Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/349505721

IAA-AAS-CU-20-01-05 AAS 20-205 TRAFFIC PREDICTION MODEL FOR


BROADBAND MICROSATELLITES CONSTELLATIONS

Article in Advances in the Astronautical Sciences · January 2020

CITATION READS

1 1,046

3 authors:

Roman Korobkov Petr Mukhachev


Skolteck Skolkovo Institute of Science and Technology
3 PUBLICATIONS 1 CITATION 8 PUBLICATIONS 4 CITATIONS

SEE PROFILE SEE PROFILE

Dmitry Pritykin
Moscow Institute of Physics and Technology
37 PUBLICATIONS 119 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Tilt-motor quadrotor control system View project

CubeSat Magnetic Attitude Control: In-Orbit Identification and Rejection of Residual Magnetization View project

All content following this page was uploaded by Dmitry Pritykin on 22 February 2021.

The user has requested enhancement of the downloaded file.


IAA-AAS-CU-20-01-05
AAS 20-205

TRAFFIC PREDICTION MODEL FOR BROADBAND


MICROSATELLITES CONSTELLATIONS
Roman Korobkov,* Petr Mukhachev,† and Dmitry Pritykin‡
This study is a part of a general constellation design project conducted in Skoltech
Space Center. The main objective of our work is to outline and validate a pro-
cedure to derive preliminary requirements for a microsatellite in a constellation to
provide global internet service. Thus, we consider an orbital motion of all satellites
within a given constellation pattern (e.g. Walker delta, Walker star, some flower
constellation line-ups, etc.), the service provided to potential global customers
(accounting for the satellite coverage geometry, world population density models,
and the internet traffic demand). The latter models are based on the 2020 NASA
Socioeconomic Data and Applications Center gridded population count map. For
more precise predictions, the following factors are taken into account for each
point in the satellite field of view: time (accurate to one hour), population, country
and internet penetration rate in that country respectively. Finally, given the esti-
mated required single-satellite link-budget, we proceed to analyze an inter-satellite
laser communication subsystem, which, in turn, drives the requirements to attitude
control and electrical power satellite subsystems.

INTRODUCTION
Broadband connectivity is widely recognized as a critical enabler of efforts to achieve the UN
Sustainable Development Goals.1 In 2019, the world crossed a number of major thresholds in
global internet adoption. The latest data estimate 21.7 billion connected devices – and this num-
bers are growing fast.1 On average in the year 2019 over 74,500 GB of data were sent over the
internet every single second.1 Growing demand for penetration and speed of the internet connec-
tion requires new ways to provide an access to the global web, and one of such ways is satellite
internet. First satellite internet projects2, 3 were proposed back in the 1990s, but in recent years
a number of mega-constellations were announced and are already being deployed. Federal Com-
munications Commission has received applications from 11 companies to deploy constellations in
non-geostationary satellite orbits with the purpose of providing broadband services. LEO satellite
constellations with global or near global coverage are no longer a far fantasy and are seriously con-
sidered as a complement to ground communication systems that may provide a backhaul for those
areas that have little or no internet infrastructure. Some of the LEO constellation projects – Star-
Link, OneWeb and Telesat to name a few – are even more ambitious than that and actually aim to
provide global broadband internet services.
It has been long stated that the performance of the upcoming satellite networks will strongly de-
pend on the performance of the data cross links to other satellites in the network or to other satellite

*
Space Center, Skolkovo Institute of Science and Technology, Moscow Oblast, Russia, 143026. E-mail:
r.korobkov@skoltech.ru.
† Space Center, Skolkovo Institute of Science and Technology, Moscow Oblast, Russia, 143026. E-mail:

p.mukhachev@skoltech.ru.
‡ Space Center, Skolkovo Institute of Science and Technology, Moscow Oblast, Russia, 143026 E-mail:

d.pritykin@skoltech.ru.

