PaddleSats: Attitude Control and Station-Keeping

for Ultra-Low Density SSP Satellites

2023 IEEE International Conference on Wireless for Space and Extreme Environments (WiSEE) | 979-8-3503-3870-6/23/$31.00 ©2023 IEEE | DOI: 10.1109/WiSEE58383.2023.10289225

Vaibhav Bhosale§ , Jonathan Dolan§ , Grishma Kalepu§ , Deeksha Manjunath§ , Gregory Durgin
Georgia Institute of Technology
Atlanta, USA
{vbhosale6, jdolan30, gkalepu3, deekshamanjunath, durgin}@gatech.edu

Abstract—‘PaddleSats‘ represent a unique class of Space Solar In this paper, we explore the use of Paddlesats, a class
Power (SSP) satellites distinguished by their ultra-low area den- of large, thin space solar power satellites made up of two
sity and distinctive design, featuring two circular disks—a solar ultra-low area density disks – one for capturing solar energy
collection disk and a microwave transmission disk—connected
by a cylindrical joint. This paper introduces a comprehensive and the other to transmit this captured power back to Earth.
framework and presents initial results for an advanced Paddle- This area density can be as low as 1.5kg/m2 [2], [3]. This
Sat attitude control algorithm. The primary objective of this class of satellites experiences non-Keplerian astrodynamics
algorithm is to emulate the behavior of traditional geostationary due to significant effects of solar radiation pressure (SRP) [4]
satellites, particularly their station-keeping procedures while owing to their low area density. We assume the deployment of
adhering to the specific requirements for efficient microwave SSP
transmission. Our results show that the PaddleSat attitude con- these satellites in the geostationary orbit and try to answer the
trol algorithm successfully achieves the desired station-keeping question of whether these satellites can be deployed without
behavior, effectively balancing the demands of stable positioning any additional fuel required for station keeping. This can
with the unique requirements of microwave SSP transmission. reduce the weight of the satellite in turn leading to lower
These findings highlight the potential of PaddleSats as a viable launch costs, and potentially, increase the lifetime of the
and efficient means of harnessing solar power in space.
Index Terms—Paddlesats, Space Solar Power, Control Algo-
satellite in orbit.
rithms, Station Keeping, Solar Radiation Pressure Through simulations, we try to understand the impact of
external forces such as solar radiation pressure (SRP) on
the satellite. We started with constantly pointing the solar
I. I NTRODUCTION
disk towards the sun in order to maximize the total energy
Solar power is considered the most efficient source of captured and observed the impact of the SRP on the orbit. The
renewable energy. While the Earth receives enough radiation satellite accelerates while it is traveling along the direction of
in a minute to cater to the annual energy requirement of the the pressure and decelerates while traveling in the opposite
entire population, most of the energy is lost in the atmosphere direction. This helped us determine ways in which we could
in the form of diffused radiation. An extension to this area manipulate the effects of SRP on the orbital dynamics to
was proposed by Peter Glazer in 1968 [1] in his key patent to provide station-keeping control authority.
use space solar power satellite (SPS) systems. The key idea is We are able to show how a PaddleSat can be deployed in
to capture the solar power in space and transmit this energy the geostationary orbit without the need for additional station-
to Earth. keeping fuel. The rest of our paper is organized as follows:
The SPS systems absorb the solar energy using the PV In Section II we take a look into the major related works
panels of the satellites and convert this energy to microwave about solar radiation pressure, some of the existing techniques
signals. Microwave signals can be transmitted to the Earth for station keeping, and some control mechanisms to perform
with much better efficiency than the radiation itself. The station keeping with just the solar pressure. Section III explains
energy conversion techniques for the SPS systems are the same the system model and problem formulation. We discuss the
as the PV panels - photovoltaic design and thermal-electric objectives and our exploration of the controller design problem
conversion. The performance of the SPS systems also depends in Section IV. Simulation results in Section V show the
on various other parameters other than the energy conversion. efficacy of our solution. Section VI concludes the paper with
Solar radiation pressure, orbital perturbations due to the moon, pointers to future research.
Sun, and Jupiter, microwave radiation recoil pressure, the
ellipticity of the Earth’s equatorial plane, rotary joint friction II. R ELATED W ORKS AND M OTIVATION
torque, magnetic field interactions, and aerodynamic drag are Station-keeping is a very important task that needs to be
some factors that may alter the orbit and hence, the absorption performed by satellites in order to correct the effect of vari-
rate of solar energy. ous orbital perturbations. For geostationary satellites, station-
keeping requirements typically place a bounding box on the
§ Equal Contribution satellite sub point (SSP) drift in the latitudinal (north-south)

