Alfredo M. Antón, GMV; Jose Miguel Lozano, GMV; Keiran McNally, GMV; Sean C. Hannon, GMV
Keywords: : Data curation, space surveillance and tracking, space debris, space domain awareness, sensor calibration, data fusion
Abstract:
Together with the increase of the number of man-made objects launched in the last few decades and the associated space debris generated by in orbit fragmentations, the concern on the space community has also grown in the same direction, giving a great importance to the space surveillance and tracking (SST) systems to protect the space assets. In the last decade, hundreds of sensors have been deployed or repurposed from the scientific domain to answer the high demand of data required by these systems to fulfil their mission.
The availability of large amounts of data, in form of sensor observations or orbital information, represents great news for the SST systems, however it opens a new discussion on the quality of the data received and the need of monitoring and validation processes to ensure the quality of the products provided. The concept of Data Curation represents the required tools and processes to ensure the quality of the data ingested or produced by the SST systems.
This paper is divided in three sections where different methodologies and processes for data curation used by GMV are presented.
First section of the paper introduces the concept of accuracy, as the union of the precision and trueness. In order to determine the accuracy of the data, a source of precise, validated and trusted data is required. These sources are usually based on GNSS satellites, laser ranging calibration spheres, altimetry or geodesy satellites, etc… They are commonly designated as calibration satellites (abbr. calsats). The paper analyzes in detail the different sources, publicly available or generated internally at GMV, by looking into concepts such as estimated accuracy, latency, availability, reliability, etc…
A second section of the paper centers on the sensors data curation processes. Two different methodologies are applied independently during the process: First one is based on sensor calibration concept; by using the previously mentioned trusted sources, it is possible to perform a parameter fitting process, estimating the desired parameters and residuals. This methodology allows to estimate time tag biases together with the biases and noise of the observables, which could be applied directly to the latter processing of the data (e.g. orbit determination) by the SST system. The data curation process automatically applies this methodology to any received observation data belonging to the defined calsats, providing a continuous monitoring of the health of the sensors. This methodology has a small drawback, the sensors are required to observe a certain number of calsats, which may not be actually required by the SST system and will slightly reduce the total effective observation time.
Second methodology for the sensors data curation is based on the observation correlation concept. Previously to the ingestion in the orbit determination processes, the observations are analyzed against the theoretical observation generated from the orbital information given a catalogue (or several catalogues) of objects, known as synthetic observations. Considering the estimated accuracy of the orbital information and the expected accuracy of the observations, it is possible to extract statistics such as the Mahalanobis or Bhattacharyya distances and therefore to perform a statistical monitoring on the quality of the observations received for each sensor. This process, although is not valid to characterize or calibrate the sensor due to the lesser accuracy, it is still very valuable to detect possible anomalies in a particular sensor without observation time on specific objects.
Finally, a third section of the paper presents the data curation of the products provided by the SST system, usually in form of orbital data, such as state vectors or ephemerides, or derived from them, such as conjunctions messages. In this case, using a similar methodology to the applied on the observation correlation, the statistical distances are computed directly between the orbital product and one or several orbital references. Two scenarios are considered: First, the reference is one of the calsats and its orbit is accurately known, biases and noise of the overall orbit determination process can be derived, including effects from the data and the algorithms used. Second scenario, involving non-calsats objects, will consider the statistical distances (including uncertainty if available) previously mentioned between the different sources to identify possible deviations, either object by object or as a statistical average to the whole population. This last scenario is particularly interesting in the case of several orbital data providers, by comparison for each source to a reference catalogue, such as the Special Perturbations from the 18th SDS, it is possible to detect and identify anomalies in the data provided, either object by object, e.g. miscorrelations, or systematic on the overall catalogue
Date of Conference: September 19-22, 2023
Track: Space Domain Awareness