Professional Documents
Culture Documents
Dynamic Mechanisms For Spacecraft Disposal
Dynamic Mechanisms For Spacecraft Disposal
DOI: 10.2514/1.G000581
This work investigates end-of-life trajectories for spacecraft in orbit about the sunEarth L1 and L2 libration
points. A decommission plan is often required during the mission design process. The spacecrafts natural departure
dynamics are studied in both the circular restricted three-body problem and an ephemeris model. In particular, the
roles of the unstable manifold and forbidden regions in determining disposal outcomes are considered. A close
correspondence is observed between the models when computing the probabilities of return to the Earth
neighborhood. A simple maneuver scheme, which uses regions partitioned by the zero-velocity surfaces to prevent
returns to the Earth vicinity, is also analyzed. Disposal maneuver costs less than 50 ms are possible for many initial
libration point orbits during a decommissioning phase lasting less than one year.
I.
heliocentric orbit, and will not return to the Earth vicinity for at least
300 years. A similar scheme was performed for Planck in
October 2013 to raise the spacecraft into a long-term parking orbit.
These disposals set a precedent for current missions, such as the
Gaia space observatory launched in December 2013 en route to a
sunEarth L2 Lissajous orbit.**
Constraints present during the decommissioning process are
distinguishable from other mission phase requirements. These
constraints play a fundamental role in the end-of-life analysis and the
determination of feasible trajectories. In the current study, the
primary requirements considered are the following:
1) For protection of the Earth and its satellite orbit zones, a spacecraft
passing through protected regimes such as low Earth or geosynchronous
orbits has strict requirements to mitigate the possibility of collisions [1].
Furthermore, if a spacecraft reenters the Earths atmosphere, a detailed
analysis is required to ensure it does so in a safe manner. For many
spacecraft, such as those using radioisotope power generation, strict
avoidance of the Earth is necessary.
2) It is required that only limited fuel remain aboard the spacecraft.
After a spacecraft completes its primary mission, a majority of its fuel
will be spent. It is expected, however, that future missions will have
some fuel budgeted specifically for the end-of-life phase. In this
analysis, 100 ms of v is assumed to be about the maximum
maneuver capability, though potentially less may be available.
3) Because of mission operation costs, it is desirable for
decommissioning to be completed in the least amount of time possible,
and so a restricted time window for final maneuvers is required. Ideally,
all maneuvers would be completed within 36 months, though this
may not always be feasible. A decommission period of about one year
is considered to be the maximum practical time available.
4) A robust end-of-life scheme is required. If very careful planning
and maneuvering are necessary for a disposal scheme, the associated
costs and risks may make it less desirable than a simpler, though
perhaps nominally more expensive, option. A less complicated
scheme is preferred to maximize the chance of successful disposal.
Individual mission needs will introduce additional constraints to
the design process. Logistically, the availability of resources, such as
ground stations and operations teams, will play a role. Furthermore,
orbit and attitude determination will be needed before and after any
disposal maneuvers, and a small clean-up maneuver may be required
Introduction
1976
1977
to correct any errors. Before the final system shutdown, the tanks
should be empty of any propellant.
The aim of this paper, however, is to give an initial analysis of the
dynamics, as well as a simple disposal technique, including
associated costs. Given the limited fuel remaining and time to
perform maneuvers, natural dynamics will drive the motion and
understanding that provide a framework for the end-of-life design.
This work could be viewed as a starting point for a particular
missions end-of-life study. Independently, recent studies have
explored the feasibility of disposal options from libration point orbits,
including the use of solar sails and alternatives such as Earth reentry
and moon collisions [6,7].
The paper is organized as follows. The primary systems of interest,
circular restricted three-body problem (CR3BP), and ephemeris model
are introduced, and computational aspects relevant for later analysis are
discussed. Natural dynamics of a spacecraft originating in a libration
point orbit are then considered. Because of the dynamic similarities
between the equilibria, the focus is on departure dynamics near L2.
Numerical verification demonstrates that the relevant dynamics are
captured by the CR3BP and that the flow away from the initial orbit is
dominated by the unstable manifold. In addition, the role of the stable
manifold and the zero-velocity surface (ZVS) for bounding the
spacecrafts motion are discussed, and investigation of the associated
probabilities of a return to Earths vicinity is performed. To reduce this
outcomes probability, performing a simple maneuver to modify the
ZVS geometry is considered. The cost associated with various
maneuver locations are presented. Because the costs are based on the
CR3BP, discussion is provided on how these maneuvers can be applied
to a more realistic ephemeris model, and validation of the long-term
orbit lifetimes is conducted.
