Uso de inteligencia artificial generativa para la optimización de parámetros de vuelo de drones aplicado a la fumigación foliar de frutales
Loading...
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
La aplicación foliar en frutales mediante el uso de drones fumigadores es una tecnología que se encuentra en constante crecimiento, de hecho, en los últimos años se han realizado numerosos estudios, principalmente en Asia con la finalidad de demostrar la eficiencia de los drones fumigadores en la agricultura, ya sea combatiendo plagas o enfermedades en específico, por lo que encontrar los parámetros óptimos de vuelo de estos equipos, estrecharía la brecha existente sobre el uso correcto de los plaguicidas, contribuyendo a la reducción del impacto ambiental y a la eficiencia en el uso de los agroquímicos en cultivos frutales. Esta investigación tiene como objetivo determinar los parámetros óptimos de vuelo para un dron DJI Agras T20 en frutales de cerezos y kiwis, por lo que se realizaron 18 vuelos por cada especie frutal en las que se combinó la altura, velocidad y densidad del fluido. Luego se cuantificaron los impactos/cm2 y el tamaño de los impactos que registraron los papeles hidrosensibles. Posteriormente se entrenó un modelo de lenguaje extenso (LLM) para predecir parámetros de vuelo utilizando los resultados de esta investigación como aprendizaje para futuras aplicaciones. Esto demostró que estos modelos pertenecientes a la Inteligencia Artificial Generativa si
sirven para optimizar vuelos de fumigaciones y son capaces de predecir parámetros de vuelo dándoles condiciones específicas. Se determino que para frutales hilerados la altura optima de vuelo es de 3 metros sobre el cultivo y para parrones es de 4 metros, además, que la velocidad dependía del objetivo de la fumigación, ya que velocidades más lentas favorecen la cobertura por el envés de las hojas. Además, se demostró que la variable mas influyente es la densidad de la mezcla, ya que, a mayor densidad disminuye drásticamente la cobertura, tanto en el haz como envés de las hojas.
The foliar application of nutrients in fruit trees using spraying drones is an emerging technology experiencing rapid growth. In recent years, numerous studies, primarily conducted in Asia, have explored the efficiency of spraying drones in agriculture, particularly in controlling specific pests and diseases. Optimizing the flight parameters of these devices could help bridge the existing gap in the proper use of pesticides, thereby reducing environmental impact and enhancing the efficiency of agrochemical applications in fruit crops." This study aims to determine the optimal flight parameters for a DJI Agras T20 drone in cherry and kiwi orchards. A total of 18 flights were conducted for each fruit species, varying altitude, speed, and fluid density. The number of impacts per cm² and the impact size recorded on hydrosensitive paper were then quantified. Subsequently, a large language model (LLM) was trained to predict flight parameters, using the study results as training data for future applications. The findings demonstrate that generative artificial intelligence models can effectively optimize spraying flights and predict flight parameters under specific conditions. The study determined that for row-planted fruit trees, the optimal flight altitude is 3 meters above the crop, while for trellis systems, it is 4 meters. Additionally, flight speed varies depending on the spraying objective, as lower speeds enhance coverage on the underside of leaves. Furthermore, the results indicate that the most influential factor is the mixture density, as higher density significantly reduces coverage on both the upper and lower leaf surfaces.
The foliar application of nutrients in fruit trees using spraying drones is an emerging technology experiencing rapid growth. In recent years, numerous studies, primarily conducted in Asia, have explored the efficiency of spraying drones in agriculture, particularly in controlling specific pests and diseases. Optimizing the flight parameters of these devices could help bridge the existing gap in the proper use of pesticides, thereby reducing environmental impact and enhancing the efficiency of agrochemical applications in fruit crops." This study aims to determine the optimal flight parameters for a DJI Agras T20 drone in cherry and kiwi orchards. A total of 18 flights were conducted for each fruit species, varying altitude, speed, and fluid density. The number of impacts per cm² and the impact size recorded on hydrosensitive paper were then quantified. Subsequently, a large language model (LLM) was trained to predict flight parameters, using the study results as training data for future applications. The findings demonstrate that generative artificial intelligence models can effectively optimize spraying flights and predict flight parameters under specific conditions. The study determined that for row-planted fruit trees, the optimal flight altitude is 3 meters above the crop, while for trellis systems, it is 4 meters. Additionally, flight speed varies depending on the spraying objective, as lower speeds enhance coverage on the underside of leaves. Furthermore, the results indicate that the most influential factor is the mixture density, as higher density significantly reduces coverage on both the upper and lower leaf surfaces.
Description
Tesis presentada para optar al título de Ingeniero Civil Agrícola
Keywords
Inteligencia artificial, Drones, Fumigación