1

 
FNWF'23 1570936547
2  
3  
4  
5   On the Optimal Antenna Beamwidth for UAV
6  
7  
8  
Localization using mmWave/THz Range Radars
9  
10   Anna Gaydamaka, Andrey Samuylov, Dmitri Moltchanov, Mateen Ashraf, Bo Tan, Yevgeni Koucheryavy
11   Unit of Electrical Engineering and Communications, Tampere University, Finland. Email: [Link]@[Link].
12  
13  
14   Abstract— A number of critical UAV applications such as rescue due to the use of wide bandwidth [6], [7]. The operational
15   missions are performed in locations, where external positioning wavelengths allow for miniaturized and lightweight electronics
16   information is unavailable. To enable situational awareness in including transceiver and antenna systems [8]. On top of it, by
these use cases, millimeter wave (mmWave, 30 − 300 GHz) and
17   utilizing directional antenna radiation patterns, they generally
terahertz (THz, 0.3 − 3 THz) bands range radars can be utilized.
18   These radars are characterized by extremely high resolution. produce less interference [9], [10]. Finally, mmWave/THz
19   However, the small antenna aperture in these bands forces system radars can partially reuse mmWave/THz communications
20   designers to utilize massive antenna arrays to reach distances of hardware, either 5G New Radio or future 6G systems, that can
21   hundreds of meters. Thus, by utilizing directional antennas, these be utilized by UAVs for communications purposes [11].
22   radars need to scan the environment around a UAV detecting
23  
the location of surrounding UAVs. In this paper, by utilizing the However, the use of directional antenna radiation patterns
tools of stochastic geometry we determine the optimal beamwidth brings additional challenges for system designers. On the one
24   required to detect all UAVs around in a minimal time. Owning to
25   hand, to minimize the time required to detect surrounding
the solution complexity we also propose a simple approximation for
26   the UAV detection probability. Our results reveal that the the use UAVss, the beamwidth needs to be large. However, the use
27   of antenna’s small half-power beamwidths (HPBW) increases the of large beamwidth increases the number of objects scanned
28   successful detection probability at the expense of the scanning time. and thus the responses from them are characterized by a
If one wants to decrease the scanning time, the only option is to higher ”overlapping” probability. On top of this, the larger
29   decrease the radar’s scanning radius. The proposed approximation
30   beamwidth implies lower gains and smaller coverage. On the
is shown to introduce errors of no more than 5-8%.
31   Index Terms—Range radars, UAV, THz, mmWave, beamwidth, other hand, smaller beamwidth increases the number of antenna
32   scanning time, order statistics. configurations that needs to be used to perform the full scan.
33   Thus, the choice of the beamwidth optimizing the time to scan
34   I. I NTRODUCTION the surroundings is a non-trivial optimization task.
35   Unmanned Aerial Vehicles (UAVs) have become increasingly The aim of this paper is to determine the optimal beamwidth
36   of the antenna radiation pattern that needs to be utilized at the
popular in recent years, owing to their diverse applications
37   UAV mmWave/THz radar to successfully detect the surrounding
across various fields. These applications include but are not
38   objects. To this aim, we propose a mathematical model based
limited to delivery services, agriculture, and search and rescue
39   on stochastic geometry and order statistics. As for the objects’
[1]–[3]. UAVs have significant advantages over traditional
40   characteristics, we utilize the distance to them. This is assumed
manned aircraft, such as their ability to access difficult-to-
41   to be determined by using the time-of-arrival metric (Time of
reach areas, lower operational costs, and reduced environmental
42   Arrival (ToA)). The metric of interest is the successful detection
impact. Further, UAVs can perform dangerous and risky tasks
43   probability of all the objects surrounding the UAV of interest.
44   without endangering human lives, making them ideal for use
45   in hazardous situations. The main contributions of the work are:
46   Situational awareness is one of the critical safety require-
47   ments for UAVs. Performing their missions UAVs often rely • mathematical model for determining the half-power an-

