- For more than 35 years, the U.S. Space
Command (USSPACECOM) and its predecessor organizations have maintained
a continuous surveillance of space and a complete inventory of trackable
Earth-orbiting objects. This space surveillance mission is rooted in the
military's need to know when hostile satellites might be in position to
target operational units or collect intelligence.
-
- But, as the number of objects - satellites
and debris - orbiting the Earth has grown, the 'space watch' has also become
increasingly important in protecting the safety of manned and unmanned
missions into space.
-
- Today, in one of the largest-scale tracking
enterprises ever undertaken, more than 9,500 objects are tracked daily
by a network of more than 20 sensors spread around the globe. 'Objects'
includes active and inactive satellites, as well as assorted 'space junk.'
And the total trackable population has been growing at a long-term rate
of about 250 objects per year.
-
- Of particular concern in recent years
is the planned deployment of large new satellite constellations such as
Iridium and Teledesic, which will increase the trackable orbiting population
by hundreds of objects in a short period of time. Additionally, concerns
about the risk to payloads from small space debris have led to proposals
to expand the catalog to include objects as small as 1 centimeter in size
- potentially over 400,000 in number. These trends threaten to overburden
our current space surveillance system.
-
- However, a number of initiatives by Naval
Space Command (NAVSPACECOM) will help ensure that the United States will
be able to maintain a much larger and more rapidly growing space catalog
in the years ahead.
-
- The Surveillance Challenge
-
- USSPACECOM's satellite catalog is a database
of orbital elements from which it is possible to calculate position and
velocity at any time for any trackable object in Earth orbit. Predicted
positions of these objects are essential data for all analyses, including
threat assessments. But since the prediction accuracy of each element set
degrades with time from the last contact with the object, all objects must
be tracked continually and element set updates calculated frequently if
the threat assessments and other analyses are to be valid.
-
- The operators of active payloads usually
use some kind of cooperative tracking scheme to determine the orbits of
their satellites. Radio beacons, on-board sensors, laser retroreflectors,
on-board GPS receivers and other systems are commonplace. They provide
accurate orbits very economically. Cooperative tracking also neatly identifies
which satellite is being tracked in a constellation of several objects,
since each payload can, for example, be assigned a unique frequency.
-
- However, for space surveillance, non-cooperative
tracking is mandatory, for at least two reasons. First, one cannot assume
that enemy satellites will emit signals that are usable for orbit determination
by friendly forces. Second, 90 to 95 percent of cataloged space objects
are inert - dead payloads, empty rocket bodies, a variety of objects deliberately
jettisoned from payloads and rocket bodies, and miscellaneous debris from
satellite explosions and other inadvertent causes. It is essential to track
and maintain current orbital elements for every object that has a reasonable
probability of detection, including debris, if the catalog is to serve
its purposes.
-
- Non-cooperative tracking tends to be
more expensive and less accurate per observation than a cooperative method,
especially for radars, the mainstay of surveillance tracking. Moreover,
with non-cooperative tracking there is the problem of determining from
which object a given measurement originates. This problem of data association
is, in fact, one of the most difficult aspects of multitarget, multisensor
tracking. It is often the limiting factor in overall system performance,
especially if the principal means of target identification is only the
positional tracking data themselves, the case in space surveillance today.
-
- In recent years, multitarget, multisensor
problems other than space surveillance have spawned a vast and sophisticated
literature. One of the first lessons of research related to ballistic missile
defense was that the processing needed for data association and sensor
coordination (command and control) always increases exponentially with
the number of objects being tracked, and rapidly becomes unfeasible for
any current technology when the number of targets rises to a few thousand.
-
- 100 Fly Balls
-
- There is an easily visualized baseball
analogy. Imagine the difficulties faced by 100 outfielders on the same
field trying to catch 100 simultaneous fly balls. Space surveillance is
actually easier than the outfielders' problem for a number of reasons:
-
- A complete catalog has been maintained
since the first satellite was launched. Satellite motion can usually be
predicted much more accurately than the motion of other high-interest objects
such as missiles, aircraft and baseballs. Very few satellites maneuver,
split into several distinct satellites or try to evade detection. The space
environment provides a low-clutter background against which to track most
satellites, so that the false-detection rate is extremely low. Large as
it is, the current catalog size is small enough to prevent frequent confusion
of targets by the sensors. Since cataloging was started almost 40 years
ago, computer and communication capabilities have always grown faster than
the catalog size.
