Modeling and Simulation for Design, Development, Testing and Evaluation of Autonomous Multi-Agent Models

ABSTRACT: The rapid continued development of unmanned air systems (UAS) is enabling new mission types, in-creased mission effects, and increased airman safety. However, these advances also present numerous challenges to airman-machine interaction, tactics development, and defense. The rapid development pace has produced a situation where new technologies are outpacing the knowledge of how best to use them. To maximize the effectiveness of automated and semi-automated systems in future conflicts we will develop a testbed that includes predictive models, which airmen can use to train, experiment with, and assess these new capabilities. The Configurable Adversary Response Prediction (CARP) system will provide predictive analytical human decision-making models that are accurate, navigable to systemati-cally explore spaces of predictions, adaptable to match realistic outcomes from data, and easy to inte-grate with existing distributed mission simulation environments. CARPs foundation rests on a sub-stantial legacy of high-fidelity tactical models developed by SoarTech. Our innovative approach will adapt model-building techniques for high-fidelity, data-driven behavior models to enable the systematic navigation of accurate and adaptable predictive behaviors spaces.; BENEFIT: Anticipated DOD Benefits: The research, development, and implementation of CARP will offer the DOD an unprecedented predic-tive what-if analysis capability for complex mission types (such as Anti-Access Area Denial, A2AD). CARPs incorporation of accurate and configurable decision-making and behavior models will support a usable and useful analytical capability that provides the following benefits: 1. Models that generate accurate predictions through a systematic exploration/navigation process. 2. Decision-making models that incorporate modern theories of human reasoning, as well as mod-ern techniques and representations for engineering human decision-making processes 3. The capability to analyze dynamically changing work, mission, and infrastructure configura-tions 4. Easy reconfigurability of red and blue forces, as well as systematic exploration of configuration settings to generate spaces of accurate predictions. 5. Adaptability of the models to increase predictive accuracy with experience and information from real-world and other data, using state-of-the-art machine learning techniques 6. Sharable and fully interoperable models and simulation environments, including existing LVC environments. Potential Commercial Applications: Accurate modeling of decision making is significant win them in corporate environments. The ability to accurately analyze and predict outcomes from decision-maker interactions is useful in training, strategy evaluation, negotiation, and numerous other business activities.