2-DOF Laser Structure for Autonomous External Quadrotor Guidance

• Autor(a) principal: Soler, Diego Pavan
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/18/18149/tde-12052020-175028/
• topic: agricultura de precisão; computer vision; cooperação robótica; precision agriculture; quadrotor; quadrotor; robot cooperation; UAV; VANT; visão computacional

This dissertation, 2-DOF Laser Structure for Autonomous External Quadrotor Guidance, has the goal of designing a novel method to attract an aerial robot to a ground vehicle using a system composed of a camera and a laser for autonomous landing. This applications is designed to be used alongside other guiding methods. This is a complex problem that requires work on different fields on both the ground and the aerial robot. On the ground station, know as Mirã, a structure was designed using 3D printed parts, and two stepper motor adjusted for 1:16 microstepping to precisely orient a camera and a laser with 2 DoF. The motor is controlled by a computer vision algorithm that uses four green LED on the aerial robot. With the combination of a light source filter and a color filter along with metric filters, the algorithm can find the LEDs position on the image. By estimating the distance of the center of the LED formation and the center of the image, the algorithm uses the camera calibration parameters to convert it to azimuth and elevation angles, allowing the structure to follow and target the laser on the aerial robot with a error of elevation = 4.96º and azimuth = 0.27º with a standard deviation of 1.03 and 0.51 degree. To call the aerial robot, the structure will perform a linear cut on the elevation motor, because of this, the azimuth precision is made much higher. On the aerial vehicle, using a 6 x 6 matrix of photodectors, a 0.5 x 0.5 m sensor was designed to be capable of estimating the position of the laser dot on the sensor at a rate of 200 Hz. The aerial robot will decode the sensor reading and acquire multiples instances of the readings, once possible, a linear approximation of the point collection is used to estimate a vector, the direction of the vector is estimated by looking at the older and newer points, resulting in a velocity vector which the aerial robot will follow. Once the vehicle is close enough, the structure will perform a cut in the opposite direction, commanding the aerial robot to stop and allowing for reliable autonomous landing.