75
networks.4 The right choice of inter-satellite link (ISL) technology responsible for forwarding traf-
fic from one satellite to a neighbouring satellite may become a key factor for the whole constellation
commercial operation and the overall quality of its service. The performance of satellite networks
layouts used to be limited by the microwave technology of ISL, which did not allow high data rates.
Optical links which use laser light as carrier frequency can transmit much high data rates. It is now
universally accepted that optical inter satellite links are better suited to address the requirements of
transmitting dozens of Gbps at the distances of a few thousand kilometers and are offering a size,
weight and power advantage over traditional RF ISL technology.5–7
It is clear that with the development of the described complex systems there arises a need in
the design tools, allowing to assess the viability of the proposed solutions in the preliminary de-
sign stages. Our study is thus aimed at developing a tool that might be used for the first estimates
of the highly loaded and complex constellation projects within the frame of the concurrent design
approach.8 We propose a framework of models that help deriving the preliminary requirements
to a satellite, which is to be a part of a prospective broadband microsatellite constellation given
high-level requirements to the mission (in terms of customer distribution, development budget and
expected revenues). The high-level requirements can then be reformulated into mission design re-
quirements such as constellation orbital pattern, coverage, and vital subsystems parameters (ISL,
electrical power system (EPS, attitude control system, etc). These parameters can further be ana-
lyzed or even optimized to find a solution that meets the stakeholders expectations.
The paper has the following structure. First section presents the model that predicts the traffic
demand, i.e. is ground-to-satellite and satellite-to-ground link budget given the orbital configura-
tion of the designed constellation and coverage model parameters (such as the field of view of the
telecom instrumentation). The implemented routine processes the information on the orbital motion
of the multiple spacecraft comprising the constellation and matches the covered area to the Earth
population density and internet penetration models, thus providing a good qualitative distribution
of the traffic generated by the potential service consumers. Section two proceeds to describe and
parametrize the laser ISL subsystem deriving the link budget requirements. Section three provides
the estimates for the EPS subsystems connecting the previously derived requirements to the power
budget and estimating the spacecraft mass budget. Finally, a few use-cases are analyzed within the
proposed framework.

SATELLITE TRAFFIC PREDICTION MODEL

In order to build satisfactory traffic prediction model, one needs to consider orbital motion prop-
agation which defines Earth area coverage, population density model, as well as internet demand
model.

Orbital motion propagation

We consider a Walker delta constellation, which is defined by the following notation: i : T /P/f ,
where i is orbit inclination, T is the total number of satellites, P is the number of orbital planes, f
is the phasing parameter. A particular example that we used as a baseline and reference example is
53◦ : 1584/24/6. The constellation (see Figure 1) is deployed at altitude h = 550 km.
Since all satellites in the constellation move along the orbits of the same altitude and inclination,
it is sufficient to propagate the orbit for any particular satellite and then change all other spacecraft
positions in accordance with the constellation design pattern. The orbital motion model of a single

76
Figure 1. StarLink Constellation

satellite requires considering an object in a circular geocentric orbit of radius R and inclination i.
The Earth’s oblateness causes the orbit’s precession with angular velocity
1/2
3J2 μG Ro2
nΩ ≈ cos i. (1)
R7/2
where Ro = 6378.245 km is the Earth’s mean equatorial radius, μG = 3.986 · 105 km3 /s2 is the
gravity parameter of the Earth, J2 = 1.082626 · 10−3 is the first zonal harmonic coefficient in the
expansion of the Earth’s gravity field.
Argument of latitude u is a linear function of time:
u̇ = ωD , (2)
where ωD = 2π/TD , TD is the draconic period of an object’s revolution around the Earth (the
time between two consecutive passages through the ascending node). Employing the formula for
draconic period we obtain:
  2 
3 Ro 2
ωD = ωo 1 − J2 (1 − 4cos i) , (3)
2 R

where ωo is the mean motion for the circular orbit of radius R in the central gravity field with
parameter μG .

