi.org/10.1155/2020/3543782.
mation, in: 2017 2nd International Conference on Computer Engineering, In- [173] X. Zhao, Q. Zong, B. Tian, B. Zhang, M. You, Fast task allocation for heteroge-
formation Science and Internet Technology, 2017, pp. 489–496. neous unmanned aerial vehicles through reinforcement learning, Aerosp. Sci.
[148] X. Du, Q. Guo, H. Li, Y. Zhang, Multi-UAVs cooperative task assignment and Technol. 92 (2019) 588–594, https://doi.org/10.1016/j.ast.2019.06.024.
path planning scheme, J. Phys. Conf. Ser. 1856 (2021), https://doi.org/10.1088/ [174] K. Xiao, J. Lu, Y. Nie, L. Ma, X. Wang, G. Wang, A benchmark for multi-UAV
1742-6596/1856/1/012016. task assignment of an extended team orienteering problem, http://arxiv.org/
[149] H. Choi, Y. Kim, H. Kim, Genetic algorithm based decentralized task assign- abs/2009.00363, 2020.
ment for multiple unmanned aerial vehicles in dynamic environments, Int. [175] M. Alighanbari, J.P. How, A robust approach to the UAV task assignment prob-
J. Aeronaut. Space Sci. 12 (2011) 163–174, https://doi.org/10.5139/IJASS.2011. lem, Int. J. Robust Nonlinear Control 18 (2008) 118–134, https://doi.org/10.
12.2.163. 1002/rnc.1268.
[150] B. Wang, J. Li, Load balancing task scheduling based on Multi-Population Ge- [176] N.N. Ei, S.W. Kang, M. Alsenwi, Y.K. Tun, C.S. Hong, Multi-UAV-assisted MEC
netic Algorithm in cloud computing, in: Chinese Control Conf., CCC 2016, system: joint association and resource management framework, in: Int. Conf.
August, 2016, pp. 5261–5266. Inf. Netw., 2021-January, 2021, pp. 213–218.

27
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

[177] X. Hu, K.K. Wong, K. Yang, Z. Zheng, UAV-assisted relaying and edge comput- [203] S. Poudel, S. Moh, J. Shen, Residual energy-based clustering in UAV-
ing: scheduling and trajectory optimization, IEEE Trans. Wirel. Commun. 18 aided wireless sensor networks for surveillance and monitoring applications,
(2019) 4738–4752, https://doi.org/10.1109/TWC.2019.2928539. https://doi.org/10.20517/jsss.2020.23, 2021.
[178] L. Zhang, Z. Zhao, Q. Wu, H. Zhao, H. Xu, X. Wu, Energy-aware dynamic [204] B.J.O. De Souza, M. Endler, Coordinating movement within swarms of UAVs
resource allocation in UAV assisted mobile edge computing over social inter- through mobile networks, in: 2015 IEEE Int. Conf. Pervasive Comput. Com-
net of vehicles, IEEE Access 6 (2018) 56700–56715, https://doi.org/10.1109/ mun. Work., PerCom Work 2015, 2015, pp. 154–159.
ACCESS.2018.2872753. [205] D. Kim, J. Lee, T.Q.S. Quek, Multi-layer unmanned aerial vehicle networks:
[179] H.X. Chen, Y. Nan, Y. Yang, Multi-UAV reconnaissance task assignment for het- modeling and performance analysis, IEEE Trans. Wirel. Commun. 19 (2020)
erogeneous targets based on modified symbiotic organisms search algorithm, 325–339, https://doi.org/10.1109/TWC.2019.2944378.
Sensors (Switzerland) 19 (2019), https://doi.org/10.3390/s19030734. [206] P. Perazzo, F.B. Sorbelli, M. Conti, G. Dini, C.M. Pinotti, Drone path planning for
[180] I. Lugo-Cardenas, G. Flores, S. Salazar, R. Lozano, Dubins path generation for a secure positioning and secure position verification, IEEE Trans. Mob. Comput.
fixed wing UAV, in: 2014 Int. Conf. Unmanned Aircr. Syst. ICUAS 2014 - Conf. 16 (2017) 2478–2493, https://doi.org/10.1109/TMC.2016.2627552.
