Automated Instantaneous Performance Assessment for Marine-Squad Urban-Terrain Training / 14th ICCRTS: C2 and Agility

Abstract

Urban operations (MOUT) are a key component of U.S. Marine training, yet assessment of performance during the later and more integrative training exercises is difficult because events happen quickly and walls impede full view. We are building a performance assessment system called BASE-IT that will collect data on positions, orientations, postures, and activities of Marines during training using a network of cameras. BASE-IT will automatically summarize performance over exercises and flag problematic behaviors for after-action review by commanders. It will then provide a virtual-reality display showing not only what happened but alternative courses of action using capabilities similar to those of video games. The work reported here focuses on measures of performance assessment for each instant of time ("instantaneous measures"). From reading of Marine documents and discussions with subject-matter experts, we identified 30 potential instantaneous performance measures for a group of Marines, and selected thirteen for implementation: dispersion between Marines, collinearity of Marines, number of clusters of Marines, number of interactions with non-Marines, danger, situational awareness measured as view coverage, path safety, relative mobility, speed of the group of Marines, safety of weapons orientation, coverage with weapons, surrounding of search targets, and contact with the leader. These measures have interesting locality properties in space and time. We show results of automated analysis of images of training at the Twenty-Nine Palms Marine base.