Advanced Adversary Forces Artificial Intelligence (3xAI) Agent

Nov. 1, 2022
How SEI is developing advanced artificial intelligence agents using VBS technology
Systems Engineering, Inc. (SEI), a military contractor in Sterling, Virginia, is working with Bohemia Interactive Simulations (BISim^TM) to develop its Advanced Adversary Forces Artificial Intelligence (3xAI) agent, automatically learning the enemy’s most dangerous course of action (COA) that best counters and challenges existing operational plans.

Operational planning challenges faced by Military Leadership

Conventional military planning requires leaders to perform multiple iterations of what/if scenarios and develop contingencies for all foreseeable enemy actions. However, the what/if scenarios can contain historical and institutional biases, which can lead to faulty assumptions of opponent behavior, overconfidence in a selected COA, and a lack of mitigation planning. History, recent and ancient, provides many examples where leadership was unable to account for the enemy’s most dangerous COA.

An innovative partnership for Defense with AI and simulation technologies

From his PhD research in Systems Engineering in AI and Competitive Simulations, Dr. Roy Hayes drove the development and partnership with BISim to answer the funded challenge from the Office of Naval Research (ONR). Using BISim’s VBS4 and World Server to learn tactics and COAs in real-world environments, an artificial intelligence developed by SEI equips military leaders to identify successful tactical approaches and critical vulnerabilities, that increase lethality and survivability.

3xAI is designed to address the challenge of anticipating the enemy’s most dangerous COA, by identifying tactics that can counter a leader’s battleplan. This enables the development of contingency plans, and enables leaders to rapidly adjust to the dynamic battlespace.

Dr. Roy Hayes, CTO, SEI, revealed, “Using VBS4 with the VBS World Server enables a commander to use 3xAI to respond rapidly to changes in live operations, to perform tactical analysis of any battlespace demanded. Additionally, VBS4 can rapidly share 3xAI’s reports through the After Action Review.”

3xAI uses online artificial intelligence (AI), which means that it learns as it fights. No longer bound to loading the huge quantities of Big Data of traditional AI, 3xAI can achieve near optimal solutions in hundreds of runs, instead of the thousands of runs required with Deep Reinforcement Learning Techniques. Working with the commander’s first input of their COA into VBS4, and through repeated trials 3xAI learns how to position and maneuver units to counter the commander’s battleplan. With 3xAI, scenario and training designers can rapidly create the Intelligence Preparation of the Battlefield (IPB) phase of an operation, identifying a plan’s critical vulnerabilities and building contingencies. Currently, 3xAI can perform the military task of seize, retain, ambush, and attack by fire.

Assessing 3xAI, David Russel, Retired Infantry Company Commander United States Marine Corps said, “You win during preparation. 3xAI would have provided me the ability to prepare for my missions more effectively, which would have increased our lethality.”

VBS4 provides a 2-dimensional overlay and a 3-dimensional rendering of the battleplan or course of action. When performing the mission rehearsal before an engagement, adjacent units can virtually place themselves in each other’s roles, mitigating miscommunication and improving coordination.

The results deliver faster and more effective battleplans

3xAI’s ability to highlight key terrain, optimal defensive/ambush locations, and the enemy’s most dangerous COA, allows the military leaders to rapidly adapt to the dynamic battlespace and outmaneuver their opponent. When SEI and ONR tested it in multiple force-on-force actions, the commander that leveraged 3xAI won every engagement. The After Action Reviews (AARs) facility in VBS4, means that users can track the AI development from situation to solution.

AI development integral to the future of VBS4

Oliver Arup, SVP Product Management at Bohemia Interactive Simulations explained, “We are at an exciting stage in the community of understanding and delivery of meaningful AI into Training and Simulation. This is an ongoing focus of NATO, and we will be looking to support this activity through VBS4’s development roadmap.”

3xAI future development

The 3xAI development pathway includes adversary prediction models, which seek to forecast an opponent’s future action. The inclusion of these models means that 3xAI can predict future states of the battlespace, enabling it to leverage more complex tactics, like deception and feinting. Additional future developments include incorporating an enemy’s doctrine into 3xAI, and limiting the possible tactics that our agent will use to those that were previously seen.