An Overview of Aerial Wireless Relay Networks for ...

1376IEICE TRANS. COMMUN., VOL.E103–B, NO.12 DECEMBER 2020

INVITED PAPER Special Section on IoT Sensor Networks and Mobile Intelligence

An Overview of Aerial Wireless Relay Networks for EmergencyCommunications during Large-Scale Disasters

Hiraku OKADA†a), Senior Member

SUMMARY In emergency communication systems research, aerialwireless relay networks (AWRNs) using multicopter unmanned aerial ve-hicles (UAVs) have been proposed. The main issue of the AWRNs is howto minimize the delay time of packet transmissions since it is not easy tosupply many multicopters to cover a wide area. In this paper, we reviewthe flight schemes and their delay time for the AWRNs. Furthermore, thenetwork has specific issues such as multicopters’ drops due to their batterycapacity depletion and inclination of moving multicopters. The inclinationof multicopters affects the received power, and the communication rangechanges based on the inclination as well. Therefore, we clarify the effect ofthese issues on the delay time.key words: aerial wireless relay networks, multicopter UAV, delay time,flight schemes

1. Introduction

The effect of the 2004 Mid Niigata Prefecture Earthquakeserves the primary research question of this paper. Themagnitude of the earthquake is M6.8, and the intensity onthe Japanese seven-stage seismic scale is 7. In Yamakoshivillage, as a lot of landslides destroyed lifelines, the vil-lage stood alone. The infrastructures of communication net-works were damaged; the villagers could not cry for helpfrom the outside of the village. It was challenging for thepeople inside and outside the village to collect informationfrom the disaster-stricken areas. Consequently, therefore,we identify the need for emergency communication systemsduring large-scale disasters.

K. Mase et al. started the project, SKYMESH, to con-struct an ad hoc network in the sky for large-scale disas-ter recovery [1]–[3]. In SKYMESH, balloons float at 50–100 m above the ground, with wireless nodes suspendedfrom the balloons. The wireless nodes communicate witheach other via wireless transmissions and construct an adhoc network in the sky. Due to the significant locations ofthe balloons, SKYMESH has advantages of good line-of-sight (LoS), long transmission distance, and low interfer-ence from wireless systems on the ground.

Nowadays, the technologies of unmanned aerial vehi-cles (UAVs) are rapidly developed and widely used in var-ious fields. For example, in [4], a fixed-wing mini UAV isused for a wireless relay system to extend communication

Manuscript received February 27, 2020.Manuscript revised April 7, 2020.Manuscript publicized July 1, 2020.†The author is with the Institute of Materials and Systems for

Sustainability, Nagoya University, Nagoya-shi, 464-8603 Japan.a) E-mail: [email protected]

DOI: 10.1587/transcom.2020SEI0001

capability to the disaster-stricken area. Further, in our pre-vious work, We proposed an aerial wireless relay network(AWRN) using multicopter UAVs to construct an ad hoc inthe sky [5]–[7]. Unlike balloons, multicopters fly within adisaster-stricken area. Also, the reason to use multicoptersis their mobility, which allows us to employ the technologyof disruption-tolerant networking (DTN) [8]. Although themulticopters cannot be supplied enough to make the con-nections among them, moving the multicopters sustains thenetwork.

The main issue of the AWRN is how to reduce the delaytime of packet transmissions since it is not easy to supplymany multicopters to cover a wide area. Unlike wirelesscommunications, the delay time by multicopters’ movementis significant. Thus, to minimize the delay time depends onhow the multicopters move, which we call a flight scheme.

Thus, in this paper, we study the AWRN using multi-copters and flight schemes to minimize the delay time. Al-though our research group has been investigating the flightschemes and their delay time [6], [7], [9], we review thesestudies and discuss the feasibility of the AWRN for emer-gency communication systems during large-scale disasters.

