Stochastic and Geometric Observability of Aided Inertial Navigators

Y. Shao, D. Gebre-Egziabher

Abstract:	The observability and accuracy of aided-inertial navigation systems which use low quality (cost) inertial sensors is evaluated. It is shown that geometric observability as defined by checking the rank of the observability matrix is a necessary but insufficient condition for predicting whether INS/GPS integration filters will be stochastically observable (that is, provide an accurate state estimate) or assessing their propensity to diverge. This is because observability as defined by checking the rank of the observability matrix underplays or ignores the combined effects of process noise and non-linearities. This is shown by recasting the standard INS/GPS Extended Kalman Filter equations into a batch covariance estimator. The resulting batch estimator clearly shows that stochastic observability depends on the interaction between measurement noise, vehicle trajectory, process noise and initial state uncertainty. These observations are validated by a series of simulation studies.
Published in:	Proceedings of the 19th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2006) September 26 - 29, 2006 Fort Worth Convention Center Fort Worth, TX
Pages:	2723 - 2732
Cite this article:	Shao, Y., Gebre-Egziabher, D., "Stochastic and Geometric Observability of Aided Inertial Navigators," Proceedings of the 19th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2006), Fort Worth, TX, September 2006, pp. 2723-2732.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In