Error Prediction for SINS/GPS after GPS Outage Based on Hybrid KF-UKF

Abstract

The performance of MEMS-SINS/GPS integrated system degrades evidently during GPS outage due to the poor error characteristics of low-cost IMU sensors. The normal EKF is unable to estimate SINS error accurately after GPS outage owing to the large nonlinear error caused by MEMS-IMU. Aiming to solve this problem, a hybrid KF-UKF algorithm for real-time SINS/GPS integration is presented in this paper. The linear and nonlinear SINS error models are discussed, respectively. When GPS works well, we fuse SINS and GPS with KF with linear SINS error model using normal EKF. In the case of GPS outage, we implement Unscented Transform to predict SINS error covariance with nonlinear SINS error model until GPS signal recovers. In the simulation test that we designed, an evident accuracy improvement in attitude and velocity could be noticed compared to the normal EKF method after the GPS signal recovered.