Then, the remainder of this paper is organized as fol-lows: We compare the features of a balloon and multicopterin Sect. 2 and discuss the AWRN in Sect. 3. The networkhas particular issues we considered, which are discussed inSects. 4 and 5. Finally, Sect. 6 presents the conclusions ofthis paper.

2. Balloon versus Multicopter

Table 1 compares the features of a balloon and multicopter.The balloon is tethered to a fixed point on the ground; there-fore, it cannot move. The power can be supplied by eithera battery in the wireless node or wire from the ground. Theweight of the balloon is about 13 kg, and its size is about 8 m× 4 m [1]. The treatment of a balloon is not easy; however,it can float for a long duration until the balloon deflates.

Whereas, the multicopter flies with a battery. Com-pared with the balloon, the multicopter can move with highspeed, the weight is light, and the size is smaller. The flightduration is short due to the battery limitation, but it is pos-sible to supply power to the multicopter from the groundusing wire. In this case, the flight duration becomes long,but mobility is lost.

Copyright c© 2020 The Institute of Electronics, Information and Communication Engineers

OKADA: AN OVERVIEW OF AERIAL WIRELESS RELAY NETWORKS1377

Table 1 Comparison between a balloon and multicopter.

Balloon Multicopter

Power Battery/wire Battery WiresupplyWeight Heavy (about 13 kg) Light (about 1 kg)

Size Large SmallFlight Long Short Long

duration (more than 10 h) (20–40 min)Mobility N/A Max. 50 km/h N/A

Fig. 1 An aerial wireless relay network.

3. An Aerial Wireless Relay Network

Figure 1 shows an aerial wireless relay network (AWRN)using the multicopters. The multicopters construct a wire-less relay network in the sky, which serves as a backbonenetwork. Also, each multicopter operates as an access pointand accommodates user nodes on the ground. Then, apacket that is generated by a source node gets transmittedto a destination node through the AWRN. If a multicopterexists in the communication range of another multicopter,they can forward packets to each other. Otherwise, the pack-ets get conveyed by the movement of the multicopters.

3.1 Flight Schemes

In the AWRN, flight schemes play an important role in min-imizing the delay time. In addition, emergency communica-tion systems are unexpectedly used when a large-scale dis-asters occurs. In this case, non-experts about networkingmight execute the operations of the emergency communi-cation systems. Taking account of this situation, the flightschemes have to satisfy the following conditions:

• autonomous movement of multicopters without sophis-ticated settings,

• consideration of multicopters’ drops due to depletionof the battery capacity, and

• ignorance of the information such as locations of theother multicopters before communications.

We use a random manner of multicopters’ movement tosatisfy the above conditions. When some multicopters drop

Fig. 2 The rebounding flight scheme.

due to the battery depletion, the other multicopters shouldcompensate for them. If the multicopters move definitively,the flight scheme is sophisticated because it has to considermulticopters’ drops, connections between multicopters, andcoverage of user nodes on the ground.

In this section, we explain three flight schemes as fol-lows.

(1) Random Waypoint

In general, the random waypoint (RWP) [10] is often usedas an accidental movement. Here, each multicopter selectsa point randomly in the disaster-stricken area and the po-sition is regarded as a destination. The multicopter moveslinearly to the destination, whereas the new destination gotrearranged when the multicopter arrives at the current one.

Because of randomness of the RWP, there are sparseor dense areas of the distribution of multicopters within thestricken area. Furthermore, the distribution of multicoptersis sparse in border areas of the disaster-stricken area due toborder effects [11], [12]. In the dense area, communicationranges of the multicopters are overlapped, and the coveringefficiency in the disaster-stricken area gets degraded.

(2) Rebounding

To avoid overlaps of communication ranges among multi-copters, we proposed a rebounding flight scheme [6], [7].Figure 2 shows the operation of the rebounding flightscheme. A multicopter does not have prior informationabout the positions of the other multicopters until informa-tion exchange between the multicopters. Like the RWPflight scheme, each multicopter moves to its destination.When a multicopter moves into the communication range ofanother multicopter, they exchange information about theirlocations. Using the current location information, they de-cide new destinations as follows: first, they draw the linethrough their positions. Next, each multicopter selects a newdestination randomly between its location and the boundaryof the disaster-stricken area opposite from the other multi-

1378IEICE TRANS. COMMUN., VOL.E103–B, NO.12 DECEMBER 2020

Fig. 3 A common flight path.

copter on the line. Finally, these multicopters move linearlyto the new destinations.