77
All other perturbations, such as higher-order harmonics of the Earth’s gravity field, atmospheric
drag, Solar radiation pressure, etc., are not taken into account in this work.
Having computed the spacecraft’s argument of latitude u and right ascension of ascending node
Ω (RAAN) from the equations (1)-(2), and assuming the rest of the orbital elements to be known
constants, we obtain the spacecraft’s position RECI (t) with respect to Earth-centered inertial (ECI)
reference frame for any given time. Then a gmst-based9 rotation matrix is employed to transform
the position vector to the Earth-centered Earth-fixed (ECEF) reference frame:
RECEF (t) = R3 (θgmst ) · RECI (t),
where R3 is the z-rotation matrix, and θgmst is the Earth rotation angle.
According to Bradley10 using gmst-based transformation can lead to position errors of 20–30 km
on the level of the Earth’s surface, however it stands to reason that for a qualitative study we do
not require a more precise conversion routine, as such errors will average out (speaking in terms
of average traffic per satellite) for longer simulation periods. It may also be worth noting that it is
sufficient to compute the position of any chosen single satellite in the constellation according to the
outlined procedure, because we assume that the constellation moves as a rigid body and the position
of all other satellites can be obtained by translating the first satellite’s position over the values of
RAAN and argument of latitude that are defined by the constellation pattern.
Thus, knowing RECEF (t) = [X Y Z]T for each spacecraft, we also know the position of
each spacecraft’s subsatellite point and can proceed with the constellation coverage analysis. Geo-
centric latitude and longitude of subsatellite point are given by:
   
Z Y
ϕgc = arctan √ , λgc = arctan . (4)
X +Y2 2 X

Coverage model
We assume that all satellites in the constellation are equipped with identical antennas that have
the same field of view (FOV). The coverage model assumes spherical shape of the Earth and the
footprint of each antenna is obtained as a part of the Earth’s surface within the inner part of the right
circular cone, whose origin is at the position of the satellite in its orbit, symmetry axis goes through
the subsatellite point on the Earth, and the aperture corresponds to the antenna’s FOV α = 56◦ (see
Figure 2, α is the half-beam angle of satellite antenna). The Earth central angle σcr , whose value
serves as a limiting condition to find all the points on the Earth surface that fall into the instantaneous
footprint of the antenna, the following formula is used:
 
π (Ro + h)
σcr = − arccos sin α − α. (5)
2 Ro

Every point on the Earth’s surface is characterized by its latitude ϕgc and longitude λgc , which
can be used along with the central angle formula (Equation (6)) to calculate the angular distance σ
from the subsatellite point (see Figure 2):
σ = arccos (sin ϕsp sin ϕgc + cos ϕsp cos ϕgc cos (λgc − λsp )), (6)
where ϕsp and λsp are the latitude and longitude of the subsatellite point, respectively.
Finally, the condition that a certain point belongs to the area on the Earth’s surface covered by
the FOV is:
σ ≤ σcr . (7)

78
Figure 2. Relationship between satellite FOV angle and Earth central angle

Earth maps
To generate the global internet traffic demand map we use two data sets provided by NASA’s
Socioeconomic Data and Applications Center — National Identifier Grid and Population Count.
Population Count (Figure 3) data set estimates the human population (number of persons per pixel),
consistent with national censuses and population registers, for the year 2020.11 National Identifier
Grid (see Figure 4) is a raster surface that uses year 2010 census data, where pixels that represent
the same source of census data have equal values.12

Figure 3. Population count

Population count map allows to determine population in specific point on Earth (one pixel on the
map), this value is represented by ω coefficient in further data volume calculations.

