Flight crew scheduling

Abstract

A new crew scheduling optimization system has been developed for United Airlines. The system was developed to permit quick response to schedule changes and to reduce crew scheduling costs. It was designed to work efficiently for both the medium sized problems (300 flights daily) and the very large problems (1,700 flights daily) that United must solve. The system has two main components, a generator and an optimizer. The generator creates pairings (candidate crew trips) which are fed as variables to the optimizer as an elastic embedded set partitioning integer programming problem. The optimizer then seeks to find a set of pairings that covers all of the flight segments exactly once with the minimal cost. Once a disjoint solution has been found, the system cycles between the generator and the optimizer to improve it. Savings of $16,000,000 annually in crew scheduling costs have been obtained.