Alejandro Pastor, GMV; Guillermo Escribano-Blázquez, Universidad Carlos III de Madrid; Diego Escobar, GMV
Keywords: maneuver detection, track association, track correlation, multi hypothesis tracking, cataloguing, optical data
Abstract:
The increasing number of resident space objects (RSOs) and congestion of the orbital debris environment makes the cataloging activities more challenging year after year. In this congested scenario, the main source of potential new object detections are the maneuvers of operational satellites, which cause those satellites to be in unexpected orbits for the Space Surveillance sensor network. There are more than 500 operational satellites only in geostationary orbit, most of which performs manoeuvers every one or two weeks. Detecting them is crucial for maintaining catalogues of RSOs, since otherwise these potential new objects would be promoted to actual new objects leading to sets of duplicated RSOs. Besides, for the continuous and reliable provision of Space Situational Awareness (SSA) and Space Traffic Management (STM) services, a challenging trade-off between detection time and maneuver characterization accuracy should be performed.
The automatic detection of maneuvering RSOs can be framed within the general multi-target tracking-association problem, in which, the tracked objects are allowed to maneuver. There are different proposed solutions in this field, ranging from Multiple Hypothesis Tracking (MHT), Joint Probabilistic Data Association or Bayes multi-target filter. All the methods listed above must be tuned for each specific application. The characteristics of the RSOs maneuvering problem with scarcity of data and large time intervals between them makes the tuning of the methods demanding. In previous years, a series of works have tried to address the gaps of the classical approaches with methods developed ad hoc for the Space Surveillance and Tracking (SST) problem. In those approaches, space maneuvers are characterized a priori, in order to incorporate more information to the problem, and make it tractable. The way in which the maneuvers are characterized differs from one author to another, but they can be broadly categorized as historic-based or optimal control-based. Regarding the former, several works are devoted to exploring and analysing the possibility of using historical data to associate new observations of an object with the catalogue entry after a maneuver. Criticisms to this approach lay in the fact that the use of historical data is based on the repetitiveness of the maneuvers, which is not always guaranteed. Regarding the latter, many works are based on the definition of a control distance metric to address the feasibility of an alleged manoeuver and the so-called Optimal Control-Based Estimator, to detect and reconstruct manoeuvers with no a priori information.
In this paper, a novel and operationally feasible algorithm is proposed to solve this correlation/association problem. An optimal control approach is proposed using the velocity increment as cost function in a Multi Hypothesis Tracking framework. In this way, maneuver detection can be understood as an association problem between an orbit of the RSO, estimated before the maneuver, and a set of Uncorrelated Tracks (UCTs), received afterwards. Associations of objects and tracks, or hypothesis, are generated, evaluated, pruned and promoted in such a way that the involved catalogued orbit and sensing data belongs to a common maneuverable RSO. This allows to considerably reduce the maneuver detection time, since, as opposed to orbit-to-orbit association, a new object detection and initiation is not required. However, a single track after the maneuver is not enough to estimate the maneuver and orbit after the maneuver with enough accuracy to correlate further tracks. Thus, this track-to-object correlation approach needs to be combined with a track-to-track method in order to associate tracks corresponding to the same RSO after the maneuver up to a number of tracks when the accuracy of orbit and maneuver and orbit estimation is similar to that of the no-maneuver scenario.
Results are presented in optical survey scenarios with both simulated and real data, providing clear association performance metrics and emphasizing on scalability and robust operational applications. The trade-off between detection time and characterization accuracy is discussed not only from the maneuver detection point of view, but considering the whole cataloguing process, from the ingestion of new tracks to the update of orbital information.
Date of Conference: September 15-18, 2020
Track: Astrodynamics