Using constraint propagation to accelerate column generation in aircraft scheduling

Abstract

We discuss how the use of Constraint Programming can help speed up a Column Generation process for the Tail Assignment problem. A generalized preprocessing technique based on constraint propagation is presented that can remove up to 98% of the connections of the flight network, much more than methods currently in use. The result is substantial speed-up and increased solution quality for the column generator. We also show how propagation can be used within fixing heuristics, and present a heuristic propagation-based preprocessing method taking cost into consideration. Computational results on data from several airlines is presented.