By selecting the new destinations in the opposite direc-tions, the multicopters move reboundingly. The reboundingmovement can mitigate the overlap of their communicationranges.

(3) Common Flight Path

The common flight path (CFP) was proposed for the effi-cient forwarding of packets in [6], [7]. Here, some multi-copters move on a predefined path and maintain connectionsbetween the neighboring multicopters. Figure 3 shows anexample of the CFP, which is constructed as a square. Themulticopters on the path fly within the communication rangeto each other to establish the connected network. Therefore,the multicopters on the CFP forward packets via wirelesstransmissions, and the delay time get minimized if the CFPis used effectively for the packet transmissions.

The multicopters on the CFP can move while keepingthe network connectivity. Whereas, if the multicopters stay,a wired power supply can be used. The multicopters notbelonging to the CFP obey the RWP or rebounding flightschemes.

3.2 Delay Time Evaluation

In this section, we evaluate the delay time of RWP, rebound-ing, and CFP flight schemes.

3.2.1 Simulation Conditions

When the CFP is introduced, some multicopters fly on theCFP, whereas others move according to the RWP or re-bounding flight scheme. Here, we ignore the buffer sizeof the multicopters and use the epidemic routing protocolas the DTN. The delay time of wireless transmissions is ig-nored, as this delay is much shorter than that of packets car-ried by the multicopters. A network simulator is not used,

Table 2 Simulation settings.

Simulation area 4000 m × 4000 mPosition of source node (500, 500)

Position of destination node (3500, 3500)

Size of CFP 2000 m × 2000 m(16 multicopters)

Moving speed 10 m/sCommunication range 500 m

Fig. 4 Average delay time of RWP and rebounding flight schemes withand without CFP.

and the simulation is programmed by the C language.Table 2 shows the simulation settings. The simula-

tion area, which corresponds to the disaster-stricken area,is 4000 m × 4000 m. The positions of the source and desti-nation nodes are (500, 500) and (3500, 3500), respectively.The shape of the CFP is rectangle, and its size is 2000 m ×2000 m. The CFP is located at the center of the simulationarea. The moving speed of a multicopter is 10 m/s. Sincethe LoS can be kept among multicopters, the communica-tion range becomes long, which we set to 500 m.

3.2.2 Numerical Examples

Figure 4 shows the average delay time of four differentcases, i.e., RWP and rebounding flight schemes with andwithout CFP, where the total number of multicopters indi-cates the quantity of multicopters that move according toRWP or rebounding flight schemes, and at the same timestays on the CFP. From the figure, we found that the re-bounding flight scheme minimizes the average delay timesince the overlap of communication ranges of multicopterscan be minimized. For the case of the rebounding flightscheme, the CFP is effective in reducing the average delaytime when the total number of multicopters is greater than40. Whereas, when the total number of multicopters is lessthan 40, the limited number of multicopters that obey therebounding flight scheme causes an increase in the averagedelay time. For the RWP flight scheme, the improvement bythe CFP is hard to come by.

Next, we discuss the effect of the rebounding flight

OKADA: AN OVERVIEW OF AERIAL WIRELESS RELAY NETWORKS1379

Fig. 5 The difference of the average delay times between RWP and re-bounding flight schemes.

scheme. Figure 5 shows the difference in the average delaytimes between the RWP and the rebounding flight schemes.For the case without the CFP, the difference becomes in-significant as the total number of the multicopters becomeslarge. A good number of multicopters makes small the dif-ference between them. For the case of using the CFP, thedifference is insignificant when the total number of multi-copters is 30. The multicopters that move according to therebounding flight scheme is only 14. Then, the sparse of themoving multicopters degrades the effect of rebounding.

