Unmanned Aerial Vehicle Communications for Civil Applications: A Review

The use of drones, formally known as unmanned aerial vehicles (UAVs), has significantly increased across a variety of applications over the past few years. This is due to the rapid advancement towards the design and production of inexpensive and dependable UAVs and the growing request for the utilization of such platforms particularly in civil applications. With their intrinsic attributes such as high mobility, rapid deployment and flexible altitude, UAVs have the potential to be utilized in many wireless system applications. On the one hand, UAVs are able to operate as flying mobile terminals within wireless/cellular networks to support a variety of missions such as goods delivery, search and rescue, precision agriculture monitoring, and remote sensing. On the other hand, UAVs can be utilized as aerial base stations to increase wireless communication coverage, reliability, and the capacity of wireless systems without additional investment in wireless systems infrastructure. The aim of this article is to review the current applications of UAVs for civil and commercial purposes. The focus of this paper is on the challenges and communication requirements associated with UAV-based communication systems. This article initially classifies UAVs in terms of various parameters, some of which can impact UAVs’ communication performance. It then provides an overview of aerial networking and investigates UAVs routing protocols specifically, which are considered as one of the challenges in UAV communication. This article later investigates the use of UAV networks in a variety of civil applications and considers many challenges and communication demands of these applications. Subsequently, different types of simulation platforms are investigated from a communication and networking viewpoint. Finally, it identifies areas of future research.


I. INTRODUCTION
Unmanned aerial vehicles (UAVs), informally known as 22 drones, have been the subject of intense research among  27 advancements in the design and production of inexpensive 28 and highly reliable unmanned aerial vehicles as well as the 29 The associate editor coordinating the review of this manuscript and approving it for publication was Marco Martalo . growing demand for commercial utilization of such low-cost 30 platforms, UAVs are now being used in a vast number of 31 civil and commercial applications [1]. In addition, UAVs' 32 unique attributes, such as ease of use, rapid deployment 33 to far-flung areas, high-mobility, maneuverability, and their 34 ability to hover, make them excellent candidates for civil 35 and commercial applications [1]. Examples of such applica- 36 tions include search and rescue missions [9], [10], [11], [12], 37 precision agriculture monitoring [13], natural disaster and 38 environmental monitoring [14], [15], delivery of goods [16], 39 [17], [18], [19], and remote sensing [20], [21]. A single UAV 40 or multiple UAVs can be used as communication relays or 41 maintain efficient communication links among the UAVs. 98 Various issues exist that need to be addressed spanning 99 from network planning, resource allocation, cell association, 100 to deployment. 101 Mozzaffari et al. [3] investigated the key challenges and 102 important trade-offs in UAV-enabled wireless networks. The 103 authors mainly considered the major UAV challenges such 104 as channel modeling, energy efficiency, three-dimensional 105 deployment and performance analysis. The authors then dis- 106 cussed open problems and potential research directions relat- 107 ing to UAV communications. At the end, they described a 108 variety of analytical frameworks and mathematical tools that 109 can be used in this domain such as stochastic geometry, game 110 theory, transport theory, machine learning and optimization 111 theory. Furthermore, they explained how to use such tools to 112 tackle UAV problems. Fotouhi et al. [5] presented a review of 113 current developments in the UAV industry that lead to smooth 114 integration of UAVs into cellular networks. Particularly, they 115 reviewed some types of consumer UAVs that are currently 116 available off-the-shelf. The authors addressed the UAVs' 117 related communication interference issues and explained how 118 the standardization bodies provided potential solutions for 119 integrating aerial vehicles with the existing terrestrial BSs. 120 They discussed the challenges and opportunities involved 121 in assisting cellular communications with UAV-based flying 122 relays and BSs. Moreover, the authors investigated the exist- 123 ing prototypes in this domain and test bed activities. 124 Until now, a few review papers have been written in 125 this domain. Two of those are very relevant to this article. 126 Hayat et al. [1] presented the requirements and character-127 istics of the UAV networks for future civil and com-128 mercial applications. The authors reviewed many research 129 articles published over the period of 2000-2015 from a 130 communications and networking viewpoint. They investi-131 gated the data requirements, quality of service requirements, 132 network-relevant mission parameters, and the minimum data 133 to be transmitted over the network for civil applications. 134 Subsequently, they examined general networking related 135 requirements such as safety, security, privacy, connectivity, 136 scalability and adaptability. Finally, the group reviewed the 137 experimental results from other projects in this field and dis-138 cussed the suitability of current communication technologies 139 for supporting reliable aerial networks. Comprehensive work 140 presented in [1] has helped expand the body of knowledge 141 on the topic. Further work would include more up-to-date 142 information that further assists with identifying the current 143 state of the technology. 144 Quy et al. [44] discussed the perspective of Vehicle Ad 145 hoc Networks (VANET) to be implemented into smart 146 cities. The authors presented a comprehensive perspec-147 tive of the techniques to enable automobile communica- 148 tion networks in urban environments. Their survey specified 149 three directions, listed as multimetric, Interne/UAV/Cloud, 150 and Intelligent, that would be needed to enhance VANETs 151 in the future. Another updated perspective was discussed 152 by Zaidi et al. [45]. Advancing the technology into the 153 VOLUME 10, 2022 future would require the Internet of Flying Things (IoFT). This document is organized as follows: Section II identifies 210 the ways in which to take advantage of the UAS technologies 211 within the realm of communications. Section III presents 212 a thorough review of the most up-to-date classification of 213 UAS technologies. Section IV provides a comprehensive but 214 selective review on FANET technologies. Application sce-215 narios are explored in Section V where relevant UAS wire-216 less applications in industry or civilian implementations are 217 identified. Challenges in UAV communications are discussed 218 in Section VI. Section VII reviews simulation platforms for 219 UAV application scenarios. Finally, Sections VIII and IX 220 offer future research directions and conclusions, respectively, 221 based on the reviewed literature in this document. 223 The UAV global market is very promising and offers an excel-224 lent prospect for further growth. The global market for civil 225 applications of UAV systems is expected to be among one of 226 the most vibrant developing sectors in the upcoming decade. 227 The market is anticipated to expand from an almost 5 billion-228 dollar annual market in 2019 to about a 14.5 billion-dollar 229 annual market by 2028. That indicates a compound annual 230 growth rate (CAGR) of 12.5 percent in constant dollars [47]. 231 The civil UAV market is predicted to grow to a total of 232 88.3 billion dollars over the next decade [47]. According to 233 Silver et al. [48], the civil UAV market is divided into the 234 following key industries: infrastructure, agriculture, trans-235 port, security, media and entertainment, insurance, telecom-236 munication, and mining. The distribution of market value is 237 represented by Figure 1. 238 By 2021, sales of consumer UAVs were expected to reach 239 29 million units, and sales of UAVs for commercial uses 240 were expected to reach 805,000 units [49]. Civil governments 241 in Europe and the United States are becoming keen to take 242 advantage of UAV systems for various applications such as 243 border control and maritime security. Moreover, peacekeep-244 ing operations conducted by United Nations (UN) entities can 245 also have impact on UAV market sales [50]. Public safety 246 deployment of UAVs for fire control and law enforcement 247 purposes has increased over the past few years. The European 248