Note that, during the course of this analysis, simple impulsive
maneuvers are used. This helps to simplify the analysis, while also
reflecting the desire to avoid complex maneuver schemes during the
mission decommissioning phase. Although other means of thrust,
particularly low-thrust schemes, are not considered, one of the
primary drivers of this analysis is the limited fuel remaining during
the end-of-life phase. Thus, natural dynamics play a fundamental role
and the particular thrust method is of lesser importance during the
preliminary design.
II.
(3)
1 1 2
x y2
r1
r2 2
(2)
The parameter 0; 0.5 relates the primaries masses, r1 and r2 are
the distances of the third body to each primary (the larger and smaller
1978
B. Ephemeris Model
(4)
(5)
n
X
q
qi t
Ri t
Gm
i
kqk3 i2
kqi tk3 kRi tk3
(6)
(7)
The term x^ shifts the origin to the sun. The scaling factor
Rt kRtk dimensionalizes the length using the ephemeris
vector Rt pointing from the sun to the Earthmoon barycenter. The
rotation matrix Ct converts the position vector from the rotating
frame to the inertial frame. Note that the rotating frame is oriented
such that the x axis is aligned with Rt, the z axis is aligned with
_
angular momentum Rt Rt,
and the y axis is selected orthogonal
to both, such that the reference frame is right-handed. Computing the
_ is simply a matter of computing the derivative of
inertial velocity qt
Eq. (7) using the product and chain rules. More detailed discussion of
transformation approaches can be found in [5]. Applying the process
in reverse, one can transform back into nondimensional rotating
frame states. This is useful for representing the ephemeris solutions in
a familiar rotating reference frame and will also be used for
determining the maneuver direction in the later results.
Transforming solely the initial state from the CR3BP to the
ephemeris model and then propagating forward in time, a trajectory
will quickly leave the vicinity of the libration point. However, by
looking at a collection of transformed states on the trajectory, they can
all be adjusted slightly such that the ephemeris dynamics are
satisfied, while still maintaining the qualitative structure of the
1979
wti
i
(8)
III.
Natural Dynamics
For the current study, the primary focus in the classification of the
natural motions is the determination of trajectories that return to the
Earth vicinity. Given the end-of-life requirements, the emphasis is on
minimizing the probability of this outcome. The existence of an
integral of motion and invariant manifolds is leveraged for this
analysis. In particular, this leads to boundaries in position space
separating the flow. The consequence of this structure is that a
spacecraft can only cross the Earths orbit if it passes through defined
gateways.
Fixing a Jacobi constant level C0 and examining Eq. (3), it is
apparent that 2Ur C0 . Regions in position space where this
inequality is violated are referred to as forbidden regions. The
boundary satisfies the relation 2Ur C0 , which requires that
v 0. This so-called ZVS is a two-dimensional manifold in the
three-dimensional position space. If one considers motion restricted
to the plane of the primaries, the boundary is a one-dimensional zerovelocity curve (ZVC) that forms a barrier in the two-dimensional
position space. The ZVC and forbidden region are illustrated in Fig. 4
for three different Jacobi constant levels. In the first illustration, there
is a bottleneck region around the Earth that allows a spacecraft to
access the interior region about the sun and the exterior region away
from the primaries. As the Jacobi constant is increased (i.e., the
energy is decreased), the opening at the L2 libration point closes as
shown in the second illustration. As it is further increased, the
opening at L1 closes. These boundaries set access routes to the Earth
vicinity and will play a role in the end-of-life design presented.
The other principal tools for analyzing the departure and return
dynamics are the stable and unstable invariant manifolds associated
with the spacecrafts libration point orbit. The unstable manifold
dominates the natural motion leaving the starting orbit. For the
periodic orbits of interest, these manifolds are two-dimensional
objects, and for the quasi-periodic orbits of interest, these manifolds
are three-dimensional objects. The benefit of taking advantage of this
structure is that it allows one to easily parameterize states departing
the initial orbit using two or three variables, without searching in the
full six-dimensional phase space.
Recall that the sign of the perturbation in Eq. (5) used to compute
the starting manifold states is free. Considering departure trajectories,
this allows one to generate two halves of the unstable manifold: one
leaving toward the Earth and one leaving away from the Earth. In
Fig. 5, the half of an L2 orbits unstable manifold departing away
1980
Fig. 4
Qualitative representations (not to scale) of zero-velocity curves and forbidden regions at sequence of Jacobi constant levels C.
