An Approach to Enhance Human Behavior Modeling for Computer-Generated Actors

Current computer-generated actors (CGAs) demonstrate several shortfalls that may enable human opponents to rapidly identify and defeat CGAs; a situation that results in an inadequate, and often negative, training experience. Our research is intended to correct some of these deficiencies by inserting into CGAs reasoning qualities that allow the portrayal of more human-like behaviors. The work we report was conducted in the context of developing a prototype Distributed Mission Training Integrated Threat Environment (DMTITE) system. To allow the portrayal of more human-like behaviors, we incorporated a combat psychology model into our CGA framework. The DMTITE combat psychology model is a mathematical, scalable, and computable approach to describing psychological functions affecting performance in combat. The ultimate goal for implementing this model within the DMTITE architecture is to provide a training arena that forces human participants to engage DMTITE CGAs using realistic tactics, thereby enhancing the overall training experience. Our approach combines relatively stable human traits with situationally determined transitory states to allow a CGA to compute human-like behaviors. When combined with multiple skill levels, the combat psychology model affects how a DMTITE CGA reasons about the virtual battlespace it inhabits by biasing values used in reasoning computations. This paper describes the combat psychology model incorporated into the DMTITE CGA architecture.