You are on page 1of 5

Coverage Ratio Optimization for HAP

Communications
Dingjie Xu∗ , Xinghui Yi∗ , Zhihang Chen† , Cheng Li∗ , Chaoxian Zhang† , Bin Xia∗
∗ Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai, China
† School of Information Science and Engineering, Tan Kah Kee College, Xiamen, China

Email: {xudingjie1993, Turtle803, lichengg, bxia}@sjtu.edu.cn,


cme13022@stu.xujc.com, zhangcx@xujc.com

Abstract—In this paper, we focus on the efficient deploy- In [7], the circle packing theory is adopted to determine
ment of high altitude platforms (HAPs). As the accuracy of the optimal deployment of UAVs such that the total
the channel model highly affects the deployment scheme, coverage can be maximized. But this work is based on
we first propose a geometry-based HAP channel model,
which considers the statistical and geometry properties of the assumption that the decoding threshold is lower than
the terrestrial environments comprehensively. Based on the 0 dB. Otherwise, there will be coverage holes on the
proposed channel model, we derive the Line-of-Sight (LoS) ground, and the circle packing theory is not applicable
transmission probability of air-to-ground communication, anymore.
and characterize the path loss analytically. We further pro- To determine an efficient multi-HAP deployment, it is
pose an algorithm to maximize the deployment efficiency
in terms of the coverage ratio, i.e., the ratio of the radius necessary to model the characteristics of HAP commu-
of the HAP to the inter-HAP distance. The accuracy of our nication channel accurately. In [8], [9], the HAP channel
theoretical analysis is validated by simulations. is modeled based on data sampling and data fitting.
Index Terms—HAP, geometry-based model, LoS proba- Although it provides a high accuracy, the heavy work
bility, path loss, optimal deployment. of data measurement for different scenarios prevents its
wide application. To avoid the complexity and high-
I. I NTRODUCTION
cost, the International Telecommunication Union (ITU)
High Altitude Platform (HAP) technique has attract- proposes a geometrical method to model the air-to-
ed increasing attention due to its advantages, such as ground communication [10]. However, the basic assump-
free-space-like path loss, short transmission delay, low tion that buildings are evenly spaced does not accord
manufacturing costs [1]. It is regarded as a potential with the real-world environment. Moreover, the impacts
supplement to current communication systems [2]. of different environmental parameters on QoS need to
Despite the numerous advantages of deploying HAPs be revealed, since no explicit expressions have been
as flying base stations, several technical challenges, obtained.
such as energy limitation, interference management, To characterize the air-to-ground communication
demand prompt solutions [3]–[5]. Among these chal- channel and determine an optimal deployment scheme,
lenges, multi-HAP deployment is of great significance, in this paper,
since an efficient deployment can greatly reduce the
• We propose a statistical geometry-based channel
energy consumption and expand the coverage. Though
model, which is characterized by statistical proper-
few works are conducted to analyze the multi-HAP
ties, including the mean and variance of the building
deployment problem, some works related to the efficient
density and the building height. Contrary to the
deployment of unmanned aerial vehicles (UAVs) have
assumption that buildings are evenly spaced [6],
been reported [6], [7]. Specifically, in [6], the authors
[10], in this paper, buildings are assumed to be
derive the implicit expression of the path loss for a
uniformly distributed at a certain density.
single UAV and obtain the optimal altitude for the
• Based on the channel model, we obtain the Line-
maximum coverage. However, HAPs are working at an
of-Sight (LoS) transmission probability of air-to-
altitude far above UAVs, which indicates that adjusting
ground communication, and derive the path loss for
the altitude has limited impacts on coverage extension.
deployment scheme design.
This work was supported in part by the National Key Research and • We propose an algorithm to determine the optimal
Development Program of China under Grant 2016YFE0121100, the multi-HAP deployment to maximize the coverage
National Nature Science Foundation of China under Grants 61531009
and 61521062, the State Major Science and Technology Special
ratio, i.e., the ratio of the radius of the HAP to the
Projects under Grant 2016ZX03001015, the Chinese National 863 inter-HAP distance, such that the coverage efficien-
Project under Grant 2015AA01A710. cy can be optimized. Simulations are conducted to
978-1-5386-3531-5/17/$31.00 
c 2017 IEEE verify the superiority of the proposed scheme.
,W ,W
LJ
Z ,Ć
 
 ,W
ŚсϮϬŬŵ

 

