Thesis HERRAMIENTAS DE INTELIGENCIA ARTIFICIAL PARA LA PLANIFICACIÓN DEL PROCESO DE COSECHA DE UVAS EN LA PRODUCCIÓN DE VINO CONSIDERANDO TURNOS Y ESTABILIDAD LABORAL
Date
2019-04
Authors
SFEIR YACHAN, IGNACIO ANDRÉS
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
El rubro de la agricultura, específicamente en el área estratégica, ha sido históricamente resistente al uso de nuevas tecnologías para el soporte de la toma de decisiones, apegándose al tradicional “buen juicio” del administrador, pero en los últimos años esto ha ido cambiando. En este trabajo se discute sobre el problema de optimización relacionado con el período de cosecha de uvas para vino, buscando una solución en base a un proceso objetivo de análisis de la información disponible. Existe trabajo previo en el área de cosecha de frutos y su planeación a nivel estratégico, pero no es un tema estudiado abundantemente. Las estrategias mayormente utilizadas se basan en algoritmos de búsqueda completa de soluciones, lo cual si bien permite encontrar la mejor solución posible requiere una simplificación del problema, alejándolo de la realidad. En este trabajo se estudia una versión del problema de cosecha de viñedos que considera el deterioro de la uva en el tiempo y la posibilidad de cosechar de forma manual o con maquinaria. Se modela y resuelve el problema usando el solver Gurobi con lo que se identifican los casos complejos de resolver. Además se presenta un algoritmo de tipo GRASP con HillClimbing que construye múltiples soluciones y las va modificando. De esta forma se explora gran parte del espacio de búsqueda, partiendo de diversos estados iniciales e intentando explotar al máximo cada uno. Se comparan los resultados del algoritmo para diversas instancias o situaciones posibles dela viña como el tamaño y la capacidad de procesamiento de la uva, incluyendo una instancia similar a trabajos previos análogos, evaluando qué condiciones son las que más fuertemente afectan al resultado final.
The agriculture market, specifically in the strategic area, has historically been resilient to the use of new technologies that support decision making. In most cases it is still attached to the traditional “good judgement” of the field’s manager. In recent years this has been changing. This thesis will discuss the optimization problem related with the grape-harvesting time-period, searching for a solution based on an objective analysis of the available information. There is previous work in the fruit harvesting area and its strategic planning, but it’s nota thoroughly studied subject. Their strategies are mostly based on exact algorithms, which are meant to find the best possible solution, but also require the problem to be simplified, distancing it from reality. In this thesis a version of the vineyard’s harvesting problem will be studied, which considers grape deterioration through time and the posibility to harvest manually or using machinery. The problem is modeled and solved using Gurobi solver to identify complex solution cases. Also, an algorithm is presented using GRASP with Hill Climbing that builds multiple solutions and then starts modifying them. In this way most of the search space is explored, by starting from different initial states and then trying to exploit each one to their fullest. The algorithm’s results are compared for diverse instances of possible scenarios of the vine-yard’s varying size and grape processing capacity. Moreover, an instance similar to previous studies is evaluated to determine which conditions are the most critical for the final result.
The agriculture market, specifically in the strategic area, has historically been resilient to the use of new technologies that support decision making. In most cases it is still attached to the traditional “good judgement” of the field’s manager. In recent years this has been changing. This thesis will discuss the optimization problem related with the grape-harvesting time-period, searching for a solution based on an objective analysis of the available information. There is previous work in the fruit harvesting area and its strategic planning, but it’s nota thoroughly studied subject. Their strategies are mostly based on exact algorithms, which are meant to find the best possible solution, but also require the problem to be simplified, distancing it from reality. In this thesis a version of the vineyard’s harvesting problem will be studied, which considers grape deterioration through time and the posibility to harvest manually or using machinery. The problem is modeled and solved using Gurobi solver to identify complex solution cases. Also, an algorithm is presented using GRASP with Hill Climbing that builds multiple solutions and then starts modifying them. In this way most of the search space is explored, by starting from different initial states and then trying to exploit each one to their fullest. The algorithm’s results are compared for diverse instances of possible scenarios of the vine-yard’s varying size and grape processing capacity. Moreover, an instance similar to previous studies is evaluated to determine which conditions are the most critical for the final result.
Description
Keywords
INTELIGENCIA ARTIFICIAL , INDUSTRIA VITIVINICOLA , UVAS