Nonlinear camera- and GNSS-aided INS for fixed-wing UAV using the eXogenous Kalman filter

Abstract

This chapter aims at applying a recently proposed estimator, called eXogenous Kalman Filter (XKF), to the navigation of a fixed-wing unmanned aerial vehicle (UAV) using inertial sensors, GNSS, and optical flow calculated from a camera. The proposed system is a cascade interconnection between a globally exponentially stable (GES) nonlinear observer (NLO) and a time-varying Kalman filter based on a local linearization of the system equations about the output of the preceding NLO. It is very well known that the linear time-varying Kalman filter is GES and optimal in the sense of minimum variance under some conditions, but when a nonlinear approximation (e.g., the extended Kalman filter) becomes necessary, generally such positive properties cannot be guaranteed anymore. On the other hand, a NLO often comes with strong, often global, stability properties, but without attention to optimality with respect to unknown measurement and process noise. The idea behind the XKF is to combine the advantages of the two composing estimators while surpassing the drawbacks from which they individually suffer. The theory is supported by tests on both simulated and experimental data, where the XKF is compared to a state-of-the-art solution based on the extended Kalman filter (EKF).