Fig. 5 Unstable manifold departing away from Earth (left) and randomly perturbed trajectories (right) from initial Lyapunov orbit.
from the Earth is shown. Also shown are trajectories starting from
random perturbations on the orbit state. The underlying influence of
the unstable manifold is apparent in these results. As a consequence,
following solely the natural dynamics, there is a 50% chance of
heading toward the Earth (which, unless done intentionally, should be
avoided), and a 50% chance of departing away from the Earth.
However, a small maneuver should be possible to switch between the
outcomes. The fundamental idea is that the projection of the
postmaneuver state onto the unstable subspace should be directed
away from the Earth.
The stable and unstable manifolds also provide a geometric means
for categorizing trajectories by their return outcomes. This structure
can be most easily understood for the case of planar motion in the
Fig. 6
right.
First intersections of stable (blue) and unstable (red) manifolds with surface of section. Expanded view of return trajectories (black) shown on
the stable manifold intersection will reenter the bottleneck after one
revolution of the ZVS. The second intersections of the remaining
unstable manifold segments can be used to find returns after two
revolutions, and so forth. The primary challenge with this approach is
that the geometry can become quite complicated, making it difficult
to determine whether segments in fact lie in the interior. Even for the
first intersections shown in Fig. 6, the closed curves are distorted into
narrow spiral-like shapes. The situation becomes even more complex
in the spatial CR3BP, but a similar structure bounds trajectories
returning to the Earth vicinity. However, in the spatial case, the stable
manifold of the center manifold should be considered.
Although the direct application of the stable manifold to the
computation of returns can be challenging, it still provides an
important geometric understanding of the return behavior. Instead of
algorithmically determining whether a segment lies in the interior of
the stable manifold curve, one can equivalently check whether the
trajectory reenters the bottleneck by monitoring its distance from the
Earth. For the results shown in Fig. 6, start with many (say N 104 )
evenly spaced initial conditions perturbed to be on the unstable
manifold from the initial periodic orbit. After propagating, one can
determine which initial conditions return that have a neighbor that
does not return. One can then iteratively locate the boundary of these
intervals between return and nonreturn orbits. In fact, these
boundaries correspond to homoclinic connections.
Some comments on this computational procedure are worthwhile.
It depends on using a stable and accurate numerical integration
scheme, and even then may only be accurate for the first two or three
revolutions of the ZVS (about 30 years). It also may miss return
segments corresponding to a departure interval on the Lyapunov orbit
less than 1N of the period. It provides, however, an initial look at
intervals on the unstable manifold that naturally return to the Earth
region. Refinements of this basic scheme are possible to make it more
robust. For example, instead of considering evenly spaced departure
trajectories from the initial orbit, the step size between the departure
trajectories can be adjusted based on the change in the final state.
Similar ideas could be applied to the study of out-of-plane orbits, but
they are not investigated now.
B. Monte Carlo Return Study
1981
Earth and about half depart away from the primaries. Because of the
end-of-life restrictions in the Earth vicinity, the focus is on this latter
set of trajectories in the exterior region.
To classify the outcome, one can check whether the exterior
spacecraft trajectory eventually returns to the Earth vicinity. This is
defined as passing within 1 106 km of the Earth. When the set of
exterior Monte Carlo trajectories first depart the libration point orbit,
none of the trajectories have entered the Earth vicinity. As longer
intervals of time are considered, an increasing percentage of the
trajectories will have crossed into this region at some point along their
path. This is illustrated in Fig. 7 over a 100 year time period. The first
spacecraft returns appear after approximately 15 years. A second set
of returns appear between 25 and 30 years from their libration point
orbit departure. This trend is seen in both the CR3BP and ephemeris
results. An additional observation is that natural trajectories
departing from larger Lyapunov orbits have a higher probability of
returning to the Earth. It should also be noted that the ephemeris
trajectories considered tend to have a 12% higher chance of return
within 100 years. This is likely due to the influence of the moon,
which is in an orbit of radius 3.85 105 km relative to the Earth
(rather than being located at the Earth system barycenter as modeled
by the CR3BP) and is a direction for future investigation.
Nevertheless, there is a close relation between the return behavior
seen in the CR3BP and ephemeris models.
The results can be understood by considering the forbidden
regions and zero-velocity surface presented earlier. At the Jacobi
constant level of the libration point orbits, the forbidden region makes
a thin circular shape centered about the sun with a small opening at
the Earth. For an exterior trajectory, the semimajor axis is larger than
the Earths and, thus, has a longer two-body period about the sun.
Therefore, in the rotating frame, the trajectory rotates clockwise
while bouncing off the ZVS. Depending on the particular initial
conditions, after one revolution, the trajectory can either enter
through the L2 gateway or make a subsequent revolution around the
ZVS. This is the basis for the jumps at about 15 year intervals seen in
Fig. 7. Note that similar behavior is observed for interior trajectories
departing from an L1 libration point orbit except that their two-body
period is less than the Earths, and thus, they rotate counterclockwise.
