Crew scheduling urban problem: an exact column generation approach improved by a genetic algorithm

Abstract

Many papers state that one of the best approaches to solve Crew Scheduling problems is by Column Generation. Generally a large number of columns must be handled, then the problem is decomposed and a subproblem is solved to generate the columns iteratively. This paper shows a successful application of genetic algorithm to solve the subproblem, improving the performance of the column generation algorithm, reaching the solution faster than using an integer programming package. The genetic algorithm is combined with an exact method, assuring the optimality of the final solution. The usual way to solve the subproblem is using integer programming. We compare this approach, the genetic algorithm, and a heuristic based on the linear relaxation of the subproblem formulation. We apply these algorithms to a crew scheduling problem that arises in the public transportation of a specific city. The results show that the genetic algorithm outperforms them.