Patrick Miga, Advanced Space; Matthew Givens, Advanced Space
Keywords: Space Domain Awareness, Low Latency, Maneuver Detection, Restricted IOD, Object Association, Reachability
Abstract:
A classic challenge in space domain awareness (SDA) is that of associating new measurements of space
objects to a previously generated catalog. Often, ballistic propagation and statistical distance metrics can
be used to make these track-to-orbit associations, but these methods alone can fail to associate all
available measurement tracks to the catalog and result in a backlog of uncorrelated tracks (UCTs). The
reasons for these association failures can be many, including uncertainties in the dynamic models, the
existence of new, previously untracked objects, and the fact that these objects could be executing
unknown maneuvers of unknown purpose and scale. Further, it is sometimes assumed that the track itself
contains enough information to instantiate an initial orbit determination (IOD) solution, which may not
be the case if the measurement arc is “too short” for a classical IOD solution to converge.
Motivated by the observation that successful correlation of a track to a maneuvering object constitutes a
successful detection of one or more unknown maneuvers, Advanced Space has developed tools for
associating uncorrelated tracks considering large uncertainties in the initial orbit solution as well as the
possibility that a space object has executed maneuvers since it was previously detected.
First, we leverage insights from the existing literature on the “too short arc” problem to create a restricted
IOD tool that uses rejection sampling to generate a probabilistic representation of the admissible region
for both electro-optical and radar measurement types. This approach is more robust than the more
common deterministic admissible region because it can easily incorporate any constraint available and
also take factors like measurement noise and timing uncertainty into account. The admissible region
samples can be mapped to any state space, such as inertial Cartesian or orbit elements, and either
propagated directly or fit with a Gaussian mixture model (GMM) and passed to a GMM propagation
routine.
Upon creation of an IOD solution generated by some measurement track, whether from the admissible
region approach or by some other method, we map the estimate into modified equinoctial elements and
backpropagate its reachability to determine the objects that could possibly explain the measurement
track then prune each object in the catalog that is outside this reachable bound. At this point we are left
with a much smaller subset of objects in the catalog.
Finally, the measurements, the IOD solution, and the remaining subset of objects in the catalog are fed
into an object association scheme which essentially finds the deviation from the future propagated states
of the remaining objects. Utilizing techniques from optimal control, a metric like fuel consumption or
change in velocity (ΔV) can be mapped to the magnitude of this deviation; it is somewhat reasonable to
assume that the object that would lead to the minimal deviation or minimal fuel consumption would be
the most likely to explain the measurements and thus would flag a possible maneuver. However, this is
not a guarantee; therefore, all objects are assigned a likelihood that is proportional to the size of the
metric which offers a likely association back to the object that is responsible for the measurements. This
offers a very quick, first glance at possible maneuver detection, and in many cases presents operators
with ample time to assess further decisions. As more time passes and upon further successful correlation
activities which confirms that other objects are not associated with these measurements, the certainty of
which object belongs to the measurements grows.
This flow is combined into a pipeline that seamlessly functions from a given measurement track input and
access to the space object catalog. Performance of this method is characterized as well as scenarios with
high applicability to the field of SDA, and future development is posed, such as integration into the SDA
TAP Lab and improvements to the methodology. Comparisons are also made to existing maneuver
detection methods.
Date of Conference: September 16-19, 2025
Track: Astrodynamics