Bus driver rostering by column generation metaheuristics

Abstract

In the Bus Driver Rostering Problem (BDRP) it is intended to define work schedules for workers such that costs are minimized. This problem has been addressed before by combining column generation and an evolutionary algorithm. In this paper, we show how this approach can be improved by including additional random constraints and limiting the time spent in column generation. Both approaches follow a general framework entitled SearchCol, Metaheuristic search by column generation, recently proposed for decomposable integer programming / combinatorial optimization problems.