IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 68, NO.

8, AUGUST 2020 5135

UAV-to-Ground Communications: Channel

Modeling and UAV Selection
Petros S. Bithas , Senior Member, IEEE, Viktor Nikolaidis , Student Member, IEEE,
Athanasios G. Kanatas , Senior Member, IEEE, and George K. Karagiannidis , Fellow, IEEE

Abstract— Unmanned aerial vehicle (UAV)-enabled communi- of academia [1]–[3], industry [4], [5], and standardization
cations have been proposed as a critical part of the beyond fifth- bodies [6]. However, although UAVs add a new degree of
generation (5G) cellular networks. This type of communications freedom to the network, they have also induced various new
is frequently characterized by line-of-sight (LoS) and dynamic
propagation conditions. However, in various scenarios, the pres- challenges regarding physical layer communication aspects,
ence of large obstacles in the LoS path is unavoidable, resulting such as channel modeling, UAV selection strategy, place-
in shadowed fading environments. In this paper, a new channel ment/trajectory planning, and interference management. This
model is proposed, in which the effects of mobility and shadowing paper attempts to investigate the first two research challenges.
are simultaneously considered. In particular, the performance of In UAV-to-ground communications, the wireless medium
a UAV-based communication system operating in a shadowed
double-scattering channel is analyzed. The new channel model is characterized by high mobility and line-of-sight (LoS)
is generic, since it models various fading/shadowing conditions, propagation conditions [7], [8]. However, in several cases,
while it is in terms of easy-to-evaluate mathematical functions. due to the presence of large obstacles, the LoS assumption
Moreover, a low complexity UAV selection policy is proposed, is not satisfied [9]. In such scenarios, shadowing (or large-
which exploits shadowing-related information. The proposed scale fading) occurs, which results in randomly vary mean
scheme offers a reduction of the signal processing complexity,
without any important degradation on the performance, as com- envelope levels. Another independent phenomenon, character-
pared to alternatives approaches. In this context, a new analytical izing the mobile communication environments, is the double-
framework has been developed for investigating the performance scattering (DSc) propagation. This type of fading is observed
of the new strategy. Finally, the main outcomes of this paper are in scenarios where the transmitter (Tx), the receiver (Rx),
also validated by empirical data, collected in an air-to-ground or important scatterers around them, are in motion [10], [11].
measurement campaign.
In this environment, the transmitted signal propagates via
Index Terms— Air-to-ground measurement campaign, compos- multiple reflections, scattering, and diffractions around both
ite fading, double scattering, goodness-of-fit tests, unmanned the local scattering regions of the Tx and the Rx. Thus,
aerial vehicles (UAVs) selection (UAV association).
two independent sums of plane waves are multiplied to form
I. I NTRODUCTION the resulting impulse response, originating what is known
as DSc model [10]–[12]. However, the combination of DSc
U NMANNED aerial vehicle (UAV)-enabled networks have
been proposed as a critical component of the beyond
fifth-generation (5G) networks, to satisfy the requirements for
and shadowing has rarely been investigated in the past. For
example, in [13], Nakagami-m distribution was assumed for
the small-scale fading and gamma for the shadowing. From
massive connectivity, ultra reliability, and increased through-
the psychical point of view, this shadowed DSc channel is a
put. By operating as aerial user equipments (UEs) or as
reasonable approach to model the random variations of the
flying base stations (BSs), UAVs provide rapid recovery of
envelopes’ mean values.
the network services as well as system offloading in crowded
Recently, the inverse-gamma (IG) distribution has been
situations. This technology has already attracted the interest
proposed as a shadowing model, since it has been found
Manuscript received November 7, 2019; revised February 13, 2020; to offer an excellent fit to measurement data for large-scale
accepted April 12, 2020. Date of publication May 4, 2020; date of current fading, e.g., [14]–[17]. In particular, in [16], it was shown
version August 14, 2020. This research has been partially funded by Project
BEAM of the University of Piraeus Research Center. The associate editor that the IG distribution offers a better fit (as compared to
coordinating the review of this article and approving it for publication was the gamma) to empirical data for shadowed fading conditions
Y. Liu. (Corresponding author: Petros S. Bithas.) in non-stationary conditions. In [18], it was also shown that
Petros S. Bithas is with the General Department, National and Kapodistrian
University of Athens, 15772 Athens, Greece (e-mail: [email protected]). the η − μ/IG distribution provides a good fit to shadowing
Viktor Nikolaidis and Athanasios G. Kanatas are with the Department modeling in body-centric communications. Finally, in [17],
of Digital Systems, University of Piraeus, 18534 Piraeus, Greece (e-mail: a new shadowed DSc model for UAV-to-ground communi-
[email protected]; [email protected]).
George K. Karagiannidis is with the Electrical and Computer Engineering cation was presented, which was based on the Nakagami-m
Department, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece and IG distributions for modeling the multipath fading and
(e-mail: [email protected]). shadowing, respectively. In that paper, it was shown that this
Color versions of one or more of the figures in this article are available
online at http://ieeexplore.ieee.org. new distribution provides an excellent fit to empirical data
Digital Object Identifier 10.1109/TCOMM.2020.2992040 obtained in air-to-ground (A2G) measurement campaign.
0090-6778 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 5136 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 68, NO. 8, AUGUST 2020

