A branch-and-price algorithm for the bilevel network maintenance scheduling problem

David Rey, Hillel Bar-Gera, Vinayak V. Dixit, S. Travis Waller

Research output: Contribution to journalArticlepeer-review

7 Scopus citations

Abstract

We address the network maintenance scheduling problem, which consists of finding the optimal schedule for the coordination of road maintenance projects in a transport network over a planning period. Road works and maintenance operations that require partial or total road closures over a period of time may considerably impact network performance and result in significant delays. In addition, the effects of conducting multiple maintenance projects simultaneously may be nonadditive, hence increasing the difficulty to anticipate congestion effects. In this paper, we propose a new bilevel mixed integer programming formulation for the network maintenance scheduling problem, which relies on the enumeration of maintenance project combinations-patterns-to incorporate congestion effects within the scheduling process. We present a new branch-andprice algorithm that relies on customized branching and bounding rules, and a tailored column generation framework to price patterns. In addition, a statistical regression model is introduced to approximate congestion effects and provide approximate lower bounds on the formulations therein. The proposed branch-and-price algorithm is implemented on instances derived from realistic transport networks and is shown to be able to solve the network maintenance scheduling problem in a reasonable time using only a fraction of the patterns.

Original languageEnglish
Pages (from-to)1455-1478
Number of pages24
JournalTransportation Science
Volume53
Issue number5
DOIs
StatePublished - 1 Jan 2019

Keywords

  • Bilevel optimization
  • Branch and price
  • Column generation
  • Mixed integer programming
  • Network maintenance
  • Scheduling
  • Traffic assignment

Fingerprint

Dive into the research topics of 'A branch-and-price algorithm for the bilevel network maintenance scheduling problem'. Together they form a unique fingerprint.

Cite this