Elsevier

Operations Research Letters

Volume 35, Issue 5, September 2007, Pages 660-668
Operations Research Letters

Interior point stabilization for column generation

https://doi.org/10.1016/j.orl.2006.11.004Get rights and content

Abstract

Interior point stabilization is an acceleration method for column generation algorithms. It addresses degeneracy and convergence difficulties by selecting a dual solution inside the optimal space rather than retrieving an extreme point. The method is applied to the case of the vehicle routing problem with time windows.

Introduction

Column generation was introduced by Dantzig and Wolfe [3] to solve linear programs with decomposable structures. It has been applied to many problems with success and has become a leading optimization technique to solve routing and scheduling problems [5], [1]. However, some column generation methods often show very slow convergence partly due to heavy degeneracy problems. One of these problems arises when multiple dual solutions are associated with each primal solution. Choosing the dual solution then becomes a crucial part of the column generation algorithm as the subproblem solution often heavily depends on the dual values. This observation led to the introduction of several stabilization methods, which attempted to accelerate convergence by implementing diverse mechanisms to control the selection of the dual solution. The method proposed here tries to achieve the same goal by using a different approach.

The next section, which gives a brief overview of the column generation framework, is followed by a short description of some existing stabilization methods that we will use for comparison purposes. Section 3 introduces interior point stabilization and results are reported in Section 4.

Future work on this technique involves its application to different problems in order to clearly assess in which contexts it can be useful.

Section snippets

Column generation

Column generation is a general framework that can be applied to numerous problems. However, since an example is often useful to give a clear explanation, all results in this paper will be presented with respect to the vehicle routing problem with time windows (VRPTW). The description of IPS is done without loss of generality nor restrictions to deal with this particular problem. Although it cannot be said that the VRPTW is representative of all domain applications solved by column generation,

Stability problems

Column generation frameworks depend heavily on marginal costs to guide the search at the subproblem level. In some cases it is possible that, during the first iterations, the marginal costs associated with each customer are not appropriately estimated by the dual values. For instance, it is possible that in some routes some customers pick up most of the total dual values. If this is the case (as illustrated in Fig. 1), then in the subproblem a path that visits each of those overweighted

Interior point stabilization

The idea behind interior point stabilization is to generate a dual solution that is in the interior of the convex hull of optimal dual solutions to the master instead of using one of its extreme points. Bixby et al. [2] have proposed to achieve this by using an interior point method to solve (M). The authors note that, although the interior point method requires less iterations than a simplex based method, each iteration takes more time. They argue that each call to the subproblem generates a

Experimental results

In this section, we attempt to demonstrate the effectiveness of interior point stabilization with regard to two criteria. We first evaluate IPS as a simple (almost parameter free) stabilization method for column generation and then compare it to the stabilization techniques described in Section 2. The problem chosen to perform experiments is the vehicle routing problem with time windows (VRPTW).

References (16)

  • J. Desrosiers et al.

    Time constrained routing and scheduling

  • O. du Merle et al.

    Stabilized column generation

    Discrete Math.

    (1999)
  • C. Barnhart et al.

    Branch-and-price: column generation for huge integer programs

    Oper. Res.

    (1998)
  • R.E. Bixby, J.W. Gregory, I.J. Lustig, R.E. Marsten, D.F. Shanno, Very large-scale linear programming: a case study in...
  • G.B. Dantzig et al.

    Decomposition principles for linear programs

    Oper. Res.

    (1960)
  • M. Desrochers et al.

    A new optimisation algorithm for the vehicle routing problem with time windows

    Oper. Res.

    (1992)
  • O. du Merle, D. Villeneuve, J. Desrosiers, P. Hansen, Stabilisation dans le cadre de la génération de colonnes,...
  • D. Feillet, P. Dejax, M. Gendreau, C. Gueguen, An exact algorithm for the elementary shortest path problem with...
There are more references available in the full text version of this article.

Cited by (73)

  • Upward scalable vehicle routing problem of automobile inbound logistics with pickup flexibility

    2023, Transportation Research Part E: Logistics and Transportation Review
  • A robust approximation for yard template optimization under uncertainty

    2022, Transportation Research Part B: Methodological
  • The benefits of autonomous vehicles for community-based trip sharing

    2021, Transportation Research Part C: Emerging Technologies
  • An exact algorithm for inland container transportation network design

    2020, Transportation Research Part B: Methodological
View all citing articles on Scopus
View full text