Download as pdf or txt
Download as pdf or txt
You are on page 1of 26

Spectrum Sharing of Drone Networks

38
Chiya Zhang, Zhiqing Wei, Zhiyong Feng, and Wei Zhang

Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1280
Spectrum Sharing of UAV Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1284
Optimal DSC Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1284
Effect of Different Path Loss Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1287
Spectrum Sharing Between DSC Network and Cellular Network . . . . . . . . . . . . . . . . . . . . . 1288
Optimal DSC Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1288
Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1292
Spectrum Sharing of UAV Networks with Directional Antennas . . . . . . . . . . . . . . . . . . . . . . 1293
3D Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1293
2D Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1295
Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1297
Mobility Pattern Cognition of UAV Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1297
Mobility Pattern Cognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1299
3D UAV Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1299
2D UAV Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1301
Summary and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1302
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1303

C. Zhang ()  W. Zhang


School of Electrical Engineering and Telecommunications, The University of New South Wales,
Sydney, NSW, Australia
e-mail: chiya.zhang@unsw.edu.au; w.zhang@unsw.edu.au
Z. Wei  Z. Feng
Key Laboratory of Universal Wireless Communications, Ministry of Education, Beijing
University of Posts and Telecommunications, Beijing, China
e-mail: weizhiqing@bupt.edu.cn; fengzy@bupt.edu.cn

© Springer Nature Singapore Pte Ltd. 2019 1279


W. Zhang (ed.), Handbook of Cognitive Radio,
https://doi.org/10.1007/978-981-10-1394-2_35
1280 C. Zhang et al.

Abstract
Drone networks are aerial base stations that can be used to support cellular
networks. The underlay spectrum sharing between the three-dimensional (3D)
drone small cells (DSCs) downlink network modeled by a 3D Poisson point
process and traditional cellular networks modeled by a 2D Poisson point process
is introduced. To maximize the DSC network throughput while satisfying the
cellular network efficiency constraint, the optimal density of DSC aerial base
stations is discussed. The maximum throughput of the DSC user increases almost
linearly with the increase of the DSC outage constraint. Effects of directional
transmission on DSC networks is further discussed. Besides density control,
power and beam control can also be applied in the spectrum sharing between
unmanned aerial vehicle (UAV) network and ground network. With the mobility
pattern information of UAVs, the delay-tolerant transmissions can be constructed
and multiple transmission modes are implemented to carry various types of
traffic. Exploiting cognition capability on mobility, UAV network can provide
high quality of information services in the highly dynamic environment with
limited resources.

Introduction

UAV, commonly known as a drone, refers to an aircraft without a human pilot


aboard. We can expect drones developing in many interesting applications, such
as policing and surveillance and scientific research. As illustrated in Fig. 1, UAVs
aerial base stations can establish communication connection between ground node A
and B. UAVs can also act as aerial sensors to collect environment data and transmit
data to the control station. UAVs can act as aerial base stations or relays, namely,
drone small cells (DSCs), to provide communication services to the areas with
natural disasters and traffic congestions. DSCs can be deployed in a high-altitude
platform (HAP) which is above 10 km height or in low-altitude platform (LAP)
below 10 km [1].
DSC network has several advantages over conventional cellular networks.
The fifth generation communication utilizing mmWave signals suffers many
propagation-related shortcomings, such as relatively short range and vulnerable to
blockage. UAVs can exploit its mobility to avoid the blockage and provide seamless
coverage. Further, aerial base stations are robust to environmental changes. Another
advantage of UAV networks is the flexibility of reconfiguration. For example, UAVs
can be deployed to help offloading cellular networks in wireless congestion events
with low cost. In [2], optimal deployment altitude of a drone providing maximum
coverage is discussed. The deployment design and performance analysis of DSCs
at LAP are further studied in [3], in which the optimal altitude maximizing ground
coverage and minimizing required transmit power for a single DSC is derived.
Their result shows that the optimal deployment altitude varies according to different
environment statistics.
38 Spectrum Sharing of Drone Networks 1281

UAV

B
A

Control station

Ve
ehicle
Vehicle

Connection path Data of ground users Sensing data

Fig. 1 Application scenario of UAV network

UAVs can be preprogrammed to fly autonomously from the starting point to


destination. However, when multiple UAVs acting as cas aerial base stations, the
ability of remote control could significantly improve the operation and safety [4].
For instance, if there is a possible collision due to traffic management, it is essential
that information is shared to the UAVs promptly. Federal Aviation Administration
(FAA) is exploring that each intended UAV use case has certain coverage, latency,
and reliability requirements to ensure proper operational control of the UAVs [5]. In
a typical multi-UAV network, the ground control station is mainly responsible for
dispatching, coordinating, charging, and collecting of UAVs [6].
In the scenario where UAVs have a relatively small transmission range, the traffic
is usually generated within the UAVs and transmits to control station via multi-
hop transmissions. Repeated data transmissions in multi-hop transmission manner
will consume radio resources and decrease network capacity. Thus environment
cognition is proposed to improve the capacity of UAV network. Also, broadband
UAV network would need to be constructed to support massive data transmission.
According to Shannon-Hartley theorem [7], the capacity increases logarithmi-
cally with the transmit power. However, the capacity increases linearly with the
bandwidth of spectrum. UAVs have limited carrying capacity and energy supply.
Thus exploiting more spectrum is a practical way to increase the capacity of UAV
network. With multiple sensors such as radar, Global Positioning System (GPS) ,
1282 C. Zhang et al.