II. CURRENT AND FUTURE MARKET OPPORTUNITIES
The global market size for UAV commercial applications is 305 predicted to reach 129.23 billion USD by 2025, registering a 306 compound annually growth rate (CAGR) of 56.5 percent over 307 the estimated period [76]. Furthermore, over 100,000 new 308 jobs within the UAS industry are expected to be created 309 by 2025 [77]. As the number of applications for UAV sys-310 tems continues to grow and as UAV technologies continue 311 to evolve, all of the preceding statistics show the economic 312 importance of UAV systems for numerous sectors of industry 313 in the near future. 315 Up to now, many different versions of UAV classifications 316 have been defined and clearly described by the scientific 317 community. Many of the existing classifications are per-318 formed to classify the use of UAVs for military and civil 319 applications, while a few of these classifications are specif-320 ically carried out to categorize the use of UAVs for civil 321 and commercial applications. Watts et al. [78] investigated 322 various UAV platforms including their sensor capabilities 323 and described the advantages of each platform and their 324 relevance to the demand of users in the scientific commu-325 nity. Authors in this paper categorized the UAV platforms 326 based on a few of their specific attributes such as flight 327 endurance, physical size, and potential capabilities. In this 328 categorization, the authors classified UAV platforms as nano 329 air vehicles (NAVs), micro/miniature air vehicles (MAVs), 330 vertical take-off and landing (VTOL), low altitude short 331 endurance (LASE), low altitude long endurance (LALE), 332 medium altitude long endurance (MALE) and high alti-333 tude long endurance (HALE). Gupta et al. [79] categorized 334 UAVs as NAVs, MAVs, mini UAVs (MUAVs), tactical UAVs 335 (TUAVs), MALE and HALE. Korchenko and Ilyash utilized 336 a different classification which took into account sixteen 337 important features of the UAVs, such as flight rules, air-338 craft types, aircraft engine types, aircraft applications, type 339 of control systems, take-off and landing directions, wing 340 types and fuel systems [80]. Weibel and Hansman [81] dif-341 ferentiated the UAVs by mass and then categorized them 342 into micro, mini, tactical, medium altitude and high altitude 343 UAVs. Cavoukian [82] classified UAVs into three major 344 types representing Micro and mini UAVs that can fly at 345 low altitudes (below 300 m), as operating in urban canyons, 346 inside buildings or along hallways. Tactical UAVs compared 347 to micro and mini UAVs are heavier, ranging from 150 to 348 1500 kg, and can fly at higher altitudes ranging from 3000 to 349 8000 m. Such UAVs currently only support military applica-350 tions. Strategic UAVs that belong to HALE classification can 351 support longer flight ranges and can reach a maximum flight 352 altitude of around 20,000 m. These types of UAVs can carry 353 much larger payloads and more sophisticated equipment, and 354 are also designed mainly for military applications. Australian 355 civil aviation safety authority (CASA) classified UAVs into 356 four groups based on their weight [83]. Micro UAVs with 357 gross weight of 100 grams or less, small UAVs with the 358 weight of less than 2 kg, medium UAVs with the weight 359 of greater than 2 kg and less than 150 kg, and large UAVs time and maximum speed, must be considered thoroughly and 404 in more detail.  . Compared to rotary-wing copters, fixed-wing 417 UAVs are intrinsically more energy efficient [96]. Although, 418 most of the existing studies on UAV systems for wireless 419 cellular coverage is focused on considering the rotary-wing 420 UAVs, fixed-wing UAVs are expected to be more suitable 421 for wireless connectivity purposes in situations where long 422 flight times are required. This is because fixed-wing UAVs 423 rely on a much more energy-efficient way of flight in contrast 424 to the rotary-wing UAVs [97]. Xie and Huang [98] evaluated 425 an UAV-enabled relaying network where a fixed-wing UAV 426 is positioned between the base station and ground users. The 427 authors proposed a method to optimize the radius of UAV cir-428 cular trajectory along with the transmission power allocated 429 with the purpose of maximizing energy-efficiency of the UAV 430 relay network. Fixed-wing UAVs are also able to utilize more 431 [120]. In fact, single rotor drones consist 508 of two rotors; one rotor is located on top and the other 509 one is positioned at the tail. The larger rotor on the top is 510 used for lift while the smaller one at the tail is used for 511 control [121]. Compared to multi-rotor systems, Single rotor 512 drones have higher endurance with longer flights and can 513 carry heavier payloads to perform a variety of tasks [122], 514 and they are often powered by gas engines [123]. Much like 515 the multi-rotor UAVs, single rotor drones are also suitable for 516 aerial photography [124] in addition to spraying agricultural 517 crops [121], [125]. Although the use of single rotor drones 518 for agricultural plant protection has been greatly appreciated, 519 various shortcomings still exist in this field. For instance, 520 one of the disadvantages of using single rotor drones in 521 agricultural plant protection is studied by Wen et al. [126]. 522 The authors showed that the rotor flow field of a single rotor 523 UAV can cause drift of the droplets, resulting in waste and 524 secondary disaster. They proved that digital simulation can 525 be useful to overcome this problem. Generally, single rotor 526 UAVs can come with higher operational risks as the rotating 527 blades positioned on the top often pose risks to human being 528 and nature [88], [125]. Therefore, trained professionals are 529 needed to fly them [122]. Hybrid fixed-wing/multi-rotor UAVs combine the aspects of 532 a multi-rotor and a fixed-wing aircraft [127], [128], [129]. 533 These aircraft utilize both an airfoil and downward thrust 534 to combine the VTOL capabilities of a multi-rotor with 535 the higher efficiency of a fixed-wing aircraft. Because of 536 this, a hybrid aircraft is able to take off and land virtually 537 anywhere and then fly long distances or for long periods 538 of time [108] (30-70km). The authors state 653 that there are UAVs over the Short Range category, but they 654 are rarely used for civil applications. According to [123], 655 UAVs can be classified as: Very low cost close-range 656 (5km), Close-range (50km), Short-range (150km), Mid-range 657 (650km), and Endurance (300km) UAVs. In terms of opera-658 tion, UAVs are classified mostly into two categories. A UAV 659 can operate autonomously or be remotely controlled by 660 a pilot. A remotely piloted UAV is required to establish 661 a reliable unidirectional/bidirectional communication link 662 between itself and its pilot. Due to the nature of the remote 663 presence of human(s) in UAV systems, communication range 664 also plays an important role to support pilot-UAV com-665 munication link. Communication protocol must be selected 666 such that it can support a variety of missions with differ-667 ent communication range requirements. The communication 668 range is defined as the maximum distance from which an 669 UAV can remotely be controlled. Communication range is 670 varied from tens of meters for small UAVs to hundreds of 671 is proportional to the distances between sensors and energy 727 of the sensors, and inversely proportional to data upload 728 requirements.

