Benoit Philbert, NorthStar Earth & Space, Inc.; Narendra Gollu, NorthStar Earth & Space, Inc.; Yann Picard, NorthStar Earth & Space, Inc.
https://doi.org/10.64861/YERT8725
Keywords: anomaly detection, data fusion, electro-optical, passive radio-frequency
Abstract:
The real-time identification and characterization of satellite behavior changes is a fundamental component of Space Domain Awareness (SDA). To detect these changes as soon as possible and maintain the custody of the objects, sensors of multiple technologies are used to gather different data phenomenologies on RSOs, such as angular measurements with electro-optical (EO) sensors [1] or Time Difference of Arrival (TDOA) / Frequency Difference of Arrival (FDOA) measurements with passive Radio-Frequency (RF) sensors [2]. Moreover, anomalies can also stem from sensors malfunctions that can alter RSOs observations and thus affecting the understanding of the space situation. The early detection of satellite behavior changes in spite of sensor-level anomalies enables the efficient and reliable tracking and supervision of Resident Space Objects (RSOs).
Recently, Chinese satellites SJ-25 and SJ-21 have conducted docking sequences for on-orbit servicing (OOS) [3]. These included events occurring during the solar exclusion period, a time span that can last up to 6 hours when EO sensors cannot observe. Meanwhile, passive RF sensors can get continuous custody, even during the blind spots of EO sensors. Such events occurred in the past [4] and are likely to happen more in the future with the growing interest of the space sector in OOS [5], thus it is essential for accurate tracking of RSOs.
Considering this, we propose an EO and RF data-fusion strategy to associate transmission-agnostic EO measurements with the continuous data flow that RF sensors can provide for active RSOs. Several research publications and operational systems have been focused on combining TDOA/FDOA passive RF measurements with EO data within a fused orbit, either to enhance orbit determination accuracy or to have a better understanding of the scenarios where both technologies can be used in a complementary way [6,7]. However, to the best of our knowledge, there is a research gap on the application of this combination regarding anomaly detection related to the satellite behavior while analyzing the resiliency to sensor-level anomalies.
Based on the limitation described above, the goal of this paper is to enhance the tracking of RSOs by fusing multiple independent data sources. Specifically, we will focus on analyzing how leveraging all available information from EO and passive RF sensors together can improve the accuracy of the RSO state estimation when the behavior of the observed RSO is non-nominal. To achieve this goal, we propose a three-step approach: test bench implementation, orbit determination process then anomaly detection through analysis of orbit determination results.
First, we will develop a test bench using the Orekit library [8]. This will allow us to emulate observations from generated trajectories. The test bench will give us control over the RSO trajectories, with the possibility to inject known maneuvers, as well as control over the modelling of the sensor network composed of EO and passive RF ground-based sensors. Specifically, we will be able to set sensor anomalies such as sensor disconnection, increased noise and outliers. The properties and realism of the maneuvers (e.g. station-keeping, rendez-vous and proximity operations) as well as the simulated observations (observations conditions, uncertainties, and limitations) will be verified against real-world data.
Then, the generated observations will be processed using a batch least-squares algorithm to estimate the state of the RSO (orbit determination process) for different observation scenarios: EO only, RF only and fused EO and RF observations (observation-level fusion). Using the output of the orbit determination process, we will compare the performances in anomaly detection and resiliency to sensor-level anomalies for each scenario using an information-based approach.
This approach shows the competitive advantage of the fusion of electro-optical and passive radio-frequency observations in maintaining the custody of maneuvering RSOs in an operational context where sensors are subject to anomalies.
Date of Conference: September 15-18, 2026
Track: Space Domain Awareness