M-M Estimation-Based Robust Cubature Kalman Filter for INS/GPS Integrated Navigation System

Abstract

In the engineering applications, Kalman filter (KF) is the most commonly used for INS/GPS integration. This KF-based integration is prone to divergence because of two typical problems. One is that the predicted state and measurement are contaminated by non-Gaussian noise. The other is that the initial attitude error is too large. Until now, there arise many algorithms dealing with only one of these two problems, but the algorithms which can deal with both two problems are few. Motived by this situation, an M-M estimation-based cubature Kalman filter (MMCKF) is proposed which innovatively combines the M-M estimation and nonlinear cubature Kalman filter (CKF) to enhance the performance of INS/GPS integration system in the case that above two problems occur at the same time. Simulation and vehicle-mounted experiment results validate its accuracy and robustness.