Integrity Monitoring of Integrated Satellite/inertial Navigation Systems Using the Likelihood Ratio


BIOGRAPHY Jan Palmqvist received his M.Sc. in Applied Physics and Electrical Engineering from Linkk oping University , Sweden in 1986. He has been with Saab Military Aircraft since 1989 and in 1994 he also joined the Automatic Control Group at Linkk oping University as a part time Ph.D. student. His current research focuses on algorithms for integrity monitoring of integrated navigation systems. ABSTRACT Global Navigation Satellite Systems (GNSS) have the ability to fullll the navigation accuracy requirements of most applications. The systems do however lack continuity and integrity to meet the requirements of high precision navigation applications. The use of a combination of Inertial Navigation Systems (INS) and GNSS information do however show promising results in fulllling these requirements. Methods for monitoring the integrity of integrated INS-GNSS systems are investigated. Integration of INS and GNSS is usually accomplished using a Kalman lter for recursive estimation of the parameters of interest. The residual used for integrity monitoring is the Kalman lter innovation. The innovation signatures of diierent types of faults are analyzed. Since two of the most likely types of faults in an integrated solution are INS sensor bias shifts and satellite range bias drifts or jumps, these additive types of changes are studied in more detail. Taking the approach of hypothesis testing of the two hypotheses un-failed and failed system, fault detection methods based on the likelihood ratio are considered and the Generalized Likelihood Ratio (GLR) test is proposed to be used. This method uses the innovations of the Kalman lter to compute the maximum likelihood estimates of the time and magnitude of an assumed change. Using these estimates, it evaluates the log-likelihood ratio of a change versus no change. The GLR test uses a linearly in time increasing number of matched lters. Diierent ways of decreasing this computational burden are discussed, showing that fast detection can be achieved even with a small and constant number of matched lters. A further advantage of the GLR test is that in addition to detecting the occurrence of a fault, it also estimates its magnitude, direction and time of occurrence, making it possible to identify the source of the fault, exclude faulty satellites and correct the Kalman lter estimate without reprocessing the affected data.

1 Figure or Table

Cite this paper

@inproceedings{Palmqvist2007IntegrityMO, title={Integrity Monitoring of Integrated Satellite/inertial Navigation Systems Using the Likelihood Ratio}, author={Jan Palmqvist}, year={2007} }