and camera implemented in UAVs, the environment cognition besides traditional


spectrum sensing can be realized to further improve the capacity and adaptability of
UAV network.
One of the most important problems on the coexistence between DSC and cel-
lular network is the interference management. Underlay spectrum sharing between
a 3D DSC network and a 2D cellular network is proposed for interference man-
agement [8, 9]. Underlay spectrum sharing refers to the simultaneous transmissions
from DSC network and cellular network as long as the interference level at the cel-
lular user side remains acceptable. The cellular signal would degrade dramatically
if DSC network exceeds the predefined tolerable interference threshold. It has been
demonstrated in traditional cellular network that spectrum sharing technologies can
provide complementary benefits to dedicated spectrum. Effort was made toward
spectrum sharing technologies of DSCs [10].
The spectrum sharing of 2D cellular wireless networks has been well studied in
recent years [11, 12]. Different interference management strategies in 2D Poisson
cognitive radio networks have been developed, for example, inter-cell interference
coordination and intra-cell diversity [13]. It is also revealed from [13] that
the general stochastic results for PPP base station deployments are independent
to the shadow fading distribution as long as the shadowing is independent and
identically distributed. The primary exclusive regions (PERs) centered at every
primary receiver are designed in [14]. PER is a fixed region centered at each
primary receiver in which no secondary transmitter is allowed to transmit signals.
Applying PERs provides extra level of control to the network performance. It is
often used to further guarantee the primary performance, e.g., when the primary
users require a very small outage. The relationship between shadowing with path
loss exponent and PERs is further analyzed in [15]. As the idea of PER effectively
guarantees the primary service under limited spectrum resources, research efforts
have been made to implement PERs into the optimization of different spectrum
sharing schemes [16, 17]. Assuming all transmitters are 2D Poisson distributed
nodes with limited feedback of channel quality information from their local receiver,
two limited feedback-based underlay spectrum sharing schemes are proposed in
[16], with and without PERs, respectively.
Further, it is discovered that there exist temporal spectrum opportunities in
PER. The architecture of three regions including black, gray, and white regions is
proposed in [18]. Black region is surrounded by gray region and gray region is
surrounded by white region [18]. In black region, SUs are not allowed to transmit.
In gray region, SUs can exploit temporal spectrum opportunities. Namely, SUs can
use the spectrum that is not used by primary users (PUs). In white region, because
SUs are far away from PUs, SUs can transmit all the time with maximum power.
When primary receivers are densely deployed, the union of PERs of all primary
receivers forms a layer. When a UAV is located below the layer, it is not allowed to
transmit. Moreover, the architecture of three regions can be applied in UAV network.
As illustrated in Fig. 2, layer 1 and layer 2 divide the 3D space into three regions.
In region 1, since UAVs are close to the ground users, UAVs are not allowed to
transmit. In region 2, UAVs are farther to ground users. Thus UAVs can exploit
the temporal spectrum opportunities. In region 3, since UAVs are sufficiently high
38 Spectrum Sharing of Drone Networks 1283

Region 3
Layer 2

Region 2
Layer 1

Region 1

Ground

Ground user Connection path

Fig. 2 PER layers of UAV network

above the ground, UAVs can transmit all the time. Regions 1, 2, and 3 are similar to
the black, gray, and white regions, respectively in [19]. It is verified in [19] that the
capacity of secondary network with three regions is larger than that with PER. Hence
the architecture of three regions can also be applied in cognitive UAV network.
Stochastic geometry theory provides effective tools to study the average behavior
over spatial realizations of a large wireless network [20]. For 2D networks, there
are several empirical and theoretical results indicating that a Poisson point process
(PPP) is an appropriate point process to model base station deployments with
tractability. PPP is widely used to model different types of networks, for instance,
cognitive radio networks [16], cellular networks [21], and wireless sensor networks
[22]. It is shown in [23] that under certain appropriate conditions, the stochastic
results obtained by modeling wireless networks as spatial PPPs are still valid, even
if in reality the positioning of transmitting nodes does not appear Poisson.
One of the future trends of cellular networks is to have smaller cells to deserve the
growing number of communications [24]. The standard concept of planar cellular
networks is extended into 3D space in [25]. Different cell shapes in 3D cellular
networks achieving full coverage are investigated in [26]. It is demonstrated that
truncated octahedral cell results in the best strategy. Applying stochastic tools to 3D
ultradense cellular networks, different coverage probability and throughput scaling
behaviors in terms of the path loss components using a dual path loss model are
discussed in [21].
In this chapter, the study of DSC network underlay spectrum sharing with cellular
network is presented. Taking advantage of the tractability of Poisson point process,
explicit expressions for the DSC coverage probability and achievable throughput
are derived by stochastic geometry. To maximize the DSC network throughput
while satisfying the cellular network efficiency constraint, the optimal density of
DSC aerial base stations is discussed. The maximum throughput of the DSC user
increases almost linearly with the increase of the DSC outage constraint. Effects
of PERs and directional transmission on DSC networks are further discussed.
1284 C. Zhang et al.

Besides density control, power and beam control can also be applied in the spectrum
sharing between UAV network and ground network. The traditional cognitive
radio technology focuses on spectrum sensing, spectrum decision and spectrum
sharing, etc. However, since UAVs are robotic system, the mobility of UAVs can
be controlled to improve network capacity. The concept of mobility cognitive UAV
network is then proposed. Mobility pattern cognition aims to discover returning
UAVs to construct the delay-tolerant transmission scheme, which carries the delay-
insensitive traffic with Store-Carry-and-Forward (SCF) mode to avoid multi-hop
transmissions consuming radio resources. The delay-sensitive traffic is still carried
by multi-hop transmission manner.

Spectrum Sharing of UAV Network

Optimal DSC Density

As shown in Fig. 3, DSC networks distribute in 3D space. Assume that DSC aerial
base stations follow a 3D-PPP fXi 2 ˚d g with density d in an infinite 3D space
V, but the height is limited to L, that is, V D f.x; y; z/ W x; y 2 R; z 2 Œ0; Lg. The
channel between any pair of DSC aerial base station and the user here is assumed
to undergo path loss and the small-scale fading. The path loss is proportional to
x ˛ , where x is the distance between the transmitting aerial base station and the
typical user and ˛ is the mean path loss exponent. The power gain of small-scale
fading channel hi is exponentially distributed with unit mean, and the noise N0 is
additive white Gaussian noise following the distribution N0  N .0; N /. DSCs only
transmit while they are static [10,27], e.g., moving DSCs will perform transmission
on certain “stop points.”
Assume that all aerial base stations transmit at the same power level Pd . For
a typical link, the received signal power is hence Pd h0 D ˛ , where D is the
distance between a typical user and a typical aerial base station. The transmission is
successful if the received SINR at a receiver is larger than a certain threshold. Set
the SINR thresholds of DSC users to . A typical DSC user at the origin O will
receive interference from other transmitting DSC base stations while receiving the
desired signal. The received SINR of a typical user at the origin O is:

