Loosely coupled visual odometry aided inertial navigation system using discrete extended Kalman filter with pairwise time correlated measurements

This paper presents an algorithm for processing pairwise time-correlated measurements in a Kalman filter where the measurement vector at an epoch is correlated only with the measurement vector at the epoch before. Time-correlated errors are usually modelled by a shaping filter, which is here realized using Cholesky factors as coefficients derived from the variance and covariance matrices of the measurement noise vectors. Results with the simulated data show that the proposed approach performs better than the existing ones and provides more realistic covariance estimates. Furthermore, the proposed algorithm was applied to visual odometry aided-INS and the results show an improvement of 7% in the position drifts in comparison with the conventional shaping filter.