The performance of UAV-enabled communications has been real-world communication conditions. Consequently, the main
analyzed in various studies in the past, e.g., [19]–[27]. Specif- contributions of this paper are threefold:
ically, in [19], a communication system was investigated, • A new composite fading distribution is proposed that
in which UAVs were used as relays between ground ter- simultaneously models the impact of DSc and shadow-
minals and a BS. In the same paper, a performance opti- ing on a UAV-enabled communication system. The new
mization algorithm was developed, through the control of distribution asymptotically includes as special cases pre-
the UAV heading angle. In [20], the outage probability (OP) viously reported ones, while it is given in a mathematical
of UAV-based communications has been analyzed by taking convenient form;
into consideration the co-channel interference, from aerial or • A new UAV-selection policy is proposed, which exploits
ground interfering nodes, in the case of LoS and non-LoS the stationarity of the shadowing. The new scheme offers
(NLoS) links. In [21], the OP is analyzed in a UAV-assisted similar performance with alternative approaches, which
relaying system operating in a composite fading environment, are based on the instantaneous CSI, with an important
modeled by the shadowed-Rice distribution. In [23], a multi- reduction on the signal processing overhead;
hop UAV relaying system has been analyzed, by taking into • Empirical data, collected in UAV-to-ground measure-
account the influence of both multipath and shadowing effects. ments campaign, are employed to support the proposed
For this system, the adopted performance metrics were the i) channel model and ii) UAV selection policy.
bit error rate and the channel capacity. In [24], considering a The remainder of the paper is organized as follows.
Rician-shadowed environment, the OP of hybrid-duplex UAV In Section II, the shadowed double-scattered channel model is
communication system was analyzed. In [26], the performance presented. In Section III, the performance of a single commu-
of a low-altitude aerial relaying platform is also investigated nication system operating over the considered channel model
in terms of the OP. is analyzed. In Section IV, the new UAV-selection strategy
As far as UAV-selection policies are concerned, various is presented and its performance is analyzed. In Section V,
approaches have been proposed in the past, e.g., [27]–[30]. empirical results and comparisons are presented and discussed,
More specifically, in [27], among other research targets, a UAV while in Section VI, the concluding remarks can be found.
selection policy was proposed, in which the UAV offering the
highest instantaneous received power was selected. In [28],
II. S YSTEM AND C HANNEL M ODEL
a UAV-selection mechanism was proposed, based on two
linear integer optimization problems, targeting to minimize We consider a communication scenario in which the Tx
the energy consumption and the operation time. In [29], is located in a UAV and the Rx is located at the ground.
three UAV relay selection strategies were proposed, namely It is also assumed that the Tx and/or the Rx and/or important
the closest-UAV relay selection, the maximum signal-to-noise scatterers around them are in motion, while the two local
ratio (SNR) UAV relay selection, and a random selection, with regions of scatterers, around both Tx and Rx, are separated
the second one offering the best performance with the cost of by a large distance. Such a communication environment is
high channel state information (CSI) feedback and complexity. expected to result in a double-scattered received envelope, due
The maximum SNR rule was also adopted in [30] as a relay to the keyhole propagation [11]. Moreover, it is reasonable to
selection policy. Nevertheless, such a policy offers a good assume that shadowing exists due to the presence of obstacles
performance with the cost of increased overhead and signal between the Tx and the Rx, e.g., buildings. Therefore, in the
processing complexity, since a continuous monitoring of all considered channel model, large scale fading (or shadowing)
available links is required. co-exists with the double-scattered propagation. A represen-
Motivated by the above, in this paper, a new shadowed tative example of the communication scenario under investi-
DSc distribution is proposed, which is able to model situations gation is shown in Fig. 1. In this figure, two UAV-enabled
where the received signal is simultaneously affected by DSc communication scenarios have been illustrated, namely one
and shadowing. In this context, the performance of a commu- with LoS conditions and one with NLoS, while the two
nication system operating in the aforementioned environment scattering regions as well as the shadowing area have also
is analyzed, employing well-known metrics, i.e., bit error prob- been identified.
ability (BEP), channel capacity, and OP. Additionally, a new
UAV-selection policy is proposed that exploits the information A. Double-Shadowing Communication Scenario
related to the shadowing. For the proposed scheme, the prob-
Next, we investigate the double-shadowing (DS) scenario,
ability density function (PDF) and the cumulative distribution
which is the case when shadowing (and as a consequence
function (CDF) of the output SNR are provided in closed form,
NLoS conditions) is present in both local scattering regions
while its performance is evaluated using the criteria of OP and
at the Tx and the Rx. Let γi denoting the received SNR from
average output SNR (ASNR). Finally, simulated results and
the ith (with i ∈ {1, . . . , L}) UAV, which is defined as
empirical data1 are also exploited to validate the analytical
results and prove the applicability of the proposed approach in 2 2
γi = Ni,1 Ii,1 Ni,2 Ii,2 , (1)
     
Mi,1 Mi,2

1 The data were collected in an A2G measurement campaign, which involved where Ni,j , Ii,j , and Mi,j , with j ∈ {1, 2}, denote the
a zeppelin-type airship [31]. multipath fading coefficients, the shadowing coefficients, and

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 BITHAS et al.: UAV-TO-GROUND COMMUNICATIONS 5137

It should be noted that Ni represents the square of a double-

Nakagami-m RV with PDF given as [34, eq. (11)]
 −

G2,0
0,2
m1 m2 y
Ω2 m1 ,m2
fNi (y) = , (5)
yΓ(m1 )Γ(m2 )
where Gm,n p,q [·|·] denotes the Meijer’s G-function [33, eq.
(9.301)], which is a built-in function in many mathematical
software packages, e.g., Mathematica, Maple, and thus it can
be directly evaluated. In (4), Ii models the double-IG (DIG)
RV with PDF given by
a1 +a2  1/2
2γ 2 a +a
− 1 2 2 −1 γ
fIi (y) = y Ka2 −a1 2 , (6)
Γ(a1 )Γ(a2 ) y

where Kv (·) denotes the modified Bessel function of the sec-

ond kind and order v [33, eq. (8.407)], while γ denotes
the normalized average SNR. By substituting (5) and (6)
in the expression of the total probability theorem, i.e., [35,
eq. (4.80)], and using [36, eqs. (14) and (21)], the PDF of γi
Fig. 1. The communication scenario under investigation. can be deduced as

