Refined Astrometry on Board a CubeSat

Abstract

Optical navigation on a CubeSat must rely on the best extraction of the directions of some beacons from on-board images. We present an experiment on OPS-SAT, a CubeSat of the European Space Agency (ESA), that will characterize an onboard algorithm to this aim, named Angle-based Correlation (AbC). OPS-SAT is a 3-unit CubeSat with an Attitude Determination and Control System (ADCS) and an imager that have proved their reliability with typical performance at CubeSat scale. We selected a few star-fields that all present enough visible stars within a 10 • field of view. When our experiment is run, OPS-SAT is pointed to the most convenient star-field at that time. There, the star-field is imaged and subwindows are extracted from the image around the expected location of each star, based on the attitude-quaternion reported by the ADCS. The AbC reconstructs the absolute direction of the central body, in principle unknown, which is the pointed known star in the experiment. The method intensively uses the quaternion algebra. The beacon location is first consolidated in the field of view with the AbC. Then, the field of view is finely positioned against the sky, again with the AbC. A covariance is associated with the found beacon direction. Our experiment with OPS-SAT manages the pointing and the imager, and processes the taken images. Then, it downloads the on-board computed absolute directions and their covariances, to be compared with the actual directions. After a campaign of intensive use of the experiment, the statistical performance of the algorithm will be established and compared to the on-board computed covariances. As a bonus, an assessment of OPS-SAT's inertial pointing stability will be available. The AbC can theoretically get rid of the Attitude Control Error (ACE) of the platform and of the Attitude Knowledge Error (AKE) estimated by the ADCS, and potentially converge to sub-arcsec measurements, whereas 15 to 90 arcsec for AKE are expected on CubeSats and imager pixel size ranges in 10 to 15 arcsec. Results will be available from 2022, thus, instead, we illustrate here possible applications for commissioning and autonomous operations.