Fig. 3 DSC networks


38 Spectrum Sharing of Drone Networks 1285

Pd h0 D ˛
SI NRd D P ˛ : (1)
NC xi 2˚d nf0g Pd hi xi

The transmission rate of a network is defined as [11]:

Td D d P .SI NRd >  / log.1 C /: (2)

Assuming that the DSC network has an outage probability constraint "d  1,
the optimal DSC base station density can be obtained by solving the following
optimization problem:

maximize Td .G1/
d

subject to P .SI NRd >  / > 1  "d :

Next, the optimal DSC base station density


P for a single-tier DSC downlink network

in 3D space can be derived. Let I D h x
xi 2˚d nf0g i i . From (1) we have the
coverage probability as
!
h0 D ˛
P .SI NRd >  / D P N > (3)
Pd
CI
    
N
D EI P h0 >  D ˛
CI jI
Pd
   
˛ N
D EI exp  D ˛
exp.D I /
Pd
 
N
D exp  D ˛ LI . D ˛ /: (4)
Pd
P
where LI . D ˛ / is the Laplace transform of the interference I D xi 2˚d nf0g hi xi˛ ,
which can be further derived as:

LI . D ˛ / D EI Œexp. D ˛ I / (5)
2 3
Y
D E˚d ;hi 4 exp.D ˛ hi xi˛ /5 (6)
xi 2˚d nf0g
2 3
Y
D E˚d 4 Ehi Œexp.D ˛ hi xi˛ /5 (7)
xi 2˚d nf0g
2 3
Y 1
D E˚d 4 5: (8)
1 C  D ˛ xi˛
xi 2˚d nf0g
1286 C. Zhang et al.

where (7) follows because of the i.i.d. distribution of hi and its further indepen-
dence from the point process ˚d . Equation (8) follows because hi is exponential
distributed with unit mean. The probability generating functional of a set V is given
by [28]:
0 1
Y  Z 
E@ A
f .x/ D exp d Œ1  f .x/dx : (9)
xi 2˚ V

Applying (9) into (8), we have


 Z   
1
LI . D ˛ / D exp d 1 dx (10)
V 1 C  D ˛ x ˛
 Z L Z 2 Z 1   
1
D exp d 1 p rdrd d z
0 0 0 1 C D ˛ . r 2 C z2 /˛
(11)
Z Z p !
1
L
 D ˛ r. r 2 C z2 /˛
D exp 2d p drd z
0 0 1 C  D ˛ . r 2 C z2 /˛
D exp.d H .L;  ; D; ˛// (12)

where
Z Z 1
p
L
2 D ˛ r. r 2 C z2 /˛
H .L; ; D; ˛/ D p drd z: (13)
0 0 1 C  D ˛ . r 2 C z2 /˛

With (4), (5), (6), (7), (8), (9), (10), (11), and (12), we are now able to solve the
optimization problem .G1/.

maximize Td .G1/
d

subject to P .SI NRd >  / > 1  "d :

Since Td in (2) is unimodal in terms of d and P .SI NRd > / is a decreasing


function of d , Karush-Kuhn-Tucker (KKT) conditions [16] are applicable. The
Lagrange function of the above optimization problem is given by:

L .d / D Td C .P .SI NRd >  /  1 C "d /: (14)

where  is the Lagrange multiplier. Then, the KKT conditions are given as:

d L .d / d Td dP .SI NRd >  /


D C : (15)
d d d d d d
38 Spectrum Sharing of Drone Networks 1287

.P .SI NRd >  /  1 C "d / D 0: (16)


P .SI NRd >  /  1 C "d  0: (17)
  0 and d  0: (18)

Solving Eqs. (15) to (18) yields the optimal of DSC aerial base stations, as follows:

1"d
Œ ln. exp.D ˛ N
/
/C
Pd
d D : (19)
H .L;  ; D; ˛/

where Π: C denotes max.:; 0/. The maximum DSC network throughput is given by
  C
1"d
 ln exp.D ˛ N
/
Pd
Td D .1  "d / log.1 C /: (20)
H .L;  ; D; ˛/

It can be seen from (19) that the optimal primary density is obtained when the
success probability just meets the outage constraint. Since "d  1, we have
ln.1  "d /  "d . Therefore, for small "d , both optimal DSC density and potential
throughput are linear with the DSC network outage constraint "d .

Effect of Different Path Loss Models

In this section, effect of a more specific air-to-ground path loss model is discussed.
Different air to ground channel path loss models have been proposed and measured
to suit in different environments [29]. Simplified mean path loss model for the
spectrum sharing analysis in this chapter is given by:

l.x/ D x ˛ : (21)

Another air-to-ground path loss model l.x;  / for an air vehicle locating at distance
x and elevation angle  from the origin is given by [1]:

l.x;  / D PLOS x ˛1 C PNLOS kx ˛1 : (22)

in which ˛1 and k are the path loss exponent and parameter corresponding to
different type of links. PNLOS D 1  PLOS . PLOS is the line of sight probability. It
is further given by:

1
PLOS D : (23)
1 C a exp.bΠ a/

where a and b are parameters determined by environment.


1288 C. Zhang et al.

20
18
16
Potential throughput
14
12
10
8
Urban
6
Dense Urban
4 Simplified

2
0
2 4 6 8 10 12 14 16 18 20
λ

Fig. 4 Potential throughput of different path loss models L D 0:1, D D 0:05

Table 1 Radio-frequency Environment (a,b) ˛.˛1 / k


propagation parameters
Simplified 4
Urban (9.61,0.16) 2 20dB
Dense Urban (12.08,0.11) 2 23dB
Radio frequency 2,000 MHz [1]

Figure 4 plots the scaling behavior of network throughput in terms of the drone
density  by simulations. The numerical values of the parameters of the air-to-
ground channels for different environments are presented in Table 1. It can be
observed that the potential throughput of drone network monotonically increases
with node density for all different path loss models. As the research on modeling the
air-to-ground channel is ongoing, the insight of underlay spectrum sharing results
in this chapter is applicable for more precise path loss models in the future.