The ZVS influences the Earth return results in an additional manner.
Larger libration point orbits correspond to Jacobi constant levels with
larger gateway openings to the Earth regime. Therefore, a trajectory
departing a larger orbit will have a greater probability of returning to
the Earth after a revolution about the ZVS. This explains the trend
observed in Fig. 7 over various orbit sizes.
The focus so far has been on trajectories that depart a libration
point orbit in a direction away from the Earth. For trajectories directed
toward the Earth, as a simple study, the same set of L2 Lyapunov
orbits in the CR3BP are considered. Their probabilities of passing
within the geosynchronous radius over a 10 year interval is
Fig. 7 Monte Carlo study of exterior trajectories returning to Earth vicinity in CR3BP (left) and ephemeris (right) models for L2 Lyapunov orbits with
specified amplitudes.
1982
investigated. In Fig. 8, one can see that, for the larger orbits
considered, well over 20% pass inside this radius within four years.
Although this coarse analysis does not consider the moon, it is
intended to illustrate that passes near the Earth are not an unusual
phenomena in the bottleneck region. The subsequent analysis will
thus focus on disposal to other regions of space.
Overall, it is apparent that there is a close correspondence between
the dynamics in an ephemeris model and the CR3BP outside of the
bottleneck region. In particular, the dynamics are strongly influenced
by the unstable manifold, which provides a set of natural end-of-life
trajectories that are easily parameterized. Some of these trajectories
have more desirable outcomes than others. If the probability of a
successful disposal is not deemed sufficient, maneuvers can be
incorporated to achieve a desired result.
IV.
(9)
p
2Ur C1
(10)
p p
2Ur C0 2Ur C1
(11)
v v0
q
v20 C1 C0
(12)
From the first relation (11), the cost decreases monotonically as Ur
increases inside the admissible region, 2Ur C1 > C0 . Considering positions away from the Earth region, the force potential (2) can
be approximated as
1 1
Ur r2
r 2
taking into account that jzj r for practical sunEarth libration point
orbits and their unstable manifolds. The maneuver magnitude is thus
approximately minimized at the perihelion in the interior region (r < 1)
and the aphelion in the exterior region (r > 1). Note that this
corresponds to maximizing the rotating-frame velocity magnitude v0 ,
which minimizes the equivalent second relation (11) for v. Roughly
speaking, the maneuver cost decreases as distance from the ZVS
increases. This discussion does not, however, consider the premaneuver
time spent along the unstable manifold, which is an important practical
consideration in the following analysis.
Maneuver costs for various families of initial libration point orbits
can now be considered. In Fig. 10, costs for various size planar L2
Lyapunov orbits with amplitudes Ay ranging from 100,000 to
600,000 km are included. Note that the unstable manifolds associated
with the periodic orbits studied are two dimensional. For presenting
the results, the manifold is parameterized in the following manner. On
the vertical axis, an angle uniformly parameterizing the initial
periodic orbit in time is used. On the horizontal axis, the time
traveling along the unstable manifold is shown. Each trajectory on the
unstable manifold from a departure point on the periodic orbit traces
out a horizontal line. The coloring corresponds to the cost computed
using Eq. (10) to close the ZVS at that point along the manifold.
Uncolored regions correspond to states on the manifold that are either
in the Earth bottleneck or the forbidden region at the new Jacobi
constant level. It should be noted that these costs are best case
results because they correspond to the idealized sunEarth CR3BP
model. In the following section, adding a safety factor to this
maneuver size will be considered to account for perturbations.
Examining Fig. 10, one can observe that for each orbit initially
there is only a narrow segment that allows the ZVS to be closed and
the spacecraft to be trapped in the exterior region. Furthermore, the
v maneuvers along this band are on the order of hundreds of meters
per second for the larger Lyapunov orbits considered. Once a
spacecraft passes approximately eight months on the unstable
manifold, however, a maneuver closing the ZVS is possible for all the
trajectories. The cost reduces to the order of meters per second within
a year.
The cost profile for L2 halo periodic orbits close to their bifurcation
with the Lyapunov orbits follows a similar trend, as shown in the
upper left of Fig. 11. For halo orbits with larger out-of-plane
components, however, the initial orbit is completely contained within
the forbidden region at the Jacobi constant level CL2 . A spacecraft
needs to get a sufficient distance along the unstable manifold away
from the initial halo orbit to close the ZVS. Nevertheless, after about
1983
nine months the scheme can be applied, and a cost of less than 50 ms
is possible at 12 months.
