Real-Time Gravity Gradiometer Utilization to Improve INS Accuracy

Abstract

This paper details a method for deriving approximate low-order filters for estimation of velocity error in an inertial navigation system. The filter inputs are obtained from two sources—a gravity gradiometer and some external velocity reference. It is the intent here to show that these approximate filters are near-optimal. Covariance comparisons are used to evaluate filter performance. The first of these comparisons reveals that the approximate filter gives very nearly the same rms estimate error as the Kalman filter. This result, however, assumes a rational gravity perturbation model which, although unrealistic, is required to implement the Kalman filter. In addition, a multiple-measu rement scheme employing the external velocity reference appears to significantly improve estimation accuracy. A second covariance study compares the rms estimation accuracy of the approximate filters obtained analytically and numerically when an irrational but more realistic gravity perturbation model is assumed. Agreement of these numbers justifies the Schuler Dominance assumption used in all the filter derivations and therefore gives greater weight to the claim of filter near-optimality.