Ě
h
>Ž^

h K ǁ Z džŚ dž

Fig. 1. Illustration of Air-to-Ground communication. Ͳů

II. HAP C HANNEL M ODELING


Fig. 2. Regions where the LoS ray might be obstructed.
The air-to-ground communication is illustrated in
Fig.1. Considering the complex terrestrial environment,
buildings or some other obstructions may block the line- When a building with coordinate of (xb , yb ) lies in
of-sight (LoS) between the HAP and the terrestrial user the rectangular region, the UE will be fully covered
equipment (UE), which causes a significant drop in the (considering the case that the user uses a mobile phone
signal quality. Therefore, channel characteristics such as in a building). In this case, no LoS exists.
path loss should be modeled for LoS and NLoS cases (2) The triangular region
respectively, which necessitates reliable methods for the When a building with coordinate of (xb , yb ) lies in the
LoS probability prediction. triangular region, OH’ will intersect with BC. Since the
Consider the HAP channel with the aerial view shown equation of the line OH’ is y = xyhh x, the coordinate of
in Fig.2. We construct a cartesian coordinate system, the intersection can be derived as ( xyhh (yb − l), yb − l).
where the UE locates at the original point O, and the When the building height h satisfies the condition
projection of HAP H on the ground plane, i.e., the h2 zh2
x − y plane, is denoted as H  , with the coordinate ≥ , (1)
x2h
(yb − l)2 + (yb − l)2 x2h + yh2
(xh , yh ) = ( zhtanθ
cosϕ zh sinϕ
, tanθ ), where zh is the height of yh2

the HAP. Buildings are assumed to be uniformly located


i.e., h ≥ zh (yyhb −l) , the LoS will be obstructed.
in a large enough region S surrounding the UE. Some
(3) The rhombic region
basic environmental parameters are defined as below.
When a building with coordinate of (xb , yb ) lies in
• ϕ: The azimuth angle of the user antenna, i.e., the
−−→ the rhombic region, OH’ may intersect with either AB
angle between the vector OH  and x-axis on the or BC. If the intersection is on AB, the coordinate
x-y plane. of the intersection should be (xb − w, xyhh (xb − w))
• θ: The elevation angle of the user antenna, i.e., the
−−→ −−→ and similar to (1), the threshold of the building height
angle between the vector OH  and OH. for the obstruction can be derived as zh (xxbh−w) . If the
• ρ: The building density, i.e., the average number of
intersection is on BC, the threshold should be zh (yyhb −l) .
buildings per square kilometer.
• 2w and 2l: The building dimension, i.e., the width B. LoS Probability Derivation
and the length of the buildings respectively. Based on the analysis in Section II-A, we can obtain
• h: The building height, which is assumed to follow the probability that a single building located around UE
the exponential distribution with parameters λ, i.e., obstructs the signals from or to the HAP as (2). The
fh (x) = λe−λx , x > 0. number of buildings surrounding UE can be calculated
A. Obstruction Cases as M = ρS, where ρ is the building density. Then the
LoS probability can be calculated as
The LoS probability refers to the probability that the
radiation from the transmitter to the receiver along the 0 Ω ρS (a) −ρΩ
PLoS = (1 − PNLoS )M = (1 − ) = e , (3)
geometrically shortest route is not blocked by any ob- S
struction. As shown in Fig.2, we can see that obstructions where step (a) is derived by using substitution of equiv-
can only happen in the region surrounded by the dotted alent infinite small, considering S is sufficiently large.
line. The region consists of three parts, i.e., a triangle, The building height follows the exponential distribu-
a rectangle and a rhombus, and we will analyze these tion with parameter λ, so the cumulative density function
three parts one by one. is
(1) The rectangular region Fh (x) = 1 − e−λx , (4)
and therefore, we have between U and A is the radius of the coverage denoted
as d. It is obvious that the distance between U and
P{h ≥ x} = 1 − Fh (x) = e−λx . (5)
B  must be greater than or equal to R − d, with the
Combining (2) with (5), we can derive Ω as (6). Since equality holds only when the user lies on the line A B  .
xh  w and substitute xh and yh with zhtanθ
cosϕ
and zhtanθ
sinϕ Therefore, the corresponding LoS probabilities can be
respecively, an approximation of Ω can be written as derived as

2 2βd
2lcosϕ 2wsinϕ A
PLoS = e− h ,
Ω̃in = + + 4wl √ (10)
λtanθ
√ λtanθ −
2 2β(R−d)

(b) 2 2wsin(ϕ + )
π (7) B
PLoS =e h ,
≈ 4
+ 4w2 , where ϕ is simplified as ϕ = π
λtanθ 4, and the received signal
where we assume w = l in step (b) without loss of power should be
generality. Pt
Pr,A = ,
In some related researches, only outdoor scenarios A
LFSL ψA
are considered, i.e., the rectangular region mentioned in (11)
Pt
Section II.A can be neglected. In these cases, we have Pr,B = B
,
LFSL ψB

2 2wsin(ϕ + π4 ) respectively, where ψ represents the shadow fading,
Ω̃out = , (8)
λtanθ whose typical values are given in [9], and LFSL is the
and the LoS probability becomes free space loss, which can be calculated as
√ √ √