979-8-3503-3870-6/23/$31.00 ©2023 IEEE 41 WISEE 2023

Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on November 26,2023 at 04:34:07 UTC from IEEE Xplore. Restrictions apply.
and longitudinal (east-west) directions to avoid interference or authors claim to have addressed this challenge to a great extent
collisions with other satellites. using the controller design proposed. The paddles of the above
Traditional geostationary orbital station-keeping relies on two systems require the presence of solar radiation pressure
small thrusters to correct orbital perturbations. Thrusters used for the spinning of the satellite. Most research in the 1970s
for station-keeping are usually considered sources of instanta- had addressed the control challenge of the satellites for both
neous thrust. The force they impart on the spacecraft is a short spinning and non-spinning satellite systems.
duration in relation to the period of the orbit. Any inaccuracies Satellite systems are nonlinear systems that are associated
in the orbital maneuver caused by misaligned thrust vectors with numerous uncertain parameters such as the pitch and yaw
and non-instantaneous duration are measured and corrected angles, and eccentricity in the orbit which serve as essential
by subsequent station-keeping burns [5]. Station-keeping algo- parameters for attitude-control and SRP counter-thrust. These
rithms allow orbital drift within a longitude-latitude bounding parameters though crucial for controlling the satellites are less
box defined by regulation and mission parameters. The SSP is likely to be available to run control simulations. This issue was
allowed to drift slowly between the extremes of the bounding addressed by Lakshmi [9] through an estimator design for the
box. Prior to violating the bounding box, the station-keeping input parameters and the output was tested using three control
thrusters are fired to correct the drift. This sends the SSP algorithms. The nonlinear adaptive control proposed uses the
drifting slowly toward the other extreme. The frequency at inverse control law and a high gain estimator to project a
which such station-keeping maneuvers must be performed trajectory for a satellite system. While the controller gives
varies based on orbital altitude, spacecraft characteristics, solar feedback and control to the system, it is highly susceptible
weather, and N-body perturbations. The satellites in the NASA to noise. The uncertainty of the assumed parameters and the
GOES constellation perform east-west station-keeping maneu- non-linearity of the system are cited to be the origin of the
vers about once every 3 months, and north-south maneuvers noise in the controller output. A finite time controller with
about once per year [6]. This on-off thruster control is a form estimated inputs from a differentiator and a higher-order slide
of bang-bang control that is optimal for geostationary station- mode controller gives a more practical approach to the control
keeping. problem. The controller simulation for these inputs has lower
Most satellite systems use a thruster to control their attitude, noise susceptibility and thus a more efficient solution to the
the thrusters pose a strain on the weight and the fuel require- control design.
ment of the satellite. Glaser’s proposal [1] of the Solar Power The development in satellite technology and the control
satellites, which use Solar radiation pressure as a thrust force, laws has addressed both the weight constraint of solar power
had revolutionized the satellite control industry. In the 1970s, satellite systems and their attitude control. Low-area density
various designs had been proposed to harvest the power and satellites have been a point of research since the Glaser
also use it for attitude-control. One such model for spinning proposal [1] but the area density of the satellites has been
satellites is the design proposed by Crocker [7]. The author reduced to the record minimum only very recently [2]. While
proposes a cylindrical arrangement of solar panels that is the SPS systems are the optimal solution to the weight and
deployed with a spinning axis pointing toward the Sun. The volume problems, they pose a huge challenge to attitude
SRP force acts on the solar paddles attached perpendicular to control as they are more susceptible to perturbations due to
the body. These paddles, according to the author, when aligned SRP. Brunett and Schaub [10] propose an attitude controller
such that angle between the plane of the spin regulator paddles for a single-plate satellite. The controller is designed for an
and the plane of the main paddles is 35 degrees, the satellite SRP model with relative motion. The uncertainties in the
would not require any electrical thrusters to drive the satellite orbital input parameters are accommodated in the controller
in the orbit. Though the mathematical formulation of the model which led to a more comprehensive and realistic set of
system proves to be a feasible design, numerous assumptions orbital results. This system uses a Linear Quadratic Regulator
are to be addressed like the movement of the Sun in an Earth- to counteract the SRP component on the satellite. But like the
centred model, torque vectors due to the gravitational pull of former controller, certain input parameters are assumed.
the Earth, and the position of the damper in the satellite and While the above models answer the question of optimal
its angular movement vector. control for station-keeping, new designs are being proposed
Some of these assumptions were addressed by Modi and and simulated for better results.
Brereton [8] through their simulations of the same model using
more real-time parameters. The controller design follows the III. S YSTEM M ODEL
same principle of altering the angle of the solar paddles to
maintain the spin along the spin axis. The angular momentum In this paper, we consider PaddleSats with two identical
of the satellite is concluded to be a variable parameter since circular discs deployed in the geostationary orbit. One of
the spin-axis is not assumed to be cyclic. The controller the discs is a solar panel whereas the other one is a high-
design has, to a great extent been successful in positioning gain antenna that is assembled with a 2-degree-of-freedom
the satellite spin-axis to the desired normal. While a large cylindrical joint (Fig 1). Since our goal for the system is to
satellite system would have a better control capability, smaller function by just using SRP, the overall weight of the system
satellite systems are intuitively less controllable systems. The is greatly reduced.

