Comparison Of Error State Kalman Filter with Sampling Importance Resampling Particle Filter In Loosely Coupled Integrated INS/GPS

Inertial Navigation System (INS) using initial states such as initial position, attitude, velocity and measuring accelerations and rotation rates in three dimensions, estimates vehicle position, attitude and velocity. Low cost inertial sensors in comparison with High cost inertial sensors have more noises and less reliability. Therefore INS output errors increase rapidly with time.For reducing these errors, INS should be integrated with other aided sensors such as GPS. INS has small short time errors while GPS has small long time errors. Thus, Integrated INS/GPS increases position accuracy. Integrating these two sensors needs an estimation and data fusion algorithms. Usual methods for this purpose are Kalman filters and Particle filters.Kalman filter linearize states and it model noises with the Gaussian distribution. Particle filter estimates states using state sampling and generating particles. Particle filter in comparison with Kalman filter has more computational load but doesn’t model noises with the Gaussian distribution. The simplest type of these methods are Error state Kalman filter (ESKF) and Sampling importance resampling particle filter (SIR-PF).In this paper, loosely coupled integrated INS/GPS for a simulated scenario in 185 seconds is implemented with ESKF and SIR-PF. At last, the position error results are compared with each other.

Comparison Of Error State Kalman Filter with Sampling Importance Resampling Particle Filter In Loosely Coupled Integrated INS/GPS Keywords:

Inertial navigations system (INS) , loosely coupled Integrated INS/GPS , Bayesian estimation , Error state Kalman filter (ESKF) , Sampling importance resampling particle filter (SIR-PF)