Enhancing Simulation-based Training Adversary Tactics via Evolution (ESTATE)

To significantly extend modern simulation-based training environments to incorporate realistic and adaptive adversary behavior in line with today's asymmetric strategies and tactics, we propose to develop a system for Enhancing Simulation-based Training Adversary Tactics via Evolution (ESTATE). The proposed system will consist of an on-line, executable, reactive adversary behavior model and an off-line adversary behavior adaptation engine for strategy and tactic discovery. On-line adaptation is performed using an intelligent agent framework to respond and adapt to the trainee's actions during a given simulation-based training exercise. Off-line adaptation is performed using evolutionary algorithms (EAs) to effectively search through the space of adversary behaviors to exploit fundamental weaknesses underlying the trainee's strategies and tactics by wargaming against a trainee model extracted from traces of simulation events from past training sessions. The first stage of this Phase I effort focuses on identifying a target domain, developing a set of representative scenarios, and evaluating and selecting a simulation-based training system. The second stage focuses on rigorous design, development and testing of a limited-scope ESTATE prototype to assess the feasibility of the approach.