−1 2,2 m1 m2 γ 1−α2 ,1−α1
fγi (γ) = γ S1 G2,2 m1 ,m2
the resulting SNR of the two local scattering regions, respec- γ
tively. It is noted that due to the large distance between (a)
= Γ(m1 + α2 )Γ(m2 + α2 )Γ(m1 + α1 )Γ(m2 + α1 )S1
the two regions, and thus the existence of the keyhole 
1
effect, the SNR coefficients Mi,1 and Mi,2 are consid- × 2 F̃1 m1 +a2 , m2 +a2 ; a1 +a2 +m1 +m2 ; 1−
ered to be independent [11]. In this paper, the generic γ̃γ
Nakagami-m distribution has been assumed as a multipath × γ −a2 −1 γ̃ −α2 , (7)
2
fading model. In this context, the PDF of Ni,j is given by
where 2 F̃1 (·, ·; ·; ·) denotes the regularized Gauss hyper-
[32, eq. (2.20)]
geometric function [37, eq. (07.31.02.0001.01)], S1 =
1 m1 m2
m  m x Γ(m1 )Γ(m2 )Γ(α1 )Γ(α2 ) , γ̃ = γ , while (a) holds due to [37,
mj j xmj −1 j eq. (07.34.03.0873.01)]. Here, it should be noted that the mean
fNi,j
2 (x) = exp − , (2)
Ωmj Γ(mj ) Ω value of γi is given by
γ
where mj is distribution’s shaping parameter, related to the E γi  = , (8)
(α1 − 1)(α2 − 1)
severity of the fading, i.e., as mj increases LoS conditions
are approximated, Ω denotes the mean square value, and where E · denotes expectation. It is noteworthy that the
Γ(·) is the gamma function [33, eq. (8.310/1)]. It should mean value of γi is independent from the (multipath) fading
be noted that a one-to-one mapping exists between the mj shaping parameters. It is also noted that as αi → 1 and
parameter and the Rician K factor, allowing the Nakagami- m1 → ∞, the DSc model approximates single scattering
m distribution to approximate the Rician distribution, which composite fading model.
is a well-established channel model for LoS propagation
environments [32]. B. Single-Shadowing Communication Scenario
The shadowing random fluctuations are modeled by the IG
distribution with PDF given by In several cases, due to the geometrical characteristics of
the propagation environment, it is expected that shadowing is
α  present in just one of the two scattering regions. This could be
γj j γj
fIi,j (y) = α +1
exp − , (3) the scenario where a strong LoS or strong multipath compo-
Γ(αj )y j y nents exist in only one of the two scattering regions, whereas
the other one is suffering from severe shadowing phenomena
where αj > 1 is the shaping parameter of the distribution, and thus NLoS conditions. As a consequence, randomly vary
related to the severity of the shadowing, i.e., lower values mean values of the received envelopes can be found in only
of αj result in lighter shadowing conditions, and γ j denotes one of the two regions. Under this single-shadowing (SS)
the scaling parameter. In order to evaluate the PDF of γi , assumption, (1) can be expressed as
the following two random-variables (RVs) are introduced
γis = (N1 N2 )2 Ii . (9)
2
Ni = (Ni,1 Ni,2 ) For evaluating the PDF of γis ,
(3) and (5) are substituted in
Ii = Ii,1 Ii,2 . (4) [35, eq. (4.80)], and by using a similar approach as the one

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 5138 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 68, NO. 8, AUGUST 2020

employed for the derivation of (7), the following simplified C. Ergodic Capacity
expression has been extracted The ergodic channel capacity quantifies the maximum rate
 1−α

of reliable communications that can be supported by the
fγis (γ) = S2 γ −1 G2,1
1,2 γ̃γ m ,m 1 2 subjected channel. Its definition is given by
(a)
= Γ(α + m1 )Γ(α + m2 )S2 γ̃ m1 ∞
C̄ = BW log2 (1 + γ)fγi (γ)dγ, (17)
× γ m1 −1 U (a + m1 ; m1 − m2 ; γ̃γ) , (10) 0

where α denotes the shaping parameter of the IG distribution, where BW denotes the signal’s transmission bandwidth.
U (·; ·; ·) denotes the confluent hypergeometric function [33, Assuming DS communication environment, (7) is substituted
1 in (17) and with the aid of [36, eq. (21)], the capacity can be
eq. (9.210/2) ], and S2 = Γ(m1 )Γ(m 2 )Γ(α)
. It is noted that (a)
holds due to [37, eq. (07.34.03.0719.01)]. The mean value of evaluated as
γ  1−α ,1−α ,0,1
γi is given as E γi  = (α−1) . −1
C̄ = BW S1 ln (2) G4,3
2 1
4,4 γ̃ m ,m ,0,01 2
. (18)

III. S INGLE UAV C OMMUNICATION S CENARIO Moreover, for the SS scenario and by employing (10), (18)
simplifies to
In this section, the performance of a communication system  1−α,0,1 
−1
operating in a shadowed DSc fading environment has been C̄ = BW S2 ln (2) G4,2
3,4 γ̃ m ,m ,0,0 . (19)
analytically investigated. Moreover, selected numerical evalu- 1 2

ated results are also presented.

D. Numerical Results
A. Outage Probability Next, using the stochastic analysis developed previously,
various numerically evaluated results are illustrated. In Fig. 2,
The OP is defined as the probability that the output SNR
the BEP of BPSK modulation scheme has been plotted as a
is below a predefined threshold γT , irrespective of the modu-
function of the average input SNR γ, for both DS and SS
lation scheme adopted. From the mathematical point of view,
channel models. For obtaining this figure, (15) and (16) have
the OP is defined as
γT
been employed, while m2 = m1 + 0.3 and α2 = α1 + 0.3. It is
Pout = Pr {γi ≤ γT } = Fγi (γT ) = fγi (γ)dγ, (11) shown that the performance improves when fading parameter
0 mi increases and/or shadowing parameter αi decreases, with
where Fγi (γT ) denotes the CDF of γi and Pr {·} denotes the former having highest impact. Moreover, it is interesting
the probability. For the DS scenario, substituting (7) in (11) to note that an important difference between the performance
and using [36, eq. (26)], results in the following closed-form of DS and SS channels is observed. This difference can easily
expression for the CDF of γi be explained if one considers that in DS scenario, shadowing
 1−α ,1−α ,1 is present in both scattering regions around the Tx and the
Fγi (γ) = S1 G2,3
2 1
3,3 γ̃γ . (12) Rx. On the other hand, in the SS scenario, shadowing is
m ,m ,0 1 2
present in only one of the two regions, which results in
For the SS scenario, by employing (10), (12) simplifies improved communication conditions. In Fig. 3, the normalized
to capacity C/BW is plotted as a function of γ for different
 1−α,1

values of αi . To obtain this figure, (18) and (19) have been
Fγis (γ) = S2 G2,2
2,3 γ̃γ m ,m ,0 . (13)
1 2 employed, with m1 = 1.5, m2 = 1.8, and α2 = α1 + 0.1.
In this figure, it is interesting to note that the performance
B. Average Bit Error Probability decreases as αi s increase, with a reducing rate. Finally, Monte
For evaluating the BEP, the CDF-based approach will be Carlo simulation results2 have been also included in Figs. 2-
adopted [38]. Assuming binary phase shift keying (BPSK) 3, depicting in all cases the correctness of the proposed
modulation scheme, the BEP can directly be evaluated using framework.
the following integral
∞ IV. UAV S ELECTION S TRATEGY
1
Pbe = √ Fγi (γ) exp (−γ) dγ. (14)
2 π 0 In scenarios where L UAVs operate as flying access points,
the users’ quality of experience might be improved by exploit-
Assuming DS communication environment, (12) is substituted
ing the additional available resources. In this context, a quite
in (14) and with the aid of [36, eq. (21)], the BEP of BPSK
efficient and practical approach, which is also proposed in
can be evaluated as
 √ −1 2,4  1−α2 ,1−α1 ,1,1/2 this paper, is to select the UAV access point to be connected
Pbe = S1 2 π G4,3 γ̃ . (15) to, according to a specific UAV association policy. A well-
m1 ,m2 ,0
known criterion that has been widely used in similar scenarios
Moreover, for the SS scenario and by employing (13), (15) is to select the UAV that offers the best instantaneous received
simplifies to
 √ −1 2,3  1−α,1,1/2
2 Values of the RVs under investigation have been generated in MATLAB™

