Capturing Military Doctrine in Behavior Trees Using LLMs

Explore how large language models can revolutionize the creation of behavior trees for military simulation applications in this 39-minute conference talk from Unreal Fest Orlando 2025. Learn from Paul Foster of Lockheed Martin and Gita Sukthankar of UCF as they demonstrate an innovative Unreal Engine plugin that integrates an interactive chat interface directly with the Behavior Tree editor. Discover how military doctrine - the fundamental guidelines governing military operations from campaigns to individual engagements - can be captured and translated into accurate behavior trees using LLM technology. Examine the collaborative research between the University of Central Florida and industry partners that leverages robotics behavior tree examples to streamline the authoring process for both individual agents and complex multi-agent systems. Understand practical techniques for maximizing the effectiveness of trained LLMs in generating realistic military scenarios for training purposes, while reducing the labor-intensive testing and validation traditionally required. Gain insights into how this cutting-edge approach makes behavior tree development more efficient and realistic within the Unreal Engine Game AI framework, opening new possibilities for military simulation and training applications.