4. Consideration of Battery Depletion

Since the battery is limited, a multicopter has to drop andchange its battery when the battery capacity gets depleted,which causes the degradation of the delay time. In thissection, we investigate the effects of considering the mul-ticopters’ drops due to their battery depletion.

4.1 Model of Multicopters’ Drops

To consider the multicopters’ drops, we introduce flight anddrop duration. Each multicopter flies for the flight dura-tion, which is decided based on the battery capacity. Whenthe flight duration expires, the multicopter drops out of theAWRN and stops flying during the drop duration to ex-change its battery. In this period, the multicopter cannotsend and receive packets. After the drop duration, the mul-ticopter flies again and joins the network.

In this paper, we consider three timings to begin thedrop duration: random, ordering, and synchronous sched-ules.

(1) Random Timing

Each multicopter drops at randomly independent timing.Hence, the multicopters need not arrange their drop timingsfor each other. In this case, the number of on-flying multi-copters changes from time to time, and the delay time mightfluctuate.

Table 3 Simulation settings for the evaluation considering multicopters’drops.

Simulation area 4000 m × 4000 mPosition of source node (500, 500)

Position of destination node (3500, 3500)

Size of CFP 2000 m × 2000 m(16 multicopters)

Moving speed 10 m/sCommunication range 500 m

Flight duration 20 minDrop duration 10 min

(2) Ordering Timing

The drop timing of each multicopter occurs at a uniforminterval; therefore, the number of flying multicopters is con-stant. The interval gets decided by dividing the sum of theflight and the drop duration by the total number of the mul-ticopters. In this case, the drop timings have to be arrangedamong the multicopters to mitigate the fluctuation of the de-lay time.

(3) Synchronous Timing

All multicopters drop at the same timing. Here, all multi-copters fly during the flight duration, or any multicopters donot operate during the drop duration. The delay time duringthe flight duration gets reduced.

4.2 Delay Time Evaluation Considering Multicopters’Drops

Here, we evaluate the delay time based on multicopters’drops. Table 3 shows the simulation settings. The flight anddrop durations are 20 and 10 min, respectively. We use thesimplified model in which the battery is exchanged on theground at the same horizontal position when the battery ca-pacity gets depleted. In actuality, the multicopters move toa recharging station before the battery depletion. The effectof the movement was evaluated in [7].

Figure 6 shows the average delay time of the RWPwithout the CFP, the rebounding without the CFP, and therebounding with the CFP by using the random drop tim-ing. The average delay time of the RWP with the CFP getsomitted since the improvement by the CFP cannot be ob-tained for the RWP. From this figure, we confirm that theconsideration of multicopters’ drops causes the increase ofthe average delay time. To clarify the rise due to multi-copters’ drops, we depict the increasing rate of the requiredtotal number of multicopters in Fig. 7, where the increasingrate is calculated by (the number of multicopters consider-ing the drops required to achieve the target average delaytime)/(that without the consideration of drops). For the caseof the RWP flight scheme, the consideration of multicopters’drops causes the reduction of multicopters only. Therefore,the increasing rate is constant for the target average delaytime. For the case of the rebounding without the CFP, theincreasing rate becomes insignificant for the short averagedelay time. It is because the rebounding flight scheme can

1380IEICE TRANS. COMMUN., VOL.E103–B, NO.12 DECEMBER 2020

Fig. 6 Average delay time considering multicopters’ drops.

Fig. 7 Increasing rate of the total number of multicopters required toachieve the average delay time.

compensate for the drops of multicopters by rebounding.For the case of the rebounding with the CFP, the increasingrate is significant for the short average delay time, whereasit is insignificant for the long average delay time. Therefore,the CFP can mitigate the effects of multicopters’ drops whenthe total number of multicopters is large.

