SBIR-STTR Award

SoarTech, an industry leader in the development of AI/ML-enabled agents that generate behavior in complex environments, proposes a proof of concept investigation into Behavior and Action-Driven Game Environmental Rewards in Task Learning (BADGER-TL.) The proposed approach fuses intelligent system design principals from goal-based cognitive models and supervised machine learning. BADGER-TL leverages SoarTech’s over two decades of experience researching intelligent system design to produce an agent design that is capable of human-level reasoning and team coordination, even in novel situations, while remaining data efficient in its task learning and memory mechanisms. SoarTech will also investigate intuitive, human-centric interface designs that allow subject-matter experts to control and adjust intelligent agent behavior to produce desired effects and validate our design in an example wargaming environment domain.

In this Phase II project, SoarTech proposes to extend and mature our Phase I research for Behavior and Action-Driven Game Environmental Rewards in Task Learning (BADGER-TL) into a flexible behavior engine that can be rapidly deployed to meet a large number of behavior-related requirements across the full spectrum of modeling and simulation (M&S) training systems available in the DoD enterprise, including the Army’s emerging Synthetic Training Environment (STE). SoarTech is an industry leader in the development of artificial intelligence and machine learning (AI/ML)-enabled agents that generate behavior in complex environments. Our BADGER-TL approach fuses intelligent system design principles from goal-based cognitive models and deep learning approaches to create a hybrid behavior architecture providing instructors and behavioral designers the ability to rapidly inject behaviors into modeling and simulation systems. BADGER-TL leverages SoarTech’s experience in intelligent systems to produce an agent design capable of human-level reasoning and team coordination, even in novel situations, while remaining data efficient in its task learning and memory mechanisms. SoarTech will also develop human-centric interface designs that allow subject-matter experts to control and adjust intelligent agent behavior to produce desired effects, such a variable skill levels.