Making a Case for Robust Optimization Models

Abstract

Robust optimization searches for recommendations that are relatively immune to anticipated uncertainty in the problem parameters. Stochasticities are addressed via a set of discrete scenarios. This paper presents applications in which the traditional stochastic linear program fails to identify a robust solution—despite the presence of a cheap robust point. Limitations of piecewise linearization are discussed. We argue that a concave utility function should be incorporated in a model whenever the decision maker is risk averse. Examples are taken from telecommunications and financial planning.