Intelligent Agents for Real-Time Story Adaptation for Training Assessment

Simulation-based training systems are developed for a broad range of warfighter and warfighter support activities, from clearing buildings to high-level command decision-making. These systems are expensive to build and they are expensive to operate. This proposal outlines a plan to decrease both types of costs, the costs to develop training systems and the costs to operate training systems. It addresses the former through the development of a general architecture for interactive storytelling in game-based simulation environments. It addresses the latter by automating the creation of scenarios appropriate to training goals, dynamically altering these scenarios to promote the training objectives, and intelligently generating after action explanations of trainee behavior. To accomplish these technical objectives, we leverage and extend tools (storytelling architectures and automated AAR systems) previously developed through U.S. Army funding. The extension of the storytelling architecture provides for a more robust architecture that can accommodate a number of different training simulation environments. The extension of the automated AAR tool provides for a novel type of feedback possible only to environments with storytelling capability. Finally, the integration of the two provides sophisticated, knowledge-intensive, agent-based instruction both during and after an exercise. This work not only applies state-of-the-art, but it extends state-of-the-art.