730
Maximum speed may be another important factor in UAV 731 civil applications. The classifications in [140] factor airspeed 732 into the categorizations of UAVs, with group 1 (small UAVs) 733 having airspeed up to 100 knots, groups 2-3 (medium to 734 large) with airspeeds up to 250 knots, and groups 4 and 735 5 which can have any airspeed. From a civil standpoint, 736 there are federal limitations to airspeed, however. The Federal 737 Aviation Administration (FAA) states that a drone operator 738 with a Part 107 license may only fly up to 87 knots [156]. 739 Wu et al. [154] investigated the use of a UAV-enabled two-740 use broadcast channel, where a UAV is used to send infor-741 mation to two users in different geographic locations. The 742 authors considered two cases with large/low flight dura-743 tion/speed, where the UAV's maximum speed and transmit 744 power were the primary constraints, and attempted to opti-745 mize the UAV's trajectory and transmit power allocations 746 over time with a fixed flight duration. In the first case, a hover-747 fly-hover trajectory with time division multiple access based 748 orthogonal multiuser transmission is able to achieve the 749 desired capacity. However in the second case, it is better 750 for the UAV to remain in a fixed location in closer prox-751 imity to the user with higher achievable rate and superpo-752 sition coding based non-orthogonal transmission is required 753 with interference cancellation at the receiver of the closer 754 user. Since there is a lack of universal regulation for the fre-758 quency utilization, a difficult issue in air-to-ground chan-759 nel modeling that needs to be addressed is the diversity of 760 suitable frequencies for UAV communication systems [157]. 761 With reference to channel modeling, taking the operating 762 frequency into account contributes to the creation of a more 763 complete model, improving the generality of the model 764 and enabling its application in a variety of situations with 765 diverse operating frequencies [157]. Latest research on air-766 to-ground channel modeling methods mostly concentrates on 767 low-frequency bands, including those of IEEE 802.11a/g/n 768 (2.4 GHz, 5.8 GHz), or L-band (1-2 GHz), or C-band 769 (4 GHz), which the International Telecommunication Union 770 (ITU) recommends for drone communications [157], [158]. 771 For instance, Asadpour et al. [159] showed through testing 772 that the 802.11n communication protocol works poorly in cir-773 cumstances involving high levels of mobility and aerial work. 774 Asadpour et al. demonstrated that as soon as drones take to 775 the air, network throughput between them drops below the 776 theoretical maximum. Schneckenburger et al. investigated the 777 properties of the L-band air-to-ground radio channel for posi-778 tioning applications, and then reported their findings in [160]. 779 Authors in [161] measured the performance of air-to-ground 780 VOLUME 10, 2022 channels over sea at the C-band with low airborne altitudes 781 (0.37-1.83 km). They showed that the likelihood of appear-782 ance of multi-path components increases as the airborne alti-783 tude decreases. Authors in [162] presented a comprehensive 784 survey regarding air-to-ground propagation channel model-  [166], [167]. For example, the 809 authors in [166] measured air-to-ground channels over cel-810 lular networks, where the UAV altitude varied from 1.5 m 811 to 120 m. The findings in [166] indicate that when the alti-812 tude increases, the path loss exponent (PLE) reduces from 813 3.7 to 2.0, which implies that the scattering environment 814 slowly becomes minimal with the height. Authors in [168]

837
In a multi-UAV system, it is possible for UAVs to work 838 together within a network; this is known as a flying ad hoc 839 network, or FANET  While most UAV systems today are comprised of only a sin-856 gle UAV, there are advantages to having systems containing 857 multiple UAVs. For instance, when comparing the use of a 858 single UAV to multiple UAVs for agricultural applications 859 in [181], the authors found, under several considerations 860 including an autonomously controlled system compared to 861 a remotely controlled system, setup time, flight time, battery 862 consumption, coverage ratio, inaccuracy of land, and etc., that 863 a multi-UAV system significantly outperforms a single UAV 864 system.

865
The authors of [182] proposed an algorithm to offer 866 dynamic repositioning of an aerial base station UAV in order 867 to improve spectral efficiency between aerial base station and 868 user equipment (UE). In their work, the authors utilized a 869 of radio communication within FANETs: UAV to UAV, 926 UAV to ground control station (GCS), and UAV to satellite 927 (SATCOM) according to [173] and [194]. UAV to UAV com-928 munications can be either direct or indirect. In other words, 929 a UAV system can directly communicate with another UAV 930 system or can relay its message through other UAVs. This 931 allows the UAV network to be more efficient both in data rate 932 as well as communication range [173].

933
UAV to GCS communications allow the UAV network 934 to communicate to ground infrastructure for information 935 relaying and connecting to the global network. For instance, 936 Chriki et al. [173] proposed a centralized data-oriented 937 communication architecture for swarm of UAVs for crowd 938 monitoring applications. The GCS was used to manage band-939 width usage within the local swarm, acting as the central 940 coordinator. Two classes of urgent messages were created: 941 important result and critical state. Using these classes along 942 with other relevant information, the GCS could then authorize 943 data transmission of UAVs within the network, and thus opti-944 mizing bandwidth usage efficiency. The third major method 945 of communications, SATCOM, can be useful in areas such 946 as over oceans or mountains where ground stations may not 947 be present, however the cost is very high according to [194]. 948 Skinnemoen et al. [195] investigated the use of UAVs for 949 obtaining live images for a variety of applications includ-950 ing search and rescue, safety and security, border patrol, 951 police operations, and disaster management. In many use 952 cases, terrestrial networks were insufficient for providing live 953 imagery, so satellite communication was required either in the 954 UAV itself or relayed through ground. As doing so generally 955 incurs high cost, is slow, and requires higher capacity than is 956 available, the authors presented a new concept for obtaining 957 live imagery from UAVs while combating these obstacles.

958
UAV to satellite communications also presents the chal-959 lenge of unstable beam tracking due to UAV navigation. 960 Zhao et al. [196] proposed a new approach for blind beam 961 tracking for Ka-band UAV-satellite communications. Using a 962 hybrid large scale antenna array, the UAV first mechanically 963 adjusted the position of the antenna in the relative orientation 964 of the target satellite using beam stabilization and dynamic 965 isolation, derived from data fusion of low-cost sensors. The 966 precision was then fine-tuned electrically by adjusting the 967 weight of the antennas, and an array structure based simul-968 taneous perturbation algorithm was created. FANET nodes have higher mobility and thus they can travel 972 faster compared with MANET nodes and sometimes VANET 973 nodes according to [197]. Authors in [197] explained that 974 the speed of MANETs are generally limited to the speed 975 of human being (about 6 km/hr). While VANET nodes can 976 travel faster (usually up to 100 km/hr), their speeds are still 977 generally restricted to maximum speed limit in roads. Due 978 to the high mobility of FANET, topology changes are more 979 frequent so the mobility of a FANET becomes an important 980 design consideration according to [198], which outlined four  [199]. Table 1   Khan et al. [201] explained that non-orthogonal multiple decoded signals from the UAVs in order to decode the weaker 1015 signals received from ground users [201].  Generally, in FANETs, appropriate selection of rout-1025 ing protocols is a challenging task as network topology 1026 is constantly changing due to the high mobility of UAV 1027 platforms [203]. There are five main requirements for design-1028 ing proper routing protocols in FANETs: high adaptability, 1029 scalability, high residual energy, low latency, and high band-1030 width as indicated in [204] and [203]. First, there must be a 1031 high amount of adaptability due to the frequent changes in 1032 network topology and low node density [205]. Adaptability 1033 is important as link disconnections will be frequent, reliable 1034 routes must be identified quickly and routing tables must be 1035 frequently updated [206], [207]. Hong et al. [207] proposed 1036 a routing scheme that was able to adapt to rapid changes in 1037 the network topology and as a result it could improve the 1038 performance of the network. The results were verified using 1039 several simulations and mobility models. Second, routing 1040 must be sufficiently scalable to accommodate the various 1041 applications of UAV networks ranging from small scale oper-1042 ations with few nodes to large scale operations with high node 1043 density [207]. Scalability is important as the coverage range 1044 of a single UAV is limited, but a network of several UAVs 1045 can easily expand the operational scalability, and adapt to 1046 many different applications [194]. Oubbati et al. [194] out-1047 lined and compared the scalability properties as well as 1048 operational features of several existing routing protocols. 1049 Third, routes must be established with high residual energy in 1050 order to reduce potential link disconnections resulting from 1051 node failure, as UAVs are primarily battery powered [208]. 1052 The authors of [206] and [209] developed a scheme to 1053 explore routing paths while considering energy consump-1054 tion, link breakage prediction, and connectivity degree of the 1055 showed that the scheme minimized the number of path fail-may be useful in situations where the network topology will 1089 remain constant throughout the mission [214]. One example 1090 of where a static routing protocol is useful is Load Carry and 1091 Deliver (LCAD), which was one of the first routing models 1092 for FANET [212]. LCAD utilizes a store-carry-and-forward 1093 technique, and can be useful for applications that are not 1094 time sensitive, such as data collection from fixed sensors or 1095 tracking. Authors in [215] showed another example of using 1096 static routing protocol, called multi-level hierarchical routing 1097 (MLHR). Using this system, a cluster head within a cluster of 1098 UAVs disseminates data traffic to the other UAVs. In a proactive routing protocol, each node maintains a table 1101 that is periodically updated and contains routing information 1102 to all nodes [216]. With this protocol, the destination path can 1103 be immediately accessed, eliminating the delay that a node 1104 may experience when packets are needed to be sent [217]. 1105 However, this method also increases the bandwidth usage and 1106 takes up network resources creating paths that may or may not 1107 be used. Authors in [218] conducted an experiment compar-1108 ing three different routing protocols for FANET: Ad-hoc On-1109 demand Distance Vector (AODV), Destination-Sequenced 1110 Distance Vector (DSDV), and Optimized Link State Routing 1111 Protocol (OLSR). The study found OLSR, a proactive routing 1112 protocol, to outperform the other two in terms of average 1113 throughput, packet delivery ratio, and end to end delay.

