Researching Drones with Visual Positioning System for Agrocultural Purposes

Nowadays, position detection systems are already utilised in certain industries due to the simple application of such technologies and their high accuracy, which stimulates high demand for such solutions. The purpose of this work is to study software that offers methods for visual positioning of agricultural unmanned aerial vehicles without using GPS. In this paper, the MRS Optical Flow algorithm is considered. This implementation of the optical flow algorithm differs from the traditional optical flow in that it does not calculate the pixel offset in the video stream, but calculates the linear and angular velocities of the drone using data from additional sensors. To test the algorithm, a software environment was assembled that allows testing in simulation and on an assembled experimental setup containing a Jetson Orin Nano board as an on-board computer, PixHawk 6X Mini with PX4 Autopilot firmware as a flight controller and RealSense D455 as an infrared camera and rangefinder. During the research, a number of improvements were applied and missing functionality was developed. A modified Nvidia Ubuntu 20.04 image is used as the operating system on the on-board computer of the installation. The original Ubuntu 20.04 is used on the computer where the flight simulation is performed. To simplify the development and testing process, the popular ROS meta-operating system was used, specially designed to provide convenient functionality for robotic solutions. The data obtained in the study confirm the relevance of this work and the high demand in the market of visual positioning systems.