A globally self-initializing attitude determination filter has been developed to estimate three-axis attitude and attitude rate from magnetometer data only. This filter can serve as the main attitude determination system for a low-weight, low-budget, low-accuracy mission or as a backup system for other missions. The filter uses a new attitude representation that is based on the minimum quaternion to align the body-axis and reference magnetic field vectors. It uses implicit quaternion dynamics and implicit Enter dynamics to propagate its attitude and attitude-rate estimates. It is based on a generalization of the iterated extended Kalman-filter concept. Its global initialization strategy employs a number of filters that cover a two-dimensional box in which the principal initial condition uncertainties fie. Hypothesis testing is used to eliminate all but one of the filters after the best one has reached steady state. Flight data from two missions have been used to test the new system. Convergence without the need for initial attitude or rate knowledge has been demonstrated in all cases, and per-axis 3-sigma accuracies on the order 5 deg in attitude and 0.03 deg/s in rate have been achieved.
展开▼
