Configuration of a Dynamic MOLS Algorithm for Bi-objective Flowshop Scheduling

Camille Pageau (1, 2) , Aymeric Blot (1, 2) , Holger Hoos (3) , Marie-Éléonore Kessaci (1, 2) , Laetitia Jourdan (1, 2)

1 ORKAD - Operational Research, Knowledge And Data
2 CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189
3 UBC - University of British Columbia

In this work, we propose a dynamic multi-objective local search (MOLS) algorithm whose parameters are modified while it is running and a protocol for automatically configuring this algorithm. Our approach applies automated configuration to a static pipeline that sequentially runs multiple configurations of the MOLS algorithm. In a series of experiments for well-known benchmark instances of the bi-objective permutation flowshop scheduling problem, we show that our dynamic approach produces substantially better results than static MOLS, and that longer pipeline (with a higher number of parameters) outperform shorter ones.

Domaines: Informatique [cs]; Recherche opérationnelle [math.OC]; Combinatoire [math.CO]

Camille Pageau, Aymeric Blot, Holger Hoos, Marie-Éléonore Kessaci, Laetitia Jourdan. Configuration of a Dynamic MOLS Algorithm for Bi-objective Flowshop Scheduling. Evolutionary Multi-Criterion Optimization, Mar 2019, Lansing, United States. pp.565-577. ⟨hal-02078581⟩