Spectrum Sharing Between DSC Network and Cellular Network

Optimal DSC Density

As shown in Fig. 5, cellular base stations are distributed on the ground as a 2D


PPP with density c . The distance between a typical cellular base station and its
associated user is d . Similar to that of DSC network, we assume that all cellular base
stations transmit with power Pc . The channel between any base station and the user
undergoes path loss and the small-scale fading. The power gain of the small-scale
38 Spectrum Sharing of Drone Networks 1289

Fig. 5 Spectrum sharing


between drone small cells
(DSCs) networks and cellular
networks

fading channel hj is exponentially distributed with the unit mean, and the noise
is N0  N .0; N /. Each cellular transmitter decides to transmit if the received
SINR is larger than a certain threshold c with a 1-bit feedback from the cellular
receiver [16]. The active cellular base stations follow a thinned 2D PPP fYj 2 ˚c g.
Considering a typical DSC user at the origin O, it will receive interference not
only from transmitting DSC base stations but also from transmitting cellular base
stations. The SINR of the typical DSC user is given by

Pd h0 D ˛
SI NRdc D P ˛
P : (24)
N C Pd i2˚d nf0g hi xi C Pc j 2˚c hj yj˛

The SINR expressions of a typical cellular user with and without the DSC network
are given by, respectively, as follows:

Pc h0 d ˛
SI NRc D P P : (25)
N C Pd i2˚d hi xi˛ C Pc j 2˚c nf0g hj yj˛
Pc h0 d ˛
SI NRc0 D P : (26)
N C Pc j 2˚c nf0g hj yj˛

The throughputs of DSC and cellular networks are:

Tdc D dc P .SI NRdc >  / log.1 C /: (27)


Tc D c P .SI NRc > c / log.1 C c /: (28)
Tc0 D c P .SI NRc0 > c / log.1 C c /: (29)
1290 C. Zhang et al.

where Tc and Tc0 are the cellular network throughput with and without DSC network,
T T 0
respectively. Define ı as the cellular efficiency loss ratio ı D cTc c and rth as
the cellular efficiency loss constraint [16]. The optimal DSC base station density
coexisting with a cellular network dc can be obtained by solving the following
optimization problem:

maximize Tdc .G2/


dc

subject to P .SI NRdc >  / > 1  "dc


ı  rth :

Next, we derive the optimal DSC base station density when coexisting with
cellular networks.
P Consider˛the DSC network coexists with a 2D cellular network. Denote Ic D
j 2˚c hj yj . The success probability of the typical DSC user is given by

!
h0 D ˛
P .SI NRdc > / D P N Pc
> (30)
Pd
CI C Pd
Ic
    
N Pc
D exp  D ˛ EI Œexp. D ˛ I /EIc exp D ˛ Ic :
Pd Pd
(31)

where I was given in (3). The second term in (31) has been evaluated by (4), (5),
(6), (7), (8), (9), (10), (11), and (12). The third term in (31) is the interference from
other active cellular base stations following a 2D-PPP and can be derived as [16]:
  
Pc
EIc exp  D ˛ Ic
Pd
  ˛2  !
2 Pc ˛ N
D exp c CD  exp d c : (32)
Pd Pc

2 2
where C D ˛ sin. 2
.
˛ /
The success probability of a typical DSC user is

P .SI NRdc >  / D A1 exp.A2 d / exp.A3 c / (33)

where
 
N
A1 D exp  D ˛ :
Pd
A2 D H .L;  ; D; ˛/:
38 Spectrum Sharing of Drone Networks 1291

  ˛2  
2 Pc ˛ N
A3 D CD  exp d c : (34)
Pd Pc

With the success probability expressions, we are able to solve the optimization
problem (G2):

maximize Tdc .G2/


dc

subject to P .SI NRdc >  / > 1  "dc


ı  rth :

Likewise, to derive the success probability of a typical cellular user, along with (28)
and (29), the cellular efficiency loss ratio is first calculated:

Tc  Tc0
ıD
Tc
   
Pd
D 1  exp H L; c ; d; ˛ dc : (35)
Pc

The Lagrange function of the optimization problem .G2/ is hence given by

L .dc / D Tdc C 1 .P .SI NRdc > //  2 .ı  rth /: (36)

The KKT conditions are listed as follows:

d L .dc / d Tdc d .P .SI NRdc >  / dı


D C 1  2 :
d dc d dc d dc d dc
1 .P .SI NRdc >  /  1 C "dc / D 0: (37)
2 .ı  rth / D 0: (38)
.1 C "dc /  P .SI NRdc >  /: (39)
ı  rth : (40)
1  0; 2  0 and dc  0: (41)

Solving the above equations yields:

If "dc < ;
h iC
 ln.1  "dc /  D ˛ PNd   A3 c
dc D (42)
A2
If "dc  ;
1292 C. Zhang et al.

ε η
ε η

ε η η

ε η

ε η
ε η

Fig. 6 Maximum throughput of a 3D-PPP with limited height coexisting with a cellular network,
˛ D 4,  D 1, c D 5, c D 1, Pd D 1, Pc D 1

 ln.1  rth /
dc D : (43)
H .L; c PPdc ; d; ˛/

A2
Pd
where D 1exp.A3 c /A1 .1rth / H .L;c Pc ;d;˛/ . We have the maximized potential
throughput:


Tdc D dc P .SI NRdc >  / log.1 C /: (44)

where P .SI NRdc > / can be evaluated by substituting dc into (33). For "dc <
, since 1  "dc  1, ln.1  "dc /  "dc . The approximated optimal DSC density
and potential throughput are both linear with the DSC network outage constraint
"dc . For "dc > , both optimal DSC density and potential throughput are constants
which are independent of "dc .

