Design of a visual servoing control strategy for a 4WD4WS path tracking robot in an agricultural field

• Autor: Avila López, Carlos Alberto
• Formato: tesis de maestría
• Fecha de Publicación: 2025
• Institución: Pontificia Universidad Católica del Perú
• Repositorio: PUCP-Tesis
• Lenguaje: inglés
• OAI Identifier: oai:tesis.pucp.edu.pe:20.500.12404/31842
• Enlace del recurso: http://hdl.handle.net/20.500.12404/31842
• Nivel de acceso: acceso embargado
• Materia: Robots móviles--Diseño; Robots móviles--Sistemas de control; Servomecanismos; Procesamiento de imágenes; Agricultura--Innovaciones tecnológicas; https://purl.org/pe-repo/ocde/ford#2.00.00

The search for automating agricultural tasks realization with the objective of increasing productivity and improving the exploit of resources using a ground mobile robot, arises the motivation to propose in the present thesis work, a visual servoing control strategy for the robot’s movement control over ridges in strawberry crop fields. In other words, the controller only uses a camera’s captured images for the robot’s path following. This robot has a four-wheel drive fourwheel steering (4WD4WS) structure on which the camera is mounted. For trajectory recognition, a U-Net convolutional neural network previously trained is used to perform semantic segmentation on captured images. From segmented images a point-extraction algorithm is applied in order to get the reference positions (set points) for the proposed controller. With this methodology, trajectory localization is feasible in less than 150 milliseconds with a root-meansquared- error (RMSE) of 6.99 pixels, using a U-Net neural network trained with 4200 crop field images. Meanwhile, with the visual servoing control algorithm it is possible to reach a settling time of around 12 seconds, being stable against robot slippage and image processing defects.