1115
Reactive routing calculates routes on demand when the need 1116 arises. This reduces the overhead of building and maintaining 1117 routes that are unused by each node, however there will 1118 be increased latency for sending data packets as the node 1119 must wait until a route is acquired. Reactive protocols are 1120 VOLUME 10, 2022 suboptimal for bandwidth utilization as the network will be flooded as the route to the destination is being determined [219]. However for highly dynamic networks with 1123 frequent network topology changes, reactive protocols can 1124 be scaled more easily. There are two primary methods for reactive routing: source routing and point-to-point routing.
With source routing, the data packet will contain the complete   In [203], authors investigated the use of a new adaptive 1176 routing protocol for FANET based on the fuzzy system. Using 1177 Network Simulator, the authors were able to determine that 1178 the new routing protocol performed 300% better in terms of 1179 throughput when compared to optimized link state (OLSR) 1180 and ad-hoc on-demand distance vector (AODV) routing pro-1181 tocols. Khan et al in [227] developed a hybrid communica-1182 tion scheme for FANET. They were able to conclude that a 1183 multi-layer FANET was the best architecture for networking 1184 a group of various UAVs. The authors also determined Blue-1185 tooth 5.0 to be favored protocol as it is low cost, consumes 1186 little power, and has a longer transmission range. Simulations 1187 using the optimized network engineering tool (OPNET) sup-1188 ported these results. Authors in [194] proposed new protocols 1189 as well, first a position-prediction-based directional MAC 1190 protocol (PPMAC), which utilizes directional antennas to 1191 overcome directional deafness problems. The authors also 1192 proposed a self-learning routing protocol using reinforcement 1193 learning (RLSRP), which evolves automatically. Together, 1194 these protocols may be able to provide an intelligent and 1195 autonomous solution for FANET communications. For addi-1196 tional developments in routing protocols for FANET, authors 1197 in [194] provided extensive reviews. have considered load balancing routing protocols to address 1201 both complicated dynamic network environments as well as 1202 network traffic increase in future, it can be concluded from 1203 the comparison that majority of routing protocols do not 1204 take traffic load balancing into account [232]. Most of the 1205 routing protocols including multi-path routing have not been 1206 able to effectively balance the load of network as well as 1207 energy utilization. There have been several route matrices 1208 suggested, including the shortest path, the freshest path, and 1209 the one with the highest link quality [232], [233]. However, 1210 the development of routing protocols without taking into 1211 consideration the properties of data packets cannot construct 1212 an efficient network; network throughput may be increased 1213 to some extent by forwarding data packets in accordance with 1214 varying traffic demands [232]. UAV networks require reliable 1215 communication to operate properly. However, since radio link 1216 connectivity between drones can be disconnected due to high-1217 speed, conventional routing protocols cannot work well in 1218 UAV networks [234]. If drones travel randomly in a multi-hop 1219 UAV network without pre-designing paths, it becomes chal-1220 lenging to select the next appropriate hop node for data 1221 relay [232]. In such scenarios, opportunistic routing proto-1222 cols such as [235], [236], and [237] can be utilized. Oppor-1223 tunistic routing protocols are used to transfer data packets 1224 in dynamic UAV networks. Currently, the hierarchical net-1225 work structure is used in research studies to investigate the 1226 routing protocols [238]. Although the hierarchical structure 1227 performs well in wired networks, it is insufficient for wireless 1228 networks [239]. However, it is argued that a cross-layer 1229 design would be preferable [240]. This is because the interac-1230 tion between OSI layers may significantly enhance network 1231 performance. For instance, one of the most vital parameters 1232 of physical layer, which is link-state information, serve as 1233 an important foundation for routing forwarding [240]. There-1234 fore, a more dependable path might be discovered using the The incessantly growing need for high-speed wireless access 1254 has been fueled by the rapid proliferation of highly capa-1255 ble mobile devices such as smartphones, tablets, and more 1256 recently, drone-UEs and IoT-style gadgets [24]. As such, the wireless devices ubiquitously [241]. UAV systems are antici-1265 pated to be considered one of the essential components of 5G 1266 and beyond 5G wireless networks which can ideally provide 1267 reliable and high data-rate wireless connectivity not only 1268 for stationary users, but also for crowds of people moving 1269 in private and public transportation networks. As opposed 1270 to existing fourth generation (4G) cellular networks, fifth 1271 generation (5G) and beyond 5G cellular networks are pro-1272 jected to be able to ubiquitously connect various types of 1273 wireless devices with varied requirements. Emergence of new 1274 technologies such as IoT has triggered a rise in the number 1275 of wireless devices in 5G cellular networks which has led to 1276 the creation of higher data traffic [242], [243]. According to 1277 Khan et al. [244], the total global mobile traffic in 2028 is 1278 estimated to exceed 1 zettabyte/mo, that is about 200 GB per 1279 month for nearly 5 billion users globally. This demonstrates 1280 how existing cellular network infrastructures are unable to 1281 provide the necessary capacity for demand. Moreover, sub-1282 stantial increase in the data traffic can impose an additional 1283 burden in terms of operational costs and capital investments to 1284 telecommunication operators [241]. Existing terrestrial wire-1285 less systems that use heavily-congested radio spectrum bands 1286 below 6 GHz are unable to significantly increase the speed 1287 of data transfer for various emerging applications. mmWave 1288 communications can use unoccupied bandwidth that is avail-1289 able at mmWave frequencies to overcome the problem asso-1290 ciated with congested frequency bands and to fulfill the 1291 requirements of 5G cellular network technology. mmWave 1292 communications technology can take advantage of UAVs 1293 to assist existing wireless networks for future 5G wireless 1294 applications [28]. Zhang et al. [28] provide a comprehensive 1295 review related to existing achievements for the incorporation 1296 of 5G mmWave communications into UAV-assisted wireless 1297 networks. The authors of [245] present an aerial base sta-1298 tion prototype working at millimeter-wave bands to provide 1299 multi-beam multi-stream communications. The authors were 1300 able to verify good stability and reliability of the system 1301 during uplink and downlink at multi-giga-bit-per-second data 1302 rates during field testing. The authors in [211] presented a 1303 mmWave distributed phased-arrays architecture and designs 1304 for user equipment and UAVs to be used in 5G. The UAVs 1305 were used as aerial BSs and were able to achieve a 2.2 Gbps 1306 aggregated peak downlink rate in real-world field testing. 1307 In the case of downlink traffic overload, aerial base sta-1308 tions can be used to complement existing cellular networks. 1309 Authors in [246] proposed a weighted expectation maxi-1310 mization algorithm to determine the distribution of users 1311 and downlink traffic demand. Additionally, contract theory is 1312 used to guarantee correct information exchange between the 1313 UAVs and the base stations. Finally, an optimization problem 1314 is derived to send the appropriate UAV to the overload area 1315 to maximize the base station utility. The authors of [247]  also compared three different types of backhaul scenarios uti-1379 lizing a 3.5 GHz link, 3.5 GHz link with carrier aggregation, 1380 and a 60 GHz link, using three different types of UAVs. The 1381 findings suggest that an optimal flight height of 80 m can 1382 meet both backhaul networks and access networks at the same 1383 time. Occasionally, in SAR missions, the pre-allocated radio 1384 spectrum is insufficient to deliver high data-rate transmis-1385 sions such as real-time video streaming. The UAV network 1386 in such scenarios can borrow a portion of the radio spec-1387 trum of a terrestrial licensed network in return for offering 1388 relaying services. With the aim of improving the performance 1389 of the UAV network and extending the network lifetime 1390 at the same time, several UAVs operate as communication 1391 relays for the primary network whereas other UAVs perform 1392 their assigned tasks. Shamsoshoara et al. [255] proposed an 1393 algorithm for team reinforcement learning to be performed 1394 by UAV's controller unit to identify the optimum allocation 1395 of the radio spectrum for sensing and relaying tasks among 1396 UAVs in addition to their relocation strategy simultaneously. 1397 In order to guarantee the accuracy of the collected data from 1398 the disaster areas, Abdallah et al.
[256] presented a security 1399 architecture for UAV networks. The proposed networking 1400 technique includes a two-tier cluster network which relies 1401 on IEEE 802.11ah to provide traffic isolation between tiers. 1402 The proposed security architecture also uses the lightweight 1403 ring-learning with errors (Ring-LWE) crypto-system to guar-1404 antee the confidentiality of the transferred information. The 1405 chances of finding survivors alive after occurring natural dis-1406 asters such as earthquakes or hurricanes is highly dependent 1407 on the rapid response time of the rescue team. Coordination, 1408 situational awareness (SA) and information sharing are the 1409 most common challenges associated with natural disaster 1410 management which can be achieved in the most efficient 1411 manner through aerial assessment-UAV networks [253]. 1412 A vision for future UAV-assisted disaster management sys-1413 tem was presented in [248], in which UAVs are not only 1414 focused to perform specific tasks such as surveying the 1415 affected area but also are assigned to assist in establish-1416 ing wireless communication links between the survivors and 1417 the closest existing cellular infrastructure. In SAR missions, 1418 to minimize the valuable time of finding and saving victims, 1419 Waharte et al. [257] investigated a number of important 1420 parameters that can have an effect on the SAR tasks including 1421 the quality of collected sensory data, energy constraints of 1422 the UAVs, environmental hazards (e.g. trees, winds) and the 1423 level of information exchange between UAVs. The authors 1424 then assessed and compared the advantage of sharing infor-1425 mation among UAVs with different search methods based 1426 on a greedy heuristic algorithm, potential fields and partially 1427 observable Markov decision technique. According to statis-1428 tics [258], during an avalanche incident, the survival proba-1429 bility of entirely buried victims can decrease to below 80% 1430 after only 10 minutes of being buried. Silvagni et al. [259]   . Current methods of 1493 wildfire detection such as satellite imaging and camera-based 1494 sensing are relatively slow and unreliable. In a UAV-IoT net-1495 work, IoT devices were used to detect fires at an early stage, 1496 and the results were broadcast to nearby UAVs. The authors 1497 studied the optimization of the density of IoT devices as well 1498 as UAVs covering a given forested area. Using discrete-time 1499 Markov chain analysis, they found that a UAV-IoT network 1500 can offer more reliable and timely detection of wildfires than 1501 satellite imaging techniques. Bushnaq et al. implemented a 1502 cloud service to enable video streaming for use with emer-1503 gency services, as well as control commands for the UAV 1504 systems within the cloud service [269]. The goal was to inte-1505 grate a web application and mobile client into the EURECOM 1506 IoT platform for the command, control, and supervision of 1507 various missions. Martinez-Caro and Cano presented a case 1508 study for the use of Long-Range (LoRa), low-power wide-1509 area network for the purposes of air quality monitoring [270]. 1510 The network consisted of UAVs equipped with sensors to 1511 measure air quality, as well as nodes incorporating LoRa 1512 for communications. The authors' goal was to determine the 1513 best mobility model for such a UAV-based IoT service. After 1514 extensive simulations, the authors determined that the ''Path-1515 way'' model was the best performing, where LoRa nodes 1516 move in an orderly fashion through a coverage area.