Proc. 2, 2014, pp. 339–346. [207] T. Alladi, V. Chamola, N. Sahu, M. Guizani, Applications of blockchain in un-
[181] K. Savla, E. Frazzoli, F. Bullo, Traveling salesperson problems for the Dubins manned aerial vehicles: a review, Veh. Commun. 23 (2020) 100249, https://
vehicle, IEEE Trans. Autom. Control 53 (2008) 1378–1391, https://doi.org/10. doi.org/10.1016/j.vehcom.2020.100249.
1109/TAC.2008.925814. [208] Z. Mao, X.G. Yan, B. Jiang, M. Chen, Adaptive fault-tolerant sliding-mode con-
[182] C. Xu, X. Liao, J. Tan, H. Ye, H. Lu, Recent research progress of unmanned aerial trol for high-speed trains with actuator faults and uncertainties, IEEE Trans.
vehicle regulation policies and technologies in urban low altitude, IEEE Access Intell. Transp. Syst. 21 (2020) 2449–2460, https://doi.org/10.1109/TITS.2019.
8 (2020) 74175–74194, https://doi.org/10.1109/ACCESS.2020.2987622. 2918543.
[183] C. Stöcker, R. Bennett, F. Nex, M. Gerke, J. Zevenbergen, Review of the cur- [209] B. Wang, Y. Zhang, J.C. Ponsart, D. Theilliol, Fault-tolerant adaptive control
rent state of UAV regulations, Remote Sens. 9 (2017) 33–35, https://doi.org/ allocation for unmanned multirotor helicopter, IFAC-PapersOnLine 50 (2017)
10.3390/rs9050459. 5269–5274, https://doi.org/10.1016/j.ifacol.2017.08.609.
[184] M. Park, S.G. Lee, S. Lee, Dynamic topology reconstruction protocol for UAV [210] S. Mahmoud, N. Mohamed, J. Al-Jaroodi, Integrating UAVs into the cloud using
swarm networking, Symmetry (Basel) 12 (2020) 1–17, https://doi.org/10.3390/ the concept of the web of things, J. Robot. 2015 (2015), https://doi.org/10.
sym12071111. 1155/2015/631420.
[185] Z. Zheng, A.K. Sangaiah, T. Wang, Adaptive communication protocols in flying [211] S. Aggarwal, N. Kumar, Path planning techniques for unmanned aerial ve-
ad hoc network, IEEE Commun. Mag. 56 (2018) 136–142, https://doi.org/10. hicles: a review, solutions, and challenges, Comput. Commun. 149 (2020)
1109/MCOM.2017.1700323. 270–299, https://doi.org/10.1016/j.comcom.2019.10.014.
[186] M.Y. Arafat, S. Moh, Bio-inspired approaches for energy-efficient localization [212] C. Zammit, E. Van Kampen, 3D real-time path planning of UAVs in dynamic
and clustering in UAV networks for monitoring wildfires in remote areas, IEEE environments in the presence of uncertainty, in: AIAA Scitech 2021 Forum,
Access 9 (2021) 18649–18669, https://doi.org/10.1109/ACCESS.2021.3053605. 2021.
[187] I.A. Meer, M. Ozger, C. Cavdar, On the localization of unmanned aerial vehicles [213] J. Gu, G. Ding, Y. Xu, H. Wang, Q. Wu, Proactive optimization of transmission
with cellular networks, in: IEEE Wirel. Commun. Netw. Conf., WCNC 2020, power and 3D trajectory in UAV-assisted relay systems with mobile ground
May, 2020. users, Chin. J. Aeronaut. 34 (2021) 129–144, https://doi.org/10.1016/j.cja.2020.
[188] A. Couturier, M.A. Akhloufi, A review on absolute visual localization for UAV, 09.028.
Robot. Auton. Syst. 135 (2021) 103666, https://doi.org/10.1016/j.robot.2020. [214] S. Zhang, R. Zhang, Radio map based path planning for cellular-connected
103666. UAV, in: 2019 IEEE Glob. Commun. Conf., GLOBECOM 2019 - Proc., 2019.
[189] F. Khelifi, A. Bradai, K. Singh, M. Atri, Localization and energy-efficient data [215] A. Moheddine, F. Patrone, M. Marchese, UAV and IoT integration: a flying gate-
routing for unmanned aerial vehicles: fuzzy-logic-based approach, IEEE Com- way, in: 2019 26th IEEE Int. Conf. Electron. Circuits Syst., ICECS 2019, 2019,
mun. Mag. 56 (2018) 129–133, https://doi.org/10.1109/MCOM.2018.1700453. pp. 121–122.