Figure 4. National identifier grid

Both datasets are represented by raster pictures in GeoTIFF format with WGS84-based coordi-
nate system and 30 arc-second resolution, which corresponds to approximately 1 km at equatorial

79
latitudes. SEDAC data are provided with respect to geodetic latitude and longitude, which are com-
puted from the geocentric latitude and longitude (Equation (4)) with the use of an iterative method.13
Internet penetration rate. Number of internet users and the amount of traffic varies from country
to country because of different levels of internet infrastructure development. Considering that at any
time several countries may lie within FOV of a single satellite and, furthermore, even neighboring
countries may have noticeable difference in data consumption, we cannot assume average per capita
internet consumption model. Thus, we must take into account which specific countries are covered
by the satellite’s FOV. For this reason, we invoke the National Identifier Grid to determine the
countries that are within the satellite’s FOV. In addition to that we include into our model the data
set of internet penetration rates for all existing countries to obtain the portion of each country’s
population that actually uses the internet services.∗
This dataset is represented by a table with country names, country ISO 3166-1 numeric codes and
internet penetration index (in percents). Together with the National identifier grid, which provides
these numeric codes it yields the internet penetration index for any given point on Earth. This value
allows to compute the internet-penetration map (Figure 5) and calculate internet penetration rate
coefficient η as
Internet penetration rate
η= . (8)
100

Figure 5. Internet penetration index (in percents) map

Time zones. Internet data consumption level significantly depends on the local time. Thus, we
need to take into account the local time in the points inside satellite FOV. We then proceed to use the
distribution of total minutes spent in the Internet during the day (Figure 6(a)) to understand which
portion of the daily traffic is generated at a specific hour.† The information is given by the hour, and
it will be further assumed that within each hour the local distribution remains the same. Thus for
the integration step Δτ , the expression for current local time coefficient φ is following:
Δτ
φ= Φ, (9)
3600
where Φ is a portion of daily internet traffic of an average internet-user at specific hour.

https://www.internetworldstats.com/list2.htm

https://www.statista.com/statistics/538789/hourly-social-media-app-usage-usa

80
(a) Distribution of total minutes spent on the internet (b) Data volume of global consumer IP traffic

Figure 6. Time distribution and data volume

Next, we use global data volume of consumer IP traffic (Figure 6(b)) for the year 2020 to calculate
what amount of internet traffic is generated by one internet user in a day.∗ In subsequent data volume
calculations, this value is represented by daily data volume of one internet user ξ given by:

DG
ξ= , (10)
30P
where DG is the global data volume per month and P is the number of internet users in the world.
This information combined, provides the means to estimate the amount of internet traffic at any
given hour and location.

INTER-SATELLITE COMMUNICATION SYSTEM


The laser communication equation is very similar with the link equation for any RF communica-
tion link and can be written as:14

Pr = Pt Lt Gt LP t LR LP r Gr Lr , (11)

where
Pr — the received power
Pt — the transmitted power
Lt — the efficiency of transmitter
Gt — the transmitter antenna gain
LP t — pointing loss of the transmitting beam
LR — free space loss
LP r — pointing loss of the receiver’s optical antenna beam
Gr — the receiver antenna gain
Lr — the efficiency of receiver


https://www.statista.com/statistics/267202/global-data-volume-of-consumer-ip-traffic

81
Free space loss
The laser beam wave front diverging as it travels through space, which causes link range loss.
Classical link range loss equation is following:
 
λ 2
LR = , (12)
4πR
where λ is the wavelength, R is the distance between transmitter and receiver.
Since the Cartesian coordinates were calculated on Orbital motion propagation step, we can use
following formula to find distance R:

R = (Xt − Xr )2 + (Yt − Yr )2 + (Zt − Zr )2 , (13)

where Xt , Yt , Zt are the coordinates of transmitter, and Xr , Yr , Zr are the coordinates of receiver.

Antenna gain
We assume that satellite use same telescope for transmit and receive data, therefore antenna gain
is equal for both satellites and calculation is given by:
 
πDt/r 2
Gt/r = , (14)
λ
where Dt/r is the aperture diameter of telescope (see Figure 7) and λ is the wavelength.