Numerical Examples

In this section, some numerical examples are given to validate the theoretical results
and discuss the effects of several parameters on the coverage and rate of DSCs and
cellular networks.
38 Spectrum Sharing of Drone Networks 1293

In Fig. 6, the simulation results of the maximal throughput of a 3D DSC network


with the coexistence of a cellular network are compared with the analytical results
derived from (44). The simulation of optimal node density is done by two steps.
The optimal DSC base station density and the maximal throughput are plotted as
functions depending on the DSC outage constraint. The values of d and Td are zero
at the small " region since the outage probability cannot satisfy its outage constraint.
In the region where the outage probability is able to satisfy its outage constraint, both
dc and Tdc increase almost linearly with the increase of "dc , until they flatten out
and coincide with (43). The flattening is because cellular network constraints are
restricting the DSC node density, even more DSC outage could be tolerated by the
DSC network. Moreover, it is noticed that is independent of the limited height of
the 3D network, as it represents when the cellular network interference constraints
become dominant. The scaling behaviors of DSC outage constraints provide us the
following insight: as we relax the restrictions for DSC network, the dependence of
optimal DSC base station deployment moves from the constraints of DSC network
to constraints of cellular network.

Spectrum Sharing of UAV Networks with Directional Antennas

UAVs are above the ground with a certain altitude, which creates opportunities for
UAV network to share the spectrum of ground network. In this section, we present
the study of UAV network spectrum sharing exploring the benefits of beam and
power control in 3D and 2D deployment scenarios, respectively.

3D Deployment

Density and Power Control


When UAVs are deployed in 3D space and share the spectrum of ground networks,
the density and transmit power of UAVs need to be controlled to mitigate the
interference to ground network. In section “Spectrum Sharing Between DSC
Network and Cellular Network”, UAV network shares spectrum with a ground
cellular network, where the density of UAV network is optimized to maximize the
network capacity with the constraints of the outage probability of UAV network and
the capacity loss of cellular network. Besides density control, power control can
also be applied in the spectrum sharing between UAV network and ground network.
UAVs with low altitude should transmit with small transmit power to mitigate the
interference to ground users. Meanwhile, UAVs with high altitude can transmit with
a relatively large transmit power because they are distant from the ground users.
With power control, the deployment density and capacity of UAVs can be improved.

Beam and Power Control


In previous sections, UAVs are implemented with omnidirectional antennas. When
UAVs are implemented with directional antenna, more efficient spectrum sharing
1294 C. Zhang et al.

C u

B
D
A
h

B
A C

Gr
Ground

Ground user Connection path


Conne

Fig. 7 Directional transmission scenario of UAV network

schemes can be explored. As illustrated in Fig. 7, UAVs are distributed in 3D space


and transmit directionally. The 3D space is modeled by a cuboid with height u
and the height of the bottom of this cuboid from the ground is h. Notice that the
UAVs with low altitude have small probability to transmit downward because their
destinations have large probability to be located above them. However, when a UAV
transmits downward, it may cause interference to ground nodes, such as UAV C in
Fig. 7. If the signal radiation of a UAV is modeled as a spherical cone, the projection
of the signal on ground is an ellipse. If the beam of signal is sufficiently narrow, the
probability that the signal of UAV causes interference to ground nodes is small.
Besides, power control can also be applied to control the interference from UAVs to
ground users.
However, the control of beam and power is not necessary for each direction.
There is a beam control angle for each UAV. When transmission direction is not in
the beam control angle, the beam and power control are not necessary. In Fig. 8,
the beam control angles of UAVs A, B, and C are ˛1 , ˛2 , and ˛3 , respectively,
where ˛3 D 0. With the increase of the altitude of UAV, the beam control angle
is decreasing. The beam control angle can be derived as follows. Take UAV A
as an example; the distance between UAV A and a ground user A0 is X1 . When
X1 is increasing, 1 is decreasing and the probability of non-line-of-sight (NLoS)
transmission between UAV A and ground node A0 is increasing. With the increase
of X1 and NLOS probability, the interference from UAV A to ground node A0 is
decreasing. When the distance X1 is increased to X1 such that the expectation of
the interference from UAV A to ground node A0 is smaller than a threshold, the
beam control angle ˛1 is determined. When a UAV is high above the ground such
that the interference to the ground node is small, the beam control angle for this
UAV does not exist.
38 Spectrum Sharing of Drone Networks 1295

α3
B
α2
A
h3
X 2* α1 X 2*
h2
h1 X 3*
X 1* X 1*

Ground γ 2 B⬘ γ 1 A⬘

Ground user Connection path


p

Fig. 8 Directional transmission in UAV network with different angles

In Fig. 8, the beam control angles of three UAVs are illustrated. Notice that if
we do not consider the NLOS transmissions, the value of X1 equals X2 . However,
because the UAV with high altitude has a larger angle of elevation compared with
the UAV with low altitude, the length of X1 is smaller than X2 . UAV can freely
transmit in the non-beam control angle. However, if a UAV transmits in the beam
control angle, more sophisticated transmission schemes need to be implemented.
The following schemes may be applicable:

1. Spectrum sensing If a UAV transmits in the beam control angle, it performs


spectrum sensing and transmits when ground users are absent.
2. Narrow beam If a UAV transmits in the beam control angle with narrow beam,
the interference probability to ground users will decrease. However, this requires
sophisticated beamforming algorithm.
3. Power control When a UAV transmits in the beam control angle, if the power
of UAV can be controlled to mitigate the interference to ground users, the
transmission opportunities can also be exploited.

With these schemes, UAVs can share the spectrum of ground users in every
direction and altitude, which will increase the number of wireless channel and
improve the capacity of UAV network.

2D Deployment

In previous section, we explore the spectrum sharing between 3D UAV network and
2D ground network. However, when UAVs are used to monitor or search specific
targets on ground, they may be distributed in an aerial 2D plane, which is illustrated
1296 C. Zhang et al.

Interference path

Gro
Ground

Ground user Connection path


Connectio

Fig. 9 2D UAV networks with 2D ground networks

in Fig. 9. In this section, we discuss the spectrum sharing between 2D UAV network
and 2D ground network.