48   upon external sources of positioning information such as Global tenna beamwidth (HPBW) of antenna radiation pattern
49   Navigation Satellite System (GNSS) and then use wireless maximizing the targets’ successful detection probability;
50   technologies to notify neighbors about their current location, • reliable and simple approximation for the targets’ success-
51   movement direction, and speed [4]. However, there are several ful detection probability characterized by the error in the
52   critical scenarios, where this information is partially or fully 5-8% range;
53   unavailable, e.g., rescue missions in deep woods, mountains, or • numerical results showing that the use of small HPBWs
54   indoors. In this case, to obtain information about surroundings is always preferable as it increases the target’s successful
55   UAVs need to rely upon local systems. One of the promising detection probability.
56   ways is to use radars to detect the distance to the surrounding
57   objects [5]. The rest of the paper is organized as follows. We introduce
60   As opposed to lower frequency bands, radars operating in the system model in Section II. The model is solved in Section
61   millimeter wave (mmWave, 30 − 300 GHz) and terahertz (THz, III. Numerical results are provided in Section IV. Conclusions
62   > 300 GHz) bands are characterized by enhanced resolution are given in the last section.
63  
64  
65  

1
 utilizes an antenna array to fully scan the environment around
it. Observe that HPBW α affects the number of antenna
configurations to perform the full scan, i.e., the smaller α is the
more configurations UAV needs to process and the higher the
full scanning time scan time T . The longest scanning interval
occurs when a UAV is on the edge of the coverage range, i.e.
tagged
UAV
R m away from the tagged UAV. Therefore, the scanning time
of one sector takes 2Rc seconds. Since there are 1/ sin2 (α/4)
sectors fitting the sphere, the full scanning time is
2Rc
Tf ull = . (2)
sin2 (α/4)
Let further Ti , i = 1, 2, . . . , N , be the ToA obtained from
the UAVs at the tagged UAV. Let qt denote the probability
that all the distances of UAVs are correctly detected in a single
spherical sector defined by HPBW α, i.e.,
Fig. 1. The considered deployment scenario. qt = ∩N
i=1 {Ti+i − Ti > ∆t }, (3)
II. S YSTEM M ODEL
where N is the number of UAVs in the spherical sector covered
In this section, we define our system model by introducing by a single antenna configuration, ∆t is the parameters that
assumptions for deployment, antenna, and system operation. depend on the emitted signal sequence. Assuming the minimal
We then define the metrics of interest. distance between the drones d, ∆t can be found according to
A. Deployment the range resolution [13], [14]

We study a 3D deployment of UAVs as shown in Fig. 1 and 2∗d

∆t = , (4)
consider time instant t. At this time instant we assume that c
UAVs are distributed according to the Poisson point process in where c is the speed of light.
ℜ3 with density λ UAV/m3 . We consider a randomly chosen
UAV. The emitted power of the UAV is assumed to be regulated D. Metrics of Interest
such that the area of radius R is covered by a single antenna The goal is to determine the HPBW of the antenna radiation
configuration and responses from UAVs are received with a pattern for a given density of UAVs in the environment such that
sufficient signal-to-noise ratio (SNR) for successful detection. the scanning interval Tf ull is minimized while the locations
The speed of UAVs is assumed to be negligible as compared of all the UAVs in the radius R from the tagged UAV are
to the scanning time, i.e. the round-trip time of a radar signal detected with a given probability, e.g., qt > 0.95.
at the maximum separation distance R.
The environment is assumed to be GNSS-denied, i.e., there III. T HE P ROPOSED G EOMETRICAL M ODEL
are no external sources of positioning information available at To determine the probabilities of events {Ti+i − Ti > ∆t }
UAVs. UAVs utilize mmWave/THz range radars to infer the we start with ToA, whose illustration is shown in Fig. 2, where
distance to the surrounding UAVs. To this aim, they utilize the τi , i = 1, 2, . . . , N − 1 denote the time intervals between
time-of-arrival ToA information. echoing sequences, ∆t is the minimum required separation
B. Antenna Model interval for correctly distinguish between echoed sequences
[13], while T is the scanning interval for a given antenna
We assume that a UE is equipped with arrays having NV × configuration. Observe that since the distances to targets are all
NH vertical and horizontal elements. The array’s Half-Power random variables (Random Variables (RVs)), the location of
Beamwidth (HPBW), α, can be determined as each echoed response in the ordered sequence is also a RVs.
±
α = 2|θm − θ3db |, (1)
A. Distance Detection Probability
where θm = cos−1 (−1/π) is the radiation pattern’s maximum, Recall that due to the PPP distribution of UAVs in ℜ3 ,
±
θ3db = cos−1 [− ± 2.782/(N(·) π)] are the ±3-dB points. Note UAVs’ locations are uniformly and independently distributed
that in practice, one may also use an approximation for vertical in the sphere of radius R around the tagged UAV [15]. For a
and horizontal planes α = 102/N(·) [12]. UAV randomly distributed in the sphere, the probability density
C. Operation function (pdf) of the distance for an arbitrarily chosen UAV is
given by [16]
The system operates as follows. At regular time instants
UAV scans the environment detecting the distance and speed 3x2
fL (x) = 3 , 0 < x < R. (6)
of surrounding UAVs. For this purpose, the UAV’s transceiver R

