Heuristics and stochastic local search for just-in-time scheduling of parallel machines against common restrictive due windows

• Ano de defesa: 2020
• Autor(a) principal: Rolim, Gustavo Alencar
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Atraso e adiantamento; Busca local estocástica; Common due window; Earliness and Tardiness; Heurísticas; Heuristics; Janelas de tempo comuns; Máquinas paralelas; Parallel machines; Scheduling; Sequenciamento; Stochastic local search
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/18/18156/tde-16112021-120034/

This study deals with earliness and tardiness scheduling around common due dates and windows. Related problems share a strong practical importance due to the endogenous and exogenous cost structures that occur within just-in-time production systems. Despite of their significance, there are no recent research that holistically address the two main variants of the problems in terms of the common due date (window) approach. We fill this gap by surveying over 200 papers, which cover more than 40 years of scheduling literature and classifying algorithms for problems where the due date (window) is a given constraint as well as the ones where it is a decision variable. New notations and a comprehensive framework that includes a wide range of earliness and tardiness scheduling problems are introduced. Moreover, a total of 26 structural properties and the computational complexity of available algorithms for single, parallel, and flow-shop machine environments are summarized. From the literature review, we selected the just-in-time parallel machine scheduling problem against common restrictive due windows that is known to be NP-hard. Since there is no procedure available for solving instances with more than 40 jobs, we propose a family of heuristics and report promising results for instances with up to 500 jobs in size. Moreover, the best performing heuristic is combined with a stochastic local search (iterated greedy algorithm) to improve the solutions previously found.