Altitude, Density, and Power Control


When UAVs are deployed in an aerial 2D plane, the altitude, density, and transmit
power can be coordinated to mitigate the interference to ground users. In [30],
assuming UAV network has air-to-ground transmissions, we derived the optimal
altitude of UAVs with fixed transmit power and density. When the altitude of UAVs
is fixed, the density and transmit power of UAVs can be optimized to maximize the
capacity of UAV network with the constraint of capacity loss of ground network.
Furthermore, the altitude, density, and power of UAVs can be jointly optimized.

Beam Control
Similar with section “Beam and Power Control”, when UAVs are implemented
with directional antennas, efficient spectrum sharing schemes can be designed.
As illustrated in Fig. 9, when UAV E transmits directionally, it will not cause
interference to ground node B. Assuming that the beam width of UAV is ˇ, when
ˇ is decreasing, the interference from UAV network to ground users is correspond-
ingly decreasing. Thus beam control can realize interference coordination between
UAV network and ground network. Intuitively, the directional transmission in UAV
network is equivalent to increasing the altitude of UAVs. Thus with beam control of
UAV network, the spectrum sharing between UAV network and ground network is
feasible.
38 Spectrum Sharing of Drone Networks 1297

Numerical Results

The spectrum sharing between an aerial 2D UAV network and a ground ad hoc
network is investigated in [30]. The scenario of Fig. 9 is simulated. Namely, UAV
network shares spectrum with a ground ad hoc network and UAVs are implemented
with directional antennas. Assume that the signal to interference and noise ratio
(SINR) threshold for signal reception is 1. In a 1000  1000 m area, 100 ground
users are uniformly deployed and 25 UAVs are uniformly deployed above the
ground users. The path loss exponent of ground-to-ground channel is 3, and the
path loss exponent of air-to-ground channel is 2. The air-to-ground channel model
is borrowed from [10], where the probability of line-of-sight (LOS) is an increasing
function of the angle of elevation.1 The transmit power of UAV is 5 W and
the transmit power of ground node is 1 W. With these parameters configuration,
the coverage probability of ground ad hoc network is provided in Fig. 10. The
interference from UAV network to ground network has an impact on the coverage
probability of ground network. With the increase of the altitude of UAVs, the length
of propagation path from UAV to ground user is increasing. Meanwhile, the angle of
elevation from ground user to UAV is increasing. The increase of propagation path
will decrease the interference from UAV to ground user. However, the increase of the
angle of elevation will enlarge the probability of LOS and increase the interference
from UAV to ground user. In Fig. 10, for UAV networks with omnidirectional
antennas, when the altitude of UAVs is smaller than h , the angle of elevation
is dominated. Thus with the increase of the altitude, the interference from UAV
network to ground user is increasing and the coverage probability of ground network
is decreasing. However, when the altitude of UAVs is larger than h , the length of the
propagation path is dominated. The interference from UAV network to ground users
is decreasing with the increase of the altitude in this situation. Thus the coverage
probability is increasing with the increase of altitude.
When the beam width of UAV, namely, ˇ is decreasing, the interference from
UAVs to ground network is reduced and the coverage probability of ground network
is improved. As illustrated in Fig. 10, when ˇ D 2=3, the improvement of coverage
probability is not significant. However, when ˇ D =6, the coverage probability
will rapidly increase to the maximum value with the increase of altitude. Thus
directional transmission in UAV network creates spectrum sharing opportunities
between UAVs and ground network.

Mobility Pattern Cognition of UAV Networks

UAVs have specific mobility pattern and the mobility of UAVs can be exploited
to improve network capacity. With the mobility pattern information of UAVs, the
delay tolerant transmissions can be constructed and multiple transmission modes

1
Reader could refer to (8) and (9) in [10] for the details of air-to-ground channel model.
1298 C. Zhang et al.

0.8
Coverage probability
0.6

0.4

Directional antenna with β = π/6


0.2
Directional antenna with β = 2*π/3
* Omnidirectional antenna
h
0
0 500 1000 1500 2000 2500 3000
Altitude of UAVs / Unit: m

Fig. 10 The coverage probability of ground ad hoc network

180

Returning path

Critical range

180
18
L
Control station

Fig. 11 UAV network with mobility pattern cognition. UAV sensors are deployed in 3D space

are implemented to carry various types of traffic. Figure 11 illustrates the scenario
that UAVs act as aerial sensors and transmit sensing data to control station.
In Fig. 11, when UAV A returns, it will encounter UAV B. The returning UAV
A can store and carry the data of UAV B to control station. Thus in addition to
multi-hop transmission mode, the delay-tolerant transmission mode is constructed
to improve the capacity of UAV networks. UAV network with multi-hop and SCF
transmission modes can adapt the demands of diverse traffic. The delay-insensitive
and delay-sensitive traffic can be forwarded through SCF transmission mode and
multi-hop transmission mode, respectively. In the following sections, we discuss
the SCF transmission modes.
38 Spectrum Sharing of Drone Networks 1299

Mobility Pattern Cognition

The process of the discovery of a returning UAV and the establishment of com-
munication link between the returning UAV and a UAV along the returning path
is illustrated in Fig. 12. Assume that UAV A will return to the control station and
UAV B is located in the returning path of UAV A. Each UAV has a state bit. When
a UAV is returning, the state bit is set as 0. Otherwise, the state bit is set as 1.
The neighbor discovery function of UAV B will detect nearby UAVs. Once UAV B
detects a UAV, it sends synchronize (SYN) signal. Meanwhile, the returning UAV A
is also detecting the UAVs along its returning path. Once UAV A discovers a UAV
and this UAV is sending SYN signal, UAV A feedbacks the state bit indicating that
it is returning and synchronize/acknowledge (SYN/ACK) signal containing channel
information for communication link establishment. Then UAV B sends ACK signal
and establishes communication connection. When the data transmission is finished,
the communication session is released and UAV A continues returning.
With mobility pattern information, the SCF transmission modes for 3D and 2D
UAV network are introduced in the following sections.