Figure 7. Scheme of satellite link

Pointing losses
The signal losses due to incorrect pointing can be culculated using following equations:

LP t = exp (−Gt θt2 ), (15)

LP r = exp (−Gr θr2 ), (16)


where Gt is the antenna gain and θ is the radial angle error.
Since the main goal of our calculations is to derive related requirement for satellite power and at-
titude determination and control systems, we assume that minimum power of the laser beam will
occur on the receiver when pointing error θ approaches to the half of the optical beamwidth Θ (see

82
Figure 7). Therefore, in our calculations we use θ = 0.5Θ.

We have minimum required number of photons on receiver n and data rate DR, therefore mini-
mum energy that required on receiver to detect all stream of information is:

hc
E0 = nDR, (17)
λ
where c is the speed of light, λ is the lasers‘s wavelength and h is plank‘s constant.
Assuming that Pr must be M times greater than E0 and taking into account Equation (11) we obtain
the following equation for minimum transmitting power:

E0 M
Pt = , (18)
L t Gt L P t L R L P r G r L r
where M is a margin, E0 is the minimum required energy on the receiver, rest like in Equation (11).

ELECTRICAL POWER SYSTEM


Assuming uniform power consumption of throughout the cycle, required end of life (EOL) battery
capacity and required EOL solar array area can be estimated using well-known method described
by Wertz:15

Tecl Pav
Ceol = , (19)
 DOD 
Pav Tsun Pav Tecl 
Ssa,eol = + Tsun , (20)
ηsun ηecl

where ηsun and ηecl are electrical power system efficiencies during sunlight and eclipse phases
respectively, Tsun and Tecl are corresponding times.
Assuming Li-Ion battery technology, we can take into account battery degradation throughout
active satellite life time (5 years) and estimate required depth of discharge to be 5%.16, 17 Solar array
degradation calculation is more straightforward and can be approximated by formula Ld = (1 − d)l ,
where d is year degradation and l is lifetime. For gallium-arsenide cells in LEO, the degradation
is about 0.0275 per year, which gives required area in the beginning of life to be Sbol = Seol Ld .
System weight is calculated in a straightforward way, assuming specific masses for solar array and
batteries to be as specified in Table 1.
According to statistical mass distribution between subsystems described by Wertz,15 typical EPS
weight amounts around 30% of the whole satellite. Thus, total weight could be roughly approxi-
mated by Msat = 3Meps .

PROCESSING
Thus, the process of predicting the amount of traffic that satellite can uplink for one subsatellite
point consists of following steps:

1. Calculate satellite position using orbital mechanics from Kepler‘s orbit parameters.

2. Find geocentric latitude and longitude using Equation (4).

83
3. Using iterative method calculate geodetic latitude, thus finding coordinates of subsatellite
point.

4. Using Equation (5) and (6) find geodetic coordinates of all points inside satellite FOV.

5. For all points found, determine Country this point belongs, local time and population.

6. For parameters from previous step find corresponding coefficients (Internet penetration rate,
percent of daily internet data volume at current hour) in related data sets and using Equa-
tions (8), (9), (10).

7. Calculate data volume for current subsatellite point using following relation:


n
D= ωi ηi φi ξi , (21)
i=0

where
D — data volume;
ω — population;
η — internet penetration rate coefficient;
φ — current local time coefficient;
ξ — daily data volume of one internet user;
n — amount of points inside satellite FOV;

8. Using calculated data volume and, according to number of inter-satellite links, obtain data
rate for each link;

9. Using satellite coordinates, fine distances to connected satellites;

10. Calculate link power budget using Equations 12-18;

11. Calculate overall satellite power budget with Equations 19 and 20;

12. Calculate approximate satellite weight.

The full scheme of simulation process is illustrated in Figure 8.18

SIMULATION RESULTS
Each satellite node in a constellation is usually assigned four satellite links:19 two in-plane links
between the adjacent satellites located in the same orbit plane and two inter-plane links between
the neighboring satellites located in the left and right orbital plane (see Figure 9). Therefore, to
calculate minimum required power for one satellite, we need to calculate link power budget for
each of these links and sum up the results.
For our simulation we use initial values that represented in Table 1.
We assume that the outcoming data volume obtained from the traffic prediction model is divided
equally between these 4 satellites, therefore data rate for each link is twice the data volume for the
central satellite. Distances to the adjacent satellites are calculated on using Equation 13. Applying

