Annals of Operations Research

, Volume 242, Issue 1, pp 133–160

Fast machine reassignment

  • Franck Butelle
  • Laurent Alfandari
  • Camille Coti
  • Lucian Finta
  • Lucas Létocart
  • Gérard Plateau
  • Frédéric Roupin
  • Antoine Rozenknop
  • Roberto Wolfler Calvo
Article

DOI: 10.1007/s10479-015-2082-3

Cite this article as:
Butelle, F., Alfandari, L., Coti, C. et al. Ann Oper Res (2016) 242: 133. doi:10.1007/s10479-015-2082-3
  • 128 Downloads

Abstract

This paper proposes a new method for solving the Machine Reassignment Problem in a very short computational time. The problem has been proposed by Google as subject of the Challenge ROADEF/EURO 2012. The Machine Reassignment Problem consists in looking for a reassignment of processes to machines in order to minimize a complex objective function, subject to a rich set of constraints including multidimensional resource, conflict and dependency constraints. In this study, a cooperative search approach is presented for machine reassignment. This approach uses two components: Adaptive Variable Neighbourhood Search and Simulated Annealing based Hyper-Heuristic, running in parallel on two threads and exchanging solutions. Both algorithms employ a rich set of heuristics and a learning mechanism to select the best neighborhood/move type during the search process. The cooperation mechanism acts as a multiple restart which gets triggered whenever a new better solution is achieved by a thread and then shared with the other thread. Computational results on the Challenge instances as well as instances of a Generalized Assignment-like problem are given to show the relevance of the chosen methods and the high benefits of cooperation.

Keywords

Generalized Assignment Adaptive Variable Neighborhood Search Simulated Annealing Hyper-Heuristic  Cooperative Parallel Search 

Copyright information

© Springer Science+Business Media New York 2015

Authors and Affiliations

  • Franck Butelle
    • 1
  • Laurent Alfandari
    • 1
    • 2
  • Camille Coti
    • 1
  • Lucian Finta
    • 1
  • Lucas Létocart
    • 1
  • Gérard Plateau
    • 1
  • Frédéric Roupin
    • 1
  • Antoine Rozenknop
    • 1
  • Roberto Wolfler Calvo
    • 1
  1. 1.LIPN, CNRS UMR 7030Université Paris 13, Sorbonne Paris CitéVilletaneuseFrance
  2. 2.ESSEC Business SchoolCergyFrance