Tightly-coupled GPS/INS Integration Using Unscented Kalman Filter and Particle Filter

Y. Yi, D.A. Grejner-Brzezinska

Abstract:	Integrated systems based on the Global Positioning System (GPS) and Inertial Navigation System (INS) are commonly used as the primary positioning sensor assembly providing direct georeferencing of the image sensors, such as Light Detection and Ranging (LiDAR) or digital cameras, used for geospatial data acquisition. The Extended Kalman Filter (EKF) is the traditional implementation to combine the GPS and INS data due to the nonlinearity of the system’s dynamics and the measurement model involved in the GPS/INS integration. This paper starts with a typical airborne LiDAR terrain mapping example, by Monte Carlo simulation, to illustrate the navigation accuracy requirements that must be provided by the GPS/INS integration as a function of the ground object positioning accuracy in the mapped terrain. The centimeter to decimeter –level target positioning accuracy requires centimeter- and around 10 arcsec-level accuracy in platform position and orientation, respectively, as provided by the GPS/INS navigation. One possible approach to improve the navigation performance of the GPS/INS integration is to investigate alternatives to the standard traditional EKF, such as nonlinear Bayesian filtering techniques. The Unscented Kalman Filter (UKF) from the Sigma Point Kalman Filter (SPKF) family, and Particle Filter (PF) have been the top choices among the alternatives to EKF in GPS/INS integration investigated over the last few years. This choice followed from the fact that UKF can provide better approximation to the nonlinearities through the deterministically selected sigma points, while PF can better handle both the system nonlinearities and the state vector posterior density assumption through the truly random sample particles. Our previous research (Yi and Grejner- Brzezinska, 2005; Yi and Grejner-Brzezinska, 2006) indicate that the loosely-coupled GPS/INS integration implemented using UKF and PF with an EKF importance proposal distribution (EKF+PF for short) can provide improved navigation performance in terms of the tolerance to GPS gaps and faster orientation convergence, as compared to the loosely-coupled GPS/INS implementation using EKF. Due to the advantage of the tightly-coupled system vs. loosely-coupled integration in some critical environments, for example, where GPS satellite number drops below four, this paper introduces the UKF and PF in the tightly coupled GPS/INS mode. The system design of such tightly-coupled integration is illustrated first; a kinematic example analysis based on the navigation-grade inertial sensor and geodetic dual-frequency GPS receivers is used next to provide the performance cross-comparisons between UKF, PF and EKF based on the tightly-coupled GPS/INS integration. The test results indicate that insignificant position and orientation differences (centimeter to millimeter level position differences and arcsec-level orientation differences) are obtained when the measurement updates from GPS carrier phase measurements are available. When GPS is blocked and INS is running in the free inertial navigation mode, around 2 to 6 centimeter 3D position improvement can be found in the UKF and PF solutions, as compared to the EKF solution; high-accuracy GPS trajectory was used as a reference for a simulated 1 minute GPS gap. The position accuracy improvement increases to 10 cm (UKF) and 35 cm (PF) for a simulated 10 minute GPS gap.
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:	2182 - 2191
Cite this article:	Yi, Y., Grejner-Brzezinska, D.A., "Tightly-coupled GPS/INS Integration Using Unscented Kalman Filter and Particle Filter," 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. 2182-2191.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In