Mark Bolden, Trusted Space, Inc.; Sydney Bonbrest, Trusted Space, Inc.; Islam Hussein, Trusted Space, Inc.; Holly Borowski, Trusted Space, Inc.; Erin Griggs, Trusted Space, Inc.; Bradford Tousley, Trusted Space, Inc.; Thomas Kubancik, Trusted Space, Inc.; Mathew Sandnas, Trusted Space, Inc.; Tim Craychee, Trusted Space, Inc.; Ryan Ross, Trusted Space, Inc.
Keywords: electro-optical systems, multi-hypothesis velocity, event based sensors, photon counting arrays, uncertainty management, hypothesis management, orbit determination
Abstract:
In the last decade there has been a significant increase in the types of generated and reported Electro-Optical (EO) Space Domain Awareness (SDA) information products. These new products can be categorized as measurements, estimates, or validated hypotheses. In this paper we will motivate the need to report and process these products with rigorous uncertainty management. We will discuss 1) additional labels and metadata to provide when disseminating these data products to operational users, and 2) best practices when leveraging estimated and hypothesized data products for Initial Orbit Determination & Orbit Determination (IOD/OD).
For several decades the SDA community has been leveraging digital imagery from electro-optical sensors to perform orbit determination. These systems integrate flux during exposures by converting a percentage of the incident photons into electrons and then recording that information in terms of integer counts. The result is a series of digital integer count images that can be calibrated, thresholded, and centroided to compute the average intensity weighted angular positions (i.e., right ascension and declination) timestamped at mid-exposure. Traditionally, the angular centroids and timestamps are next associated between frames to create a time series of angular centroids, commonly referred to as “tracks”. These tracks are used to perform statistical OD by leveraging the relative motion between centroids and assumptions about conservative and non-conservative forces. This results in time series estimates of spacecraft positions and velocities with uncertainties, commonly referred to as state estimates.
Over the last several decades optical systems have been coupled with new detectors that collect time series of events rather than digital integer count images. These are known as Two-dimensional (2D) array photon counters and Event-based sensors. The 2D array photon counters measure flux into lists of asynchronous photo-electron events, that are then calibrated, associated, and centroided to compute the average intensity weighted angular positions and asynchronously sampled timestamps. Event-based sensors rapidly perform analog integration of the change in flux (i.e., the flux with a small delay) and generate asynchronous polarity events when the positive or negative change in flux has exceeded on-chip thresholds. In theory and in practice, the positive polarity events can be calibrated, associated, and centroided to compute the average flux weighted angular positions and timestamps sampled asynchronously. Note: In theory only, the negative polarity events can also be calibrated and leveraged for OD, however this is challenging due to large biases introduced by temporal “wake events” in SDA collections. The asynchronous centroided event lists from both 2D-array photon counters and event-based sensors are then associated over time to create asynchronous time series of angular centroids that can be processed by the same conventional OD tools.
The processes described above are relatively well understood with community established data formats with labels and metadata. In the last decade, new techniques have been developed to increase detection sensitivity and to decrease latency for IOD. This includes techniques that 1) leverage stacks of images to hypothesize velocities and evaluate the integrated signal-to-noise to improve sensitivity, known as Multi-Hypothesis Velocity (MHV) stacking, and that 2) leverage time series of centroids to estimate angular rates and accelerations prior to IOD/OD statistical filtering. MHV techniques hypothesize possible velocities and evaluate the probability of detection. These high probability hypotheses are sometimes reported as velocity measurements, which they are not. Rapid IOD techniques use polynomial fitting techniques to estimate velocities and accelerations prior to IOD and are also sometimes labeled as measurements, which they are not. While these techniques can be revolutionary for SDA, it is important to properly label these information products and to provide the relevant metadata when disseminating these data products across the community. This ensures that downstream processing and exploitation capabilities can leverage these products without corrupting existing solutions.
To properly leverage these products, operational users must understand whether data products are 1) true measurements, 2) estimates based on one or multiple assumptions, or 3) hypotheses that have been evaluated as high probability. True measurements and some estimates, such as centroids, can be easily leveraged in conventional IOD/OD tools; however, most estimates and hypotheses require more rigorous uncertainty management. Failure to properly account for the assumptions made and the associated uncertainties of estimates can corrupt the solutions generated by conventional approaches. This paper and presentation will 1) discuss the different types of EO data products, 2) categorize them as measurements, estimates, and/or high probability hypotheses, 3) describe the operational impacts of misusing these data types with simulated examples, 4) recommend additional labels and metadata to provide when disseminating these data products to operational users, and 5) describe recommended best practices when leveraging estimated and hypothesized data products for IOD/OD.
Date of Conference: September 16-19, 2025
Track: Space Domain Awareness