[190] Q. Pan, X. Wen, Z. Lu, L. Li, W. Jing, Dynamic speed control of unmanned aerial [216] A. Israr, G.E.M. Abro, M. Sadiq Ali Khan, M. Farhan, S.U.A. Bin Mohd Zulkifli,
vehicles for data collection under Internet of things, Sensors 18 (11) (2018) Internet of things (IoT)-enabled unmanned aerial vehicles for the inspection
3951, https://doi.org/10.3390/s18113951. of construction sites: a vision and future directions, Math. Probl. Eng. 2021
[191] N. Delavarpour, C. Koparan, J. Nowatzki, S. Bajwa, X. Sun, A technical study (2021), https://doi.org/10.1155/2021/9931112.
on UAV characteristics for precision agriculture applications and associated [217] E.A. Marconato, J.A. Maxa, D.F. Pigatto, A.S.R. Pinto, N. Larrieu, K.R.L.J.C. Branco,
practical challenges, Remote Sens. 13 (2021) 1–25, https://doi.org/10.3390/ IEEE 802.11n vs. IEEE 802.15.4: a study on communication QoS to provide safe
rs13061204. FANETs, in: Proc. - 46th Annu. IEEE/IFIP Int. Conf. Dependable Syst. Networks,
[192] D.W. Matolak, Unmanned aerial vehicles: communications challenges and fu- DSN-W 2016, 2016, pp. 184–191.
ture aerial networking, in: 2015 Int. Conf. Comput. Netw. Commun., ICNC [218] B.S. Morse, C.H. Engh, M.A. Goodrich, UAV video coverage quality maps and
2015, 2015, pp. 567–572. prioritized indexing for wilderness search and rescue, in: 5th ACM/IEEE Int.
[193] W. Khawaja, I. Guvenc, D.W. Matolak, U.C. Fiebig, N. Schneckenburger, A sur- Conf. Human-Robot Interact, HRI 2010, 2010, pp. 227–234.
vey of air-to-ground propagation channel modeling for unmanned aerial vehi- [219] W. Liu, P. Si, E. Sun, M. Li, C. Fang, Y. Zhang, Green mobility management in
cles, IEEE Commun. Surv. Tutor. 21 (2019) 2361–2391, https://doi.org/10.1109/ UAV-assisted IoT based on dueling DQN, in: IEEE Int. Conf. Commun., 2019-
COMST.2019.2915069. May, 2019, pp. 14–19.
[194] C. Yan, J. Wang, L. Fu, C. Jiang, M. Chen, Y. Ren, Timing synchronization and [220] A. Sacco, F. Esposito, G. Marchetto, P. Montuschi, Sustainable task offloading in
ranging in networked UAV-aided OFDM systems, J. Commun. Inf. Netw. 3 UAV networks via multi-agent reinforcement learning, IEEE Trans. Veh. Tech-
(2018) 45–54, https://doi.org/10.1007/s41650-018-0037-y. nol. 70 (2021) 5003–5015, https://doi.org/10.1109/TVT.2021.3074304.
[195] Y.J. Kim, Y.B. Kim, H.J. Dong, Y.S. Cho, H.L. Lee, Compact switched-beam array [221] J. Xiong, H. Guo, J. Liu, Task offloading in UAV-aided edge computing: bit al-
antenna with a butler matrix and a folded ground structure, Electronics 9 location and trajectory optimization, IEEE Commun. Lett. 23 (2019) 538–541,
(2020), https://doi.org/10.3390/electronics9010002. https://doi.org/10.1109/LCOMM.2019.2891662.
[196] W. Pi, J. Zhou, Multi-UAV enabled data collection with efficient joint adaptive [222] M. Mozaffari, A.T.Z. Kasgari, W. Saad, M. Bennis, M. Debbah, Beyond 5G with
interference management and trajectory design, Electronics 10 (2021) 1–37, UAVs: foundations of a 3D wireless cellular network, IEEE Trans. Wirel. Com-
https://doi.org/10.3390/electronics10050547. mun. 18 (2019) 357–372, https://doi.org/10.1109/TWC.2018.2879940.