Pbe = S2 2 π G3,3 γ̃ . (16) using their definitions, i.e., (1) and (9) for the DS and SS scenarios,
m1 ,m2 ,0 respectively.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 BITHAS et al.: UAV-TO-GROUND COMMUNICATIONS 5139

since in the latter case an index is fed back to the Tx in order

to proceed with the signal transmission.
With the above in mind, the received SNR of the proposed
shadowing-based UAV-selection mechanism is given by

γout = Ni max {Ii } = Ni Imax , (20)

1≤i≤L

where Ni depends on the multipath fading of the selected UAV.

In order to analytically evaluate the statistics of γout , the PDF
of Imax should be evaluated.

A. Double-Shadowing Scenario
As far as the DS scenario is concerned, the PDF of Imax
can be evaluated with the aid of the following theorem.
Theorem 1: The PDF of Imax = max {Ii } is given as
1≤i≤L
Fig. 2. BEP as a function of the average input SNR.
L−1
  i1
2L
fImax (γ) =
Γ(α1 )Γ(α2 ) i =0 i =0
1 2
  i2
L − 1 i1 γ q/2
× A
i1 i2 n ,··· ,n γ q/2+1
0 2α1 −1 =0
n0 +···+n2α1 −1 =i2
 1/2  1/2
γ γ
× exp −2i2 Kα2 −α1 2 ,
γ γ
(21)

where
i2
 i2
 i2
 i2

= ··· ,
n0 ,··· ,n2α1 −1 =0 n0 =0 n1 =0 n2α1 −1 =0
n0 +···+n2α1 −1 =i2 n0 +n1 +···+n2α1 −1 =i2

2a1 −1  nj+1
i2 ! 2j
while A = n0 !n1 !···n2a1 −1 ! j=1 j! (−1)i1 +i2 , and
2a1 −1
q = a1 + a2 + j=1 jnj+1 .
Fig. 3. Normalized capacity as a function of the average input SNR. Proof: See Appendix.
Since γout is actually the product of two independent RVs, its
PDF and CDF expressions can, respectively, be defined as
SNR per coherence time (CT) interval, e.g., [39]. Apparently,
1 ∞ γ 
this approach is mainly based on the assumption of accu-
fγout (γ) = fNi fImax (x)dx
rate estimations of the received SNR values. However, these 0 x x
estimates might be inaccurate especially in highly dynamic ∞ γ 
communication environments and/or in cases with increased Fγout (γ) = FNi fImax (x)dx. (22)
0 x
distance between the Tx and the Rx. It is safe to say that both
these scenarios characterize the UAV-enabled communications. The PDF of γout can be derived by substituting (5) and (21)
An alternative approach, which is also adopted in this paper, in (22) and using [36, eq. (21)] yielding to
is to select the associated UAV by exploiting the information
that is available for the shadowing behavior [40], [41]. This L−1
  i1   i2

L − 1 i1
approach is capitalizing on the larger decorrelation distance of fγout (γ) =
i1 =0 i =0
i1 i2 n0 ,··· ,n2α1 −1 =0
the large-scale fading as compared to the small-scale fading 2
n0 +···+n2α1 −1 =i2
CT. In this context, the selected UAV is the one providing the 
γ −1 ALS1 γ̃γ 3/2−q 5/2−q
, 2
maximum averaged received power, i.e., shadowing variable, × G2,2
2,2
2
.
over a predetermined time interval. It is noted that this (i2 + 1)q−1/2 (i2 + 1)2 m1 ,m2

approach may be applied in both the uplink and downlink, (23)

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 5140 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 68, NO. 8, AUGUST 2020

Moreover, the corresponding CDF is given by

L−1
 i1   i2
L − 1 i1
Fγout (γ) =
i1 i2 n ,··· ,n
i =0 i =0
1 2 0 2α1 −1 =0
n0 +···+n2α1 −1 =i2

ALS1 γ̃γ 1, 3/2−q , 5/2−q
× G2,3
3,3
2 2
.
(i2 + 1)
q−1/2 (i2 + 1)2 m1 ,m2 ,0

(24)
For evaluating the ASNR, (23) is substituted in the correspond-
ing definition, i.e.,
∞
γ̄out = γfγout (γ)dγ, (25)
0
and using [33, eq. (7.811/4)], the following closed-form
expression is deduced
L−1
  i1  
L L − 1 i1
γ out =
Γ(α1 )Γ(α2 ) i =0 i =0 i1 i2
1 2 Fig. 4. OP as a function of the normalized outage threshold for different
i2 √ architectures.
Aγ27/2−α1 −α2 πΓ (q − 5/2)
× q−5/2
.
n0 ,··· ,n2α1 −1 =0 (i2 + 1)
n0 +···+n2α1 −1 =i2 where this particular channel model can be applied. In the
(26) next section, it will be shown that the proposed channel
model provides an excellent fit to real A2G communication
An important insight that stems out from (26) is that the shap- environments.
ing parameters of the double-Nakagami distribution, i.e., the
behavior of the multipath fading, have no impact to the ASNR
of the proposed system. C. Numerical Results
In Fig. 4, assuming the proposed UAV-selection strategy,
B. Single-Shadowing Scenario the OP is plotted as a function of the normalized outage
As far as the SS scenario is concerned, the PDF of Imax is threshold (γ/γT ), for both channel models and different values
given in [41, eq. (15)]. Based on this expression and following of the number of UAVs L. For obtaining this figure, (24)
a similar approach as the one presented in Section IV-A, and (28) have been employed, with m1 = 1.5, m2 = 1.8,
the following closed-form expression for the PDF of γout is α1 = 2, α2 = 2.5. In this figure, it is depicted that the
deduced performance improves as the number of UAVs increases.
L−1
  Moreover, the performance gain degrades as L increases.
BS2 2,1 γ̃γ 1−p Similar outcomes can be extracted in Fig. 5, where the ASNR
fγout (γ) = G1,2 , (27)
γ n ,··· ,n =0
L m1 ,m2 is plotted as a function of γ also for both channel models and
1 α
n1 +···+nα =L−1 different values of αi . For obtaining this figure, (26) and (29)
where have been employed, with α2 = α1 + 0.5. Once more, it is
α  nj shown that the performance is better in DS model and as αi s
−p L! 1
B=L , increase. It is worth noting that the gap among the curves is
n1 !n2 ! · · · nα ! j=1 (j − 1)! higher when severe shadowing conditions exist, i.e., α1 = 3,
 as compared to the mild shadowing scenarios.
and p = α + α j=1 (j − 1)nj . Moreover, the corresponding
expression for the CDF of γout is given by
 V. E MPIRICAL R ESULTS AND C OMPARISONS
L−1