2
 emitted echoed sequences The CDF of the ith response takes the form of
sequence n! 3c3i
FTi (t) = ×
(i − 1)!(n − i)! (8R3 )n
Z t
× x3i−1 (8R3 − c3 x3 )n−i dx =
−∞
3 3
1 n!c3i R3n−3i t3i 2 F1 (i, i − n; i + 1; c8Rt3 )
= ,
Γ(1 − i + n) 8i R3n Γ(i + 1)
0 < t < 2R/c. (12)
Consider now the sequence of times between ordered
responses, τi = Ti+1 − Ti , i = 0, 1, . . . , j − 1. The probability
of detection of all UAVs can be written as
j−1
Y j−1
Y
Fig. 2. Illustration of emitted and echoing radar sequences. qj = P r{Ti+1 − Ti > ∆t} = P r{τi > ∆t}, (13)
To determine the response time from the target, the RVs i=0 i=0
transformation technique [17] is utilized. The distance between where j is the number of UAVs in the spherical sector.
the tagged and an arbitrarily chosen UAV is a RV. Knowing Note that although originally the locations of UAVs are
the pdf (6) of this RV and the functional relationship y = 2xc , independently distributed in the spherical sector, the ordered
one may determine the pdf of the response time as sequence of responses is no longer independent as (i + 1)th
response Ti+1 depends on ith response, Ti . Using (11) and
3c3 t2 (12), the joint pdf of these responses can be presented as (5).
fT (t) = , 0 < t < 2R/c. (7)
8R3 The pdf of the differences τi = Ti+1 −Ti , i = 0, 1, . . . , j −1
can be found again by RVs transformation technique [17].
By integrating (7), we obtain the Cumulative Distribution Recall that knowing the joint pdf wξ1n (x1 , x2 , ..., xn ) of the
Function (CDF) of the response time initial set of RVs ξ1n = (ξ1 , . . . , ξn ) and the functional
relationship yk = fk (x1 , . . . , xn ), k = 1, . . . , m, one may
c3 t3
FT (t) = , 0 < t < 2R/c. (8) derive the pdf Wη1m (y1 , y2 , ..., ym ) of the resulting set of
8R3 RVs η1m = (η1 , . . . , ηm ). To counter (7), which had only
The next step is to find the number of UAVs in the spherical one RV as input, the calculation of pdf of τi involves a set
sector. Given that the cone angle is α, we obtain the volume of two RVs, Ti and Ti+1 , i.e. ξ12 = (ξ1 , ξ2 ). The joint pdf
of the sector wξ12 (x1 , x2 ) = fTi ,Ti+1 (ti , ti+1 ) is shown in (5). Following
the RV transformation technique’s notation, we denote RV Ti
4πR3 sin2 (α/4) as x1 , Ti+1 as x2 , and the sought τi as y1 . Then, the functional
VS = . (9) relationship yk = fk (x1 , x2 ) looks like
3
(
Observe that the number of UAVs N in the spherical sector is y1 = x2 − x1 , 0 < x1 < 2R 2R
c , 0 < x2 < c , (14)
a RVs having Poisson distribution with parameter λVS , where y2 = x2 .
λ is the number of UAVs per cubic meter, VS is the volume From (14) we have an inverse function ϕ(y1 , y2 )
of the sector, i.e., (
x1 = y2 − y1 , 0 < y2 − y1 < 2R
c , 0 < y2 < c ,
2R

(λVS )j −λVS (15)

