DART Project: A High Precision UAV Prototype Exploiting On-board Visual Sensing

This work explores a way to achieve high precision in the positioning of a 250-class aerial drone by means of only onboard sensors. The proposed technology is still in development, and its basic idea is to compensate the errors of the sensors by fusing together strongly correlated data streams. The main players are a 6 Degrees of Freedom (DoF) Inertial Measurement Unit (IMU) and a computer vision system, arranged to work together as a “virtual sensor” providing the pose of the drone relative to one or more markers acting as reference points of known position and orientation. The proposed advance sensing exploits complementary filters to merge inertial and visual data. Such a refined positioning is then used to feed a custom control strategy acting as auto-pilot for implementing autonomous navigation. Preliminary results on the developed technologies are reported.