To evaluate the difference among drop timings, Fig. 8shows the average delay time of each drop timings for therebounding flight scheme. We are to find a significant dif-ference in the average delay time between the random andthe ordering timings. Therefore, we do not have to ar-range drop timings to keep the number of on-flying mul-ticopters constant. When the CFP is not used, the aver-age delay time of the random schedule is shorter than thatof the synchronous timing. However, the synchronous tim-ing achieves the shortest average delay time using the CFP.Then, if some multicopters on the CFP drops, the connectiv-ity of the CFP cannot be kept. Therefore, we should arrangethe drop timings of the multicopters on the CFP to maintainthe connectivity.

Fig. 8 Average delay time of the random, ordering, and synchronousdrop timings for the rebounding flight scheme.

Fig. 9 Dynamic model of a flying multicopter.

5. Consideration of Multicopters’ Inclination

A multcopter flies while inclining forward. This inclinationaffects the received power, and the communication rangechanges based on the inclination as well. In this section, weevaluate the delay time by formulating the received powerand considering the multicopters’ inclination.

5.1 Relationship between Moving Speed and Inclination

Here, we derive the relationship between the moving speedand the inclination of a multicopter. For simplification, amulticopter is represented by a cuboid, and Fig. 9 shows adynamic model of a flying multicopter. A multicopter ismoving in the x-axis direction with speed vx. The axes ξand ζ is defined as longitudinal and vertical directions ofthe multicopter, respectively. The pitch angle θη is an anglebetween the x-axis and the ζ-axis. In the dynamic model,the resultant force Fr by rolling propellers, the aerodynamicforce Fa, and the gravitational force Fg are considered.

From the dynamic model, the motion equations in thex and y axes is derived by:

mx = Fr cos θη − Faζ cos θη − Fa

ξ sin θη= (Fr − Fa

ζ ) cos θη − Faξ sin θη, (1)

mz = Fr sin θη − Faζ sin θη + Fa

ξ cos θη − Fg

= (F − Faζ ) sin θξ + Fa

ξ cos θη − Fg, (2)

where Fai represents the i-axis component of the aerody-

namic force. The gravitational force is derived by Fg = mg,

OKADA: AN OVERVIEW OF AERIAL WIRELESS RELAY NETWORKS1381

where m and g are the weight of the multicopter and thegravitational acceleration, respectively. We assume that themulticopter moves in the horizontal plane with a constantspeed, whereas the acceleration in the x and z axes is 0. Thenfrom (1) and (2), the following equation is obtained:

mg cos θξ = Faξ . (3)

Let Ci, vi, Ui, and Ai be the resistance coefficient, mov-ing speed, wind speed, and swept area of the i axis, respec-tively, where i ∈ {ξ, ζ}. The aerodynamic force is obtainedby [13] as:

Fai =

12ρCi(vi − Ui)2Ai, (4)

where ρ is air density. The wind speed is assumed to be0. From (4) and vξ = vx sin θη, the following equations isderived:

Faξ =

12ρCξ(vx sin θη)2Aξ. (5)

From (3) and (5), the pitch angle θη of the movingspeed vx is obtained by:

θη = cos−1

√(

mgρCξAξv2

x

)2

+ 1 −mg

ρCξAξv2x

. (6)

The angle θv between the multicopter’s inclination and thehorizontal plane is expressed as:

θv = 90◦ − θη. (7)

5.2 Calculation of Received Power Considering Inclina-tion

The multicopter’s inclination affects the received power be-cause of the antenna directivity and the deviation of the po-larization plane. When the transmitted power is Pt, the re-ceived power Pr can be modeled by:

Pr = Pt + Gt + Gr + Gp + Gd dBm, (8)

where Gt and Gr are the relative antenna gains of the direc-tivity at a transmitter and a receiver, respectively, Gp is thegain of the polarization plane deviation, and Gd is the pathloss gain.

5.2.1 Relative Antenna Gain

The multicopter deploys an omnidirectional antenna be-cause it can connect the neighboring multicopters in thewhole directions. The omnidirectional antenna does nothave directivity in the horizontal plane, whereas it has direc-tivity in the vertical plane. The antenna gain degrades whenthe multicopter inclines, which we call the relative antennagain.

