Development of navigation techniques and algorithms for small uavs (unmanned aerial vehicle) able to follow trajectories with centimeter precision

This thesis work presents project ``Dart'', i.e. the development of an UAV prototype able to perform autonomous navigation using only on-board systems and sensors. The main objective is to achieve positioning and path following with centimetric precision by using only mass market technologies, in order to ascertain the gap between a completely autonomous ultra-high precision drones and commercial products with reasonable costs. To this aim, Dart core has been designed to feature a high precision on-board vision-based positioning system exploiting the following components: a small camera carried by a gimbal, an embedded electronic board, and an open-source computer vision library. The scenery captured by the camera is processed by a custom software which computes the drone position with respect to known markers. This information is then fused with data read from a IMU to estimate the state of the drone. This system, therefore, acts as a virtual sensor that is able to pilot the UAV along a reference trajectory.