2,2 γ̃γ 1,1−p
Fγout (γ) = BS2 G2,3 . (28) In this section, empirical results obtained in a measurement
n ,··· ,n =0
L m1 ,m2 ,0 campaign will be exploited. These results will be used to
1 α
n1 +···+nα =L−1
justify the good fit of the considered channel model to empir-
Finally, for the scenario of SS, the ASNR is given by ical data as well as to prove the performance improvement
L−1
 achieved by the proposed UAV-selection strategy in real-world
Lγ
γ out = BΓ (p − 1) . (29) scenarios. The measurement campaign took place in Prague,
Γ(α) n1 ,··· ,nα =0 in an urban pedestrian environment, with the Rx positioned on
n1 +···+nα =L−1 the street level in the middle of a crossroad, while the Tx was
Here, it should be noted that the analytical results presented mounted on the bottom part of a zeppelin-type airship. The
in the previous sections allow the system designer to predict airship moved at a constant speed of 6.2 m/s, in a pre-defined
the system performance in various communication scenarios, route over the position where the Rx was located. The height

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 BITHAS et al.: UAV-TO-GROUND COMMUNICATIONS 5141

TABLE I
K-L AND K-S G OODNESS - OF -F IT

Fig. 5. ASNR as a function of the average input SNR for different

architectures.

above the ground was maintained at approximately 200 m

while the Rx was kept stationary, mounted on a tripod at
1.7 m above the ground. The airship was equipped with a
GPS sensor, enabling the Tx to point the antennas constantly
towards the Rx by using a positioner attached to the bottom Fig. 6. Empirical and theoretical PDFs and CDF comparisons of a route
part of the ship and allowed the distance between the Rx and segment.
the Tx to be calculated. The transmitted signals were received
by a single dual-linearly polarized rectangular patch antenna where Fe (x) and Ft (x) denote the empirical and theoretical
(collocated). More details for the measurement campaign can CDFs, respectively. By assuming a 95% confidence interval
be found in [31]. to all candidate distributions, a satisfactory fit can be ensured.
Next, based on the statistical framework developed on the The results are tabulated in Table I. In this table, it is
previous section and the empirical data collected in [31], shown that the distance K-L remains below 10%, with the
various comparative results are presented. More specifically, models incorporating the IG distribution holding always the
three composite DSc distributions have been investigated, smallest values for this criterion. Moreover, the passing rates
namely the double-Nakagami double-gamma, which is widely of K-S tests are quite encouraging in all cases, with the DIG
known as double-generalized K (dkG ) [13], the proposed distribution providing the best performance.
double-Nakagami double IG (denoted as dNI (DIG)), and the The good fit can be also verified in Fig. 6, where plots with
double-Nakagami with single IG (denoted as dNI (SIG)). It is the theoretical and empirical PDFs and CDFs are provided.
noted that the method of moments was employed for estimat- The estimated values of the distributions’ parameters, used for
ing the parameters of all distributions. In this context, in order obtaining these two figures, are provided in the Table included
to investigate the suitability of the proposed distributions in in Fig. 6. A general observation is that the investigated
fitting the empirical data sets, well-known goodness-of-fit tests scenarios are characterized by NLoS conditions (for the end-
have been applied. In particular, the Kullback–Leibler (K-L) to-end link) and strong scattering phenomena at the Rx side,
divergence criterion [42] and the Kolmogorov–Smirnov (K-S) due to the moving vehicles and surrounding buildings. This
test [35, eq. (8.320] were used. As far as the K-L distance is can be easily verified by the shapes of the PDFs and CDFs
concerned, it is given by [43] in Fig. 6, in which the direct LoS component is almost totally
    blocked. In all scenarios, the double-scattered signals may
1 pi qi emanate from reflections along the vehicles and the multiple
dKL = pi log + qi log , (30)
2 i
qi i
pi building facades from the structures surrounding the Rx.
The collected data, related to the received signal power,
where pi and qi are the sets of the simulated and empirical were split in 2 (or 3) segments, in order to emulate the
PDF values, respectively. The distribution that fits best to the received power from 2 (or 3) independent UAVs. Each of
measured data is the one that minimizes the K-L distance. these segments has been assumed to belong to a “virtual” UAV
Moreover, the K-S is defined as (indicated with specific colors in Figs. 7 and 8, i.e., blue for
UAV 1, green for UAV 2, and yellow for UAV 3). Fig. 7 depicts
dKS = max |Fe (x) − Ft (x)| , (31) the “virtual” UAV’s routes. In this context, we were able
x

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 5142 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 68, NO. 8, AUGUST 2020

Fig. 7. Measurement environment and the airship route. The red pin indicates
the position of the Rx. Fig. 9. Rx Power from all UAVs along with the proposed selection strategy
and selection per sample.

What is of particular interest to be noted is that the proposed

strategy offers a comparable performance with the CT-based
selection. However, for this communication system, the CT
was equal to 0.7λ (or 170 samples). Thus, it is quite interesting
to note that comparing shadowing-based selection, performed
per stationarity region, i.e., 40λ, with CT selection, performed
in every 0.7λ, results in 57 times less channel comparisons
required for the proposed approach, without any important loss
on the performance. Therefore, this reduction on the channel
comparisons is expected to reduce i) the signal processing
complexity, ii) the switching operations (among the UAV
links), and iii) the requirements for the feedback overhead.