42

Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on November 26,2023 at 04:34:07 UTC from IEEE Xplore. Restrictions apply.
 relative to the sun. α = 0 indicates the panel is normal to
the sun and generating max power. Conversely, α = 90.0 or
α = −90.0 yields no power generation. The communication
dish is assumed to always point nadir. The controller assumes
the PaddleSat joint responds to commands instantaneously and
requires no power to move.

IV. C ONTROLLERS
Unlike traditional thruster-based geostationary orbit station-
keeping, SRP-based station-keeping relies on a small but
continuous solar pressure force applied over a large area. The
design of a PaddleSat attitude controller attempts to maximize
ground station received power by controlling the angle of the
Fig. 1: A schematic of the paddlesat design consisting of the solar panel relative to the sun.
two circular panels - communication and solar connected by Ground station received power is expressed as a function
cylindrical hinge joint. of the instantaneous solar power generation efficiency s,
transmission efficiency t, and satellite peak power generation
P in Watts,
The solar panel has a reflectivity of 0.21 and the antenna
power = s ∗ t ∗ P (3)
has 0.3 [11], [12]. We consider the effects of the SRP on both
the equivalently sized solar panel and communications dish where we use the value of P 1W which allows easy compar-
with their respective reflectivity coefficients. We use the solar ison between controller efficiencies.
pressure value of 4.56 µN/m2 [12] as observed at a distance of Solar power generation efficiency is a function of the
1 astronomical unit (AU). We assume a geostationary satellite commanded incidence angle of the solar panel and can be
orbit where the orbits of the sun, earth, and satellite are expressed as s = cos(α). Transmission efficiency is a function
coplanar. The SRP force experienced by the PaddleSat will of the antenna gain pattern and the current SSP drift from the
therefore be in the orbital plane, yielding orbital perturbations ground station. The communications dish is designed to have
that only influence the longitudinal east-west drift of the SSP. a parabolic fall-off to a half-power beam width equivalent to
This setup results in the forces acting on the satellite as the size of the east-west bounding box. The PaddleSat is not
follows: allowed to transmit if the current SSP drift is beyond the east-
• Gravitational Pull by the earth west bounding box. Therefore, the transmission efficiency is
a piecewise function as shown in Fig 2.
GME MS
F =− r̂ (1)
r2
• and from [12], the Solar Pressure force is
F = −P cos(θ)A[(1 − ϵ)ê + 2ϵcos(θ)n̂]. (2)
where P is the solar radiation pressure, A is the area of
the panel, θ is the angle of incidence, ê is the direction of
the solar radiation pressure, and n̂ is the normal direction
of the panel.
With a lot of GEO satellite infrastructure already in space,
the satellite service providers and engineers must abide by cer-
tain station-keeping requirements. For example, the ITU levies Fig. 2: Comm transmission efficiency function for max drift
strict requirements on the orbital drift of satellites in the fixed- of ±0.1 deg
satellite service to ±0.1 degree of east-west drift. Other geosta-
tionary satellites, such as those in the GOES constellation op- We designed and tested a wide spectrum of controller
erated by National Oceanic and Atmospheric Administration designs. The most simple controller is the static incidence
(NOAA), maintain ±1 degree longitude boxes [6]. PaddleSat that commands the solar panel to continuously track the sun
is assumed to have strict east-west drift requirements similar to at a given α offset. This controller requires active solar panel
that of the fixed-satellite service due to communications design alignment throughout the orbit. The static˙incidence(α = 0.0)
parameters and safety. Therefore, PaddleSat control algorithms controller maximizes power generation throughout the or-
