Automatic sensitivity analysis for an Army modernization optimization model

Abstract

This is an extremely turbulent period for the post-cold-war Army. Even though the tempo of operations has increased dramatically worldwide, a peace dividend is demanded, consequently placing substantial constraints on the Army's capital budgeting process. In spite of this, the senior leadership is determined to maintain a first-rate force capable of meeting the challenges of the future. Currently, the Office of the Deputy Chief of Staff for Operations and Plans (DCSOPS), United States Army, is reviewing a decision tool known as the Research, Development and Acquisition Alternative Analyzer (RDA3) to support the development of the Army Modernization Plan (AMP). RDA3 is a mixed integer optimization model formulated in the General Algebraic Modeling System (GAMS) by Donahue (1992). It prioritizes modernization actions and optimally allocates scarce research and development funds. The goal of this thesis work is to enhance RDA3 to provide the user with a more robust decision tool capable of providing a complete analysis of the entire decision space. Specifically, this study focuses on the unfunded investment projects in the RDA3 solution, which are collectively called the losers list. The idea is to automatically provide explanatory information as to why each project on the losers list is unfunded. This study uses techniques developed by Chinneck (1993) for identifying infeasibilities in linear programming models. Chinneck's techniques are specialized for the RDA3 context and extended to integer programming. Additionally, the idea of controlling the amount of change from one model run to another, known as persistence, is applied to RDA3.