The application of Kalman filter-based attitude determination, with simultaneous gyro bias estimation, to the Earth Observation System (EOS) AMI mission with stellar measurements is analyzed. The study examines the predictive capability of covariance analysis as compared with Monte Carlo analysis, showing that thrice the square root of the Kalman filter covariance matrix diagonal is a reasonable prediction for the 99.7% Monte Carlo results, but not of worst-case performance. The study also establishes further insight into the sensitivity of EOS-AMI attitude determination performance to simulated stellar-lunar geometries by comparing Monte Carlo performance predictions using a statistically generated star field (including statistical lunar blockage gaps) with those using a physical model (real star field and lunar and solar ephemerides). This study further demonstrates that, with EOS-AMI parameters, the performance is driven by the sensor noise rather than the gaps in the star field and confirms these conclusions by Monte Carlo assessments of the orbits with short and long star gaps.
Read full abstract