Cognitive Modeling Approach for Generating Authentic Tactical Agent Behavior

Aerial Encounter, Agent, Attention Model, Behavior Tree, Beyond Visual Range, Cognition, Computer Generated Forces, Flight Dynamics, Long-Term Memory, Nonlinear Model Predictive Control, Polynomial Chaos Expansion, Situational Awareness, Tactical Simulation, Surrogate Modeling, Synthetic Training, Tactical Autopilot, Working Memory

This paper presents a synthesis of a tactical cognitive agent that utilizes advanced information processing models to enhance the human-like behavior of Computer Generated Forces (CGFs). The agent's Situational Awareness (SA) representation is enriched through the use of working memory, attention synthesis, and decision-making based on machine-learned models to accurately replicate human cognitive processes and behavior. This improved representation allows the agent to demonstrate human-like behavior under increased workload, a common condition in tactical flying scenarios. The innovative model structure is based on established concepts in cognitive and neurosciences and implemented using an Artificial Intelligence (AI) technique called the Behavior Tree (BT). The execution of SA driven behavior utilizes a purposely designed tactical autopilot based on Nonlinear Model Predictive Control (NMPC), enabling human-like maneuver specification and validation. The resulting cognitive agent was integrated into a ground-based simulation platform for synthetic tactical pilot training.