We study linear programming models that contain transportation constraints in their formulation. Typically, these models have a multi-stage nature and the transportation constraints together with the associated flow variables are used to achieve consistency between consecutive stages. We describe how to reformulate these models by projecting out the flow variables. The reformulation can be more desirable since it has fewer variables and can be solved faster. We apply these ideas to reformulate two well-known workforce staffing and scheduling problems: the shift scheduling problem and the tour scheduling problem. We also present computational results.
IBM Research Tech. Report. Jan, 2002