Estratégias algorítmicas exatas e híbridas para problemas de escalonamento em máquinas paralelas com penalidades de antecipação e atraso

Abstract

This research investigates scheduling problems with earliness and tardiness penalties on single and parallel machine environments. This problem is also known in the literature as Just-in-Time scheduling, system widely used in industries to reduce inventories and costs, in order to lead product to be produced according to demand. In this work we present a hybrid exact-heuristic algorithmic strategy, based on an arc-time indexed integer programming formulation and a generalized evolutionary heuristic based on a strong local search, to better solve classical parallel machine scheduling problems involving weighted earliness-tardiness penalties, with independent jobs and arbitrary processing times. Selected arcs from local optima solutions generated by a genetic algorithm based on a strong local search (GLS) with generalized pairwise interchanges are given as input to the arc-time formulation, to produce better solutions than those obtained by both methods when used isolated. Computational experiments present competitive results according to the literature. Our proposed method also solves large instances up to 500 jobs in identical parallel machines.