Feasability Study of Variance Reduction in the THUNDER Campaign-Level Model
Abstract
As an Air Force Chief of Staff endorsed topic, Air Force Studies and Analyses Agency (AFSAA) requested an effective and efficient way to reduce the variance in analysis results from THUNDER. THUNDER is a large-scale discrete event simulation of campaign-level military operations and is used to examine issues involving the utility and effectiveness of air and space power in a theater-level, joint warfare context. Given the large number of stochastic components within THUNDER, results are produced with highly variable measures of effectiveness (MOEs), causing difficulties in evaluating alternative force structures, weapon systems, etc. This work responds to AFSAA's request by examining the application of Common Random Numbers (CRN), Antithetic Variates (AV), Control Variates (CV), and a combination of AVs and CVs. The difference between the standard output and variance reduced output halfwidths for 95% confidence intervals were examined. Analysis of the correlation between MOEs and the random inputs in the CV technique provided insight into the workings of THUNDER. A new, state of the art combined multiple recursive generator was incorporated into THUNDER to synchronize the random inputs for CRN and AV. The result is methodology for implementing all four variance reduction techniques.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2005
- Accession Number
- ADA441639
Entities
People
- Earl M. Bednar
Organizations
- Air Force Institute of Technology