Next, consider the Lissajous orbits about the L2 libration point
shown in Fig. 2. As before, a spacecraft can depart these orbits along
the unstable manifold and then perform a maneuver to close the zerovelocity surface at the L2 libration point. However, the manifold is
three dimensional, and so three variables are needed to parameterize it.
The sequence of images shown in Fig. 12 all correspond to a single
initial Lissajous orbit of amplitude 100,000 km. Figure 13 corresponds
to a single initial orbit of amplitude 600,000 km. Each plot uses two
angles to parameterize the initial state on the torus from which the
manifold trajectory departs (subject to a small perturbation). The
sequence is for different amounts of time along the manifold. This
allows one to explore the full space of maneuvers for this particular
initial Lissajous orbit. Initially, there are just small regions that
correspond to potential maneuvers (analogous to the situation with the
Lyapunov orbits). These maneuvers can have a prohibitively high cost.
However, the space of potential maneuvers increases as the spacecraft
departs along the manifold. Furthermore, the associated cost decreases.
For the smaller Lissajous orbit in Fig. 12, the cost is on the order of a
few meters per second at six months. For the larger Lissajous orbit in
Fig. 13, maneuvers on the manifold are possible from any initial
departure location after about eight months. The cost ultimately
decreases to less than 50 ms at 12 months.
For illustrative purposes, as a modification of the current scheme,
consider the half of the unstable manifold heading toward the Earth
rather than away from it. In this case, one can consider trajectories
from an L2 orbit that pass through the L1 gateway and into the interior
region. A maneuver can then be performed to trap the spacecraft in
this region away from the Earth. The results for a 500,000 km L2
Lyapunov orbit are shown in Fig. 14. The situation is much more
complex in this region. Only certain segments on the manifold have
maneuvers possible and the time for these maneuvers is at least one
year after departing the initial orbit. In red highlight, trajectories
passing within the geosynchronous radius are shown, which could
lead to additional end-of-life constraints. Furthermore, the moons
influence is not included, which could significantly affect these
transit trajectories. This example serves to emphasize the simplicity
gained by considering trajectories that do not pass through the Earth
bottleneck where the dynamics are much less predictable.
V.
1984
Fig. 10 Maneuver costs (meters per second) along unstable manifold for L2 Lyapunov orbits.
may not fully capture the orbit characteristics, and this approach is
not used.
The alternative approach that is pursued is to begin by assuming
that some of the structure persists, as was observed in the study of the
natural dynamics in Sec. III, and use costs from the CR3BP for the
ephemeris maneuver. This assumption can then be verified by
checking the orbit lifetimes. The approach begins by associating the
initial ephemeris orbit with the nearby orbit in the CR3BP. The costs
are then obtained at the corresponding time along the unstable
manifold from the CR3BP. To simplify the procedure and remove the
1985
Fig. 11 Maneuver costs (meters per second) along unstable manifold for L2 halo orbits.
1986
Fig. 12 Maneuver costs (meters per second) along unstable manifold for 100,000 km L2 Lissajous orbit.
1987
Fig. 13 Maneuver costs (meters per second) along unstable manifold for 600,000 km L2 Lissajous orbit.
1988
VI.
Conclusions
This study provides a look at the dynamics applicable to the endof-life disposal for libration point spacecraft. It is apparent that the
unstable manifold is the primary driver of motion away from the
orbit. Selecting an appropriate departure phase, a trajectory not
returning for a few revolutions about the zero-velocity surface can be
followed without any additional operations. By including a disposal
maneuver on the unstable manifold, however, the bottleneck can be
closed. This is observed to have a relatively low cost (less than
50 ms), though the maneuver must be performed after
approximately eight months. Nevertheless, science objectives may
still be met during part of this time period due to the initial shadowing
of the nominal libration point orbit. Ephemeris results suggest that
with a safety factor this disposal scheme can be transitioned to
realistic mission scenarios. With sufficient budgeting of both
propellant and time, this is a simple option to dispose of a libration
point spacecraft.
Acknowledgments
The authors would like to acknowledge funding from the Marie
Curie Initial Training Network PITN-GA-2011-289240, AstroNetII, and Spanish grants MTM2010-16425, MTM2013-41168-P (G.
Gomez), MTM2012-31714, and 2014SGR504 (J. J. Masdemont).
The authors are also grateful for the insightful comments and
suggestions from the reviewers.
References
Fig. 18 Percentage of trajectories returning to bottleneck region for safety scaling factors (plot on right is an enlargement).
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
1989