Content uploaded by Hoang D. Le
Author content
All content in this area was uploaded by Hoang D. Le on Nov 08, 2022
Content may be subject to copyright.
IEICE Communications Express, Vol.11, No.10, 649–654
TCP over hybrid FSO/RF-based
satellite networks in the
presence of cloud coverage
Thang K. Nguyen1, Chuyen T. Nguyen1, Hoang D. Le2,
and Anh T. Pham2, a)
1School of Electrical and Electronic Engineering,
Hanoi University of Science and Technology,
Hanoi 100000, Vietnam
2Computer Communications Laboratory, The University of Aizu,
Aizuwakamatsu 965–8580, Japan
a) pham@u-aizu.ac.jp
Abstract: Cloud coverage is one of the critical concerns for reliable
hybrid free-space optic (FSO)/radio frequency (RF) based satellite communi-
cations. This adverse issue poses various challenges to the performance of the
physical layer and different upper-layer protocols. This paper addresses the
performance of transmission control protocol (TCP), which is one of the most
essential transport protocols, over hybrid FSO/RF-based satellite networks
in the presence of clouds. Specifically, the TCP throughput performance is
analyzed under the impact of different cloud types. Also, different frequency
bands of both FSO and RF used for satellite communications are investigated.
The obtained results quantitatively demonstrate the severe impact of clouds
on TCP performance over hybrid FSO/RF-based satellite networks.
Keywords: hybrid FSO/RF satellite networks, TCP throughput, cloud
coverage
Classification: Satellite Communications
References
[1] H.D. Le, P.V. Trinh, T.V. Pham, D.R. Kolev, A. Carrasco-Casado, T. Kubo-Oka,
M. Toyoshima, and A.T. Pham, “Throughput analysis for TCP over the FSO-
based satellite-assisted Internet of Vehicles,” IEEE Trans. Veh. Technol., vol. 71,
no. 2, pp. 1875–1890, 2022. DOI: 10.1109/TVT.2021.3131746
[2] T.K. Nguyen, C.T. Nguyen, H.D. Le, and A.T. Pham, “TCP performance
over satellite-based hybrid FSO/RF vehicular networks: Modeling and analy-
sis,” IEEE Access, vol. 9, pp. 108426–108440, 2021. DOI: 10.1109/ACCESS.
2021.3101903
[3] F. Nadeem, V. Kvicera, M.S. Awan, E. Leitgeb, S.S. Muhammad, and G. Kandus,
“Weather effects on hybrid FSO/RF communication link,” IEEE J. Sel. Areas
Commun., vol. 27, no. 9, pp.1687–1697, 2009. DOI: 10.1109/JSAC.2009.091218
[4] “Attenuation due to clouds and fog,” ITU-R P-840-6, 2013.
[5] H.D. Le, T.V. Nguyen, and A.T. Pham, “Cloud attenuation statistical model for
satellite-based FSO communications,” IEEE Antennas Wireless Propag. Lett.,
vol. 20, no. 5, pp. 643–647, 2021. DOI: 10.1109/LAWP.2021.3058641
©IEICE 2022
DOI: 10.1587/comex.2022XBL0104
Received June 25, 2022
Accepted July 8, 2022
Publicized July 15, 2022
Copyedited October 1, 2022
649
IEICE Communications Express, Vol.11, No.10, 649–654
[6] T.V. Nguyen, H.D. Le, T.V. Pham, and A.T. Pham, “Link availability of satellite-
based FSO communications in the presence of clouds and turbulence,” IEICE
Commun. Express, vol. 10, no. 5, pp. 206–211, Feb. 2021. DOI: 10.1587/comex.
2021XBL0009
[7] T. Nguyen, H.D. Le, N.T. Dang, and A.T. Pham, “On the design of rate adaptation
for relay-assisted satellite hybrid FSO/RF systems,” IEEE Photon. J., vol. 14,
no. 1, pp. 1–11, Feb. 2022. DOI: 10.1109/JPHOT.2021.3130720
[8] M.S. Awan, Marzuki, E. Leitgeb, B. Hillbrand, F. Nadeem, and M.S. Khan,
“Cloud attenuations for free-space optical links,” Int. Workshop Satellite Space
Commun., pp. 274–278, 2009. DOI: 10.1109/IWSSC.2009.5286364
1 Introduction
Recent years have witnessed the rapid development of low Earth orbit (LEO) satellite
networks in the provision of the Internet from space [1]. For the last-mile access
of Internet of Vehicles (IoV), hybrid free-space optical (FSO)/radio frequency (RF)
communication has recently attracted research efforts worldwide [2]. FSO offering
extremely high-speed connections can be used as the primary link, while RF link
serves as a backup link in case of FSO link failure. One of the critical concerns
on the hybrid FSO/RF-based satellite links is the cloud coverage [3]. This adverse
issue poses various challenges to the performance of the physical (PHY) layer and
different upper-layer protocols.