2 2ρwsin(ϕ+ π )
4 −
2 2βsin(ϕ+ π )
4  4πfc h2 + d2 2
PLoS = e λtanθ =e tanθ , (9) LFSL =
A
,
ρw c
where β = λis a factor characterizing the surrounding  4πfc h2 + (R − d)2 2
environment. A larger β usually indicates an environ- FSL =
LB ,
c
ment with higher buildings and a larger building density.
III. O PTIMAL M ULTI -HAP D EPLOYMENT Algorithm 1 Optimal Deployment
Suppose that there are N HAPs deployed above a 1: Initialization: h, Pt , fc , β, ψLoS , ψNLoS , N , γ th ;
certain area providing wireless coverage for ground 2: for R = R; R ≤ R; R = R + ΔR do
users. On one hand, if HAPs are densely distributed, 3: for d = R2 ; d ≥ 0; d = d − Δd do
the interference will be significant. On the other hand, if 4: if (13) is satisfied then
HAPs are scattered, users located in the coverage hole 5: Result = R d
;
will not be able to get access to the network. Therefore, 6: Break;
how to deploy the HAPs efficiently becomes a vital issue. 7: end if
We assume that the distance between two adjacent 8: end for
HAPs is R, and all HAPs transmit signal to terrestrial 9: end for
users with a fixed transmission power Pt . The coverage 10: return Result;
of a HAP is defined as the area where the user’s SINR
is greater than the decoding threshold γ th . As shown where fc is the carrier frequency, and c is the speed of
in Fig.1, we take two adjacent HAPs A and B as an light in a vacuum. Based on the analysis above, we can
example, with their projections on the ground denoted derive the SINR of the received signal as
as A and B  respectively. We focus on the user located Pr,A
at the edge of the coverage of HAP A, i.e., the distance SINR = , (12)
σn2 + Pr,B

  yh    xyh (xb +w)+l  


1 xh xh (xb −w)+l zh (xb − w) zh (yb − l)
0 h
PNLoS = P h≥ dyb + P h≥ dyb dxb
S w
yh
xh (xb −w)−l
xh yh
xh (xb −w)+l
yh
 w  xyh (xb +w)+l    w  l
h zh (yb − l) Ω
+ P h≥ dyb dxb + dyb dxb =
−w l y h −w −l S
(2)

2lxh
z
−λ h (x −w)
x y

h h
z
−λ h (x +w)
z
−λ h (x −w)
2wy
h
Ω= 1 − e xh h + 2 2 e xh h − e xh h + + 4wl (6)
λzh λ zh λzh
LoS Probability for Selected Environments Coverage Ratio for Different Environments
100 0.35
Suburban
Urban
90
0.3 Dense Urban
Urban High−Rise
80

0.25
70
LoS Probability [%]

Coverage Ratio
60
0.2

50

0.15
40

30
Urban (ITU)
0.1
Dense Urban (ITU)
20 Urban (Geo)
Dense Urban (Geo) 0.05

10 Urban (Sim)
Dense Urban (Sim)
0 0
0 50 100 150 200 250 300 350 400
10 20 30 40 50 60 70 80
Elevation Angle of Terrestrial Antenna [ ° ] The Distance between Two HAPs [km]

Fig. 3. Comparison between the geometrical model and the ITU model. Fig. 4. Optimal multi-HAP deployment in different areas.

