File:A comparison study of JANUS and HLA Warrior (IA acomparisonstudy109459258).pdf

The Training & Doctrine Command (TRADOC) Analysis Center (TRAC), Monterey, California, re-engineered the Janus simulation as a technology demonstration. The completed simulation, HLA Warrior, applied modern technologies including an object-oriented design and state-of-art user interfaces. The project also re-wrote the Janus source code in C++. The purpose of this thesis was to assess HLA Warrior's fidelity, defined as its ability to replicate Janus results, by conducting a statistical comparison of Janus and HLA Warrior. Given that Janus has high \"face-validity,\" Janus results acted as the baseline from which HLA Warrior results were compared. The comparison involved executing identical scenarios in Janus and HLA Warrior, gathering results, and conducting a rigorous statistical comparison of Janus and HLA Warrior results. Statistical tests included the paired t-test and non-parametric Wilcoxon Signed Ranks Test. Results from the tests showed differences between Janus and HLA warrior. Investigation into the causes of the differences found two source code errors in HLA Warrior. Re-evaluation of HLA Warrior following correction of the errors resulted in a reduction in magnitude of the differences. Probable causes due to algorithm implementation differences were also identified. While differences exist, HLA Warrior appears to have face-validity and generally produces outcomes similar to Janus.