Robust innovation-based adaptive Kalman filter for INS/GPS land navigation

Abstract

The integration of Inertial Navigation System (INS) and Global Positioning System (GPS) is a most frequent method for land navigation. Conventional Kalman Filter (CKF) is an optimal estimation algorithm widely used in INS/GPS integration. CKF assumes that the covariance of the system process noise and measurement noise are given and constant. The performance of the CKF degrades seriously, when the GPS measurement noise changes. Researchers introduced an Innovation-based Adaptive Estimation Adaptive Kalman Filter (IAE-AKF) algorithm to keep the filter stable. However, under some extreme condition, the measurement noise may vary tremendously, which will lead to the degradation and divergence of the IAE-AKF. A robust IAE-AKF algorithm is presented in this paper, which evaluates the innovation sequence with Chi-square test and revises the abnormal innovation vector. Simulation and vehicle experiment results show that the new algorithm performs higher accuracy and robustness, and also has the ability to prevent the filtering from being diverged even in a rigorous GPS measurement environment.