Nonlinear, deterministic sampling based state estimation for highly-maneuverable small unmanned aerial vehicles: A Comparative Development

This paper examines the nonlinear state estimation problem for small Unmanned Aerial Vehicles (UAVs). Two deterministic sampling based solutions are developed in parallel: one based on the Unscented Kalman Filter (UKF) and one novel solution based on the Central Differences Kalman Filter (CDKF). The attitude is represented by the SO(3) constrained attitude quaternion, enforcing various modifications to the UKF and CDKF algorithms that include updating the quaternion and expressing its covariance through rotation vectors, whereas the position is represented and propagated in curvilinear coordinates using the WGS84 earth model in order to provide the highest possible fidelity while allowing unconstrained global operation. Apart from the adoption of the quaternion and the use of the full earth model, the requirement for high vehicle maneuverability imposes additional restrictions and changes from the usual treatment of the subject found in the literature. The accuracy, the performance and the error behavior of the two estimators are verified and compared through extensive Monte Carlo simulations for normal and abnormal measurement conditions.