VI. C ONCLUSION
In this paper, a new generic shadowed DSc channel model
has been proposed for describing the received signal behavior
Fig. 8. Rx Power from 3 UAVs and the proposed selection strategy. in UAV-enabled communication scenarios where DSc coexists
with shadowing. Capitalizing on the convenient mathematical
form of the new composite fading model, important stochastic
to compare the performance of the proposed UAV selection metrics for the received SNR of a UAV communication system
strategy with the corresponding of the “3 independent UAVs”. have been analytically derived. These expressions are then
First of all, in Fig. 8, the received power acquired from the 3 used to study the performance of the subjected system in
“virtual” UAVs is depicted. In the same figure, it is also terms of the OP, the BEP, and the channel capacity. Moreover,
inlcuded the corresponding power of the proposed scheme, a new UAV-selection policy has been proposed that intends to
in which the Rx selects the UAV that offers the best SNR, offer improved performance with reduced signal processing
on average, per stationarity region. This average SNR value complexity. This new scheme exploits the slow variations of
is evaluated with the help of a sliding window of W = the signal mean values for the UAV association. In this sense,
40λ, where λ = 15cm, which corresponds to 242 samples. novel closed-form expressions have been derived for the new
In Fig. 9, the empirical CDF of the received power is plotted scheme’s statistics of the output SNR and then used to study its
for the virtual UAVs, the proposed scheme, and an alternative performance. The accuracy of the channel model as well as the
UAV-selection approach. In the latter one, the UAV selection applicability of the proposed selection strategy have been also
process is performed per channel’s CT interval and is the verified by empirical data collected in an A2G measurement
most ordinary strategy that is frequent employed in scenarios campaign, proving also the usefulness of the derived results
where the best SNR principle is adopted. The performance to real-world scenarios. In our future work, various interesting
improvement due to the proposed shadowing-based selection issues will be addressed, including the impact of interfering
strategy is apparent from Fig. 9. This observation holds also effects in a multi-UAV network as well as the adoption of
for the case where selection is performed between 2 UAVs. appropriate interference mitigation techniques.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 BITHAS et al.: UAV-TO-GROUND COMMUNICATIONS 5143

A PPENDIX [4] I. B. Times. (2016). Nokia and EE Trial Mobile Base Stations
P ROOF FOR T HEOREM 1 Floating on Drones to Revolutionise Rural 4G Coverage. [Online].
Available: http://www.ibtimes.co.uk/nokia-ee-trial-mobile-base-stations-
In this Appendix, the proof for Theorem 1 is provided. With floatingdrones-revolutionise-rural-4g-coverage-1575795
the help of the order statistics of independent RVs, the PDF [5] (2017). Connected Aerial Vehicle Live. [Online]. Available: https://www.
huawei.com/en/industry-insights/outlook/1129mobile-broadband/xlabs/
of Imax is defined as use-cases/mbbf2017-connected-aerial-vehicle-live
[6] Study on Enhanced LTE Support for Aerial Vehicles. Technical
fImax (γ) = LfIi (γ)FIi (γ)L−1 , (A-1) Specification Group Radio, document TR 36.777 V1.1.0, 3GPP,
2018. [Online]. Available: https://portal.3gpp.org/desktopmodules/
where fIi (γ) is given by (6) and FIi (γ) is given by Specifications/SpecificationDetails.aspx?specificationId=3231
 [7] D. W. Matolak and R. Sun, “Unmanned aircraft systems: Air-ground
1 γ 1 channel characterization for future applications,” IEEE Veh. Technol.
FIi (γ) = 1 − G2,1
1,3 Mag., vol. 10, no. 2, pp. 79–85, Jun. 2015.
Γ(α1 )Γ(α2 ) γ 1 2 ,0
α ,α
1−2α1
√   [8] B. Li, Z. Fei, and Y. Zhang, “UAV communications for 5G and beyond:
(a) 2 π γ Recent advances and future trends,” IEEE Internet Things J., vol. 6,
= 1− γ 2α1 , 2 no. 2, pp. 2241–2263, Apr. 2019.
Γ(α1 )Γ(α2 ) γ
√    [9] A. A. Khuwaja, Y. Chen, N. Zhao, M.-S. Alouini, and P. Dobbins,
1−2α1 1/2 “A survey of channel modeling for UAV communications,” IEEE Com-
(b) 2 πΓ(2α1 ) γ
= 1− 1 − exp −2 mun. Surveys Tuts., vol. 20, no. 4, pp. 2804–2821, 2018.
Γ(α1 )Γ(α2 ) γ [10] J. B. Andersen and I. Z. Kovacs, “Power distributions revisited,” in Proc.
3rd Manage. Committee Meeting (COST), 2002, pp. 17–18.
2α
1 −1  1/2 k [11] J. Salo, H. M. El-Sallabi, and P. Vainikainen, “Statistical analysis of
1 γ
× 2 . (A-2) the multiple scattering radio channel,” IEEE Trans. Antennas Propag.,
k! γ vol. 54, no. 11, pp. 3114–3124, Nov. 2006.
k=0
[12] P. S. Bithas, A. G. Kanatas, D. B. da Costa, P. K. Upadhyay, and
In (A-2), (a) holds due to [37, eq. (07.34.03.0732.01)] (with U. S. Dias, “On the double-generalized gamma statistics and their
application to the performance analysis of V2 V communications,” IEEE
γ (·, ·) denoting the lower incomplete gamma function [33, Trans. Commun., vol. 66, no. 1, pp. 448–460, Jan. 2018.
eq. (8.350/1)]), while (b) holds due to [33, eq. (8.353/1)], [13] I. Trigui, A. Laourine, S. Affes, and A. Stephenne, “On the performance
under the assumption of 2α1 ∈ N. Next, in order to provide a of cascaded generalized k fading channels,” in Proc. IEEE Global
simplified expression for the FIi (γ)L−1 , the binomial identity Telecommun. Conf. (GLOBECOM), Nov. 2009, pp. 1–5.
[14] S. K. Yoo et al., “The κ − μ/inverse gamma and η − μ/inverse
has been employed twice resulting in gamma composite fading models: Fundamental statistics and empirical
i1  
L 
validation,” IEEE Trans. Commun., early access, Dec. 6, 2017, doi: 10.
 L i1 1109/TCOMM.2017.2780110.
FIi (γ)L−1 = (−1)i1 +i2 [15] P. C. Sofotasios et al., “Error analysis of wireless transmission over
i =0 i =0
i 1 i 2 generalized multipath/shadowing channels,” in Proc. IEEE Wireless
1 2

