Implementación de técnicas basadas en tiempos de inactividad para incrementar la robustez ambientes inactividad para incrementar la robustez en ambientes Job Shop Sheduling Job Shop Sheduling
Archivos
Fecha
2015-07-04
Autores
Freire Gómez, Marlon
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Central “Marta Abreu” de Las Villas
Resumen
Un horario o secuencia de tareas es usualmente definido como ‘un plan para ejecutar un trabajo o alcanzar un objetivo, especificando el orden y el tiempo asignado para cada parte’. La secuenciación en los procesos de manufactura no es más que un conjunto de trabajos que tiene que ser ejecutado en un conjunto de máquinas, con el objetivo de encontrar la mejor secuencia.
Las investigaciones en el área de la secuenciación de tareas se han enfocado en el desarrollo de procedimientos para la generación de una solución base asumiendo que el ambiente en el que se ejecutan se comporta de manera estática. Sin embargo, el mundo real no es tan estable, los proyectos pueden ser sujetos a eventos inesperados durante su ejecución como roturas o mantenimientos de las máquinas lo cual puede llevar a numerosas interrupciones.
Para lidiar con dichas eventualidades se generan soluciones robustas, donde robustez se refiere al desempeño de un algoritmo bajo incertidumbre. El objetivo es obtener planificaciones que sean capaces de absorber cierto nivel de incertidumbre sin tener que replanificar aun cuando esto implica tiempo extra de procesamiento. Para lograr esto existen diferentes técnicas como por ejemplo los métodos basados en tiempos de inactividad cuyo objetivo es adicionar tiempos extra de procesamiento a aquellas operaciones que lo requieran.
En el presente trabajo se exponen dos técnicas de este tipo y ambas son implementadas usando aprendizaje reforzado.
A schedule is commonly defined as ‘a plan for performing work or achieving an objective, specifying the order and allotted time for each part’. Manufacturing scheduling is defined as an optimization process that allocates limited manufacturing resources over time among parallel and sequential manufacturing activities with the objective of finding the best sequence. The vast majority of the research in scheduling has focused on the development of exact and suboptimal procedures for the generation of a baseline schedule assuming complete information and a deterministic environment. However, the real world is not so stable; projects may be subject to unexpected events during execution, which may lead to numerous schedule disruptions. One approach to deal with such disruptions is to generate robust schedules, where robustness refers to the performance of an algorithm under uncertainties. The objective is to generate schedules that are able to absorb some level of uncertainty without having to reschedule, even if this implies extra processing time. Some techniques can deal with such problems, for example the slack-based techniques, where the main idea is to provide each activity with some extra execution time (slack time) in such a way that part of the uncertainty can be absorbed. In this paper, two techniques of this type are disclosed and both are implemented using reinforcement learning.
A schedule is commonly defined as ‘a plan for performing work or achieving an objective, specifying the order and allotted time for each part’. Manufacturing scheduling is defined as an optimization process that allocates limited manufacturing resources over time among parallel and sequential manufacturing activities with the objective of finding the best sequence. The vast majority of the research in scheduling has focused on the development of exact and suboptimal procedures for the generation of a baseline schedule assuming complete information and a deterministic environment. However, the real world is not so stable; projects may be subject to unexpected events during execution, which may lead to numerous schedule disruptions. One approach to deal with such disruptions is to generate robust schedules, where robustness refers to the performance of an algorithm under uncertainties. The objective is to generate schedules that are able to absorb some level of uncertainty without having to reschedule, even if this implies extra processing time. Some techniques can deal with such problems, for example the slack-based techniques, where the main idea is to provide each activity with some extra execution time (slack time) in such a way that part of the uncertainty can be absorbed. In this paper, two techniques of this type are disclosed and both are implemented using reinforcement learning.
Descripción
Palabras clave
Secuencia deTareas, Robustez, Aprendizaje Reforzado, Técnicas Basadas en Tiempos de Inactividad