Benjamin Reifler, University of Texas Austin; Cedric Petion, University of Texas Austin; Brandon Jones, University of Texas Austin
https://doi.org/10.64861/QDVE2773
Keywords: space situational awareness, conceptual spaces, soft information
Abstract:
As low Earth orbit becomes increasingly crowded and interactions between space objects (SOs) become more complex, soft information, such as an SO’s operator and ambiguous maneuvering capabilities, can augment the hard information that is traditionally used in space situational awareness (SSA), such as sensor measurements and conjunction assessments, enabling more responsive sensor tasking and planning of collision avoidance maneuvers. For example, SSA data providers may want to prioritize observing conjunctions that are more likely to result in fragmentation events and deprioritize those that are actually rendezvous for on-orbit servicing. That sensor tasking decision may be informed by soft information, e.g., that the objects both have the same operator and appear to be actively controlled. In this paper, we leverage the theory of conceptual spaces (CSs), which translates abstract concepts into geometry, to fuse hard information about SO orbits with soft information and determine how pairs of SOs are or are not interacting. Finally, we use our space situation assessment tool to analyze different SO interactions in simulated scenarios.
Traditional SO collision warning systems use estimated probability of collision (PC), based on the assumption that objects continue along their established orbits. We have previously demonstrated that an analysis based on reachability from one SO’s orbit to another and soft data regarding an SO’s capabilities and behavior can provide earlier warning of potential collision risks than PC analysis. Our previous situation assessment methodology used established CS methods, in which a single similarity value (SV) is computed for each concept based on observations consisting of property SVs. A property SV is a number between zero and one that describes how apparent a property is, and a set of mutually exclusive properties is called a domain. For example, the set of possible satellite operators could be a domain, with each operator being a property in that domain. In our previous work, we accounted for uncertainty by modifying property SVs in a heuristic manner, with higher uncertainty yielding more uniform SVs over a given domain, but the final concept SVs did not have any associated uncertainty. Now, we improve our approach by propagating uncertainty in hard and soft data observations through the CS analysis, yielding a probability density function for each concept SV and correlation coefficients between concept SVs, which can help reveal ambiguity in distinguishing between different concepts. This approach also allows our tool to ingest observations in the form of belief functions, as used in Dempster-Shafer theory, by sampling property SV observations between the belief and plausibility of that property produced by the belief function for the given domain. This enables us to fuse multiple sources of soft information via belief fusion rules before beginning the probabilistic CS analysis.
We define a CS with concepts corresponding to different SO interactions, including pending collisions, on-orbit servicing, and active debris removal. We then apply our space situation assessment tool to simulated orbit data and analyze the tool’s sensitivity to observation uncertainty. We build on our previous work by simulating more SOs in a greater variety of orbits with different soft data-informed properties and varying levels of uncertainty in the hard and soft data.
Date of Conference: September 15-18, 2026
Track: Conjunction/RPO