A Graphical Game Theoretic Asymmetric Tactic and Strategy Generation for Simulation and Training

We propose a highly innovative data fusion with data mining approach for asymmetric adversary tactics and strategy generation in synthetic training environment. Our approach has two parts: 1) Data fusion module. Sensor data are fused to obtain the situation awareness. A graphical dynamic game model is used to generate the Course of Actions (COAs) of two sides (Blue-trainees, and Red-asymmetric adversary strategy generator). The COAs of red will be implemented as the asymmetric adversary tactics and strategies; and 2) Dynamic/adaptive feature recognition module. Adaptation (online-learning) and pattern/feature recognition are carried out to dynamically select (or mine) appropriate features or feature sets of blue side so that the algorithm parameters in the Data Fusion Module can be dynamically, intelligently, automatically tuned. Our multiplayer non-zero sum game theoretic approach is effective because it takes into account the fact that both the adversary and the non-neutral players are intelligent. We integrate the deception concept in our game approach to model the action of purposely rendering partial information to hide the asymmetric threats. With the consideration that an asymmetric threat may act like a neutral or white object, we also model the actions of white objects in our non-zero sum graphical game framework.