Generating Crew Pairings for Very Large Flight Networks

Airline operations depend on the complex control of many expensive, tightly-constrained, interdependent resources such as crews and aircraft. Operations research (OR) has played an important role in developing planning tools for this complex system. In recent years, however, the already intense pressure on the airline industry has dramatically increased due to rising fuel costs, competitive pricing, increased congestion, and security concerns. Important research opportunities still exist in the areas of integrated planning, robust planning, and optimization-based recovery tools. One outstanding challenge is that all of these require the simultaneous consideration of aircrafts and crews, which are typically treated independently in current systems. The result is far larger instances of the crew scheduling problem than are currently being solved, because this problem can no longer be decomposed by fleet type. Existing methods, which typically rely on enumeration-based approaches to solve the imbedded crew pairing problem, become intractable for these larger problem instances. In this paper we introduce an alternative approach, based on integer programming, to generate crew pairings over very large flight networks. Computational results based on a major U.S. carrier are provided.