Toward Automated Aerial Refueling: Relative Navigation with Structure from Motion
Technical Report,01 Sep 2014,24 Mar 2016
AIR FORCE INSTITUTE OF TECHNOLOGY WRIGHT-PATTERSON AFB OH WRIGHT-PATTERSON AFB United States
Pagination or Media Count:
The USAFs use of UAS has expanded from reconnaissance to hunterkiller missions. As the UAS mission further expands into aerial combat, better performance and larger payloads will have a negative correlation with range and loiter times. Additionally, the Air Force Future Operating Concept calls for formations of uninhabited refueling aircraft...that enable refueling operations partway inside threat areas. However, a lack of accurate relative positioning information prevents the ability to safely maintain close formation fight and contact between a tanker and a UAS. The inclusion of cutting edge vision systems on present refueling platforms may provide the information necessary to support a AAR mission by estimating the position of a trailing aircraft to provide inputs to a UAS controller capable of maintaining a given position. This research examines the ability of SfM to generate relative navigation information. Previous AAR research efforts involved the use of differential GPS, LiDAR, and vision systems. This research aims to leverage current and future imaging technology to compliment these solutions. The algorithm used in this thesis generates a point cloud by determining 3D structure from a sequence of 2D images. The algorithm then utilizes PCA to register the point cloud to a reference model. The algorithm was tested in a real world environment using a 17 scale F-15 model. Additionally, this thesis studies common 3D rigid registration algorithms in an effort characterize their performance in the AAR domain. Three algorithms are tested for runtime and registration accuracy with four data sets.
- Military Aircraft Operations