A3GENT: Army Adaptive Adversary Generation Technology

Project Summary

Principal Investigators

Grace Bochenek, University of Central Florida (UCF)

Faculty

Bulent Soykan, Ghaith Rabadi, UCF

Students

Government

Victor Paul, U.S. Army GVSC

Industry

Project #2.A127 began in 2025.

Virtual Crew Members (VCMs) and Digital Advisors (DAs) will become integral components of ground vehicle platforms, providing real-time tactical support and strategic guidance during missions. VCMs and DAs represent advanced virtual entities designed to emulate and enhance the roles of human operators, providing real-time tactical support and strategic guidance during missions. However, their effectiveness is inherently linked to the quality of training they receive. Current simulation systems lack the ability to emulate adversaries that can adapt their strategies, limiting the development of robust decision-making skills. This research project centers on leveraging adaptive adversaries to augment the training of VCMs and DAs within ground vehicle platforms. The foundational phase of training these VCMs involves human experts who impart essential knowledge, decision-making frameworks, and operational protocols. However, to achieve optimal performance and adaptability, VCMs and DAs require extensive and diverse training data that human-led training alone cannot sufficiently provide. To address this gap, the project employs simulation environments that generate a vast array of dynamic and realistic combat scenarios. Within these simulations, these adversaries function as Red Forces (REDFOR), embodying intelligent and adaptive adversarial agents capable of responding to and challenging the VCMs and DAs in real-time.

The primary objective of this project is to develop ADTs that can realistically simulate intelligent and adaptive adversaries in real-time, thereby enhancing the training of VCMs and DAs within ground vehicle platforms. By advancing the scientific and technical frontier in AI, ML, and simulation technologies, this research aims to create adversarial agents capable of dynamic adaptation, strategic decision-making, and realistic multidimensional threat simulation. To achieve this goal, the project is structured around three key research tasks, each addressing fundamental research questions that contribute to advancing the state-of-the-art in AI, simulation technologies, and military training methodologies.

Research Task 1 (RT1): Development of Adaptive AI Algorithms for Realistic Adversary Simulation
To create AI and ML algorithms that enable adversarial agents to learn, adapt, and respond dynamically to the actions of VCMs and DAs within simulation environments.

Research Task (RT2): Real-Time Data Integration and Automated Scenario Generation for Dynamic Simulation Environments
To develop methods for integrating real-time intelligence and operational data into ADTs, ensuring that adversarial behaviors and scenarios dynamically adjust based on evolving operational environments.

Research Task 3 (RT3): Multidimensional Threat Simulation and Game-Theoretic Strategy Integration
To simulate complex, multidimensional threats by employing multiagent systems and integrating game-theoretic approaches, enabling adversarial agents to plan and execute strategies with long-term objectives and simulate threats across various domains.

#2.A127