Toward finding driving communications factors in the system of systems survivability simulation model

Abstract

The System of Systems Survivability Simulation (S4) was created by the Army Research Laboratory's Survivability and Lethality Analysis Directorate in cooperation with the New Mexico State University Physical Science Laboratory. S4 is a multi-level, agent-based, time-stepped, high resolution, stochastic combat model with a focus on survivability and lethality of equipment and forces. There are over 300 factors (or input parameters) used to define the elements on the simulated battlefield. This thesis explores a factor screening method using a supersaturated design that could be used to eliminate insignificant design parameters for given scenarios. Eliminating insignificant parameters could reduce the run-time of an experiment, thereby allowing a more robust design to be used only on the significant factors that are selected. The ability of the method to properly identify significant parameters is analyzed by creating a model in which the significant factors are already known and determining how well the method identifies the significant factors. The results of the analysis show that the method is effective when the factors are moderately to highly significant and for a small number of significant factors. Additional research comparing this method with other factor screening methods may lead to the use of this method when there are more factors than design points.