3D UAV Deployment

As illustrated in Fig. 11, when UAV A returns, it can store and carry the data of UAV
B to the control station. Assume that UAVs return in straight line path. In Fig. 11,
when the UAVs in the shaded region return, they will assist UAV B to deliver data
to control station. With the increase of the distance between the control station and
UAV B, the volume of shaded region is decreasing and the number of returning
UAVs assisting the data transmission of UAV B is also decreasing.
In [6], a critical range is discovered, which is illustrated in Fig. 11. Within critical
range, the number of returning UAVs is sufficiently large, such that the capacity of
UAVs within critical range contributed by SCF mode is higher than that outside
of critical range. Thus the mobility of the returning UAVs outside of critical range
needs to be controlled to improve the capacity. Mobility control schemes are as
follows:

1. Returning path control: The returning path outside of critical range should be
fold line such as zigzag curve to increase the length of returning path, such that
more data can be delivered with SCF mode.
2. Pause time control: The returning UAVs outside of critical range pause for a time
to gather the data of neighbor UAVs, such that the capacity of SCF mode can be
improved.

However, all the mobility control scheme aiming to improve the capacity of SCF
mode will increase the energy consumption of UAVs because the returning time
will be longer than that without mobility control. Hence there is a trade-off between
energy consumption and capacity improvement.
1300 C. Zhang et al.

UAV B detects
UAV A returns
nearby UAVs

Set state bit as 0

UAV A detects UAV


on the returning path

no no
Discover a Discover a
UAV? UAV?

yes
yes Obstacle
avoidance
Is there data no
to transmit?

yes
UAV B sends SYN signal
no
Obstacle
avoidance
Is UAV
returning?

no yes
Is their SYN
signal?

yes Feedbacks state bit, SYN/ACK signal


containing channel information, etc.

Sends ACK and establishes


communication connection
Finish data transmission
and send confirmation

Continue Release
returning connection

Fig. 12 The establishment of SCF transmission link


38 Spectrum Sharing of Drone Networks 1301

A
Critical range
8
80
L

Returning path h

180

Control station

Fig. 13 UAV network with mobility pattern cognition. UAV sensors are deployed in an aerial 2D
plane

2D UAV Deployment

When UAVs are deployed in an aerial 2D plane, the returning path of UAV is
redesigned. As illustrated in Fig. 13, the UAVs are deployed in a circle above the
ground. When UAV A returns, it will firstly flies to the center of the circle. Then
UAV A flies downward to control station. With this returning path, the returning
UAV A can deliver the data of UAV B to control station with SCF mode. In 2D
deployment of UAVs, the critical range also exists and the mobility of the returning
UAVs outside of critical range should be controlled to improve the capacity of SCF
mode.
The previous discussions consider the scenario that UAVs act as aerial sensors. In
the scenario that UAVs act as aerial base stations, the mobility pattern information
can also be applied to construct SCF transmission mode. As illustrated in Fig. 14,
UAVs are deployed in the sky to provide communication connections for ground
nodes. For example, when ground node C transmits data to ground node D, it will
firstly transmit to UAV A, then UAV A will transmit the data with multi-hop manner
to UAV B. Finally, the data is forwarded from UAV B to ground node D. However,
with SCF mode, if UAV A is a returning UAV passing ground node C, ground
node C will forward data to UAV A. Then the data is brought to control station
via UAV A. When UAV B is departing from control station and passing ground
node D, UAV B will bring the data of ground node C and forward the data to
ground node D. In this way, the multi-hop transmission can be avoided, such that the
wireless channels will not be consumed by repeated data transmission in multi-hop
transmission.
1302 C. Zhang et al.

B
A

Returning path Departure path

C D

Control station

Fig. 14 UAV network with mobility pattern cognition. UAVs act as base stations

Summary and Future Work

This chapter presents the spectrum sharing between the drone network and tradi-
tional ground cellular network under different scenarios. It has provided explicit
solutions of the optimal density of DSCs modeled by a 3D Poisson point process
with limited height. It was demonstrated that larger deployment height limit results
larger optimal DSC density. For DSC network underlay spectrum sharing with
cellular network, as we relax the restrictions for DSC network, the dependence of
optimal DSC base station deployment moves from the constraints of DSC network
to constraints of cellular network. Effects of directional transmission on DSC
networks are further discussed. Exploiting cognition capability on mobility, UAV
network can provide high quality of information services in the highly dynamic
environment with limited resources.
The spectrum sharing of drone networks is an emerging field with numerous
interesting applications. We conclude by discussing some fruitful avenues for future
research:

• UAV Channel Modeling As mentioned in the chapter, the characteristics of UAV


wireless channels are very different from those of terrestrial wireless channels.
More empirical measurements can be expected for developing more accurate
channel models.
• Multi-UAV Scheduling Based on the analytical framework for drone spectrum
sharing presented, it would be interesting to further design the scheduling scheme
for multiple UAVs. The cooperation between UAVs can further improve the
network performance.
• UAV to Infrastructure/Vehicle Spectrum Sharing We have focused on the spec-
trum sharing between UAVs and conventional cellular networks. There are
38 Spectrum Sharing of Drone Networks 1303

also ongoing thrusts on the coexisting between UAV and infrastructure/vehicle


networks. Since UAVs can provide extra coverage, how to integrate the aerial
base stations and relays with the existing ground networks deserves further
study.

