The U.S. Army lacks a common governing policy, methodology and architecture for generating, collecting, reducing and analyzing the training data available today. As described in the Army’s latest Learning Concept for 2030-2040, collecting training data is critical to diagnose, prescribe, and facilitate effective training. Current methods to collect and store training data are disparate and require Army units to manually input data into the Army’s current data tracking system. This paper will describe a Data Collection, Reduction and Analysis (DCRA) research effort, sponsored by Army Futures Command - University Technology Development Division, to support the U.S. Army’s broader Data Strategy, and Combat Capabilities Development Command - Soldier Center’s more specific research on new data strategies for the U.S. Army’s Synthetic Training Environment (STE) program. The research effort’s goal is to produce a prototype data model that will enable the U.S. Army to automatically collect existing data being “lost” from current training systems. We discuss challenges to collect, reduce, and align different data sources, types, and formats needed to diagnose, prescribe, and support the US Army’s warfighter qualifications; including facilitating machine learning algorithms that could support future STE AI-based training and assessment capabilities. We discuss collaborative efforts occurring with commands and facilities at Fort Cavazos, Texas to baseline current data collection processes and practices.  We also provide a high-level description of the DCRA-STE model, and capabilities being applied to a US Army, Integrated Weapons Training Strategy (IWTS) use case. We conclude by discussing best practices and data standards we have found or developed to address existing data challenges, and how they may apply to data collection at scale.

Keywords
ANALYTICS;DATA;EVALUATION;HUMAN PERFORMANCE;PROFICIENCY

Additional Keywords
Weapons Training, Synthetic Training Environment