Enriched metaheuristics for the resource constrained unrelated parallel machine scheduling problem

• Published in: Computers & operations research Vol. 111; pp. 415 - 424
• Main Authors: Vallada, Eva, Villa, Fulgencia, Fanjul-Peyro, Luis
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 01.11.2019; Pergamon Press Inc
• Subjects: Additional scarce resource; Constraints; Enrichment; Heuristic methods; Job shops; Maintenance; Makespan; Metaheuristics; Operations research; Optimization; Production scheduling; Scattering; Scheduling; Search algorithms; Unrelated parallel machine; Unrelated parallel machine; Scheduling; Makespan; Additional scarce resource; Metaheuristics
• ISSN: 0305-0548; 1873-765X; 1873-765X; 0305-0548
• DOI: 10.1016/j.cor.2019.07.016

•The problem of unrelated parallel machines with an additional resource is considered.•Enriched metaheuristic algorithms based on a Scatter Search and an Iterated Greedy are proposed.•A comprehensive computational and statistical evaluation is carried out.•The enriched methods are shown to outperform previous results from the literature.•The enriched methods are able to solve large problems efficiently and effectively. A Scatter Search algorithm together with an enriched Scatter Search and an enriched Iterated Greedy for the unrelated parallel machine problem with one additional resource are proposed in this paper. The optimisation objective is to minimise the maximum completion of the jobs on the machines, that is, the makespan. All the proposed methods start from the best known heuristic for the same problem. Non-feasible solutions are allowed in all the methods and a Repairing Mechanism is applied to obtain a feasible solution from a resource constraint point of view. All the proposed algorithms apply different local search procedures based on insertion, swap and restricted neighbourhoods. Computational experiments are carried out using an exhaustive benchmark of instances. After analysing the results, we can conclude that the enriched methods obtain superior results, outperforming the best known solutions for the same problem.