Extended Quest Attitude Determination Filtering

The quaternion estimation (QUEST) batch attitude determination algorithm has been extended to work in a general Kalman-filter framework. This has been done in order to allow the inclusion of a complicated dynamics model and to allow the estimation of additional quantities beyond the attitude quaternion. The QUEST algorithm, which works with vector attitude observations, serves as a starting point because it is able to work with a poor (or no) first guess of the attitude. It is able to do this because its nonlinear estimation problem can be solved exactly by solving an eigenvalue/eigenvector problem. This paper's extended version of QUEST uses square-root information filtering techniques and linearization of the dynamics to handle all of the non-QUEST parts of the estimation problem. The remaining QUEST-type part of the problem can be solved by a technique that is an extension of the original QUEST algorithm's eigenvalue/eigenvector solution. The paper shows that two previously-proposed iterative QUEST techniques are special cases of the present algorithm. It also demonstrates the new algorithm's performance on an attitude determination problem that uses star-tracker and rate-gyro measurements to estimate the attitude time history and the rate-gyro biases. The new algorithm is able to converge from initial attitude errors of 180 and initial rate-gyro bias errors as large as 2,400/hour.