UTILIZING ADAPTIVE DESIGN OF EXPERIMENTS WITH LOGISTICS BATTLE COMMAND TO DETERMINE THE MOST IMPORTANT FACTORS FOR A DECISION MAKER

The Army makes extensive use of simulation models to obtain data on the operational effectiveness of future capabilities to inform acquisition decisions. A key step in this process is understanding how the key performance parameters of a new system impact the system’s performance in the operational environment. This enables decision makers to set threshold and objective values of these parameters to provide to industry as system requirements. Another challenge faced with new technology is understanding how tactics are potentially impacted by the new capability. This requires the exploration of new methods of employment. Efficiently exploring the interaction between key performance parameters and tactics within its simulation models remains a challenge for the analytic community. The analytic community makes use of some methods from the design of experiments literature but continues to require more efficient methods of searching the design space. Methods from the adaptive sequential design of experiments community or the field of machine learning could potentially be leveraged to address this need. These methods make use of the results of previous experimentation to inform the subsequent trials, potentially resulting in fewer costly simulation experiments. TRAC has identified the Logistics Battle Command (LBC) model as a candidate for proof-of-principle work in this area. The end result is a metamodel of LBC for the selected factors with significant factors highlighted.http://archive.org/details/utilizingadaptiv1094563507Major, United States ArmyApproved for public release; distribution is unlimited