1517
UAV platforms suffer from limitations related to weight 1518 and autonomy, which impact their effectiveness for 1519 remote sensing when capturing and processing data for 1520 the use of collision avoidance and obstacle detection. where one UAV in a fixed position will communicate with 1597 the ground UE, and other UAVs within the vicinity do not 1598 communicate with the UE but are able to move about in 1599 3-D space, these UAVs will be seen as interference. First 1600 characterizing the interference received by the ground UE, 1601 then evaluating the coverage probability, the authors pro-1602 posed both random and uniform waypoint mobility models 1603 to characterize the UAV movement process. In their work, 1604 Zhang et al. investigated UAV-based emergency communi-1605 cation networks where ground power systems are not oper-1606 ational after a disaster and UE energy is limited [279]. The 1607 authors consider this UE energy limitation as well as physical 1608 obstacles to UAV flights to develop a trajectory optimization 1609 solution by simplifying this problem as a constrained Markov 1610 decision-making process and propose a Lyapunov-based deep 1611 learning trajectory design algorithm, where the UAV is the 1612 agent. The authors' work shows convergence in the uplink 1613 throughput in simulation results, with satisfactory trade-off in 1614 energy consumption. This work can be extended to multiple 1615 UAV deployment in larger disaster areas with UAVs as UEs. 1616 Another UE application of UAVs in disaster response is 1617 described in an earlier publication by Yin et al. who ana-1618 lyzed uplink performance of UAV UEs in dense cellular 1619 networks [280]. The group investigated system performance 1620 with respect to parameters such as with (non-line-of-sight) 1621 and without (line of sight) flight obstacles, antenna height 1622 difference between UAVs and base stations, and idle mode 1623 capabilities that affect inter-cell interference. They found that, 1624 as is intuitive, the probability of coverage can be improved 1625 by idle mode capability; when distance between antennas 1626 and base stations increases system performance degraded, 1627 and finally when this distance is large, the fractional power 1628 control factor is not that impactful on UAVs' performance. 1629 Pai and Sainath [281] presented their study on tethered 1630 UAV-assisted hybrid cooperative communication to improve 1631 the performance of the links between BSs and UEs through 1632 a UAV selection policy without the channel state information 1633 (CSI), and a link switching policy based on a hybrid PHY 1634 layer (RF or mmWave or FSO). The Authors' simulations 1635 resulted in a recommended selection policy for a single UAV 1636 from a swarm of UAvs. The group also investigated combin-1637 ing selections in PHY layer links with appropriate switching 1638 thresholds. This work could potentially serve as an analytical 1639 benchmark for UAV-assisted wireless systems as UAVs are 1640 used as BS and UEs in multiple applications.  the purposes of monitoring road traffic in a decentralized 1697 navigation scheme [285]. In this work, the UAVs performed 1698 four actions, including initial tasks, searching, accumulating, 1699 and monitoring. When the UAV network detects blockage, 1700 the UAVs can then move to the area for further visual investi-1701 gation of ground vehicles. The UAVs capture measurements 1702 from the scene, and share their location with one another. The 1703 simulations are implemented in a single plane, which can be 1704 expanded to 3D movement and potentially be implemented 1705 in real time. In their parallel paper, Savkin and Huang dis-1706 cussed navigation of a UAV network for surveillance using 1707 a distributed navigation algorithm. Each UAV in the network 1708 uses individual local information to determine its movement 1709 with minimal involvement from the central controller, and 1710 converge to an optimal location [286].

