Parallel integer optimization for crew scheduling

Performance aspects of a Lagrangian relaxation based heuristic for solving large 0-1 integer linear programs are discussed. In particular, we look at its application to airline and railway crew scheduling problems. We present a scalable parallelization of the original algorithm used in production at...

Full description

Saved in:
Bibliographic Details
Published in:Annals of operations research 2000-01, Vol.99 (1), p.141
Main Authors: Alefragis, Panayiotis, Sanders, Peter, Takkula, Tuomo, Wedelin, Dag
Format: Article
Language:English
Subjects:
Airlines
Algorithms
Heuristic
Linear programming
Mathematical programming
Operations research
Optimization
Scheduling
Studies
Work stations
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Performance aspects of a Lagrangian relaxation based heuristic for solving large 0-1 integer linear programs are discussed. In particular, we look at its application to airline and railway crew scheduling problems. We present a scalable parallelization of the original algorithm used in production at Carmen Systems AB, Goteborg, Sweden, based on distributing the variables. A lazy variant of this approach which decouples communication and computation is even useful on networks of workstations. Furthermore, we develop a new sequential active set strategy which requires less work and is better adapted to the memory hierarchy properties of modern RISC processors. This algorithm is also suited for parallelization on a moderate number of networked workstations. [PUBLICATION ABSTRACT]
ISSN:0254-5330
1572-9338
DOI:10.1023/A:1019293017474