Fathoming rules for biobjective mixed integer linear programs: Review and extensions

We consider the class of biobjective mixed integer linear programs (BOMILPs). We review fathoming rules for general BOMILPs and present them in a unified manner. We then propose new fathoming rules that rely on solving specially designed LPs, hence making no assumption on the type of problem and only using polynomial-time algorithms. We describe an enhancement for carrying out these rules by performing a limited number of pivot operations on an LP, and then we provide insight that helps to make these rules even more efficient by either focusing on a smaller number of feasible solutions or by resorting to heuristics that limit the number of LPs solved. (C) 2016 Elsevier B.V. All rights reserved.