Scaleout Systems, a Swedish AI startup founded in 2018 by researchers from Uppsala University, successfully demonstrated autonomous target engagement using a low-cost loitering munition from BAE Systems Bofors. The mission, conducted under the Affordable Loitering Modular Ammunition (ALMA) program, required only a button press to start — the designated pilot remained as a failsafe controller but manual input was optional. Human-set parameters governed which type of target to engage.
All processing was handled onboard the drone, with no external communication needed during the mission, supporting the company's claim of resilience against electronic warfare. The drone used a small computer-vision model — YOLOv8 Nano — running on Nvidia's Jetson Orin Nano to detect and geolocate targets, then ranked an armored engineering vehicle highest before autonomously flying to it and dropping an explosive.
"With the war in Ukraine and a shifting world, we realized that this technology can be very important to operationalize edge data and sensor data for machine learning to make sure that NATO allies have found that strategic advantage," Andreas Hellander, cofounder and CEO of Scaleout Systems, told Ars Technica.
The company originally focused on deploying machine learning models on commercial trucks and other vehicles but pivoted toward defense applications after Russia's full-scale invasion of Ukraine in 2022. Scaleout's project, FEDAIR, is part of NATO's DIANA accelerator, receiving €100,000 of development funding, training, and test access.
Scaleout combines its Scaleout Edge platform with federated learning in a "Tactical Computer Vision Network (TCVN)." Devices train AI locally and share model updates for rapid adaptation. A separate "arctic strike demonstration" in February at BTC Karlskoga, Sweden, showed the same YOLOv8 Nano model achieving 30 fps at about 20 m/s with a target latency of 30 ms or less, in temperatures of -18°C. Ranging without a depth sensor is possible using a pinhole camera model combined with known object size.
A follow-up demonstration in June with the Swedish Air Force tested how the system handles disrupted communications: a ground node whose connection was degraded and then cut kept inference and active learning at full frame rate while offline, logging detections locally and backfilling on reconnect in priority order for heartbeat, critical alerts, drift, model updates and telemetry.