Proposta de uma arquitetura de posicionamento visual para aplicações em VANTS autônomos colaborativos

Abstract

Autonomous UAVs offer advantages in industrial, agriculture, environment inspection, and logistics applications. Sometimes using cooperative UAVs is important to solve specific demands or achieve productivity gains in these applications. An important technical challenge is the precise positioning between two or more UAVs in a cooperative task flight. Some techniques provide solutions, e.g. the GNSS positioning, visual and LIDAR SLAM, and computer vision intelligent algorithms, but all these techniques present limitations that must be solved to work properly in specific environments. The proposal of new cooperative position methods is important to face these challenges. The present work proposes an evaluation of a visual relative positioning architecture between two small UAV multi-rotor aircrafts working in a master-slave operation, based on an Augmented Reality tag tool. The simulation results showed absolute error measurements lower than 0.2cm in avarage and 0.01 standard deviation for X, Y and Z directions. The real-world experiments executing flight with the slave UAV commanded by the master UAV achieved success in 8 out of 10 experiment rounds, indicating that the proposed architecture is a good approach to build cooperative master-slave UAV applications.