Fig. 10 A model of intending and target multicopters to calculate relativeantenna gain.

Figure 10 shows a model of intending and target mul-ticopters to derive the relative antenna gain. The target mul-ticopter is a receiver if the intended multicopter is a trans-mitter, and vice versa. In this model, we use three vectors:the moving direction vector, the antenna direction vector d,and the position vector p of the target multicopter. Let θdbe the angle between the moving direction vector and thex-axis. The angle between the antenna direction vector andthe position vector of the target multicopter is θa.

The antenna direction vector of the non-inclination ofthe multicopter is defined as e = (0 0 1). The antenna di-rection vector d can be calculated by rotating e in the x-axisdirection with −θv around the y-axis and re-rotating it with−θd around the z-axis, i.e.,

d=e

cos θv 0 − sin θv0 1 0

sin θv 0 cos θv

cos θd sin θd 0− sin θd cos θd 0

0 0 1

=(

sin θv cos θd sin θv sin θd cos θv)

, (9)

where θd is positive when the intended multicopter moves tothe positive direction in the y-axis and negative otherwise.Then, the angle θa is obtained by:

θa = cos−1 d · p|d||p|

. (10)

When the multicopter uses a dipole antenna, the rela-tive antenna gain Gk in the vertical plane is expressed by[14] as:

Gk = 10log10

cos(π2 cos θa

)sin θa

2

dB, (11)

where k represents a transmitter (t) or a receiver (r).

5.2.2 Gain of Polarization Plane Deviation

Since the antenna pattern in the horizontal plane is omni-directional, the antenna is vertically attached to the multi-copter. Therefore, the polarization plane is vertical. Whenthe multicopter inclines, the polarization plane has a tilt, too.The deviation of the polarization planes between a transmit-ter and a receiver loses the gain defined by Gp.

1382IEICE TRANS. COMMUN., VOL.E103–B, NO.12 DECEMBER 2020

Fig. 11 A model of intending and target multicopters to calculate gain ofpolarization plane deviation.

Figure 11 shows a model of intending and target multi-copters to calculate the gain of the polarization plane devia-tion. In the figure, new axes x′ and y′ are defined instead ofthe x and y axes. The y′ axis represents the direction fromthe intending multicopter to the target one, and its origin isthe location of the intending one. The x′ axis is the directionorthogonal to the y′ axis.

The angles of the position vector of the target multi-copter and the moving direction vector are φp, and the an-gle of the z-axis and the projection vector of an antenna di-rection vector on the x′-z plane is φk, where k indicates atransmitter (t) or a receiver (r). When the magnitude of theantenna direction vector is 1, the x′ and z components of thevector are sin θv sin φp and cos θv, respectively. Therefore,the angle φk is derived by:

φx = tan−1 sin θv sin φp

cos θv. (12)

The angle of the polarization plane deviation can be derivedfrom the projection vectors on the x′ − z plane of transmit-ter’s and receiver’s antenna direction vector, which can beexpressed as:

φ =

|φt − φr | (both multicopters move

in the same direction)|φt + φr | (otherwise)

. (13)

The gain Gp of the polarization deviation is obtained by:

Gp = 10log10 cos2 φ dB. (14)

5.2.3 Path Loss Gain

The multicopters fly in the sky, and the received signal ishardly affected by the ground reflection waves. Therefore,the path loss gain is assumed to obey the free space propa-gation model, in which the path loss gain Gd of the distanced is derived by [15] as:

Gd = 20 log10

(λ

4πd

)dB, (15)

where λ is the wavelength.

Table 4 Simulation settings for the evaluation considering multicopters’inclination.