1711
Other groups have also implemented the use of UAV net-1712 works for traffic monitoring and surveillance. Khan et al. pro-1713 posed a UAV-based smart traffic surveillance system [287]. 1714 The proposed technique was introduced as a smart system 1715 that made use of 5G technology. The UAV is designed to 1716 track speeding vehicles on the highway. Layer 1 involves 1717 the UAV which is deployed for traffic monitoring. Layer 2 1718 represents a communication bridge between base station 1719 and layer 1. Layer 3 is the monitored traffic. Alioua et al. 1720 considered UAV data processing as applied for multi-UAV 1721 traffic monitoring [288]. The authors' approach involved 1722 computation offloading and sharing related decision making 1723 problems to reduce computational delay and optimization of 1724 energy overhead and computation/communication cost. The 1725 authors use a theoretical game approach as a three-player 1726 sequential game seeking Nash equilibrium, with simulation 1727 results showing improvements over previously used algo-1728 rithms. Deep learning approaches were employed by Gupta 1729 and Verma for urban traffic surveillance using imagery from 1730 low-flying UAVs [289]. Ahmed et al. looked into modeling 1731 mobility of multiple UAVs in urban traffic surveillance [290]. 1732 Araujo et al. described observer (UAV) and target (road vehi-1733 cle robots) for a monitoring application in a cooperative UAV 1734 scheme for urban traffic monitoring scenario [291]. Pedes-1735 trian traffic monitoring is also described by multiple authors 1736 including Huang and Savkin [292] and Wang et al. [293]. 1737 In both traffic monitoring and surveillance applications, the 1738 final goal is to improve the safety of traffic under efficient 1739 and effective UAV path planning, image processing, commu-1740 nication as well as energy considerations.

1741
Crowd monitoring and control at large public events is vital 1742 since it guarantees safety of individuals and also improves 1743 public security. An increase in crowd density and also abnor-1744 mal behavior of individuals in the crowd may lead to unpleas-1745 ant incidents [294]. Strict spatiotemporal restrictions, such 1746 as those used in religious festivals including Hajj, increase 1747 the likelihood of dangers [295], [296]. In addition, potential 1748 public health hazards in such large crowds may even be 1749 more serious, including the spread of infectious illnesses, 1750 heat-related disorders, the potential for terrorist attacks, and 1751 aggressive mob behavior brought on by alcohol and/or drug 1752 usage [297]. UAVs can be utilized for crowd control and monitoring activities [298]. DeMoraes et al. in [298] introduced a multi-UAV based crowd monitoring system that 1755 utilizes UAVs to regularly monitor moving individuals. The

1809
In agriculture, there are many uses for UAVs. For example, 1810 Christiansen et al. used a UAV to measure the height of 1811 crops on a wheat farm in order to determine the correct 1812 level of nitrogen treatment [300]. By combining the data 1813 obtained from a Light Detection and Ranging (LiDAR) unit, 1814 Global Navigation Satellite System (GNSS), and an Inertial 1815 Measurement Unit (IMU), the authors could then generate a 1816 point cloud, which was recorded, mapped, and analyzed using 1817 functionalities within the Robot Operating System (ROS) as 1818 well as the Point Cloud Library (PCL). The authors could also 1819 estimate crop volume from this data as well.

1820
UAVs can also be used for inspecting pipelines, power 1821 transmission lines, wind turbines, and more. As stated earlier, 1822 Gammill et al. report that drones can be 97% more efficient 1823 in solar farm inspections when compared to manual inspec-1824 tions [67] [303]. 1830 Using magnetic field data, and combining a metaheuristic 1831 algorithm and interior point method into their own algorithm, 1832 the authors were able to reconstruct the position and cur-1833 rent parameters of the transmission lines. The algorithm was 1834 shown in experimentation to be useful for transmission line 1835 monitoring and controlling the trajectory of the UAV for such 1836 purposes. 1837 Elmokadem et al. [304] provide a comprehensive review 1838 of some of the recent advancements in the field of UAVs in 1839 regard to safe autonomous navigation. A significant portion 1840 of this article is focused on the state-of-the-art techniques 1841 capable of producing three-dimensional avoidance maneu-1842 vers and safe trajectories. UAVs for an automated delivery system in an urban envi-1847 ronment [305]. The study identified scheduling of multiple 1848 UAVs and multiple flights to be problematic within the sys-1849 tem. They proposed a multiple objectives decision-making 1850 method and special encoding method to tackle the prob-1851 lem, and was able to experimentally determine that the pro-1852 posed algorithms were able to solve the problem on a small 1853 scale.

1854
Another application of multi-UAV systems is explored by 1855 Maza et al. where an architecture for a cooperative system 1856 of UAVs used for joint payload delivery is presented [306]. 1857 A control system was proposed to enable several UAVs to 1858 work together to transport a single load.

