ROBOTICS AND AUTONOMOUS SYSTEMS INDUSTRY PERSPECTIVE: Vision-Based AI at the Tactical Edge a Force Multiplier By David Guffey iStock illustration In Afghanistan and Iraq, white Toyota Hilux trucks were synonymous with the Taliban and many other terrorist groups that U.S. and coalition forces pursued daily. Ubiquitous and covered in dust, nearly indistinguishable from thousands of identical vehicles on the same roads, complex hyperspectral sensors were used to look for minute spectral differences. Sensors were flown, data gathered and analyzed, and over time, automated algorithms developed to provide a level of autonomy and precision. But the challenge existed not just in finding a truck, but the right truck. U.S. forces needed the one with a dent on the rear panel, a cracked headlight or another small detail buried somewhere in tens of thousands of hours of Predator and Reaper footage. Unique spectral signatures across tens of thousands of vehicles served as additional discriminators. The military mastered collection, but processing at scale was dramatically more complex. As new sensors, unmanned systems and space assets expanded the data flow, much of that data went unreviewed. Regardless of how much data flowed into servers and storage, most of it went unanalyzed. Often, data was collected but took days or even weeks to be analyzed at the forward operating base or in garrison in the United States. Now, the use of vision-based artificial intelligence at the most disadvantaged edge will become a tactical discriminator and a force multiplier. Compute power has evolved to the point where handheld devices can capture, consume and analyze data at the most disadvantaged edge. Much of today’s AI discussion centers on large data centers and models trained on data from everywhere. Those systems are invaluable, providing leaders and decision-makers with strategic answers to broad, complex problems across multiple domains and