-
- The essential problem of space surveillance
is data association. The space object catalog is maintainable only to the
extent that sensor observations can be associated with the correct satellite.
-
- All of the above factors combine to simplify
the problem of data association enough for straightforward processing techniques
to work. The current system does, in fact, depend critically on these simplifying
factors.
-
- If any of the above fortuitious circumstances
were to change for the worse in the long term, we would have to both re-think
the current concept of operations and invest in major technical improvements
throughout the system.
-
- Sometimes one or more of these circumstances
does change temporarily and in those cases we do face special difficulties
maintaining the catalog. For example, solar storms disturb the upper atmosphere
and temporarily degrade our ability to predict near-Earth orbits accurately.
On at least one occasion during March 1989, the integrity of the catalog
was endangered because of unpredictable atmospheric perturbations affecting
an unprecedented number of orbits.
-
- In another example, satellite break-ups
may instantly increase the trackable orbiting population by 500 or 600
objects. A substantial fraction of the objects detected by a sensor may
be uncataloged and therefore unknown to any other sensors.
-
- Once the break-up pieces have separated
sufficiently, typically after a few hours, we can begin to determine orbits
for all trackable pieces. It usually takes from a few days to several weeks
to find all the pieces, depending on the size of the break-up.
-
- The Surveillance 'Fence'
-
- The space surveillance sensor operated
by NAVSPACECOM is unique in the worldwide inventory of space tracking assets.
It is particularly valuable for catalog maintenance and high-altitude unalerted
detection.
-
- The 'fence,' as it has come to be known,
is a continuous-wave, multistatic radar interferometer deployed along a
great-circle arc across the southern United States. Raw signals detected
at the receiver stations are sampled at an effective rate of 55 Hz for
up to one second as a satellite passes through the sensor beam, and then
forwarded in real time to NAVSPACECOM's computational center at Dahlgren.
-
- Although most satellites are in the near-Earth
regime, the fence routinely makes detections at ranges of more than 25,000
kilometers and occasionally at ranges of more than 40,000 kilometers. Well
over 60 percent of the entire space catalog is visible to the fence.
-
- Objects in low-inclination orbits, or
extremely small objects, are essentially the only ones not routinely detected,
and at present there are more than 100 satellites that no other sensor
but the fence detects routinely.
-
- Predictions based on the most recent
element set for each satellite are computed for 36 hours into the future
every time the element set is updated, or at least once per day. These
predictions are sorted in time order and merged with the predictions for
all other satellites in a single prediction database. The fence data association
then proceeds in time order as the observations arrive.
-
- Normally, at least 97 percent of the
fence observations can be associated automatically with known satellites.
This is a noticeably higher percentage than can presently be associated
by the other tracking sensors in USSPACECOM's Space Surveillance Network,
which makes the NAVSPACECOM combined fence-database system an especially
important asset for analyzing any observation or track that has not been
associated with a known satellite.
-
- Known as 'uncorrelated targets,' or UCTs,
these observations must be disposed of by either associating them with
a cataloged element set or an existing analyst element set, or by creating
a new analyst element set from the UCT data.
-
- Satellite break-up events present special
challenges because many unassociated observations and an unknown number
of new satellites are involved. Whenever the automatic system cannot associate
observations with known satellites, orbital analysts must use special software
tools and manual processing to make the proper data association.
-
- Recently, advanced computation and parallel
processing techniques have made it possible for analysts to consider many
more association hypotheses than in the past. Although the analyst work
is inherently slower than real-time, it is still essential to make the
correct data associations as soon as possible. Human expertise built up
through long experience has always been indispensible for this analysis.
