Adaptive Multi-Objective Local Search Algorithms for the Permutation Flowshop Scheduling Problem

Aymeric Blot (1, 2) , Marie-Éléonore Kessaci (1, 2) , Laetitia Jourdan (1, 2) , Patrick de Causmaecker (3)

1 ORKAD - Operational Research, Knowledge And Data
2 CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189
3 Department of Computer Science - K.U.Leuven

Automatic algorithm configuration (AAC) is an increasingly critical factor in the design of efficient metaheuristics. AAC was previously successfully applied to multi-objective local search (MOLS) algorithms using offline tools. However, offline approaches are usually very expensive, draw general recommendations regarding algorithm design for a given set of instances, and does generally not allow per-instance adaptation. Online techniques for automatic algorithm control are usually applied to single-objective evolutionary algorithms. In this work we investigate the impact of including control mechanisms to MOLS algorithms on a classical bi-objective permutation flowshop scheduling problem (PFSP), and demonstrate how even simple control mechanisms can complement traditional offline configuration techniques.

Domaines: Recherche opérationnelle [math.OC]

Aymeric Blot, Marie-Éléonore Kessaci, Laetitia Jourdan, Patrick de Causmaecker. Adaptive Multi-Objective Local Search Algorithms for the Permutation Flowshop Scheduling Problem. Learning and Intelligent Optimization Conference (LION 12), Jun 2018, Kalamata, Greece. ⟨hal-01868401⟩