References
1. Al-Hourani A, Kandeepan S, Jamalipour A (2014) Modeling air-to-ground path loss for
low altitude platforms in urban environments. In: IEEE global communication conference
(GLOBECOM), Austin, pp 2898–2904. https://doi.org/10.1109/GLOCOM.2014.7037248
2. Al-Hourani A, Kandeepan S, Lardner S (2014) Optimal LAP altitude for maximum coverage.
IEEE Wirel Commun Lett 3(6):569–572. https://doi.org/10.1109/LWC.2014.2342736
3. Mozaffari M, Saad W, Bennis M, Debbah M (2015) Drone small cells in the clouds:
design, deployment and performance analysis. In: IEEE global communication conference
(GLOBECOM), San Diego, pp 1–6. https://doi.org/10.1109/GLOCOM.2015.7417609
4. Lin X et al (2018) The sky is not the limit: LTE for unmanned aerial vehicles. IEEE Commun
Mag 56(4):204–210. https://doi.org/10.1109/MCOM.2018.1700643
5. Federal Aviation Administration (FAA) (2017) UAS traffic management research transition
tea plan. Technical report. https://www.faa.gov/uas/research/utm/media/FAA_NASA_UAS_
Traffic_Management_Research_Plan.pdf
6. Wei Z, Wu H, Huang S, Feng Z (2017) Scaling laws of unmanned aerial vehicle network with
mobility pattern information. IEEE Commun Lett 21(6):1389–1392. https://doi.org/10.1109/
LCOMM.2017.2671861
7. Cover T, Thomas J (2012) Elements of information theory. Wiley, New York
8. Zhang C, Zhang W (2017) Spectrum sharing for drone networks. IEEE J Sel Areas Commun
35(1):136–144. https://doi.org/10.1109/JSAC.2016.2633040
9. Zhang C, Zhang W (2016) Spectrum sharing in drone small cells. In: Proceedings of
IEEE global communication conference (GLOBECOM), Washington. https://doi.org/10.1109/
GLOCOM.2016.7842290
10. Mozaffari M, Saad W, Bennis M, Debbah M (2016) Unmanned aerial vehicle with underlaid
device-to-device communications: performance and tradeoffs. IEEE Trans Wirel Commun
15(6):3949–3963. https://doi.org/10.1109/TWC.2016.2531652
11. Andrews JG, Baccelli F, Ganti RK (2011) A tractable approach to coverage and rate in cellular
networks. IEEE Trans Commun 59(11):3122–3134. https://doi.org/10.1109/TCOMM.2011.
100411.100541
12. Bai T, Heath RW Jr (2015) Coverage and rate analysis for millimeter wave cellular networks.
IEEE Trans Wirel Commun 14(2):1100–1114. https://doi.org/10.1109/TWC.2014.2364267
13. Zhang X, Haenggi M (2014) A stochastic geometry analysis of inter-cell interference
coordination and intra-cell diversity. IEEE Trans Wirel Commun 13(12):6655–6669. https://
doi.org/10.1109/TWC.2014.2339273
14. Vu M, Devroye N, Tarokh V (2009) On the primary exclusive region of cognitive networks.
IEEE Trans Wirel Commun 8:3380–3385. https://doi.org/10.1109/TWC.2009.080454
15. Bagayoko A, Tortelier P, Fijalkow I (2010) Impact of shadowing on the primary exclusive
region in cognitive networks. In: European wireless conference (EW), Lucca, pp 105–110.
https://doi.org/10.1109/EW.2010.5483402
16. Wang Z, Zhang W (2014) Opportunistic spectrum sharing with limited feedback in Poisson
cognitive radio networks. IEEE Trans Wirel Commun 13(12):7098–7109. https://doi.org/10.
1109/TWC.2014.2363676
17. Dahama R, Sowerby KW, Rowe GB (2013) Estimating protection distances in spectrum
sharing systems. IEEE Trans Signal Process 61(17):4284–4295. https://doi.org/10.1109/TSP.
2013.2269901
1304 C. Zhang et al.

18. Wei Z, Feng Z, Zhang Q, Li W (2015) Three regions for space-time spectrum sensing and
access in cognitive radio networks. IEEE Trans Veh Technol 64(6):2448–2462. https://doi.org/
10.1109/GLOCOM.2012.6503290
19. Wei Z, Feng Z, Zhang Q, Li W (2015) Three regions for space-time spectrum sensing and
access in cognitive radio networks. IEEE Trans Veh Technol 64(6):2448–2462. https://doi.org/
10.1109/TVT.2014.2342612
20. Guo A, Haenggi M (2015) Asymptotic deployment gain: a simple approach to characterize the
SINR distribution in general cellular networks. IEEE Trans Commun 63(3):962–976. https://
doi.org/10.1109/TCOMM.2014.2387170
21. Gupta AK, Zhang X, Andrews JG (2015) SINR and throughput scaling in ultradense urban
cellular networks. IEEE Wirel Commun Lett 4(6):605–608. https://doi.org/10.1109/LWC.
2015.2472404
22. Lee S, Zhang R, Huang K (2013) Opportunistic wireless energy harvesting in cognitive radio
networks. IEEE Trans Wirel Commun 12(9):4788–4799. https://doi.org/10.1109/TWC.2013.
072613.130323
23. Keeler HP, Ross N, Xia A (2015) When do wireless network signals appear Poisson. Available
via DIALOG. https://arxiv.org/abs/1411.3757
24. Andrews JG, Buzzi S, Choi W, Hanly SV, Lozano A, Soong ACK, Zhang JC (2014) What
will 5G be? IEEE J Sel Areas Commun 32(6):1065–1082. https://doi.org/10.1109/JSAC.2014.
2328098
25. Carle J, Myoupo JF, Seme D (2001) A basis for 3-D cellular networks. In: Proceedings of the
15th international conference on information networking, Beppu City, pp 631–636. https://doi.
org/10.1109/ICOIN.2001.905525
26. Alam SM, Haas ZJ (2006) Coverage and connectivity in three-dimensional networks. In:
Proceedings of the 12th annual international conference on mobile computing and networking,
Los Angeles, pp 346–357. https://doi.org/10.1145/1161089.1161128
27. Zeng Y, Zhang R, Lim TJ (2016) Wireless communications with unmanned aerial vehi-
cles: opportunities and challenges. IEEE Commun Mag 54(5):36–42. https://doi.org/10.1109/
MCOM.2016.7470933
28. Streit RL (2010) Probability generating functional. In: Poisson point processes: imaging,
tracking, and sensing. Springer Science & Business Media, Berlin, p 27
29. Zhou L, Yang Z, Zhou S, Zhang W (2018) Coverage probability analysis of UAV cellular
networks in Urban environments. In: Proceeding IEEE international conference on communi-
cation (ICC), Kansas City
30. Guo Z, Wei Z, Feng Z, Fan N (2017) Coverage probability of multiple UAVs supported ground
network. Electron Lett 53(13):885–887. https://doi.org/10.1049/el.2017.0800

You might also like