Monte Carlo Simulation-Based Methods for Establishing Solution Quality in Stochastic Programs

University Of Texas At Austin, Austin TX

Investigators

Abstract

This grant concerns the development of a statistical procedure for testing the quality of a feasible candidate solution for an important class of stochastic programs. Quality is defined via the so-called optimality gap and the output of the procedure is a confidence interval on this gap. Monte Carlo simulation-based tools will be employed to make the procedure efficient and to tighten the width of the resulting confidence interval. The method is not tied to specific solution algorithms. Hence, success will result in an approach that can be widely adopted for solving real-world stochastic programs. Stochastic programming is a powerful tool for making decisions in complex systems that contain significant uncertainties. Stochastic programming problems arise in many areas of application including financial planning, production capacity planning, managing natural resources, communications networks, and national security. Success with the proposed work will enable analysts to efficiently assess the quality of candidate solutions to such problems.