1859
Logistics carriers attempt to perform the last-mile parcel 1860 delivery through the air to customers to benefit from its 1861 flexibility and convenience. However, there are still some 1862 challenges to belong categorically to area and target cover-1918 age; path planning, collision avoidance in swarming, swarm 1919 formation and energy planning; collection, analysis and visu-1920 alization of visual data; network design, network connec-1921 tivity, quality of service, and general safety and security; 1922 and flight control and controllers and learning-based meth-1923 ods [36]. All these challenges create opportunities to enable 1924 cyber-physical applications in large UAV networks. The use of drones has been considered as a complement 2034 to the existing cellular networks, in order to obtain higher 2035 transmission efficiency with improved communication cov-2036 erage and channel capacity. However, the extensively used 2037 microwave frequency bands below 6 GHz employed by con-2038 ventional wireless networks cannot sufficiently provide a 2039 significant improvement in terms of data rates for many 2040 upcoming emerging applications. Using the vast amounts of 2041 unutilized bandwidth present at millimeter wave frequen-2042 cies (over 30-300 GHz) is one possible solution to the 2043 spectrum crunch dilemma and to address the needs of 5G 2044 and beyond for mobile communications [28]. By consider-2045 ing the use of UAV-assisted cellular networks in mmWave 2046 spectrum, an important challenge is very high propaga-2047 tion loss at millimeter wave. In other words, the mmWave 2048 spectrum's propagated signals suffer from significant prop-2049 agation loss and susceptibility to obstruction, which can 2050 lead to a high likelihood of outages and a low signal-to-2051 noise ratio (SNR) [329]. Nevertheless, multiple antennas 2052 can be built into a small UAV due to the short wave-2053 length of mmWave signals which can help in mitigat-2054 ing the propagation loss issue [330]. In addition, many 2055 works have been done to model the multiple-input multiple-2056 output (MIMO) channel for mmWave communications. For 2057 instance, Ma et al. in [331] investigated a Non-Stationary 2058 geometry-based MIMO channel model for millimeter-Wave 2059 UAV networks. Multiple antenna technologies have shown 2060 to have promising future. Zhang et al. [332] provided a com-2061 prehensive review regarding three novel multiple antenna 2062 technologies that might be significant and play important 2063 roles in beyond 5G networks: These technologies are cell-free 2064 massive MIMO [333], beamspace massive MIMO [334], and 2065 intelligent reflecting surfaces [334]. Another approach that 2066 can be used to deal with high propagation loss is to beam-2067 forming technique. In this method, directional antennas or 2068 antenna arrays are used to obtain high beam gains in order to 2069 increase the communication coverage [335]. Xiao et al. [335] 2070 provided a comprehensive survey on mmWave Beamforming 2071 enabled UAV communications. Moreover, Zhang et al. [336] 2072 presented a novel D2D-based UAV mmWave communica-2073 tion framework where the flying drones had severe energy 2074 limitations. The authors showed that there is a need to use a 2075 duty cycling mechanism such that drones' radio can only be 2076 turned on when it is necessary and also demonstrated that it 2077 is necessary to overcome the beam misalignments that caused 2078 by the radio OFF periods. Authors then suggested a new fast 2079 beam tracking discontinuous reception method to deal with 2080 these issue. Santana et al. [349] provided a comprehensive overview of 2147 the technology trends that involve the mix of UAVs and CR. 2148 Specially the authors in [349] provided a more up-to-date 2149 perspective in which the main concern was focused on how 2150 UAVs that were operating in unlicensed frequency spectrum 2151 bands were able to compete with mobile communication 2152 technologies. The main perspective was how to implement 2153 CR into the UAVs. Allowing UAVs to utilize PU resources as 2154 SU shows a good opportunity for the emerging technology. 2155 Another proposed approach of implementing CR into UAVs 2156 involves the employment of energy harvesting techniques. 2157 The paper by Xiao et al. [335] introduced the perspective of 2158 UAV-assisted energy harvesting wireless networks. In their 2159 paper, they claimed that they could obtain a significant fre-2160 quency spectrum and energy efficiency through UAV-assisted 2161 energy harvesting cognitive radio network (UAV-EH-CRN). 2162 This work showed how a drone could adjust its communi-2163 cation transmissions to a dedicated receiver based on the 2164 positive identification of a PU in the frequency spectrum 2165 band. The authors in [350] proposed a technique to integrate 2166 the capabilities of spectrum sharing within drones to assist 2167 with mission-critical services. In addition, CR can be utilized 2168 in natural disaster scenarios. During the lack or destruction 2169 of network resources due to a disaster, it might be possi-2170 ble for Cognitive Radio Networks (CRN) to use UAVs as 2171 relays. Nguyen et al. proposed a technique to optimize the 2172 implementation of such drone relays for both PUs and SUs 2173 within the CR schema. Another interesting research work for 2174 radio spectrum resource optimization can be seen in [351]. 2175 Wang et al. suggested the implementation of a UAV relay 2176 network that could assist with the communication between a 2177 secondary base station and a SU. As a result, the SU could 2178 coexist with the PU at the same band. Nobar et al. [352] 2179 developed an updated perspective into the resource allocation 2180 with CR enabled UAV communications. They presented a 2181 similar perspective to the work that Wang et al. proposed 2182 but with further results and an optimized algorithm. It can be 2183 seen, that a significant work has been done for implementing 2184 CR into UAVs to assist with the spectrum scarcity. An SU 2185 capability to accessed licensed PU spectrum without affecting 2186 the integrity of its communication is a great capability that can 2187 be enhanced with UAVs.  characteristics. Therefore, for better and more cost-effective 2249 design and also improvement in performance of UAV com-2250 munication, it is critical to be accurately investigated some 2251 of the most important features of the UAV channels. Several 2252 challenges still exist for modeling of UAV channels. For 2253 instance, in non-stationary channels, the propagation prop-2254 erties of channels for temporal and spatial fluctuations are 2255 still under investigated. Furthermore, airframe shadowing 2256 characteristics of tiny UAVs with rotary-wing has yet to be 2257 studied [138]. The following are the most distinct proper-2258 ties that differentiate UAV communication from traditional 2259 wireless communication: 1) highly dynamic properties of 2260 communication channel of UAV for radio propagation of air-2261 to-air and air-to-ground that is caused as a result of UAV high 2262 mobility [162], [362]; 2) Airframe shadowing which is one of 2263 the less-studied characteristics of the air-to-ground channels. . This method is especially valuable 2287 when there are non-stationary properties in the air-to-ground 2288 channel. Lastly, geometric-based stochastic models can pro-2289 vide effective tools for assessing temporal-spatial features 2290 in a geometric simulation environments. For describing the 2291 air-to-ground channels in a 3D plane with less environmen-2292 tal factors, these methods are preferred [370], [371]. The 2293 air-to-air propagation channel, unlike the air-to-ground 2294 channel, is primarily used in multi-hop UAV networks for the 2295 purpose of autonomous coordinating and managing between 2296 UAVs, as well as supporting back-haul radio connectivity 2297 to complement current communication systems [138], [372]. 2298 Furthermore, the propagation properties of air-to-air chan-2299 nels are comparable to propagation characteristics in free 2300 space and are heavily reliant on line-of-sight propagation 2301 and ground reflection effects [138]. Authors in [373] investi-2302 gated the propagation characteristics of air-to-air channels in 2303 urban environments. Authors in [374] proposed a wideband 2304 munications. In this work, authors suggested to use a 2306 three-dimensional (3D) non-stationary geometry-based 2307 stochastic model for air-to-air channels in UAV communi-2308 cation. In current literature, Broadly used low-power radios 2309 that are designed based on IEEE 802.15.4 [375], IEEE 2310 802.11 [376], [377] and LoRa standards [378] have been 2311 used to experimentally characterize the air-to-air propaga-2312 tion channel [379]. However, the influence of the Doppler  As commercial drones become ever more popular and their 2370 operational range grow rapidly, their security issues become 2371 more important. Typically, during the flight, drones require to 2372 operate within a wireless communication network to achieve 2373 their operational goals [382]. Drones may also be controlled 2374 remotely, in which case, can lead to an unique opportunity 2375 for the cyber-attacks (e.g., taking over control or denial-of-2376 service (DoS)) [383]. Ly et al. [384] provided a comprehen-2377 sive review on different types of cyber threats. The types of 2378 cyber-attacks reviewed in this work were categorized into 2379 three groups: model of threats, the type of challenges they 2380 pose, and the required tools for the attack.

2381
Many scientists have investigated various security vul-2382 nerabilities of wireless protocols. For instance, authors 2383 in [385] investigated the security vulnerabilities imposed 2384 by the use of wireless protocols and then proposed effec-2385 tive methods for increasing the wireless network security. 2386 Pelechrinis et al. [386] introduced different mechanisms for 2387 detection of jamming attacks in wireless networks and then 2388 proposed various techniques to defend network from these 2389 attacks.

2390
In cellular networks, drones can either be utilized as aerial 2391 base stations to complement the terrestrial base stations in 2392 order to provide wireless services to ground users; or be used 2393 as independent aerial users within the network consisting 2394 of terrestrial base stations. However, since drones are only 2395 operational at high altitude in cellular networks, they are 2396 able to effectively establish direct line-of-sight communica-2397 tion links with other terrestrial users, which in return can 2398 pose new challenges for security of cellular networks [387]. 2399 On the one hand, jamming attacks and eavesdropping by 2400 malevolent nodes on the ground are more likely to occur dur-2401 ing UAV-ground communications. On the other hand, mali-2402 cious drones are more capable of intercepting and disrupting 2403 ground communications than malicious ground nodes [387]. 2404 Wu et al. [387] explored aforementioned new concerns from 2405 the perspective of physical-layer security and provided cre-2406 ative solutions to effectively address them. Figure 8 shows an 2407 example of occurring eavesdropping and jamming attacks by 2408 malicious nodes on the ground. Drones can also be integrated 2409 into WSNs to deal with potential threats and attacks such as 2410 jamming attacks as shown in [388].