assume a ±0.1 degree east-west bounding box from nominal bit. Conversely, the static incidence(α = 90.0) controller
ground station position. does not generate any power. Interestingly, the orbit of
Constrained to planar motion, the solar incidence angle α, static incidence(α = 90.0) controller is still perturbed by the
valid in the range [−90, 90], is the tilt of the solar panel effects of the SRP on the communications dish.

43

Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on November 26,2023 at 04:34:07 UTC from IEEE Xplore. Restrictions apply.
 The zenith controller varies the angle of the solar panel such
that it is always pointing nadir by setting the commanded angle
to equal the orbital angle. This controller requires no active
solar panel alignment as it is the natural way the solar panel
changes orientation throughout an orbit.
The zenith off 90 controller varies the angle of the solar
panel offset such that it is offset 90 degrees from the current
orbital angle. Conversely, the negative˙zenith off 90 controller
performs a -90 degree phase shift. Both these controllers
require active solar panel alignment throughout the orbit.
The clipped zenith off 90 controller has the same operation
as the zenith off 90 controller except the output is clipped to
a specified angle range. The output of the controller will never
exceed +angle and never be below -angle.
The smart controller is designed to maximize power gen-
eration while maintaining east-west station-keeping based on
the results of the simulations presented in Section V.
V. S IMULATION E XPERIMENTS AND R ESULTS
A. Simulation Setup
For our simulations, we incorporate the gravitational force
along with the force exerted by the solar radiation pressure. We Fig. 3: 30 days of static incidence(α = 0.0)
perform our analysis in an earth-centered frame with the earth,
sun, and satellite lying in the same plane. Along this plane,
the earth rotates around itself whereas the satellite revolves
around the earth with the time period being one sidereal day.
We also move the sun around the earth with a time period of
one sidereal year. For almost all our simulations, we assume
the area density of the Paddlesat as 5kg/m2 .
We run our simulations for a total period of 30 days studying
a wide variety of controller scenarios. For each such scenario,
we show the orbital diagram, the longitude value of the
satellite sub-point, the power received by the ground station,
and the power transmitted by the communication panel. For
the power being transmitted, since our goal is to study the Fig. 4: 30 days of static incidence(α = 60.0)
differences caused by the drift, we ignore free space path loss
instead focusing on the trends observed in the absolute values
of the power transmitted.
B. Analysis
The static incidence controller provided insight into how
the orbit of PaddleSat behaves with SRP effects on the com-
munications dish and solar panel. static incidence(α = 0.0)
maximizes power generation but the drift is large and will
quickly push PaddleSat out of the bounding box after only 7
days as shown in Fig 3.
When we compared static incidence(α = 60.0) in Fig 4 and
static incidence(α = −60.0) in Fig 5, we found that positive Fig. 5: 30 days of static incidence(α = −60.0)
and negative incidence angles yield symmetrical effects with
drift rates that trend in the same direction.
It is clear that static incidence alone does not provide static incidence controller will not be 0 degrees because it
enough control authority. However, one can still characterize optimizes the power generation and the time spent inside the
the optimal alpha for the static incidence controller given bounding box. However, this result largely is an artifact of
an area density. Fig 6a and Fig 6b show that the optimal the time span chosen. The PaddleSat controller will make
angle changes for changing area densities. Lower area density state changes on time spans far smaller than where this effect
ratios lead to more drift and given sufficient time the optimal becomes pronounced.