-
- Near-Term Improvements
-
- In order to improve our cataloging process,
we must improve data association. This, in turn, calls for better accuracy
and precision of observations, better accuracy of predictions and faster
processing to accommodate the growing catalog. Some aspects of these problems
are now being addressed in research efforts at NAVSPACECOM.
-
- The distributed computing system at NAVSPACECOM
has the basic computing capacity to use a fairly sophisticated special
perturbation (SP) numerical integration orbit model for catalog maintenance.
Most satellite programs use SP models for operations because they are the
most accurate methods of prediction, though they are computation-intensive.
-
- Space surveillance is one of the few
space activities that still relies mostly on general perturbation (GP)
analytical models which can execute much faster than SP models, though
with reduced prediction accuracies. Though NAVSPACECOM,s computer network
was not specifically designed for the extra processing load of SP, it is
a very capable system and efforts are underway to maximize the network,s
capacity for such computation. Parallel processing seems to be the most
practical and economical way to achieve this.
-
- For the past several years, NAVSPACECOM
has been developing, in collaboration with researchers at the Naval Research
Laboratory, parallel versions of three important space surveillance applications:
-
- * COMBO (Computation of Miss Between
Objects) computes when close approaches between selected satellites will
occur. It has several uses, though perhaps the most important is for collision
avoidance for the Space Shuttle. To date, during more than 500 days of
on-orbit Shuttle operations, there have been seven occasions when cataloged
objects came within a 'warning zone' extending 4 kilometers radial, 10
kilometers along-track and 4 kilometers cross-track, centered on the Shuttle.
On four of those occasions, the Shuttle has maneuvered to avoid the risk
of collision.
-
- COMBO analysis is the means by which
U.S. Space Command provides notice to NASA sufficiently far in advance
for the Shuttle to be able to maneuver. The parallel COMBO algorithm successfully
demonstrated at NAVSPACECOM uses up to 23 medium-performance workstations
to analyze the list of all U.S. satellites of military interest (198 satellites
at this writing) for close approaches to any cataloged object for a period
of seven days, in a total run time of well under an hour.
-
- * SAD (Search and Determine) is used
by NAVSPACECOM orbital analysts to associate sparse UCT data that are widely
separated in time and to generate candidate element sets from the database
of UCT observations. The software works by generating candidate element
sets from every possible pair of UCT radar observations and comparing each
of these candidates against the entire database of observations.
-
- A parallel version for up to 23 workstations
is now available to NAVSPACECOM analysts and has been in use for several
months. It is capable of processing 30 days' worth of UCT data in about
six to 12 hours. However, some work remains to improve the efficiency of
the implementation and to integrate it into routine operations.
-
- * Parallel SP-based catalog maintenance
is in development, and initial demonstration runs on the NAVSPACECOM network
have begun.
-
- Additionally, the problem of parallel
database management to support all these processes is receiving intense
investigation.
-
- Far-Term Improvements
-
- In the more distant future, we will have
to consider how advanced computation - especially parallel processing -
allows us to re-engineer the whole catalog maintenance system. For example,
should the catalog have to include small debris, we will need a more robust
catalog maintenance process, and we will need it before the catalog begins
to grow rapidly, because of the large numbers of UCTs that will have to
be processed.
-
- Ultimately it is desirable to be able
to reconstitute the entire catalog without any prior element sets or data
associations. At present and for the forseeable future, we have neither
the resources nor the know-how to do this task, meaning that the satellite
catalog should be considered a national treasure.
-
- Rebuilding the catalog ab initio on demand
in a reasonable amount of time is the 'grand challenge' problem of space
surveillance. It is probably out of reach to solve this problem in practical
terms anytime soon, but by pursuing it we will develop the more robust
catalog maintenance process that we do need soon.
-
- In particular, we need detailed modeling
and simulation of the entire space cataloging operation in order to estimate
system sensitivities, risks, performance boundaries, failure modes and
relative merits of proposed improvements. High-fidelity simulation of the
whole system at both sensor level and command-and-control level is probably
the only way that a full reconstitution capability can be developed and
tested.
-
- Author Dr. Paul Schumacher is technical
advisor for Naval Space Command's Logistics and Information Systems Division.
|