The attenuation models due to cloud coverage have been widely studied in both
RF [4] and FSO [5, 6] communications. These models are then used to evaluate
the PHY layer performance, e.g., in hybrid FSO/RF networks [7]. Nevertheless,
the impact of clouds on the performance of upper-layer protocols has not been well
investigated. On the other hand, transmission control protocol (TCP) is, by far, one of
the most essential transport protocols for many of the Internet’s popular applications
that require reliable connectivity. It is worth noting that many TCP variants, e.g.,
loss-based variants, tend to perform poorly as it misinterprets the losses due to bad
channel conditions as an indication of network congestion. Its performance was
considerably degraded under the impact of hybrid FSO/RF fading channels [2].
Therefore, it is of importance and necessary to evaluate the performance of TCP
under the impact of clouds in hybrid FSO/RF-based satellite networks.
This paper addresses the performance of TCP over hybrid FSO/RF-based satellite
networks in the presence of cloud coverage. In particular, the TCP throughput per-
formance is analyzed under the impact of different cloud types. Different frequency
bands of both FSO and RF used for satellite communications are also investigated.
2 Network descriptions
The end-to-end network scenario considered in this paper is depicted in Fig. 1, where
the last-mile connection is the hybrid FSO/RF link between an LEO satellite and
a vehicle (e.g., unmanned aerial vehicle (UAV), self-driving car, and high-speed
train). We assume that the satellite-to-vehicle link experiences the cloud coverage.
In the presence of clouds, a hybrid FSO/RF link using a hard-switching scheme is
©IEICE 2022
DOI: 10.1587/comex.2022XBL0104
Received June 25, 2022
Accepted July 8, 2022
Publicized July 15, 2022
Copyedited October 1, 2022
650
IEICE Communications Express, Vol.11, No.10, 649–654
Fig. 1. End-to-end network scenario with last-mile hybrid
FSO/RF satellite access in presence of clouds.
employed. The high-speed FSO link is considered a primary one, while the RF
link acts as the backup link in case of FSO link failure. If we denote Sas the link
selection for hybrid FSO/RF scheme, it is then expressed as [2, (17)]
S=
FSO, γFSO
th ≤γf<∞,
RF,0≤γf< γFSO
th ,
Outage,0≤γr< γRF
th ,
(1)
where γFSO
th and γRF
th are the signal-to-noise ratio (SNR) thresholds of FSO and
RF link, respectively. Additionally, γfand γrare respectively the instantaneous
received SNRs of FSO and RF link. As for the FSO link, cloud attenuation (hFSO
c)
and atmospheric turbulence (hFSO ) modeled by Gamma-Gamma distribution are
investigated, in which γf=ρ2PFT2h2
F
σF2. Here, ρis the detector responsivity, PFT
is the satellite’s transmitted power for FSO link, σ2
Fis the noise variance, and
hF=hFSOhFSO
c.As for the RF link, the cloud attenuation (hRF
c) and Nakagami
fading model (hRF) are considered, where γr=PRThR
σR2. Here, PRT is the satellite’s
transmitted power for RF link, σ2
Ris the noise variance, and hR=hRF hRF
c.
On the other hand, TCP CUBIC is employed at the transport layer to maintain
the reliable end-to-end connection between the TCP source (server) and end-user
(i.e., vehicles) over the Internet section and last-mile hybrid FSO/RF link. The TCP
throughput model incorporating both congestion loss and transmission errors can be
found in [2], in which the average TCP throughput was obtained in [2, (51)].
3 Cloud attenuation models
This section presents the attenuation models due to clouds for both FSO and RF links.
The clouds are characterized by their height, cloud droplet distribution (including
droplet size and droplet number concentration), and cloud liquid water content
(CLWC).
©IEICE 2022
DOI: 10.1587/comex.2022XBL0104
Received June 25, 2022
Accepted July 8, 2022
Publicized July 15, 2022
Copyedited October 1, 2022
651
IEICE Communications Express, Vol.11, No.10, 649–654
3.1 Cloud attenuation model for FSO link
The liquid water particles present in clouds cause the scattering phenomenon of laser
beam propagation. This leads to a decrease in visibility and significant attenuation
of signal power. To compute the cloud attenuation, it is necessary to determine the
visibility depending on the cloud characteristics. It is given as [5, (1)]
V=1.002
(L×Nc)0.6473 ,(2)
where L(g/m3) is the average CLWC, Nc(cm−3)is the cloud droplet number concen-
tration, and Ncfor the mono-dispersed distribution of droplets is found in [5, (2)].
According to the visibility range-dependent empirical model for the optical extinc-
tion coefficient for scattering, the cloud liquid water-specific attenuation coefficient
is estimated as [8, (9)]
ζ=3.91
V[km](λ
550 [nm])−q
[dB/km],(3)
where λis the optical wavelength, and qis the size distribution of the scattering
particles determined by using Kruse model as [8]. Based on the Beer-Lambert law,
the cloud attenuation on FSO link can be described as
hFSO
c=exp (−ζ
∆h
sin (θ)),(4)
where ∆his the vertical extent of the liquid cloud, and θis the satellite elevation
angle.