Parameters of aerial wireless relay networkSimulation area 4000 m × 4000 m

Position of source node (500, 500)Position of destination node (3500, 3500)

Moving speed 10, 15, 20 m/sTransmitted power 10 dBm

Threshold of received power−85 dBmfor successful transmissions

Aerodynamic force parametersAir density ρ 1.293 kg/m3

Resistance coefficient Cξ 0.40Swept area Aξ 0.03 m2

Fig. 12 Average delay time considering inclination of multicopters.

5.3 Delay Time Evaluation Considering Inclination

Table 4 shows the simulation settings. All multicoptersmove according to the rebounding flight scheme. The CFPis not deployed. The moving speed is 10, 15, and 20 m/s.The transmitted power is 10 dBm. The received power iscalculated by (8), (11), (14), and (15). A packet is trans-mitted successfully if the the received power is more thana predefined threshold, −85 dBm. The aerodynamic forceparameters are decided by [13] based on the specificationsof DJI Phantom 4.

Figure 12 shows the average delay time considering theinclination of multicopters. We confirm that the considera-tion of multicopters’ inclination causes an increase in theaverage delay time. When the inclination is not considered,fast movement of multicopters can reduce the average de-lay time. However, this relationship is not always valid forthe case of considering the inclination. The delay time withmoving speed 20 m/s takes the longest when the number ofmulticopters is more than 60. For the case of dense mul-ticopters, frequent rebounding causes the stay of the mul-ticopters in a small region. In this case, the average delaytime increases because of the reduction of communicationrange due to the multicopters’ inclination.

To evaluate the effect of moving speed on the average

OKADA: AN OVERVIEW OF AERIAL WIRELESS RELAY NETWORKS1383

Fig. 13 Increasing rate of average delay time considering the inclinationof multicopters.

delay time more clearly, Fig. 13 shows the increasing rate ofaverage delay time by considering the multicopters’ incli-nation. Thus, the faster the moving speed is, the larger theincreasing rate of the average delay time is. In other words,the effect of the tilt is significant when the moving speed isfast. Furthermore, the increasing rate becomes more signif-icant as the number of multicopters is large. Therefore, wereconfirm that the effect of rapid movement is unattainablefor the case of dense multicopters.

6. Conclusions

We have reviewed the flight schemes in the AWRN usingmulticopters. From the simulation results, we found that therebounding and CFP flight schemes are sufficient to reducethe average delay time. As the issues particularized for theAWRN, we have investigated the effects of drops due to bat-tery capacity depletion and the inclination of moving multi-copters. The rebounding flight scheme can compensate forthe drops of the multicopters. From the result, there is anoptimum moving as we consider the effects of the inclina-tion.

The flight schemes discussed in this paper is based onsimple operations. To reduce the delay time, especiallyfor the small number of multicopters, we shall further im-prove the flight schemes in the future work while keepingautonomous movement [16].

Acknowledgments

The author would like to thank Mr. Jyo Suzuki, Mr. HirokiYanai, Prof. Kentaro Kobayashi, Prof. Takaya Yamazato,and Prof. Masaaki Katayama of Nagoya University. Thiswork is supported in part by JSPS KAKENHI Grant Num-ber 19K04392.

References

[1] H. Suzuki, Y. Kaneko, K. Mase, S. Yamazaki, and H. Makino,

“An ad hoc network in the sky, SKYMESH, for large-scale disas-ter recovery,” IEEE Vehicular Technology Conference, Sept. 2006.DOI:10.1109/VTCFall.2019.8891275.

[2] T. Umeki, H. Okada, and K. Mase, “Evaluation of wireless channelquality for an ad hoc network in the sky, SKYMESH,” IEEE Inter-national Symposium on Wireless Communication Systems, pp.585–589, 2009.

[3] H. Okada, H. Oka, and K. Mase, “Network construction manage-ment for emergency communication system SKYMESH in largescale disaster,” IEEE International Workshop on Management ofEmerging Networks and Services, pp.875–880, 2012.

