Solving very large crew scheduling problems to optimality

Tallys H. Yunes, Arnaldo V. Moura, Cid C. De Souza

Research output: Chapter in Book/Report/Conference proceedingConference contribution

13 Scopus citations


In this article, we present a hybrid methodology for the exact solution of large scale real world crew scheduling problems. Our approach integrates mathematical programming and constraint satisfaction techniques, taking advantage of their particular abilities in modeling and solving specific parts of the problem. An Integer Programming framework was responsible for guiding the overall search process and for obtaining lower bounds on the value of the optimal solution. Complex constraints were easily expressed, in a declarative way, using a Constraint Logic Programming language. Moreover, with an effective constraint-based model, the huge space of feasible solutions could be implicitly considered in a fairly efficient way. Our code was tested on real problem instances arising from the daily operation of an ordinary urban transit company that serves a major metropolitan area with an excess of two million inhabitants. Using a typical desktop PC, we were able find, in an acceptable running time, an optimal solution to instances with more than 1.5 billion entries.

Original languageEnglish (US)
Title of host publicationProceedings of the 2000 ACM Symposium on Applied Computing, SAC 2000
PublisherAssociation for Computing Machinery
Number of pages6
ISBN (Print)1581132409, 9781581132403
StatePublished - Jan 1 2000
Externally publishedYes
Event2000 ACM Symposium on Applied Computing, SAC 2000 - Como, Italy
Duration: Mar 19 2000Mar 21 2000

Publication series

NameProceedings of the ACM Symposium on Applied Computing


Other2000 ACM Symposium on Applied Computing, SAC 2000


  • Column generation
  • Constraint programming
  • Crew scheduling
  • Hybrid algorithms
  • Mathematical programming

ASJC Scopus subject areas

  • Software


Dive into the research topics of 'Solving very large crew scheduling problems to optimality'. Together they form a unique fingerprint.

Cite this