Swedish startup Scaleout has demonstrated autonomous loitering munitions capable of identifying and engaging targets without human intervention or external communication. The system utilizes the Nvidia Jetson Orin Nano module to run the YOLOv8 Nano object-detection model directly on the drone hardware.
During a February test in Sweden, the drone operated at -18 degrees Celsius, achieving a framerate of 30 fps with a latency of approximately 30 ms. The mission, which involved engaging an armored vehicle, required 200 seconds for reconnaissance and completed in under 320 seconds.
The company employs a Tactical Computer Vision Network that combines local AI training with federated learning. This allows devices to adapt rapidly while maintaining functionality during communication blackouts, as demonstrated in a June test involving nodes with cut connections.
Scaleout’s FEDAIR project is currently part of the NATO DIANA accelerator, which provided €100 000 in development funding. While the technology is currently deployed under an active license, it has not yet received an official NATO procurement order for combat use.
