| Abstract [eng] |
The objective of this work is to develop and test a fixed-wing unmanned aerial vehicle (UAV) which would autonomously drop items in the intended spot by capturing a QR code with on board camera, analysing it in real-time and making decisions with an integrated artificial intelligence algorithm. The literature review part of the thesis introduces the different UAV designs, their strengths and weaknesses, 3D scanning overview, modelling and aerodynamics analysis importance and principles. Further, it discusses 3D printing materials, processes, and the importance of parameters for this project. Also, sensors and artificial intelligence integration in drones and autonomous object detection. Lastly, real-life testing and delivery accuracy are important in such a project. In this project, a fixed-wing UAV is scanned with a HandySCAN 700 3D scanner and a digital twin of the drone's structure is created for the best possible cargo compartment modelling. After that, components are picked for the intended task of delivering a small parcel in a visual line-of-sight flight. The cargo bay is modelled with an integrated Sony 8MP 77° IMX219 autоfоcus camera mоdule and servo motor for the autonomous dropping of the cargo. Camera is connected to the Raspberry Pi Zero 2 W computer for AI-Based QR cоde detectiоn from computer vision libraries tо autonmously drop cargo without a pilot intervention. Lastly, everything is tested to evaluate the delivery accuracy and limitations of the proof-of-concept fixed-wing UAV system. During real-life testing it was identified that out of ten flights, four of them identified the target and initiated the payload dropping sequence. With a used 35.5 x 35.5 cm ArUco QR code on the ground and 2-3 m flight height, when the code was detected, delivery accuracy varied from 4.13 m to 5.64 m when measured from the centre of the QR code. |
