Landmark database selection for vision-aided inertial navigation in planetary landing missions

Abstract

Vision-aided inertial navigation (VAIN) system offers a means to enable high-accuracy autonomous navigation for planetary landing. The VAIN system provides precise state estimates for the landing vehicle by combining measurements from an inertial measurement unit (IMU) with visual observations of the mapped landmarks contained in a predetermined terrain database. Due to the constraints of the onboard storage and computation capabilities of the landing vehicle, a terrain database containing the minimum number of landmarks should be found to guarantee the requirements of the navigation task. In this paper, a landmark database selection algorithm based on linear covariance (LinCov) technique is presented. The algorithm selects landmarks into the terrain database by minimizing the uncertainties of the chosen state parameters at each imaging time point along the nominal trajectory. The uncertainties are calculated by LinCov which is an efficient approach to analyze the effect of landmark databases on the navigation performance and allows for considering the visibility constraints of landmarks prior to actual flight. Extensive simulations are performed to evaluate the proposed landmark selection algorithm, and the results show that the proposed method is effective.