Machine vision algorithms for autonomous aerial refueling for UAVs using the USAF refueling boom method

Abstract

The purpose of this chapter is to provide an extensive review of a research effort by a team of researchers from West Virginia University and the University of Perugia focused on the design of a Machine Vision (MV)-based system for the Autonomous Aerial Refueling of Unmanned Aerial Vehicles (UAVs) using the US Air Force refueling boom set-up (as opposed to the probe-drogue system used by the US Navy). Following an "Introduction" section with a description of the UAV aerial refueling problem, another section provides an overview of a detailed Simulink-based simulation environment specifically developed for reproducing the UAV/tanker docking maneuver. Next, a section describes the specific approach followed in this effort based on breaking down the problem in a sequence of a Feature Extraction (FE) task (for the purpose of detecting the corners of the tanker from the images on the UAV camera), Detection and Labeling (DAL) task (for the purpose of introducing a tracking for specific corners during the docking), and Pose Estimation (PE), for the purpose of estimating the tanker-UAV relative position during the docking phase. The methodology has been labeled as the FEDALPE approach. The following sections - relative to the Feature Extraction, the Detection and Labeling, and the Pose Estimation - provide comparative studies for a number of methods for each of the above tasks, leading to the selection of the method with the best performance. Another section highlights the advantages of introducing a sensor fusion scheme blending GPS and Machine Vision data for improving the docking performance. A final section summarizes the document providing general conclusion. © 2010 Springer-Verlag Berlin Heidelberg.