[4] Z.M. Fadlullah, D. Takaishi, H. Nishiyama, N. Kato, and R. Miura,“A dynamic trajectory control algorithm for improving the commu-nication throughput and delay in UAV-aided networks,” IEEE Netw.,vol.30, no.1, pp.100–105, Jan. 2016.

[5] A. Mori, H. Okada, K. Kobayashi, M. Katayama, and K. Mase,“Construction of a node-combined wireless network for large-scaledisasters,” IEEE Consumer Communications and Networking Con-ference, pp.219–224, 2015.

[6] H. Yanai, H. Okada, K. Kobayashi, and M. Katayama, “Flightschemes considering breaks for wireless relay networks using dronesduring large-scale disasters,” IEEE Vehicular Technology Confer-ence, Aug. 2018. DOI:10.1109/VTCFall.2018.8690913.

[7] H. Yanai, H. Okada, K. Kobayashi, and M. Katayama, “Flight mod-els in wireless relay networks using drones for large-scale disasters,”IEICE Trans. Commun., vol.J103-B, no.2, pp.57–66, Feb. 2020 (inJapanese).

[8] M.J. Khabbaz, C.M. Assi, and W.F. Fawaz, “Disruption-tolerantnetworking: A comprehensive survey on recent developments andpersisting challenges,” IEEE Commun. Surveys Tuts., vol.14, no.2,pp.607–640, 2nd Quarter 2012.

[9] H. Okada, J. Suzuki, H. Yanai, K. Kobayashi, and M. Katayama,“Inclination of flying drones in aerial wireless relay networks,”IEEE Vehicular Technology Conference, Sept. 2019. DOI:10.1109/

VTCFall.2019.8891275.[10] T. Camp, J. Boleng, and V. Davies, “A survey of mobility models for

ad hoc network research,” Wirel. Commun. Mob. Comput., vol.2,no.5, pp.483–502, Aug. 2002.

[11] C. Bettstetter, G. Resta, and P. Santi, “The node distribution of therandom waypoint mobility model for wireless ad hoc networks,”IEEE Trans. Mobile Comput., vol.2, no.3, pp.257–269, July–Sept.2003.

[12] C. Bettstetter, “Mobility modeling in wireless networks: Categoriza-tion, smooth movement, and border effects,” ACM Mobile Comput.Commun. Rev., vol.5, no.3, pp.55–66, July 2001.

[13] T. Nomura and T. Kobayashi, “A fundamental study on flight mod-elling of drone,” Summaries to Technical Papers of Annual Meeting,Japan Association for Wind Engineering, pp.149–150, 2017.

[14] S. Silver, ed., Microwave Antenna Theory and Design, McGraw-Hill, 1949.

[15] J.G. Proakis and M. Salehi, Digital Communications, 5th ed.,McGraw-Hill, 2007.

[16] H. Yanai, H. Okada, K. Kobayashi, and M. Katayama, “An emer-gency communication system using drones for large-scale disas-ter — An study on flight model using estimation of drones’ po-sitional distribution —,” IEICE Technical Report, SeMI2019-138,March 2020 (in Japanese).

1384IEICE TRANS. COMMUN., VOL.E103–B, NO.12 DECEMBER 2020

Hiraku Okada received the B.S., M.S.and Ph.D. degrees in Information ElectronicsEngineering from Nagoya University, Japan in1995, 1997 and 1999, respectively. From 1997to 2000, he was a Research Fellow of the JapanSociety for the Promotion of Science. He was anAssistant Professor at Nagoya University from2000 to 2006, an Associate Professor at NiigataUniversity from 2006 to 2009, and an AssociateProfessor at Saitama University from 2009 to2011. Since 2011, he has been an Associate Pro-

fessor at Nagoya University. His current research interests include wirelesscommunication systems, wireless networks, inter-vehicle communications,and visible light communication systems. He received the Inose ScienceAward in 1996, the IEICE Young Engineer Award in 1998, the IEICE Com-munications Society ComEX Best Letter Award in 2014, and the IEEECCNC Best Paper Award in 2020. Dr. Okada is a member of IEEE andACM.