2411
The Internet of Drones (IoD) is relatively new architecture 2412 designed recently for providing managed access to controlled 2413 airspace for UAVs [389], [390]. Internet of Drone Things 2414 VOLUME 10, 2022 FIGURE 8. Eavesdropping and jamming attacks conducted by malicious nodes on the ground in a cellular network [387].
(IoDT) is also expected to be the potential future path of UAVs backend through IoT, big data, cloud computing, smart 2416 computer vision, advanced wireless protocols, and high-end 2417 security methods [391]. The main goal of the IoDT is to make To do this, a programmable 2490 control layer has taken the role of the division between the 2491 network's control structure and communication infrastruc-2492 ture, enabling setting of the network's behavior. However, 2493 in conventional networking practices, the network itself is in 2494 charge of both communication and control operations. In con-2495 trast to the conventional networks, in which, the whole sys-2496 tem must be reconfigured in order to upgrade the system, 2497 in the SDN, only the software requires an update, which 2498 is a more efficient approach for upgrading the system and 2499 reduction of the overall cost [398], [400]. SDN has shown 2500 to be a flexible platform and it can also be programmed by 2501 high-level programming languages [401]. In order to enhance 2502 overall network performance and also identify defects, SDN 2503 enables network parameters to be adjusted based on the 2504 operating environment. SDN  The support for robotic swarms is also and added enhance-2611 ment for developers that have to be taken in consideration. 2612 The support or ArduPilot is also critical for researchers 2613 developing their own drone for specific research tasks. Sim-2614 ilar to jMAVSim the utilization of this software requires 2615 the researcher to be more experience with Linux systems. 2616 Figure 9 shows an example of working with Gazebo simulator 2617 with swarm drones. 2618 VOLUME 10, 2022 FIGURE 9. Gazebo simulator with swarm drones [420].

5) MICROSOFT AirSim
simulator for drones, cars, and more, built on Unreal Engine. applications. Figure 10 shows an example scenario of work-2648 ing with MathWorks simulator environment.

2649
Researchers can use this toolbox to design autonomous 2650 flight algorithms, UAV missions, and flight controllers. 2651 An accompanying Flight Log Analyzer app lets developers 2652 to interactively analyze 3D flight paths, telemetry informa-2653 tion, and sensor readings from common flight log formats. 2654 Users can also generate and simulate UAV scenarios with 2655 an HITL testing of autonomous flight algorithms and flight 2656 controllers. Sensors such as camera, lidar, IMU, and GPS 2657 can be incorporated within the simulation in a photorealis-2658 tic 3D environment. The UAV Toolbox also provides refer-2659 ence application examples for common UAV usages, such 2660 as autonomous drone package delivery with multirotor UAV. 2661 The toolbox supports C/C++ code generation for rapid pro-2662 totyping, HITL testing, and standalone deployment to hard-2663 ware, such as the Pixhawk R Autopilot.

2664
Mathworks has been a constant innovator and producer 2665 of tools for academia, scientific and industry development 2666 of technology. Their UAV Toolbox has many options and 2667 the company keeps investing on upgrading the tools trying 2668 to meet the user needs. The mathematical capability and 2669 flexibility to produce results and models with their MATLAB 2670 and Simulink components still very critical in the scientific 2671 community. Routing protocols play an important role in UAV networks. 2749 Although research on routing protocols for ad hoc net-2750 works has grown significantly in recent years, they cannot 2751 be directly applied to drones. Designing an effective rout-2752 ing protocol to manage mobility, specifically for high-speed 2753 drones, is a difficult challenge [232]. The repeated change 2754 of topology as well as disconnection of radio links owing to 2755 high-speed result in the UAV network routing issues [426]. 2756 Therefore, there should be a routing protocol that efficiently 2757 resolves these issues such as [203], [429], and [430]. Lit-2758 tle research has been done on cross-layer design routing 2759 protocols. Cross layer design enables interaction between 2760 OSI layers and assists in obtaining numerous routing met-2761 rics [429], [430]. New cross-layer routing protocols such 2762 as [432], [433], and [434] has recently been introduced for 2763 improvement of routing protocols. Authors in [240] pro-2764 vided a comprehensive review on routing protocols from a 2765 cross-layer design perspective. Furthermore, security risks 2766 are not taken into account by present routing protocols [433]. 2767 For improving UAV communication security in both the 2768 physical and network levels, authors in [435] and [436] pro-2769 vided a thorough evaluation of the security countermeasures 2770 already in place. Ensuring efficient QoS is a challenging issue 2771 in UAV communication networks. Therefore, there should be 2772 a requirement for a system that can enhance the performance 2773 of the UAV Communication Network to guarantee efficient 2774 QoS [426]. Future study may be focused on minimizing 2775 the ratio of packet loss or routing failure caused by the 2776 intermittent connectivity as a result of the rapid mobility of 2777 UAVs. For instance, geographic position mobility-oriented 2778 routing (GPMOR) utilizes a prediction algorithm in order to 2779 designate the next forwarding UAV based on a Gauss-Markov 2780 mobility model [435]. In addition to the Generic algorithms, 2781 further QoS algorithms might be investigated by leverag-2782 ing hybrid routing protocols to find the best path [436]. 2783 Delay is another important factor that affects the QoS. The 2784 researchers might investigate the QoS-preserving and delay-2785 minimizing routing strategies, nevertheless [426]. In con-2786 trast to on-ground old-style transmitters and receivers that 2787 were powered by external power sources, UAVs are pow-2788 ered by batteries with limited capacity, which means that 2789 the energy available for carrying out different operations 2790 such as sensing information, on-board computation, wireless 2791 data transmission and flight control is limited. According  Therefore, it needs to be properly addressed [241].
ated an optical quantum channel that uses several drones  In future, UAV-assisted wireless networks will possibly be 2852 combined with mmwave communications to not merely 2853 obtain higher transmission efficiency, increasing coverage 2854 range and network capacity but also be utilized to provide 2855 support to a broad range of 5G and beyond wireless appli-2856 cations [28]. With high data transmission throughput, ultra-2857 fast speed, large wireless bandwidth, super-low transmission 2858 latency and increased connectivity, these new applications 2859 are projected to unleash a gigantic IoT ecosystem. Consid-2860 ering these interesting opportunities and new applications, 2861 it will be difficult and challenging to design, control and 2862 optimize UAV-assisted wireless networks incorporated with 2863 mmWave communications [28]. Machine learning algorithms 2864 can be used to assist in intelligent decision making. Many 2865 machine learning algorithms have been used to support 2866 UAV-assisted wireless networks. As an example, based on 2867 the prediction of users' mobility information, a framework 2868 was proposed by [  To be able to offer wireless communication services to ground 2880 users over a substantially large geographical area, a swarm 2881 of UAVs is required to form a multi-hop wireless network. 2882 Information packets will then be sent to different UAVs with 2883 different trajectories. Although, UAVs must keep their radio 2884 communication links close to the ground users, however, 2885 because of fast mobility, the radio links between nearby 2886 UAVs are interrupted frequently. As a result of these inter-2887 ruptions, many current conventional routing protocols will 2888 not work properly in FANETs. Hence, the main challenge 2889 is the manner in which flight of UAVs are controlled to 2890 provide acceptable services. Furthermore, when UAVs decide 2891 to collaborate with each other, avoidance of collisions also 2892 become a major issue and needs to be considered in order to 2893 guarantee UAVs safe operation. On the other hand, cutting-2894 edge satellite-to-UAV channel characteristics require detailed 2895 information regarding the propagation effects. The develop-2896 ment of cutting-edge propagation models for satellite-to-UAV 2897 communication is yet in its early stages and will be a subject 2898 for future research [442].

2900
Not very long ago, unmanned Aerial Vehicles (UAVs), also 2901 known as drones, were a technology primarily used for 2902