Pj = P r{N = j} = e , j = 0, 1, . . . . (10) x2 = y2 .
j!
Then, having the integration limits and the joint pdf, one
Let us now fix the number of UAVs in the sector at some may obtain the sought pdf of τi
constant j and consider the ordered sequence of responses, Ti ,
Z
i = 1, 2, . . . , N . The pdf of the ith response is given by [18] fτi (y1 ) = wξ1n [ϕ(y1 , y2 , .., yn ), y2 , ..., yn ] ×
Y (n−1)
∂ϕ(y1 , y2 , .., yn )
n(n − 1)! × ∂y2n =
fTi (t) = [FT (t)]i−1 [1 − FT (t)]n−i fT (t) = ∂y1
(i − 1)!(n − i)! Z
 3 3 i−1  n−i 2 3 ∂ϕ(y1 , y2 )
n! c t c3 t3 3t c = fTi ,Tj [ϕ(y1 , y2 ), y2 ] ∂y2 =
= 1− = Y 1 ∂y1
(i − 1)!(n − i)! 8R3 8R3 8R3 Z
n! 3c3i t3i−1 (8R3 − c3 t3 )n−i = fTi ,Tj [y2 − y1 , y2 ] |−1| ∂y2 =
= , ZY 1
(i − 1)!(n − i)! (8R3 )n
= fTi ,Tj [y2 − y1 , y2 ] ∂y2 . (16)
0 < t < 2R/c. (11) Y1

3
 n!
fTi ,Ti+1 (ti , ti+1 ) = [FTi (ti )]i−1 (1 − FTi+1 (ti+1 ))n−i−1 fTi (ti )fTi+1 (ti+1 ) =
(i − 1)!(n − i − 1)!
c3 t3i
!(i−1)
(n!)3 1 n!c3i R3n−3i t3i i 2 F1 (i, i − n; i + 1; 8R 3 )
= ×
((i − 1)!)3 ((n − i − 1)!)3 (n − i)2 Γ(1 − i + n) 8i R3n Γ(i + 1)
 (n−i−1)
c3 t3i+1
1 n!c 3i 3n−3i 3i
R t F
i+1 2 1 (i, i − n; i + 1; 8R3 ) 3c3i t3i−1
i (8R3 − c3 t3i )n−i
× 1 − i 3n
 ×
Γ(1 − i + n) 8R Γ(i + 1) (8R3 )n

3c3i t3i−1 3 3 3
i+1 (8R − c ti+1 )
n−i
× , 0 < t < 2R/c. (5)
(8R3 )n

Once fτi , i = 0, 1, . . . , j − 1 is found, we can calculate the y1 = x1 − x2 , x2 − x1 < 0, and (ii) y1 = x2 − x1 , x2 − x1 > 0.
probabilities of components in (13) as This leads to the following branches of the inverse function
Z ∞ 
2R
t
qi,j = fτi (t)∂t, i = 0, 1, . . . , j, (17) x1 = y1 + y2 , 0 < y1 + y2 < c ,

∆t x2 = y2 , 0 < y2 < 2R c ,
where j is the number of UAVs in the spherical sector. 

y1 > 0.
Now, recalling that the number of UAVs N in the spherical 
2R
sector follows Poisson distribution in (10) the probability that x1 = y2 − y1 , 0 < y2 − y1 < c ,

we detect all UAVs correctly is given by x2 = y2 , 0 < y2 < 2R c ,
(21)

y1 > 0.

