Algoritmo multi-objetivo para la solución de problemas de secuenciación de tareas tipo Job Shop
Fecha
2016-11-10
Autores
Rodríguez Bazán, Erick David
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Central “Marta Abreu” de Las Villas
Resumen
Durante las últimas décadas, los algoritmos multi-objetivo han captado considerable atención. El problema de secuenciación tipo Job Shop, uno de los principales problemas de optimización, también ha generado el interés de la comunidad científica. Sin embargo, la mayoría de los algoritmos multi-objetivo que se han desarrollado para este problema están basados en enfoques no profesionales, es decir, la mayoría de ellos combinan sus objetivos y resuelven el problema multi-objetivo a través de enfoques basados en un solo objetivo, exceptuando a un reducido grupo de investigadores que utilizan algoritmos basados en la Frontera de Pareto. Además, gran parte de la literatura solo permite optimizar entre dos y tres objetivos, y estos están predefinidos de antemano. En esta tesis se implementa un algoritmo multi-objetivo usando sistemas multi-agente y Aprendizaje Reforzado basado en la Frontera de Pareto que permite optimizar el makespan, el tiempo de inactividad de las máquinas, la tardanza total y la robustez. El algoritmo fue implementado de manera genérica para si se desea, agregar un nuevo objetivo, solo sea necesario incluir una nueva clase. El algoritmo propuesto se aplicó a un conjunto de instancias para optimizar los objetivos tratados y finalmente los resultados experimentales muestran que el algoritmo implementado en esta tesis es capaz de superar a otros propuestos en la literatura y desde el punto de vista multi-objetivo cumple con las métricas utilizadas para evaluarlo.
In recent decades, the multi-objective algorithms have captured significant attention. Job shop scheduling problems are one of the most complex scheduling problems and are commonly encountered in manufacturing industries. Most of the studies are based on optimizing one objective, excepting a reduced group of researchers that use the Pareto Front approach. Besides, generally, in these approaches, it is only possible to optimize two or three objectives which are predefined with anticipation. In this thesis, we implemented an Multi-objective and Multi-agent Reinforcement Learning algorithm based on the Pareto Front that optimizes the makespan, machines idle times, tardiness and robustness. The algorithm is generic because it allows to add a new objective if you desire, and for that it only necessary to include a new class. The proposed algorithm was applied to a set of instances to optimize different objectives and finally experimental results show that the algorithm implemented in this thesis is able to improve other algorithms proposed in the literature and it gives good results when multi-objective metrics are used to evaluate it.
In recent decades, the multi-objective algorithms have captured significant attention. Job shop scheduling problems are one of the most complex scheduling problems and are commonly encountered in manufacturing industries. Most of the studies are based on optimizing one objective, excepting a reduced group of researchers that use the Pareto Front approach. Besides, generally, in these approaches, it is only possible to optimize two or three objectives which are predefined with anticipation. In this thesis, we implemented an Multi-objective and Multi-agent Reinforcement Learning algorithm based on the Pareto Front that optimizes the makespan, machines idle times, tardiness and robustness. The algorithm is generic because it allows to add a new objective if you desire, and for that it only necessary to include a new class. The proposed algorithm was applied to a set of instances to optimize different objectives and finally experimental results show that the algorithm implemented in this thesis is able to improve other algorithms proposed in the literature and it gives good results when multi-objective metrics are used to evaluate it.
Descripción
Palabras clave
Algoritmo Multiobjetivo, Solución de Problemas, Secuenciación, Tareas Tipo Job Shop