Abstract
This paper introduces an efficient adaptation of the branch-and-bound technique that solves real-world rostering problems for airline crews. The efficiency of the algorithm is based on the exploitation of rostering-specific properties (e.g. variable selection, branching strategy and cutting-planes). This approach shortens the solution process and outperforms standard techniques. Furthermore, we formally introduce a general concept of downgrading that makes it possible to solve certain rostering problems that might otherwise have no solution. This paper also computes a sample monthly schedule on the basis of a medium-sized European airline's real data. Scope and purpose The scheduling of airline crews and the assignment of crews to flights are important and difficult planning functions that most airlines undertake on a monthly basis. Solving a so-called rostering problem includes the construction of individualized schedules that take into account various pre-assignments (like training or observer flights), as well as crew requests (such as days off or preferred flights). The European rostering approach implies large scale and complex integer problems with 10 000 variables and several hundred constraints. This paper develops a new algorithm (SWIFTROSTER) that incorporates several strategies that exploit problem-specific knowledge in order to solve even large problems in very short runtimes, thus outperforming commercial solvers. Using real data from a medium-sized European airline the article demonstrates that our approach generates efficient solutions that can be applied in the real world to produce crew schedules.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have