Ankit Agrawal, Digantara; Thamim Ansari, Digantara
Keywords: Conjunction assessment, Data fusion, Bayesian inference, Space traffic management, Space situational awareness
Abstract:
With the ever-increasing population of resident space objects (RSOs) around Earth, collisions pose a growing concern to satellite operators, reducing safety and depleting their resources for analysis, management, and maneuver fuel. Space Situational Awareness (SSA) practices have undergone significant improvisations, directed towards providing them with a more realistic and accurate assessment of the collision threat, thus, allowing them to incorporate appropriate measures for risk mitigation.
Based on various conjunction assessment (CA) techniques employed by multiple originators, the nature of information related to conjunction of the objects concerned may be highly inconsistent. Thus, ensuring data interoperability is the foremost step before ingesting the data for further processing. In addition to that, the CDMs provided by different originators may sometimes also provide mixed insights about the possible conjunction event. The ambiguity in the time of conjunction, state and uncertainty estimates of the respective objects, and the collision probability renders the operator with incomplete decision-making capability for the counter action. In the absence of a unified situation picture, an operator would not be able to reliably bifurcate between an actual dangerous close approach and a false alarm.
The growing diversity of space missions, the ensuing collision threat, and the resulting space safety risk together amplify the importance and requirement of data fusion. It can be regarded as one of the most efficient strategies for bringing multiple datasets together to significantly improve the completeness, and accuracy of SSA products, which form the basis of Space Traffic Coordination (STC) and Space Traffic Management (STM).
The presented work proposes a novel methodology to combine CDMs received from multiple agents in order to extract unambiguous and refined analytics about the collision threat. An approximate Bayesian fusion technique, known as the product of experts (PoE) rule has been used, where local posteriors from individual agents are multiplied together, followed by normalization, to produce a global posterior estimate.
The resulting CDM fusion framework, when tested on synthetically generated data, yielded a better collision probability computation with significant reduction in the respective state uncertainties. Further testing on real datasets would help us realize the practical applicability of the proposed methodology.
Date of Conference: September 16-19, 2025
Track: Conjunction/RPO