Adaptive Behavior Modeling of Asymmetric Tactics (ABMAT)

In order for simulation based training to help prepare warfighters for modern asymmetric tactics, opponent models of behavior must become more dynamic and contemporary and challenge trainees with adaptive threats consistent with those increasingly encountered by the military. The proposed Adaptive Behavior Modeling of Asymmetric Tactics (ABMAT) system will apply artificial intelligence machine learning techniques to generate adaptive adversary behaviors in a synthetic training environment. The adaptive behaviors developed for the ABMAT system will be based on real-world example scenarios collected from subject matter experts with recent operational experience. The action and reaction cycle that typifies recent operations against asymmetric adversaries will be captured from these scenarios and incorporated into the virtual experience base of the synthetic adversary models. The resulting behaviors will adapt tactics in a given exercise both in response to events in the current exercise and in previous exercises, using a combination of reinforcement learning and evolutionary algorithms. This system will build on, and enhance, two commercial products: the OLIVE massively multiplayer virtual environment, and the SimBionic behavior modeling and execution tool. The ABMAT system will be developed and validated with the OLIVE environment, but also designed for platform portability with minimal cost.