44

Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on November 26,2023 at 04:34:07 UTC from IEEE Xplore. Restrictions apply.
 (a) Area Density=5kg/m2 (b) Area Density=0.1kg/m2
Fig. 6: static incidence controller at varying alpha for differ-
ent area densities

Fig. 10: Varying cutoff angles of clipped zenith controller

Fig. 7: 30 days of zenith controller

Fig. 11: 30 days of clipped zenith(angle=8.0) provides desired

west drift control authority

shows that a PaddleSat attitude controller can correct a drift

rate occuring in either east or west directions individually. We
found zenith off 90 to be the first controller that demonstrated
a westward drift. Both controllers remain continuous such that
there are no sudden jumps in commanded attitude throughout
the simulation.
Fig. 8: 30 days of zenith off 90 controller We then look at ways through which power generation can
be maximized by looking at clipped˙zenith control by restrict-
ing the range of the zenith off 90. The clipped zenith off 90
controller allows for more optimal power generation by re-
stricting the range the zenith off 90. Fig 10 shows the optimal
cutoff angle for the controller to maximize power generation,
and Fig 11 shows that control authority is still sufficient at
this cutoff angle to correct east-ward drift.
Using the bang-bang control mechanism to switch be-
tween the static incidence(α = 0.0) controller and
clipped zenith(angle=8.0) controller, the PaddleSat is able to
control orbital drift within the east-west bounds. The bang-
bang control law allows for maximum power generation and
Fig. 9: 30 days of negative zenith off 90 controller maximum negative longitude drift with static incidence(α =
0.0) until that drift leads to a bounding box violation, at which
point the controller switches to the clipped zenith to perform a
We next look at the family of zenith controllers. While period of correction (Fig 12). This behavior is analogous to the
The zenith controller does not help reduce the orbital drift, it way traditional thruster-based geostationary station-keeping
however shows the effect of the SRP on the communications corrections are performed.
dish exclusively. Therefore, we looked at the zenith off 90 and We observed that the orbital perturbations PaddleSat ex-
negative zenith off 90 controllers that show distinctly unique periences cause the orbit to rotate. Over time, this rotation
behavior as seen in Fig 8 and Fig 9 respectively. This behavior grows and is only compounded by the effect of regular time