84
Figure 8. Simulation process

Table 1. Initial values


Parameter Value Unit Parameter Value Unit
n 133 ηsc 0.28
λ 780 nm kbat 125 Wh/kg
M 5 dB ksc 2.8 kg/m2
Gt/r 10 cm
Lt/r 0.8

data rate and distances at a certain point in orbit to Equation 11 we obtain the power consump-
tion required to transmit information to the adjacent satellites with the corresponding data rate and
distances.
To show how the developed model can be used for constellation analysis, we describe two cases.

Case 1
For example we want to know how much power the inter-satellite communication system will
consume if the constellation is aimed to replace regular internet services. In this case we do not
create additional constrains like limit of maximum uplink/downlink speed. After the simulation for
20 orbits is over we obtain results that are shown in the Figures 10 and 11.
Simulation allows to determine power consumption of the inter-satellite communication system
and design power supply system according to these results. But to make assumption about power

85
Figure 9. Connections in constellation

Figure 10. Data volume for 20 orbits

Figure 11. Instant required transmitter power

supply system, we need to take into account power that is required for downlink. Using same
equations that is for inter-satellite link, but with different initial values we obtain required power
for downlink. Figure 12 illustrates power consumption of downlink and inter-satellite link systems

86
together. Calculations were made with assumption that half of the data volume is for downlink, and
other half is for inter-satellite communications.

Figure 12. Communication system power consumption

Case 2-4

According to Figure 10, uplink/downlink speed should be up to 40 Tbit/s, which is far away
from the value of this parameter in existing and proposed solutions of satellite internet. For more
realistic approach we can decrease satellite antenna FOV, but it will lead to increasing of number of
satellite up to 15 thousands, or we can limit uplink/downlink speed with 50 Gbit/s and loose some
consumers. In this case the results for data rate and power consumption are shown in Figures 13, 14.

Figure 13. Data rate with narrowed down FOV and limited speed

Figure 14. Instant required transmitter power with limited speed

87
Figure 15. Communication system power consumption with limited speed

And power consumption for communication system is represented in Figure 15. Free-space loss
for laser link is largely determined by the distance between satellites, which varies with sinusoid-
like shape. That results in the same shape of lasercom energy consumption timeseries, which could
be clearly seen on Figure 15.
Parameters of the constellations for cases 1-4 as well as StarLink constellation parameters are
shown in Table 2 for comparison.

Table 2. Case scenarios comparison


Variable Case 1 Case 2 Case 3 Case 4 StarLink Unit

FOV constraint 20 30 40 56 56
Number of satellites 6500 3200 1600 432 1584
Average power required 250 615 1400 5100 N/A W
EPS mass 30 75 170 625 N/A kg
Satellite mass (est.) 90 250 500 1900 270 kg

It is seen that 40◦ FOV is enough to provide full coverage, however StarLink uses 56 in order to
provide distribution of peak traffic between different satellites. Other parameter incompatibility is
considered due to extremely low fidelity of employed weight models.

CONCLUSION
A basic procedure for delivering preliminary technical requirements for LEO telecommunica-
tion satellite constellation was presented in this work. Preliminary models for the most important
subsystems were used to conduct constellation preliminary design study, including laser telecom-
munication for inter-satellite communication as well as electrical power system. As a case study
a comparison for four different scenarios was conducted based on different constraints on FOV of
each satellite, which essentially determines total number of satellites to provide full coverage. The
model is considered sufficiently stable for using subsystems models of higher fidelity.

REFERENCES
[1] State of Broadband Report 2019. Geneva: International Telecommunication Union and United Nations
Educational, Scientific and Cultural Organization, 2019.
[2] I. d. Portillo, B. G. Cameron, and E. F. Crawley, “A technical comparison of three low earth orbit
satellite constellation systems to provide global broadband,” Acta Astronautica, Vol. 159, 2019, pp. 123
– 135.
[3] A. K. Majumdar Optical Wireless Communications for Broadband Global Internet Connectivity, Else-
vier, 2019.