3.2 Cloud attenuation model for RF link
Contrary to the optical band, the wavelength of the RF link is often larger than the
typical size of cloud droplets. This results in less significant attenuation than the
FSO one. The small cloud liquid water droplet (i.e., vary from several micrometers
to hundreds of micrometers on cloud types) may cause scattering and significant
attenuation for backup radio frequencies greater than 10 GHz. In the case of cloud
liquid water droplets with size lesser than 0.01 cm, as recommended by ITU-P840 [4],
the Rayleigh scattering model is valid for frequencies up to 200 GHz, and the cloud
attenuation coefficient for RF link can be given as
ξ=L×0.819 f
ε′′(1+η2)[dB/km],(5)
where L(g/m3) is the average CLWC, fis the RF frequency carrier (GHz), and
η=2+ε′
ε′′ , with ε′and ε′′ are the real and imaginary parts of permittivity of water
expressed in [4]. From ξ, the cloud attenuation on RF link can be given as
hRF
c=10−ξ∆h
10 sin (θ),(6)
where ∆his the vertical extent of the liquid cloud, and θis the satellite elevation
angle.
©IEICE 2022
DOI: 10.1587/comex.2022XBL0104
Received June 25, 2022
Accepted July 8, 2022
Publicized July 15, 2022
Copyedited October 1, 2022
652
IEICE Communications Express, Vol.11, No.10, 649–654
4 Numerical results and discussions
This section presents several numerical results of TCP throughput performance
in hybrid FSO/RF-based satellite networks. Different cloud types are considered,
where their parameters are found in [8, Table I]. Monte Carlo simulations are also
performed to validate the analytical model. The available link margins for the cloud
attenuation of both FSO and RF links are 15 dBm, the liquid cloud vertical extent
∆h=2km, and the UAV altitude of 50 m. Other parameters used in the analysis can
be found in [2, Table 2].
4.1 Impact of cloud types on TCP performance
The impact of different cloud types in TCP throughput performance is illustrated in
Fig. 2. It is observed that the TCP throughput performance considerably deteriorates
for high CLWC values.
Fig. 2. TCP throughput performance versus CLWC for differ-
ent cloud types.
This demonstrates the severe impact of clouds on the performance of upper-
layer protocols. On the other hand, the achievable TCP throughput is different with
different cloud types. This is because different types of clouds have different liquid
particle sizes and droplet number concentrations, which determine the amount of
energy subtracted from the incoming FSO signals.
4.2 Impact of frequency bands on TCP performance
Next, we investigate the impact of the Nimbostratus cloud on the TCP throughput
performance in Fig. 3(a). Different optical wavelengths are taken into account. As is
evident, the higher the optical wavelengths are, the better achievable TCP throughput
is. This can be explained from Eq. (3), which implies that for any meteorological
conditions, there will be less cloud attenuation for the higher optical wavelength.
©IEICE 2022
DOI: 10.1587/comex.2022XBL0104
Received June 25, 2022
Accepted July 8, 2022
Publicized July 15, 2022
Copyedited October 1, 2022
653
IEICE Communications Express, Vol.11, No.10, 649–654
Fig. 3. TCP throughput versus CLWC for (a) different optical
wavelengths, (b) different RF frequencies.
Finally, Fig. 3(b) analyzes the TCP throughput performance with different RF
frequency bands over a range of CLWC values of the Nimbostratus cloud. It is
observed that the TCP throughput retains below 200 Mbps for high CLWC values,
i.e., CLWC >100 mg/m3. The reason is that when the CLWC increases, the FSO
link is significantly degraded and then becomes an outage for such high CLWC
values. In this case, the backup RF link operating at a much lower data rate is
activated, in which the relationship between TCP performance and RF frequency
bands is quantitatively illustrated in this figure. From this figure, we can observe
that for RF links using millimeter-wave bands (e.g., Ka, V, Q bands), the attenuation
due to clouds becomes significant. This leads to the degradation of TCP throughput
over RF-based last-mile access.
5 Conclusion
This paper has investigated the TCP performance over hybrid FSO/RF-based satellite
networks in the presence of clouds. In particular, the TCP CUBIC throughput
was analyzed under the impact of different cloud types. Also, different frequency
bands of both FSO and RF were investigated. The numerical results quantitatively
demonstrated the severe impact of clouds on TCP performance, especially with
satellite networks using optical bands (FSO link), and Ka, V, and Q bands (RF link).
Monte Carlo simulations were performed to validate the analytical derivation.
©IEICE 2022
DOI: 10.1587/comex.2022XBL0104
Received June 25, 2022
Accepted July 8, 2022
Publicized July 15, 2022
Copyedited October 1, 2022
654