UAV Pose Estimation using Sensor Fusion of Inertial, Sonar and Satellite Signals

Abstract: In this thesis a system for pose estimation of a quadcopter is developed. The implemented approach is to use a complementary Kalman filter to combine the information from an IMU and sonar sensors on-board the quadcopter with GPS. This approach does not require a dynamic model of the quadcopter, therefore modifications of the current quadcopter or a change to another one is facilitated. Experiments indicate that the system provides accurate pose estimates and are robust in the sense that it can handle loss of GPS signals during shorter time periods. Provided that further work is spent on testing and improving the stability of the estimation, especially the heading estimation, the system may be used to enable autonomous navigation of a quadcopter. The system is implemented using ROS (Robot Operating System) and tested on the low-cost quadcopter Parrot AR.Drone 2.0.