Daniel Dowd, HQ Space Operations Command, USSF
Keywords: SSN, radar, error model, measurement uncertainty
Abstract:
The purpose of this study is to demonstrate potential operational improvements in the Space Surveillance Network (SSN) characterization of radar measurement uncertainty. This report describes the development of an empirical measurement uncertainty model for a selection of twelve SSN phased array radars. The model is implemented in a look angle optimization algorithm designed to minimize measurement noise on the satellite catalog. A framework is developed to use the model for the improvement of satellite state accuracy and covariance realism.
The SSN performs sensor calibration using independent satellite laser ranging (SLR) precision ephemerides computed for calibration satellites (calsats). Sensor data quality is described in terms of mean offset (bias), standard deviation (sigma), and root-mean-square (RMS) error. These sensor error and uncertainty parameters are captured and implemented in orbit determination and state propagation algorithms. Therefore, the realism of these parameters is fundamental to many critical space domain awareness (SDA) mission areas.
Phased-array radar detections are generally subject to increased noise off boresight due to beam width, effective aperture, and atmospheric effects. Other factors such as phase errors and thermal effects can also contribute to non-uniform noise over a radars field of view (FOV). Therefore, observation errors and uncertainty can vary significantly with changes in look angle. Capturing this variation accurately would lead to more realistic estimated satellite covariance and improved state accuracy.
Calsat SLR residuals are mapped by their azimuth/elevation coordinates in each radar FOV. A k-nearest neighbor regression algorithm is implemented to estimate the measurement uncertainty at any arbitrary look angle in the FOV. A look angle optimization algorithm is developed and implemented on approximately 1 million observations on 20,000 space objects. For each track, the associated two-line element set (TLE) is propagated and additional look angles are generated over the length of the pass. The measurement uncertainty is calculated at each look angle, and the optimum set of look angles is selected for each pass to minimize the total measurement uncertainty. Trends in both real and optimal look angles are investigated in relation to their relevant measurement uncertainty models. The cumulative reduction in measurement uncertainty for each radar is presented.
Further analysis is required to demonstrate additional operational impacts of this method of radar uncertainty characterization and look angle optimization. Potential improvements in both state accuracy and covariance realism will be investigated by estimating satellite states and uncertainties using both optimized look angles and look-angle-dependent uncertainty values.
Date of Conference: September 27-20, 2022
Track: SSA/SDA