AutoBenchmarking

The selection of model fidelity in military simulations, particularly for the Missile Defense Agency (MDA), presents a challenge due to the high computational cost of high-fidelity models. These costs limit the exploration of operational scenarios and inhibit rapid prototyping of new technologies in digital environments. Introducing a range of model fidelities could significantly enhance the analytic research environment, providing the flexibility to choose the appropriate model for each study, thereby reducing computational costs and expanding the evaluation range. The development of benchmarks to measure model accuracy against a trusted data source is central to this approach as it enables automatic model tuning to optimize simulations. Under Phase I of Benchmarking Analysis for Rapid Transition (BART), Infinity Labs successfully developed methods to automatically benchmark and tune computationally fast ballistic missile threat models. This Phase II solution amplifies and optimizes benchmarking techniques for a greater impact to MDA’s ability to develop and field new technologies. The development of a software toolkit, leveraging the Advanced Framework for Simulation and Integrated Modeling (AFSIM), will enablie MDA to (1) intelligently benchmark digital models of interceptors, threats, and sensors (2) Identify system performance gaps and limitations (3) Visualize, understand, and analyze performance of digital models. The result of this technology reduces MDA’s cost to rapidly innovated, prototype, and test current and future capabilities against a dynamic threat environment. Approved for Public Release | 25-MDA-12220 (23 Sep 25)