Steven Paligo, a.i. solutions, Inc.
Keywords: Non-traditional sensors, data quality, calibration
Abstract:
Over the last several years, we have seen a significant shift in the SDA sensor landscape. Before then, space surveillance was primarily dominated by militaries and research organizations. They used a small number of sensor types (mainly electro-optical, conventional radar, and phased arrays), each confined to a narrow set of configurations and use cases. For example, the electro-optical (EO) sensors had high magnification, narrow fields of view (FOV), and were steerable. Those primarily tracked targets in deep space only, slewing around their fields of regard as necessary.
Assessing the accuracy of these sensors’ metric observations (time and positional measurements) was a relatively straightforward process that mimics sensor calibration. They tracked satellites for which we have highly accurate reference ephemerides. These included the ILRS satellites in LEO and the IGS constellations in MEO. The differences between the reference ephemerides and the measured positions yielded the error residuals. From those, noise and bias values were calculated, and their magnitudes were a key factor in the overall evaluation of a sensor’s quality.
That process was repeated periodically (e.g., every two weeks), giving the assessors time to gather the pristine reference ephemerides, some of which are not published frequently. They could afford infrequent quality assessments, because the exquisite care of these sensors resulted in consistent noise and bias values over time.
Surely most organizations involved in SDA adhered to a set of quality standards defining noise and bias thresholds for acceptable metric observations (though this is difficult to verify, as few standards documents have been made public). In 2019 the United States Air Force (USAF) published theirs in a document called, “Space Situational Awareness Metric Data Integration Guidelines for Non-Traditional Sensors” (AFSPCI 10-610). In it, they define thresholds for both “traditional” sensors (those controlled by the U.S. Department of Defense) and “non-traditional” sensors (all others). However, those thresholds assume the non-traditional sensors match the same narrow set of configurations and use cases seen in the traditional sensors.
Today, SDA sensor types, their configurations, and the data they produce are more varied. In the last few years, dozens of companies emerged with commercial SDA sensor offerings. Many do space surveillance differently than it was done in the past, bringing new capabilities. For example, passive radio frequency (RF) sensors gather data regardless of weather or lighting conditions. They are also effective at tracking objects in cislunar space. Short-wave infrared (SWIR) sensors can observe during local daylight. EO sensors are being used for LEO tracking, wide-field surveys, or are constantly staring at a single patch of sky.
The quality of some new sensors cannot be easily assessed using conventional methodologies, threshold values, and reporting formats. For example, EOs tracking LEO objects may not be able to detect the small, dim ILRS satellites. Staring sensors with narrow FOVs may not get consistent observations of the IGS satellites, since there are relatively few of them. SWIR sensors observing during daylight might require more permissive noise thresholds than the radars that traditionally had exclusive coverage of the LEO orbital regime.
Organizations with quality standards too rigid to support these new sensors risk missing out on the valuable capabilities they provide. Without a standard, quality cannot be assessed, which leads to lack of trust, and ultimately underutilization. In this congested, contested, and competitive space environment, the SDA community cannot afford to ignore new capabilities.
We propose several methods for extending metric observation quality standards to accommodate these new non-traditional sensors. First is to define the noise and bias thresholds in terms common to all sensor and measurement types. Instead of azimuth thresholds in degrees and right ascension in arcseconds, the quality standards would simply state the maximum contribution each measurement can make to the positional error in each orbital regime. This provides flexibility to support new measurement types, like time difference of arrival (TDOA), without having to revise the standards. It also decouples the sensor type from the orbital regime, an assumption made in the USAF document.
Our second proposal allows the use of slightly less accurate reference orbits. For example, satellite owner/operator ephemerides, or the special perturbations (SP) state vectors calculated from traditional, well-calibrated sensor observations. Their use could considerably expand the number of assessment satellites available to staring sensors, as well as open access to larger and brighter LEO targets.
Another example of acceptable, yet less accurate reference orbits are the IGS “Rapid” ephemeris products. They are published with significantly lower latency as compared to the “Final” products, but still have meter-level accuracy. Their use enables more real time quality assessment, which is necessary to quickly identify the sudden noise and bias changes experienced by some commercial sensors.
Our experiments show it is still possible to generate accurate satellite states from observations made by sensors assessed using these less accurate reference orbits.
Next, to prevent problems caused by using non-traditional reference orbits, we propose amending the transmission formats for noise and bias values to accommodate ephemeris quality indicators and data provenance statements. That would provide recipients the context to understand how those values were calculated and the follow-on impacts to their own results.
Finally, a common occurrence for the wider FOV sensors is to serendipitously capture observations they are not optimized to detect. For example, EO sensors tuned to stare at a portion of the GEO belt often track MEO satellites that happen to pass through their field of view. They appear as blurry (less accurate) streaks in the images. When evaluating quality, operators are forced to either throw away these less accurate tracks, or combine them with the more accurate GEO observations, which diminishes the quality scores of the GEO data.
We propose updating the quality standards and transmission formats to allow publication of multiple noise and bias value pairs for staring sensors, each for a different type of observed target. Thus, allowing more nuanced and consistent assessments.
In summary, we recommend that organizations in the SDA community update their data quality standards to include this new generation of non-traditional sensors. We believe these proposals are a starting point that will enable increased data utilization and maximization of available capabilities.
Date of Conference: September 17-20, 2024
Track: Space Domain Awareness