Computationally difficult instances for the uncapacitated facility location problem

Abstract

Design and analysis of computationally difficult instances is one of the promising areas in combinatorial optimization. In this paper we present several new classes of benchmarks for the Uncapacitated Facility Location Problem. The first class is polynomially solvable. It has many strong local optima and large mutual pair distances. Two other classes have exponential number of strong local optima. Finally, three last classes have large duality gap and one of them is the most difficult for metaheuristics and the branch and bound method.