DOCTRINAIRE ATR for Rotorcraft Targeting (DART)

The goal in this effort is to increase the lethality of U.S. forces by providing AI-enabled targeting systems that ingest passive imaging (RGB, IR) data, run in real-time on low-SWaP hardware, and are easy to understand and update in the field. These systems will help warfighters achieve their mission and destroy enemy targets more quickly and reliably. Our novel approach to target classification and aim-point identification enables us to build ATR systems that infer not only the target class, but also the target poses and a 3-D representation of the target CAD model even if the ATR system was never exposed to any images of the target class during training. We call our approach DOCTRINAIRE because it is based on warfighter doctrine. Just like the warfighter is trained to identify enemy tanks through their hull, armaments, turret, and suspension (HATS), enemy aircraft through wings, engine, fuselage, tail (WEFT), and enemy ships through their silhouettes, DOCTRINAIRE was designed to work on these same principles. The goals of this Phase I effort are to mature the capabilities of the DOCTRINAIRE specifically for the classification of rotorcraft (e.g., helicopters), to measure performance, and determine suitability for operational use.