A GLOBAL APPROACH TO CREW-PAIRING OPTIMIZATION

The problem addressed in this paper is crew-pairing optimization in airline flight planning: finding tours of duty (pairings) that are legal and cover every flight leg at the least cost. The legal rules and cost of a pairing are determined by complex Federal Aviation Agency and contractual requirements. In addition, the problem is made more difficult by the hub-and-spoke system used by airlines that multiplies the possible ways a pairing can link flight legs. The state-of-the-art crew-pairing TRIP system of American Airlines uses subproblem optimization and, as is true for other crew-scheduling systems, may not be able to improve a solution even though a better one exists. We report on the methodology developed during a joint study by IBM and American Airlines Decision Technologies to use the IBM Optimization Subroutine Library in conjunction with TRIP to improve on crew-pairing solutions by taking a global approach. The resulting improvements have been a reduction of 5 to 11 percent in excess crew cost. Estimated total savings are five million dollars per year.