[197] K.A. Swieringa, C.B. Hanson, J.R. Richardson, J.D. White, Z. Hasan, E. Qian, A. [223] S. Rangan, T.S. Rappaport, E. Erkip, Millimeter-wave cellular wireless net-
Girard, Autonomous battery swapping system for small-scale helicopters, in: works: potentials and challenges, Proc. IEEE 102 (2014) 366–385, https://
Proc. - IEEE Int. Conf. Robot. Autom, 2010, pp. 3335–3340. doi.org/10.1109/JPROC.2014.2299397.
[198] K.A.O. Suzuki, P. Kemper Filho, J.R. Morrison, Automatic battery replacement [224] X. You, C.X. Wang, J. Huang, X. Gao, Z. Zhang, M. Wang, Y. Huang, C. Zhang,
system for UAVs: analysis and design, J. Intell. Robot. Syst. Theory Appl. 65 Y. Jiang, J. Wang, M. Zhu, B. Sheng, D. Wang, Z. Pan, P. Zhu, Y. Yang, Z. Liu, P.
(2012) 563–586, https://doi.org/10.1007/s10846-011-9616-y. Zhang, X. Tao, S. Li, Z. Chen, X. Ma, I. Chih-Lin, S. Han, K. Li, C. Pan, Z. Zheng,
[199] D. Lee, J. Zhou, W.T. Lin, Autonomous battery swapping system for quadcopter, L. Hanzo, X.S. Shen, Y.J. Guo, Z. Ding, H. Haas, W. Tong, P. Zhu, G. Yang, J.
in: 2015 Int. Conf. Unmanned Aircr. Syst., ICUAS 2015, 2015, pp. 118–124. Wang, E.G. Larsson, H.Q. Ngo, W. Hong, H. Wang, D. Hou, J. Chen, Z. Chen, Z.
[200] N.K. Ure, G. Chowdhary, T. Toksoz, J.P. How, M.A. Vavrina, J. Vian, An auto- Hao, G.Y. Li, R. Tafazolli, Y. Gao, H.V. Poor, G.P. Fettweis, Y.C. Liang, Towards
mated battery management system to enable persistent missions with multi- 6G wireless communication networks: vision, enabling technologies, and new
ple aerial vehicles, IEEE/ASME Trans. Mechatron. 20 (2015) 275–286, https:// paradigm shifts, Sci. China Inf. Sci. 64 (2021), https://doi.org/10.1007/s11432-
doi.org/10.1109/TMECH.2013.2294805. 020-2955-6.
[201] L. Cheng, L. Zhong, S. Tian, J. Xing, Task assignment algorithm for road patrol [225] M. Chen, U. Challita, W. Saad, C. Yin, M. Debbah, Artificial neural networks-
by multiple UAVs with multiple bases and rechargeable endurance, IEEE Ac- based machine learning for wireless networks: a tutorial, IEEE Commun. Surv.
cess 7 (2019) 144381–144397, https://doi.org/10.1109/ACCESS.2019.2944881. Tutor. 21 (2019) 3039–3071, https://doi.org/10.1109/COMST.2019.2926625.
[202] S. Poudel, S. Moh, Medium access control protocols for unmanned aerial [226] U. Challita, A. Ferdowsi, M. Chen, W. Saad, Machine learning for wireless con-
vehicle-aided wireless sensor networks: a survey, IEEE Access 7 (2019) nectivity and security of cellular-connected UAVs, IEEE Wirel. Commun. 26
65728–65744, https://doi.org/10.1109/ACCESS.2019.2917948. (2019) 28–35, https://doi.org/10.1109/MWC.2018.1800155.

28
S. Poudel and S. Moh Vehicular Communications 35 (2022) 100469

[227] V. Chamola, V. Hassija, S. Gupta, A. Goyal, M. Guizani, B. Sikdar, Disaster and [228] M.Y. Arafat, S. Moh, A Q-learning-based topology-aware routing protocol for
pandemic management using machine learning: a survey, IEEE Int. Things J. flying ad hoc networks, IEEE Int. Things J. 4662 (2021) 1–16, https://doi.org/
(2020) 1–30, https://doi.org/10.1109/JIOT.2020.3044966. 10.1109/JIOT.2021.3089759.

29