In this paper, we describe a model of intelligent agent that learns BLUEFOR's mission structure and develops highly effective counteractions for Opposing Forces over time. The model consists of three main components. First, the agent aggregates data about movements, actions, and interactions of BLUEFOR actors to infer the sequence and types of operations that BLUEFOR is conducting. Second, the agent develops a plan to counteract BLUEFOR's operations or adjusts its mission to improve the rate of success. Finally, the agent learns over time the effects of its actions on the BLUE operations, incorporating terrain constraints and tactical effects. The main distinction of the model from standard AI techniques is in how local and global information about multiple BLUE actors and terrain features are used to make estimates about space-time activities composing coordinated BLUE tactics and learn effects of agent's actions with every experience.
Our model has been integrated with 3D virtual world to control OPFOR avatars. It provides a unique capability to train adaptation skills in urban combat. This technology can also be utilized during intelligence analysis, Wargaming, and mission rehearsals, allowing more accurate estimation of enemy courses of action and reduction of OPFOR manning footprint.