 1/2 2α
1 −1
 k/2
i2 Commun. Netw. Conf. (WCNC), Apr. 2018, pp. 1–6.
γ 2k γ [16] P. S. Bithas, A. G. Kanatas, and D. W. Matolak, “Shadowing-based
× exp −2i2 . antenna selection for V2 V communications,” in Proc. IEEE 29th Annu.
γ k! γ Int. Symp. Pers., Indoor Mobile Radio Commun. (PIMRC), Sep. 2018,
k=0
(A-3) pp. 106–110.
[17] P. S. Bithas, V. Nikolaidis, and A. G. Kanatas, “A new shadowed double-
scattering model with application to UAV-to-ground communications,” in
Finally, by employing the multinomial identity, [44, Proc. IEEE Wireless Commun. Netw. Conf. (WCNC), Apr. 2019, pp. 1–6.
eq. (24.1.2)] [18] S. K. Yoo, P. C. Sofotasios, S. L. Cotton, M. Matthaiou, M. Valkama,
n n n and G. K. Karagiannidis, “The η − μ/inverse gamma composite fading
n
   n!xn1 1 xn2 2 · · · xnLL model,” in Proc. IEEE Annu. Int. Sym. Per. In. Mob. Rad. Commun.,
(x1 + x2 + · · · xL ) = ··· Aug. 2015, pp. 166–170.
n1 =0 n2 =0 n =0
n1 !n2 ! · · · nL ! [19] P. Zhan, K. Yu, and A. L. Swindlehurst, “Wireless relay communications
L
n1 +n2 +···+nL =n with unmanned aerial vehicles: Performance and optimization,” IEEE
(A-4) Trans. Aerosp. Electron. Syst., vol. 47, no. 3, pp. 2068–2085, Jul. 2011.
[20] M. Kim and J. Lee, “Outage probability of UAV communications in
and using (A-1) and [33, eq. (8.468)] finally yields (21) and the presence of interference,” in Proc. IEEE Global Commun. Conf.
(GLOBECOM), Dec. 2018, pp. 1–6.
also completes this proof. [21] L. Yang, J. Chen, M. O. Hasna, and H.-C. Yang, “Outage performance
of UAV-assisted relaying systems with RF energy harvesting,” IEEE
ACKNOWLEDGEMENT Commun. Lett., vol. 22, no. 12, pp. 2471–2474, Dec. 2018.
[22] M. M. Azari, F. Rosas, K.-C. Chen, and S. Pollin, “Ultra reliable UAV
The authors would like to acknowledge Prof. N. Hatzidia- communication using altitude and cooperation diversity,” IEEE Trans.
mantis for his valuable help on implementing the K-S tests. Commun., vol. 66, no. 1, pp. 330–344, Jan. 2018.
P. S. Bithas dedicates this article to Spilios G. Bithas. [23] X. Chen, X. Hu, Q. Zhu, W. Zhong, and B. Chen, “Channel modeling
and performance analysis for UAV relay systems,” China Commun.,
R EFERENCES vol. 15, no. 12, pp. 89–97, Dec. 2018.
[24] T. Z. H. Ernest, A. S. Madhukumar, R. P. Sirigina, and A. K. Krishna,
[1] Y. Zeng, R. Zhang, and T. J. Lim, “Wireless communications with “A power series approach for hybrid-duplex UAV communication
unmanned aerial vehicles: Opportunities and challenges,” IEEE Com- systems under Rician shadowed fading,” IEEE Access, vol. 7,
mun. Mag., vol. 54, no. 5, pp. 36–42, May 2016. pp. 76949–76966, 2019.
[2] S. Hayat, E. Yanmaz, and R. Muzaffar, “Survey on unmanned aerial [25] T. Z. H. Ernest, A. S. Madhukumar, R. P. Sirigina, and A. K. Krishna,
vehicle networks for civil applications: A communications viewpoint,” “Hybrid-duplex communications for multi-UAV networks: An out-
IEEE Commun. Surveys Tuts., vol. 18, no. 4, pp. 2624–2661, 2016. age probability analysis,” IEEE Commun. Lett., vol. 23, no. 10,
[3] E. T. Michailidis, N. Nomikos, P. S. Bithas, D. Vouyioukas, and pp. 1831–1835, Oct. 2019.
A. G. Kanatas, “Optimal 3-D aerial relay placement for multi-user [26] L. Yang, J. Yuan, X. Liu, and M. O. Hasna, “On the performance
MIMO communications,” IEEE Trans. Aerosp. Electron. Syst., vol. 55, of LAP-based multiple-hop RF/FSO systems,” IEEE Trans. Aerosp.
no. 6, pp. 3218–3229, Dec. 2019. Electron. Syst., vol. 55, no. 1, pp. 499–505, Feb. 2019.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.
 5144 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 68, NO. 8, AUGUST 2020

