Robust direct georeferencing of lightweight UAV

点击翻译按钮获取中文摘要

The goal of this subproject is the development of methods, which allow the real-time determination of the position and the orientation of a lightweight aerial vehicle (<5kg) with high accuracy (<1cm, <1°) and high robustness in order to provide direct georeferencing information. A high robustness is necessary for the autonomous navigation of the vehicle. A high accuracy is needed for the georeferencing of the maps, which has been reconstructed from the different mapping sensors on the platform. Therefore no ground control points are needed. The accurate position and orientation information also provide a high quality approximate solution for the image based 3D reconstruction algorithms. The real-time capability is necessary for the autonomous operation and also allows a fast near real-time georeferencing of the maps, in order to adapt the flight plan to the current situation. The challenges are the technical realization of the multi-sensor system due to the size and weight constraints of the vehicle and the development of the algorithms, which determine the position and orientation based on various sensors modalities. After the first phase of the project it is possible to estimate the position and orientation under ideal conditions in accordance with the intended specification. This estimation is based on GNSS carrier phase observations, inertial sensor data and a three axis compass module. The calculation is performed during the flight in order to provide real time information to the process of autonomous navigation and to the 3D reconstruction algorithms.The main goals of the second phase of the proposal are the robustification of the GNSS based position determination by adding observation from the Glonass system as well as the location and environment dependant prediction of the GNSS observation conditions. Another goal is the robustification of the position determination in situations without any or with only a few available GNSS observations. This is done by extending the estimation algorithms in a way, that any additional sensor information, e.g. coming from the cameras or the laserscanner, can be used to bridge these situations. Another goal is the evaluation of the georeferencing accuracy during the flight.