IMU-GPS Integration in Presence of S/A Errors

Zeev Berman

Abstract:	The integration of INS (Inertial Navigation System) with C/A code GPS receiver is a challenge for every Kalman filter designer. This work focuses on the design considerations for an airborne system. Our performance criterion is the attitude (leveling) accuracy. The actual performance depends on particular trajectory and actual sensor models, therefore, in order to make general conclusions some simplifications are required. Our approach is to view the attitude estimation as a fusion of two processes. The first is called filtering and it aims to calculate high order derivatives of position or velocity errors. The second is called separation and it is viewed as a calculation of attitude errors based on the filtering results. This representation is best matched to a flight with straight legs. Moreover, often a general trajectory can be approximated by an equivalent one which consists of straight legs and provides similar estimation results. It turns out that the separation process deals with the estimation of random constants like initial attitude error, or accelerometer bias, while filtering process is an estimation facing short term random errors, like random walk, sensor instabilities and S/A errors. We observe that to understand fundamental filtering limitations, which are obtained as the optimal performance, the detailed sensor model is not required, only a rough “order of quality” is needed. Since the S/A error model can not be known perfectly to the Kalman filter designer (moreover it can be changed in the future), the tool of “true” covariance analysis is used. The questions of accuracy degradation due to both changes in error models and due to model reduction are studied. Our conclusion is that a first order Markov random process model yields a minor performance degradation with a good robustness to error model changes. The filtering process is very sensitive to unmodeled error jumps, especially while using Markov random process models. Therefore, first order Markov random process should be applied separately to every satellite in view or to every channel of GPS receiver. A special attention should be paid to treat the switches in satellites that participate in GPS solution. This type of model is easy to implement in tightly coupled system, nevertheless the implementation is possible in loosely coupling scheme as well. We describe the modifications required to apply the S/A error model in loosely coupling system.
Published in:	Proceedings of the 53rd Annual Meeting of The Institute of Navigation (1997) June 30 - 2, 1997 Albuquerque, NM
Pages:	277 - 286
Cite this article:	Berman, Zeev, "IMU-GPS Integration in Presence of S/A Errors," Proceedings of the 53rd Annual Meeting of The Institute of Navigation (1997), Albuquerque, NM, June 1997, pp. 277-286.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In