88
[4] K. Pribi, “Optical Crosslinks for Broadband Satellite Networks,” Mobile and Personal Satellite Com-
munications 3 (M. Ruggieri, ed.), London, Springer London, 1999, pp. 290–300.
[5] M. Motzigemba, H. Zech, and P. Biller, “Optical Inter Satellite Links for Broadband Networks,” 2019
9th International Conference on Recent Advances in Space Technologies (RAST), June 2019, pp. 509–
512, 10.1109/RAST.2019.8767795.
[6] H. Kaushal and G. Kaddoum, “Optical Communication in Space: Challenges and Mitigation
Techniques,” IEEE Communications Surveys Tutorials, Vol. 19, Firstquarter 2017, pp. 57–96,
10.1109/COMST.2016.2603518.
[7] O. Ivanenko, A. Kharlan, and V. Teplyakov, “Optical intersatellite laser communication link for Low-
Orbit communication satellite systems,” Vol. 5, 2015, pp. 3662–3668.
[8] A. Kharlan, N. Veliev, S. Biktimirov, N. Mullin, and A. Ivanov, “A university-based facility for evalua-
tion and assessment of space projects,” Vol. 2018-October, 2018.
[9] D. A. Vallado and W. D. McClain, Fundamentals of astrodynamics and applications, Vol. The space
technology library. Microcosm Press, 4th ed ed., 2013.
[10] B. Bradley, A. Sibois, and P. Axelrad, “Influence of ITRS/GCRS implementation for astrodynamics:
Coordinate transformations,” Advances in Space Research, Vol. 57, No. 3, 2016, pp. 850 – 866.
[11] Center for International Earth Science Information Network - CIESIN - Columbia University,
“Gridded Population of the World, Version 4 (GPWv4): Population Count, Revision 11,”
https://sedac.ciesin.columbia.edu/data/set/gpw-v4-population-count-rev11, 10.7927/H4JW8BX5.
[12] Center for International Earth Science Information Network - CIESIN - Columbia University,
“Gridded Population of the World, Version 4 (GPWv4): National Identifier Grid, Revision 11,”
https://sedac.ciesin.columbia.edu/data/set/gpw-v4-national-identifier-grid-rev11, 10.7927/H4TD9VDP.
[13] T. Kelso, “Orbital Coordinate Systems, Part III,” Satellite Times, Vol. 2, No. 3, 1996, pp. 78–79.
[14] D. G. Aviv, Laser Space Communications. ARTECH HOUSE, Inc., 2006.
[15] J. R. Wertz, D. F. Everett, and J. J. Puschell, Space Mission Engineering: The New SMAD. Microcosm
Press, 2011.
[16] R. V. Bugga, M. C. Smart, and L. Whitcanack, “Storage Characteristics of Lithium-Ion Cells,” Vol. 25,
No. 36, pp. 297–306, 10.1149/1.3393865.
[17] S. S. Choi and H. S. Lim, “Factors That Affect Cycle-Life and Possible Degradation Mechanisms of a
Li-Ion Cell Based on LiCoO2,” Vol. 111, No. 1, pp. 130–136, 10.1016/S0378-7753(02)00305-1.
[18] D. J. Pate, J. Gray, and B. J. German, “A Graph Theoretic Approach to Problem Formulation for Mul-
tidisciplinary Design Analysis and Optimization,” Vol. 49, No. 5, pp. 743–760, 10.1007/s00158-013-
1006-6.
[19] Q. Xiaogang, M. Jiulong, W. Dan, L. Lifang, and H. Shaolin, “A survey of routing techniques for
satellite networks,” Journal of Communications and Information Networks, Vol. 1, No. 4, 2016, p. 66.

89

View publication stats

You might also like