LRF Assisted SLAM for Airborne Platforms
Kerem Eyice, Onur Çulha, Defense Industries Research and Development Institute of The Scientific and Technological Research Council of Turkey
Location: Big Sur
This paper presents a one-dimensional Laser Range Finder (LRF) assisted visual Simultaneous Localization and Mapping (SLAM) technique in order to aid Inertial Navigation System (INS) for airborne platforms. Landmark initialization is a critical process of SLAM algorithm, which is registering features to the map by estimating their position. Due to scale ambiguity, precise undelayed landmark initialization may not be possible by using monocular 2D vision sensor. Although several delayed initialization methods have been proposed and used, number of observed frames of a scene does not suffice to both determine the position of landmarks accurately and aid INS for high speed airborne platforms with a 2D narrow field of view vision sensor. To achieve precise mapping of features and localization simultaneously within a few frames, we further develop LRF assisted visual SLAM which estimates the position of newly detected landmarks instantaneously providing more accurate updates to INS sooner. Fusing LRF and optical measurements improved accuracy of navigation solution by more than 95 percent over INS only mode for 15 minutes flight simulations.