∞ j
X (λVS )j −λVS
Y
t
qt = e qi,j . (18)
j! Combining these branches, the pdf of the time difference
j=1 i=0
between responses of two UAVs is obtained by integrating
B. Approximation
As one may notice, several factorials and gamma functions f∆t (y1 ) =
Z
make the calculation of (5) challenging. To reduce the compu- = fTi ,Tj [ϕ(y1 , y2 , .., yn ), y2 , ..., yn ] ×
tational complexity, we propose an approximation. The main Y (n−1)
idea of is to omit the order of the received signals but still ∂ϕ(y1 , y2 , .., yn )
× ∂y2n =
account for their number. Thus, instead of the time difference ∂y1
between two sequential responses, any two responses within a Z
∂ϕ(y1 , y2 )
spherical sector are compared. By choosing the minimal time = fTi ,Tj [ϕ(y1 , y2 ), y2 ] ∂y2 =
Y 1 ∂y1
difference between responses, the probability that all UAVs Z −y1 + 2R c 9c6 (y + y )2 y 2
are detected correctly is approximated. 1 2 2
= |1| ∂y2 +
Consider two random responses Ti and Tj within a spher- 0 64R6
Z 2Rc 9c6 (y − y )2 y 2
ical sector. By utilizing the random variables transformation 1 2 2
technique and assuming that the responses are independent, + |−1| ∂y2 =
y1 64R6
we obtain the joint pdf of the time difference
9c 9c2 y1 3c3 y12 3c6 y15
9c6 t2i t2j = + + + , 0 < y1 < 2R/c. (22)
2R 2R 5R 4R 2 4R 3 320R6
fTi ,Tj (ti , tj ) = , 0 < ti < , 0 < tj < . (19)
64R6 c c In the original notation, the pdf of the time difference
Using the RV transformation technique’s notation, we denote between responses of two UAVs is computed as
the time instant of the response from UAV i as x1 , and the
time instant of the response from UAV j as x2 . The time 9c 9c2 t 3c3 t2 3c6 t5
f∆t (t) = + + + , 0 < t < 2R/c.
difference between two responses, ∆t, is denoted as y1 . As 5R 4R2 4R3 320R6
(23)
we are interested in the absolute value of the time differences,
the functional relationship is y1 = f (x1 , x2 ) = |x2 − x1 |. The The CDF can be obtained by integrating (23)
system of equations for transformation is therefore
1152cR5 t − 720c2 R4 t2 + 160c3 R3 t3 − c6 t6
F∆t (t) = ,
(
y1 = |x2 − x1 | , 0 < x1 < 2R 2R
c , 0 < x2 < c , 640R6
(20)
y2 = x2 . 0 < t < 2R/c. (24)

Next, we need an inverse function ϕ(y1 , y2 ) of (20). Since y1 If the minimal time difference exceeds a certain value, it
is a multivalued function, we have the following branches: (i) guarantees that all UAVs in the sector are detected correctly.

4
 1.0 1.0 1.0

Probability of succesful detection, qt

Probability of succesful detection, qt

Probability of succesful detection,

0.8 0.8 0.8

0.6 0.6 0.6

50 m 50 m
0.4 100 m 0.4 100 m 0.4
150 m 150 m
0.2 0.2 0.2

0.0 0.0 0.0

10 20 30 40 50 60 0.0002 0.0004 0.0006 0.0008 0.0010 60 80 100 120 140
HPBW angle, α UAV density, λ Coverage radius,

(a) HPBW angle, α (b) UAV density, λ (c) Coverage radius, R

Fig. 3. UAV detection performance as a function of the HPBW angle, UAV density, and coverage radius.

Thus, according to [17], we have the following CDF for the We start our discussion with Fig. 3 visualizing the depen-
minimal time difference dencies of the probability of successful detection of all UAVs
in the sector, qT , on the HPBW angle α, UAV density λ,
F∆tmin (t; n) = 1−
n and coverage radius, R. By analyzing Fig. 3(a) showing the
1152cR5 t − 720c2 R4 t2 + 160c3 R3 t3 − c6 t6

− 1− , considered probability as a function of α, one may observe
640R6 that the smaller the HPBW angle the higher the probability
0 < t < 2R/c, (25) of successful detection. The rationale is that smaller HPBWs
lead to a smaller volume of the spherical sector and thus lower
where n is the number of UAVs in the spherical sector. number of UAVs on average. Additionally, we see that for
Recall that we are interested in cases when the time differ- all HPBWs smaller than approximately 10◦ the considered
ence is equal or greater than a certain value. This probability probability plateaus at approximately 1 for all the considered
is provided by a Complementary Cumulative Distribution system parameters.
Function (CCDF)
Fig. 3(b) further highlights that the impact of the UAV density
F̄∆tmin (t; n) = 1 − F∆tmin (t; n), 0 < t < 2R/c. (26) is more profound as the considered metrics drops sharply for
By choosing a certain probability, e.g. q > 0.95, one may all the considered distances R. Nevertheless, the rationale is
find a corresponding time difference t using (26). Finally, the similar – increasing the density of UAVs leads to a higher
probability that all UAVs are detected correctly is given by number of UAVs in a single spherical sector for fixed values
of HPBW angle, α, and coverage radius R. Finally, the impact
∞
X (λVS )n −λVS of coverage radius R, shown in Fig. 3(c), demonstrates that
qt = e−λVS + λVS e−λVS + e F̄∆tmin (t; n).
n! the successful detection probability qT decreases exponentially
n=2
(27) with the increase in R.
Note that by decreasing the value of HPBW angle the
IV. N UMERICAL R ESULTS successful detection probability increases. However, smaller
In this section, we study the system’s performance. We will HPBW angles lead to higher overall scanning time as the
start showing the UAV detection performance as a function number of antenna configurations that need to be scanned
of three main parameters – HPBW angle α, UAV density λ, 1.0
Probability of succesful detection, pt