TABLE I
T YPICAL VALUES OF E NVIRONMENTAL PARAMETERS 60% when the elevation angle is 50◦ , which is equivalent
Environment ρ (buildings/km2 ) w/l (m) λ β
to a coverage area with a radius 23.8km. This result
Suburban 7.5 × 10−4 11.55 1/8 0.069 demonstrates the immense potential of HAPs to provide
Urban 5 × 10−4 24.49 1/15 0.184 better communication quality and extensive coverage. In
Dense Urban 3 × 10−4 40.82 1/20 0.245 the figure, we also compare the simulation results with
Urban High-Rise 3 × 10−4 40.82 1/50 0.612
the LoS probability derived by using the ITU model.
* Without loss of generality, we assume that w = l.
It is obvious that the ITU model’s prediction greatly
deviates from the simulation results, in another word,
where σn2 is the variance of the zero mean additive white it overestimate the LoS probability. The reason lies in
Gaussian noise. the fact that in the ITU model, buildings are assumed to
We try to determine the optimal multi-HAP deploy- be evenly spaced and the cases where UEs are located
ment such that the coverage ratio can be maximized, close to a building are neglected. Therefore, the proposed
and the corresponding optimization problem can be model provides a higher accuracy compared with the
illustrated as ITU model.
d Fig.4 depicts the coverage ratio of different de-
max ployment schemes in different environments. The SNR
d,R R (13) threshold is set as 10dB, the communication bandwidth
s.t. SINR ≥ γ th . is taken as 10MHz, and the carrier frequency is chosen as
We propose an iterative algorithm to solve the optimiza- 2GHz. Other typical values of environmental parameters
tion problem as shown in Algorithm 1, where ΔR and are given in Table I. We can see that as the distance
Δd are the searching steps. What should be noticed is between two adjacent HAPs increases, the coverage ratio
that, when γ th > 0dB, the coverage of HAPs can not first increases and then decreases after reaching its peak
intersect with each other, i.e., there will be coverage value. The reason lies in the fact that as the distance
holes, which reveals the limitation of the deployment gets larger, the interference decreases, so the coverage
based on the packing theory [7]. for each HAP expands. However, when the distance is
large enough, the interference between HAPs can be
IV. S IMULATION R ESULTS AND D ISCUSSION neglected and the noise becomes the predominant factor,
As shown in Fig.3, we simulate the terrestrial envi- i.e., the coverage will not expand anymore even when
ronment and calculate the LoS probability by statistics. the distance increases, so the coverage ratio decreases.
The theoretical results derived according to the pro- Another thing that can be noticed is that the optimal
posed geometrical model match well with the simula- coverage ratio decreases as the building density gets
tion results. It is easy to see that the LoS probability larger, and the coverage for each HAP also decreases.
increases monotonically with the elevation angle of the This is due to the decrease of the LoS probability.
user antenna. Besides, the LoS ray is more likely to be
blocked in areas with a larger β, that is, the increase V. C ONCLUSION
of the building density, the building width, and the Since most existing geometric models are based on
building height will all result in the decrease of the LoS the impractical assumption that buildings are evenly
probability. In urban area, the LoS probability reaches spaced, in this paper, we first construct a novel statistical
geometry-based model, and analyze the impacts of dif-
ferent environmental parameters on the LoS probability
of air-to-ground communication. The proposed model
provides a higher accuracy compared with the ITU mod-
el. Based on the LoS probability, we further analyze the
path loss, and propose an efficient algorithm to determine
the optimal multi-HAP deployment to maximize the
coverage ratio. Simulation results indicate that we can
achieve a larger coverage ratio in suburban area than in
high-rise area with the same number of HAPs.
R EFERENCES
[1] G. M. Djuknic, J. Freidenfelds, and Y. Okunev, “Establishing
wireless communications services via high-altitude aeronautical
platforms: a concept whose time has come?” IEEE Commun.
Mag., vol. 35, no. 9, pp. 128–135, 1997.
[2] D. Grace and M. Mohorcic, Broadband Communications via
High-Altitude Platforms. John Wiley & Sons, 2011.
[3] F. Dong, H. Han, X. Gong, J. Wang, and H. Li, “A constellation
design methodology based on qos and user demand in high-
altitude platform broadband networks,” IEEE Trans. Multimedia,
vol. 18, no. 12, pp. 2384–2397, Dec. 2016.
[4] C. Li, Z. Chen, Y. Wang, Y. Yao, and B. Xia, “Outage analysis
of the full-duplex decode-and-forward two-way relay system,”
IEEE Trans. Veh. Technol., vol. 66, no. 5, pp. 4073–4086, 2017.
[5] M. Sharma, D. Chadha, and V. Chandra, “High-altitude platform
for free-space optical communication: Performance evaluation
and reliability analysis,” IEEE/OSA J. Opt. Commun. Netw.,
vol. 8, no. 8, pp. 600–609, Aug. 2016.
[6] A. Al-Hourani, S. Kandeepan, and S. Lardner, “Optimal lap al-
titude for maximum coverage,” IEEE Wireless Commun. Letters,
vol. 3, no. 6, pp. 569–572, 2014.
[7] M. Mozaffari, W. Saad, M. Bennis, and M. Debbah, “Efficient
deployment of multiple unmanned aerial vehicles for optimal
wireless coverage,” IEEE Commun. Letters, vol. 20, no. 8, pp.
1647–1650, Aug. 2016.
[8] E. Lutz, D. Cygan, M. Dippold, F. Dolainsky, and W. Papke, “The
land mobile satellite communication channel-recording, statistics,
and channel model,” IEEE Trans. Veh. Technol., vol. 40, no. 2,
pp. 375–386, 1991.
[9] A. Al-Hourani, S. Kandeepan, and A. Jamalipour, “Modeling air-
to-ground path loss for low altitude platforms in urban environ-
ments,” IEEE GLOBECOM’14, pp. 2898–2904, Dec. 2014.
[10] “Propagation data and prediction methods required for the design
of terrestrial broadband millimetric radio access systems,” ITU
Rec.P.1410-2, 2003.

You might also like