| Abstrakt: |
Computer generated forces (CGFs) are autonomous, computer controlled, entities employed to model human actors in many simulation-based training and decision support tools. In this work we present a CGF modeling approach targeted towards modeling aggregate entities representing groups and organizations such as civilians, armed forces, etc. in military conflict zones. The proposed modeling approach uses Bayesian networks and rule-based methods that operate on state variables that have been carefully selected to represent the characterizing properties of groups and organizations. Specifically, state variables are used to represent the CGF's knowledge or beliefs about itself, other actors and the environment; Bayesian net- works are used to model the behavior and action selection mechanism of the CGF; and rules are used to model actions and their effects on the CGF's state variables. Furthermore, to reduce authoring complexity, this work also presents a sampling-based analysis method capable of analyzing the behavior and real-time performance of a model even prior to embedding the model in its intended simulation environment. [ABSTRACT FROM AUTHOR] |
