A Heuristic Branch-and-Price Approach for the Airline Crew Pairing Problem

We describe a methodology for nding near-optimal solutions to airline crew pairing problems. We use a dynamic column generation scheme to identify crew work schedules combined with a customized branch-and-bound procedure that allows column generation to be performed at each node of the search tree. Our approach provides an approximation to optimality since we only solve the column generation subproblems approximately and we do not necessarily consider all of the unexplored nodes in the search. We present computational results for both a research implementation and a production implementation of the algorithm on test problems from a major domestic carrier. We test the innuence of various algorithmic design choices with the research implementation. These results were used to build a production implementation capable of nding good solutions to problem instances with over 2000 ight legs.