Master thesis defence by Erik Bærentsen
Acquiring Accurate State Etimates For Use In Autonomous Flight
Autonomous flight of unmanned aerial vehicles (UAV) demands continuous and accurate estimates of both orientation and position. This thesis presents three methods for obtaining accurate orientation estimates from angular velocity measurements obtained with sensors mounted on the UAV combined with measurements of the Earth's magnetic and gravitational fields. The difficulties stemming from these sensors' inherent noise characteristics are discussed, and methods for reducing their effects are presented. It is shown that procedures continuously estimating the bias present in measurements of the angular velocity can lead to significantly lowered estimation errors. This thesis also presents an implementation of a Kalman filter for obtaining accurate position estimates. Using the estimated orientation the filter fuses acceleration measurements with GPS position and velocity estimates to produce accurate position estimates at a high rate. This project has hereby examined a range of aspects for converting noisy and diverse measurements to useful and accurate data and thus laid the groundwork required in order to build an autopilot for UAVs.