[27] B. Galkin, J. Kibilda, and L. A. DaSilva, “A stochastic model for UAV Viktor Nikolaidis (Student Member, IEEE) received
networks positioned above demand hotspots in urban environments,” the B.Sc. degree in technology education and digital
IEEE Trans. Veh. Technol., vol. 68, no. 7, pp. 6985–6996, Jul. 2019. systems from UNIPI, Greece, in 2005, and the
[28] N. H. Motlagh, M. Bagaa, and T. Taleb, “UAV selection for a UAV- M.Sc. degree in mobile and satellite communications
based integrative IoT platform,” in Proc. IEEE Global Commun. Conf. from the University of Surrey, U.K., in 2007. He is
(GLOBECOM), Dec. 2016, pp. 1–6. currently pursuing the Ph.D. degree with the Depart-
[29] P. K. Sharma and D. I. Kim, “Secure 3D mobile UAV relaying for hybrid ment of Digital Systems, UNIPI. From 2008 to 2009,
satellite-terrestrial networks,” IEEE Trans. Wireless Commun., vol. 19, he was with the European Space Agency, Hands-On
no. 4, pp. 2770–2784, Apr. 2020. Projects Section, Education Office, European Space
[30] P. K. Sharma, D. Deepthi, and D. I. Kim, “Outage probability of 3-D Research and Technology Centre, The Netherlands.
mobile UAV relaying for hybrid satellite-terrestrial networks,” IEEE In 2009, he attended a Space Studies Program with
Commun. Lett., vol. 24, no. 2, pp. 418–422, Feb. 2020. the NASA Ames Research Center, Mountain View, CA, USA, organized by
[31] V. Nikolaidis, N. Moraitis, and A. G. Kanatas, “Dual-polarized narrow- the International Space University, Strasbourg, France. In 2013, he was with
band MIMO LMS channel measurements in urban environments,” IEEE the Telecommunication Systems Laboratory, University of Piraeus (UNIPI),
Trans. Antennas Propag., vol. 65, no. 2, pp. 763–774, Feb. 2017. as a Researcher. Since 2013, he has been involved in a number of research
[32] M. K. Simon and M.-S. Alouini, Digital Communication over Fading projects, including channel measurements and characterization for V2V and
Channels, 2nd ed. New York, NY, USA: Wiley, 2005. air-to-ground communication systems. His current research interests include
[33] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and MIMO techniques and channels, channel measurements, characterization, and
Products, 6th ed. New York, NY, USA: Academic, 2000. simulation for wireless and satellite communication systems. He served as a
[34] P. S. Bithas, G. P. Efthymoglou, and A. G. Kanatas, “V2V cooperative Treasurer for the local IEEE Student Branch, UNIPI, from 2003 to 2006.
relaying communications under interference and outdated CSI,” IEEE
Trans. Veh. Technol., vol. 67, no. 4, pp. 3466–3480, Apr. 2018.
[35] A. Papoulis and S. Pillai, Probability, Random Variables, and Stochastic
Processes, 4th ed. New York, NY, USA: McGraw-Hill, 2002.
[36] V. S. Adamchik and O. I. Marichev, “The algorithm for calculating inte- Athanasios (Thanasis) G. Kanatas (Senior Mem-
grals of hypergeometric type functions and its realization in REDUCE ber, IEEE) received the Diploma degree in electrical
system,” in Proc. Int. Symp. Symbolic Algebr. Comput. (ISSAC), Tokyo, engineering from the National Technical University
Japan, 1990, pp. 212–224. of Athens in 1991, the M.Sc. degree in satellite com-
[37] (2020). The Wolfram Functions Site. [Online]. Available: http:// munication engineering from the University of Sur-
functions.wolfram.com rey, U.K., in 1992, and the Ph.D. degree in mobile
[38] Y. Chen and C. Tellambura, “Distribution functions of selection com- satellite communications from NTUA in 1997. From
biner output in equally correlated Rayleigh, Rician, and Nakagami-m 2013 to 2017, he has served as the Dean for the
fading channels,” IEEE Trans. Commun., vol. 52, no. 11, pp. 1948–1956, ICT School, University of Piraeus, Greece. He is
Nov. 2004. currently a Professor with the Department of Digital
[39] B. Ji, Y. Li, B. Zhou, C. Li, K. Song, and H. Wen, “Performance analysis Systems, University of Piraeus (UNIPI), Greece.
of UAV relay assisted IoT communication network enhanced with energy He has published more than 200 articles in international journals and confer-
harvesting,” IEEE Access, vol. 7, pp. 38738–38747, 2019. ence proceedings. He has authored six books in the field of wireless and satel-
[40] A. Yilmaz, F. Yilmaz, M.-S. Alouini, and O. Kucur, “On the performance lite communications. He has been the technical manager of several European
of transmit antenna selection based on shadowing side information,” and National Research and development projects. His current research interests
IEEE Trans. Veh. Technol., vol. 62, no. 1, pp. 454–460, Jan. 2013. include the development of new waveforms and digital techniques for next
[41] P. S. Bithas, A. G. Kanatas, and D. W. Matolak, “Exploiting shadowing generation wireless systems, wireless channel characterization and modeling,
stationarity for antenna selection in V2 V communications,” IEEE Trans. antenna design and array beamforming for next generation wireless systems,
Veh. Technol., vol. 68, no. 2, pp. 1607–1615, Feb. 2019. and V2V communications and cybersecurity issues for V2V communications.
[42] S. Kullback and R. A. Leibler, “On information and sufficiency,” Ann. In 1999, he was elected as the Chairman of the Communications Society of
Math. Statist., vol. 22, no. 1, pp. 79–86, 1951. the Greek Section of IEEE.
[43] I. Sen and D. W. Matolak, “Vehicle–vehicle channel models for the
5-GHz band,” IEEE Trans. Intell. Transp. Syst., vol. 9, no. 2,
pp. 235–245, Jun. 2008.
[44] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Func-
tions: With Formulas, Graphs, and Mathematical Tables, vol. 55.
George K. Karagiannidis (Fellow, IEEE) was born
North Chelmsford, MA, USA: Courier Corporation, 1964. in Pithagorion, Samos Island, Greece. He received
the University Diploma (5 years) and Ph.D. degrees
Petros S. Bithas (Senior Member, IEEE) received in electrical and computer engineering from the
the Diploma degree (5 years) in electrical and com- University of Patras, in 1987 and 1999, respectively.
puter engineering and the Ph.D. degree in wire- From 2000 to 2004, he was a Senior Researcher
less communication systems from the University with the Institute for Space Applications and Remote
of Patras, Greece, in 2003 and 2009, respectively. Sensing, National Observatory of Athens, Greece.
From 2009 to 2018, he was with the Department In June 2004, he joined the Faculty of Aristotle
of Electronics Engineering, Technological Educa- University of Thessaloniki, Greece, where he is cur-
tional Institute of Piraeus, Greece. Since Novem- rently a Professor with the Electrical and Computer
ber 2010, he has been an Associate Researcher with Engineering Department and the Head of the Wireless Communications
the Department of Digital Systems, University of Systems Group (WCSG). He is also an Honorary Professor with South West
Piraeus (UNIPI), Greece, where he participates in Jiaotong University, Chengdu, China. His research interests are in the broad
a number of Research and development projects. He is currently an Assistant area of digital communications systems and signal processing, with emphasis
Professor with the General Department, National and Kapodistrian University on wireless communications, optical wireless communications, wireless power
of Athens, Greece. He serves on the Editorial Board for International Journal transfer and applications, and communications and signal processing for
of Electronics and Communications (ELSEVIER) and Telecom (MDPI). biomedical engineering.
He has been selected as an Exemplary Reviewer of IEEE C OMMUNICATIONS Dr. Karagiannidis has been involved as a general chair, a technical program
L ETTERS and IEEE T RANSACTIONS ON C OMMUNICATIONS in 2010 and chair, and a member of technical program committees in several IEEE and
2020, respectively, while he was a co-recipient of Best Paper Awards at the non-IEEE conferences. In the past, he was an editor in several IEEE journals.
IEEE International Symposium on Signal Processing and Information Tech- From 2012 to 2015, he was the Editor-in Chief of IEEE C OMMUNICATIONS
nology, 2013. He has published 39 articles in International scientific journals L ETTERS . He currently serves as an Associate Editor-in Chief for IEEE O PEN
and 39 articles in the proceedings of International conferences. His current J OURNAL OF C OMMUNICATIONS S OCIETY. He is one of the highly-cited
research interests include stochastic modeling of wireless communication authors across all areas of Electrical Engineering, recognized from Clarivate
channels and the design and performance analysis of vehicular communication Analytics as Web-of-Science Highly-Cited Researcher in the five consecutive
systems. years 2015–2019.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on November 21,2023 at 05:19:01 UTC from IEEE Xplore. Restrictions apply.