Flight path reconstruction – A comparison of nonlinear Kalman filter and smoother algorithms

Abstract

This paper is concerned with the choice of a state-estimation algorithm to perform the flight path reconstruction (FPR) procedure. Both simulated data and experimental data collected from a sailplane aircraft are used to illustrate the methods investigated. In both cases the unscented Kalman filter (UKF) is employed to determine the biases associated to each accelerometer and gyro in the inertial measurement unit (IMU), together with high sampling-rate trajectory reconstruction from low frequency sampled GPS data. A comparison between UKF and the well-known extended Kalman filter (EKF) shows that the former outperforms the latter in the presence of high noise level and large initialization error, sensitivity to the tuning of covariance matrices, and joint estimation of IMU bias terms. Furthermore, substantial accuracy improvement due to the adoption of the unscented Kalman smoother (UKS) for high sampling-rate FPR is quantified, whereas the same cannot be claimed on the use of an iterated filtering approach.