Evaluation of Automatic Landmark Selection Strategies for Navigation of Unmanned Aircraft

Abstract

This paper presents an extension of an approach for integrating pan-tilt camera systems into the state estimation of unmanned aircraft, thereby introducing an additional source of navigation data. The state estimation algorithm, based on a Unscented Kalman Filter (UKF), and the measurement model are presented in order to provide motivation for the presented landmark selection strategies. A variety of landmark selection strategies are put forth for consideration. Some are founded upon heuristic methodologies, while one employs the concept of the UKF by minimising state variance through the anticipation of potential landmark measurements. In order to evaluate the efficacy of the proposed strategies, a Matlab/Simulink simulation environment has been developed. This environment provides a range of measurements, including data from inertial, magnetic and barometric sensors, as well as emulated landmark tracking data. The subsequent section presents the evaluation of the various landmark selection strategies based on the mean position error over a range of different parameter sets. Finally, an outlook on prospective flight tests employing an unmanned helicopter weighing 85 kg is provided.