45

Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on November 26,2023 at 04:34:07 UTC from IEEE Xplore. Restrictions apply.
 ACKNOWLEDGMENTS
This project was partially supported by grants from the
Space Solar Power Institute, Cisco, and the Georgia Smart
program.
R EFERENCES
[1] P. E. Glaser, “Power from the sun: Its future,” Science, vol. 162, no.
3856, pp. 857–861, 1968.
[2] P. Jaffe, “24 - space solar,” in Future Energy (Third Edition), third
edition ed., T. M. Letcher, Ed. Elsevier, 2020, pp. 519–542.
[3] E. Gdoutos, C. Leclerc, F. Royer, M. D. Kelzenberg, E. C. Warmann,
P. Espinet-Gonzalez, N. Vaidya, F. Bohn, B. Abiri, M. R. Hashemi
Fig. 12: 30 days of smart controller provides desired east and et al., “A lightweight tile structure integrating photovoltaic conversion
and rf power transfer for space solar power applications,” in 2018 AIAA
west drift control authority Spacecraft Structures Conference, 2018, p. 2202.
[4] G. D. Durgin, “Quasi-geostationary earth orbits for space solar power
infrastructure and other low area density satellites,” IEEE Journal of
Radio Frequency Identification, 2022.
interval controller actuations that do not rotate with the orbit. [5] T. Douglas, C. Kelly, and A. Grisé, “On-orbit stationkeeping with
Similar in concept to the method used to derive the other ion thrusters telesat canada’s bss702 experience,” in Space OPS 2004
Conference, 2004, p. 301.
controllers, if the smart controller’s actuation intervals rotated [6] D. Chu, S. Chen, D. Early, D. Freesland, A. Krimchansky, B. Naasz,
with the orbit (with respect to time) the negative feedback loop A. Reth, K. Tadikonda, J. Tsui, and T. Walsh, “Goes-r stationkeeping and
may be eliminated (as in zenith˙controller). If the actuation momentum management,” in 29th Annual AAS Guidance and Control
Conference, no. AAS-06-046, 2006.
intervals are rotated ”beyond” the orbit (with respect to time), [7] M. C. CROCKER, “Attitude control of a sun-pointing spinning space-
an improved smart controller may work to dampen the high- craft by means of solar radiation pressure,” Journal of Spacecraft and
frequency oscillations. Rockets, vol. 7, no. 3, pp. 357–359, 1970.
[8] V. Modi and K. Kumar, “Attitude control of satellites using the solar
radiation pressure,” Journal of Spacecraft and Rockets, vol. 9, no. 9, pp.
711–713, 1972.
VI. C ONCLUSION AND F UTURE W ORK [9] L. Srinivasan, “Robust and adaptive attitude control of spacecraft using
solar radiation pressure,” 2013.
In this paper, we study different classes of attitude control [10] E. R. Burnett and H. Schaub, “Spacecraft formation and orbit control
using differential attitude-dependent solar radiation pressure,” Advances
algorithms that can be used for Paddlesats. We are able to in Space Research, vol. 67, no. 11, pp. 3396–3408, 2021.
show that the average orbital drift rate can be corrected [11] J. Van der Ha and V. Modi, “Long term solar radiation effects upon an
by using the zenith off 90 controller by using the bang- orbit in the ecliptic,” Acta Astronautica, vol. 4, no. 7-8, pp. 813–831,
1977.
bang control mechanism in our smart controller. Despite its [12] O. Montenbruck, E. Gill, and F. Lutze, “Satellite orbits: models, meth-
susceptibility to larger high-frequency oscillations compared ods, and applications,” Appl. Mech. Rev., vol. 55, no. 2, pp. B27–B28,
to other controllers, this represents a significant advancement 2002.
in research for PaddleSat attitude control. Our hypothesis is
that these high-frequency oscillations can be dampened by
identifying mechanisms to destructively interfere as part of
the bang-bang control mechanism.
Furthermore, we hypothesize that these high-frequency os-
cillations can be mitigated by incorporating mechanisms for
destructive interference within the bang-bang control mecha-
nism. This suggests a potential avenue for future exploration
in this direction.
To further advance the field, it is recommended to conduct
a comprehensive analysis to determine the optimal times for
activating the bang-bang controller through simulations with
varying controller periodicity. Additionally, an additional layer
of logic can be developed to enable automatic actuation of
the controller when bounding box violations are predicted,
enhancing the efficiency of PaddleSat operations. These simu-
lations will provide valuable insights for designing a prediction
algorithm aimed at maximizing power generation.
By addressing these areas of future work, we can continue
to refine the attitude control algorithms for PaddleSats, paving
the way for improved power generation and the realization of
their full potential in space-based solar energy applications.

46

Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on November 26,2023 at 04:34:07 UTC from IEEE Xplore. Restrictions apply.