U.S. Army researchers have developed a new method called deep compressive offloading that could help facilitate artificial intelligence processing and enable commanders to accelerate decision-making in combat.
The technique developed by Army Combat Capabilities Development Command Army Research Laboratory and its partners from the Internet of Battlefield Things Collaborative Research Alliance can compress and offload data from infrared sensors, cameras, radars and other battlefield devices to remote processing machines even in constricted communication environments, the service said Thursday.
“Compressive offloading, and other sensing and processing research being conducted in the IoBT CRA, are essential to the Army in effectively preparing for the future battlefield,” said Maggie Wigness, Army researcher and deputy collaborative alliance manager of the IoBT CRA.
Researchers applied the deep compressive offloading approach to an image offloaded from a device at White Sands Missile Range in New Mexico and a server in Massachusetts. They were able to reduce the size of the image while preserving during the compression critical data that AI needs for further processing.
“For example, in an application where the goal is to recognize different types of vehicles in an image, what cues are used by the AI to distinguish the different types of vehicles?” said Tarek Abdelzaher, a professor at the University of Illinois Urbana-Champaign and the academic lead of the lab’s IoBT CRA.
“Those cues should be preserved by compression when data is sent from the field. Other irrelevant information can be compressed away to improve compression abilities," Abdelzaher added.
