A virtual conjunctions generator for testing spacecraft collision detection strategies

Paper number

IAC-24,A6,IP,35,x85427

Author

Mr. Giacomo Curzi, Alma Mater Studiorum - University of Bologna, Italy

Coauthor

Prof. Dario Modenini, Alma Mater Studiorum - University of Bologna, Italy

Coauthor

Prof. Paolo Tortora, Alma Mater Studiorum - University of Bologna, Italy

Year

2024

Abstract

Miss detection rate in spacecraft collision avoidance is of great concern when selecting an actionable threshold to minimize unnecessary maneuvers while remaining safe. Such an index, characterizing any detection strategy, is exceptionally difficult to evaluate in the context of conjunctions analysis because the scarcity of the satellite collisions that occurred would undermine any statistical inference based on data from actual conjunctions. To support the selection of a proper actionable threshold, some authors derived useful analytical and/or approximate theoretical bounds on the expected miss detection and false alarm rates when decisions are based upon the probability of collision or the miss distance in Mahalanobis space.

As an alternative, this paper presents a novel, high-fidelity virtual conjunctions generator on which generic detection strategies can be tested. The underlying idea is to profile a publicly available database of anonymized Conjunction Data Messages (CDMs), namely ESA’s Collision Avoidance Challenge dataset; by fitting statistical distributions to the population of CDM parameters affecting the encounter geometry and positional uncertainty, a generator of random, yet representative, conjunctions is created. In doing so, however, the collision rate affecting the virtual conjunctions cannot be considered realistic because collisions are strongly underrepresented in CDMs. Collision events are thus randomly introduced by enforcing an a-priori collision rate based on available debris population studies.

Then, the collision detection strategies based on the probability of collision and Mahalanobis miss distance are compared: for every encounter, the risk metrics are evaluated, the detection strategy is applied, and eventually, the miss detection and false alarm rates are computed. The main result is that all the metrics have similar detection performances, i.e. they can work at similar miss detection rates with an appropriate decision threshold. Because the latter is strategy-dependent, care must be taken in choosing the threshold to avoid working at unexpected miss detection levels. 

Although applied to characterize well-known collision risk metrics, the virtual conjunctions generator developed in this work can be exploited to support the validation of new detection strategies and guide the selection of their actionable threshold.

Abstract document

IAC-24,A6,IP,35,x85427.brief.pdf

Manuscript document

IAC-24,A6,IP,35,x85427.pdf (🔒 authorized access only).

To get the manuscript, please contact IAF Secretariat.