Realistic evaluation of terrain by intelligent natural agents (RETINA)

Abstract

US Army and Joint constructive simulations require human operators to observe the exercise in progress, conduct analysis of the results, and provide a realistic reports and assessment of the action presented on their screens to the desired training audience. Current software tools provide excellent mathematical assessments (such as center of mass calculations, optimal routes, and sensor ranges) but poor human-like assessment of data (most likely route, probable enemy intention, etc.). This Thesis presents an artificial intelligence architecture specifically designed to reduce that manpower requirement by describing a concept for computer modeling that can produce realistic human-like assessment results. Specific concepts described are approaches for conducting a digital terrain assessment, development of avenues of approach, deployment of templated forces to a specific piece of terrain, and then a method of adjusting the templated force to react to actual sightings and known information. Also included are more detailed discussions and implementation details for use of gas diffusion as a method of analyzing avenues of approach through digital terrain. This approach seems quite promising as a method of modeling human movement tendencies and appears superior to classic path finding or optimal route selection methods.