Improved spacecraft attitude filter using a sequentially correlated magnetometer noise model

Measurement of the geomagnetic field provides a reliable and economical basis for determining the attitude of low-Earth orbiting three-axis stabilized spacecraft. Magnetometers are often part of the momentum management system, working in parallel with the onboard attitude control system (ACS). They may also be used as primary ACS sensors for missions with relatively coarse pointing requirements or for acquisition or contingency control modes on fine-pointing star tracker-based missions. When used in combination with strap-down gyroscopes for rate determination, magnetometer data can be filtered to obtain surprisingly accurate attitude estimates. The attainable accuracy depends on the specifics of the gyroscope quality, prior magnetometer calibration, the length of the data set, and to a lesser extent, the orbital details. It has been found, though, that even with very clean sensor data the magnetometer residuals still show systematic variations arising from small errors in the reference Earth magnetic field model. Two filters have been developed to estimate and compensate for the reference field errors along with the attitude and gyroscope biases. The field errors are modeled as sequentially correlated random variables. In the first filter, the autocorrelation of the error is assumed to decay exponentially in time. In the second, the autocorrelation is a decaying sinusoid. The resulting attitude estimates are improved by 40 percent (to 126 arcsec) and the bias estimates by a factor of two. With these improvements, autonomous onboard attitude control using magnetometers and gyroscopes becomes practical for a much broader class of missions.