and coverage radius, R. Then, we will proceed assessing the

trade-off between the detection probability and time it takes 0.8
to scan the whole region around UAV. Finally, we show the
performance of the proposed approximation. The default system 0.6
parameters are provided in Table I.
TABLE I 0.4
THE DEFAULT SYSTEM PARAMETERS 50 m
0.2 100 m
Notation Description Value
150 m
λ Number of UAVs per cubic meter 0.0001...0.001 units/m3
α HPBW 1◦ ...60◦ 0.0
0 1 2 3 4 5
c Speed of light 299,792,458 m/s
Scanning time, ms
d Minimal distance between UAVs 1m
R Coverage radius 50, 100, 150 m
Fig. 4. UAV detection performance as a function of the scanning time.

5
 1.0 increased scanning time. If one wants to decrease the scanning

Probability of succesful detection, qt

Exact, 100 m time while keeping the detection probability intact, the only
0.8 Approximation, 100 m
option is to decrease the coverage radius.
Exact, 150 m
In our future work, we will complement the developed model
0.6 Approximation, 150 m
by accounting for the received SNR. Furthermore, we will also
investigate how interference from other UAVs impacts the
0.4 detection performance.

0.2 ACKNOWLEDGMENT
The work in this paper has been funded by the Academy
0.0 of Finland within ACCESS (Autonomous Communication
10 20 30 40 50 60
HPBW angle, α Converged with Efficient Sensing for UAV Swarms) project.
R EFERENCES
Fig. 5. Comparison of the exact and approximation approaches.
[1] M. Bacco, A. Berton, A. Gotta, and L. Caviglione, “IEEE 802.15.4
increases. To understand the trade-off between these two Air-Ground UAV Communications in Smart Farming Scenarios,” IEEE
metrics, Fig. 4 shows the probability of successful detection Communications Letters, vol. 22, no. 9, pp. 1910–1913, 2018.
of all UAVs in the coverage radius computed as pT = (qT )n , [2] A. Ranjha, G. Kaddoum, and K. Dev, “Facilitating URLLC in UAV-
Assisted Relay Systems With Multiple-Mobile Robots for 6G Networks:
where n is the number of antenna configurations that explicitly A Prospective of Agriculture 4.0,” IEEE Transactions on Industrial
depends on HPBW angle, as a function of the scanning Informatics, vol. 18, no. 7, pp. 4954–4965, 2022.
time, Tf ull . Here, we observe that for all the considered [3] A. Albanese, V. Sciancalepore, and X. Costa-Pérez, “SARDO: An
Automated Search-and-Rescue Drone-Based Solution for Victims Local-
coverage radii the probability of successful detection increases ization,” IEEE Transactions on Mobile Computing, vol. 21, no. 9, pp.
as the scanning time becomes larger. The rationale is that 3312–3325, 2022.
the successful detection probability in a single sector is an [4] Y. Dang, C. Benzaı̈d, B. Yang, T. Taleb, and Y. Shen, “Deep-Ensemble-
Learning-Based GPS Spoofing Detection for Cellular-Connected UAVs,”
exponentially increasing function of the HPBW angle while IEEE Internet of Things Journal, vol. 9, no. 24, pp. 25 068–25 085, 2022.
the scanning time is a power function of the HPBW angle. [5] S. Hu, X. Yuan, W. Ni, and X. Wang, “Trajectory Planning of Cellular-
This implies that the use of small HPBW is always preferable Connected UAV for Communication-Assisted Radar Sensing,” IEEE
Transactions on Communications, vol. 70, no. 9, pp. 6385–6396, 2022.
from the point of view of detection probability. If one wants to [6] A. V. Lopez, A. Chervyakov, G. Chance, S. Verma, and Y. Tang, “Oppor-
decrease the scanning time while keeping the successful UAV tunities and Challenges of mmWave NR,” IEEE Wireless Communications,
