Trafalgar Ph II

Today’s After Action Review (AAR) systems are often tied to particular training or operational systems. This approach is insufficient for current Navy/Marine training needs, which increasingly require different platforms and mission capabilities interacting and operating together. Stand-alone systems require instructors to gather data manually to deliver an AAR. They must resort to hand-written notes, chatlogs, and review of trainee work products to connect events and observations across systems. The current process is tedious and slow, does not scale, and results in less effective training. In Phase I, we researched and designed a novel software architecture to deliver more automated and more effective AAR. The Trafalgar architecture exploits and integrates well-structured, consistent data, but also anticipates the inclusion of unstructured documents and semi-structured data. All of this data is required to put exercise events into a coherent, causal context to support evidence-based debrief and AAR principles. Importantly, meaningful integration of heterogenous data requires artificial intelligence (AI) to integrate, process, and personalize AAR. In Phase I, we implemented several feasibility demonstrations using AI to capture meaningful learner and performance data, even from stand-alone systems, and used this captured data in a proof-of-concept AAR Delivery environment that tailors its presentation based on evidence-based principles and the individual performance of individuals and teams. For Phase II, we propose to develop the AAR architecture designed in Phase I and then to instantiate and to test the architecture for a specific application. We also propose to investigate the generality of the architecture via application in additional LVC training domains.