detection probability intact, the only option is to decrease the vol. 26, no. 2, pp. 4–6, 2019.
[7] C. Zhang, W. Zhang, W. Wang, L. Yang, and W. Zhang, “Research
coverage radius. Challenges and Opportunities of UAV Millimeter-Wave Communications,”
Recall that the use of direct calculations according to IEEE Wireless Communications, vol. 26, no. 1, pp. 58–62, 2019.
Section III-A is computationally intensive. Thus, we now [8] J. Zhang, X. Ge, Q. Li, M. Guizani, and Y. Zhang, “5G Millimeter-Wave
Antenna Array: Design and Challenges,” IEEE Wireless Communications,
proceed assessing the approximate approach elaborated in vol. 24, no. 2, pp. 106–112, 2017.
Section III-B. To this aim, Fig. 5 shows the comparison of [9] V. Petrov, J. Kokkoniemi, D. Moltchanov, J. Lehtomäki, M. Juntti, and
the exact and approximate approaches. As one may observe, Y. Koucheryavy, “The Impact of Interference From the Side Lanes on
mmWave/THz Band V2V Communication Systems With Directional
the approximation curve follows the pattern of the exact one Antennas,” IEEE Transactions on Vehicular Technology, vol. 67, no. 6,
- the growth of the HPBW angle is followed by the drop in pp. 5028–5041, 2018.
the detection probability. If one aims at lower computational [10] M. Ashraf, B. Tan, D. Moltchanov, J. S. Thompson, and M. Valkama,
“Joint optimization of radar and communications performance in 6g
complexity and faster calculation, the approximation approach cellular systems,” IEEE Transactions on Green Communications and
might be a viable option to consider. Networking, vol. 7, no. 1, pp. 522–536, 2023.
[11] V. Petrov, G. Fodor, J. Kokkoniemi, D. Moltchanov, J. Lehtomaki,
V. C ONCLUSIONS S. Andreev, Y. Koucheryavy, M. Juntti, and M. Valkama, “On unified
vehicular communications and radar sensing in millimeter-wave and
Motivated by the use of mmWave/THz communications low terahertz bands,” IEEE wireless communications, vol. 26, no. 3, pp.
systems in 5G/6G systems, we explored the use of THz band 146–153, 2019.
[12] C. Balanis et al., “Antenna theory: analysis and design,” Microstrip
for UAV positioning in locations, where external positioning Antennas, John Wiley & Sons, 2005.
information is unavailable, e.g., deep woods, mountains, in- [13] A. Annan, “Electromagnetic principles of ground penetrating radar,”
doors. Specifically, we utilized the tools of stochastic geometry Ground penetrating radar, vol. 1, pp. 1–37, 2009.
[14] M. Ashraf, A. Gaydamaka, D. Moltchanov, A. Mohammad, and B. Tan,
to characterize the successful UAV detection probability in the “Maximizing detection of target with multiple direction possibilities to
proximity of the target UAV as a function of the beamwidth support immersive communications in metaverse,” in Proceedings of the
and scanning time. We also proposed a simple approximation 2nd Workshop on ISAC for Metaverse, 2023, pp. 30–35.
[15] D. Stoyan, W. S. Kendall, S. N. Chiu, and J. Mecke, Stochastic geometry
whose error lies in the range of 5-8 %. and its applications. John Wiley & Sons, 2013.
Our results reveal that the successful detection probability [16] D. Moltchanov, “Distance distributions in random networks,” Ad Hoc
is an exponentially decreasing function of the HPBW angle, Networks, vol. 10, no. 6, pp. 1146–1166, 2012.
[17] S. M. Ross, Introduction to probability models. Academic press, 2014.
coverage radius, and UAV density. Considering the whole [18] H. A. David and H. N. Nagaraja, Order statistics. John Wiley & Sons,
scanning area around a UAV the use of small HPBWs increases 2004.
the probability of successful detection at the expense of the