CapabilityUA · independent

Russian engineers develop 50 g onboard NN module for UAV relay under degraded link

Published 23 October 2026

Russian developers have produced a 50-gram hardware-software module embedding a neural network directly on a UAV. When channel throughput falls below 25 kbps and continuous video is no longer viable, the system automatically compresses the feed and, if necessary, shifts to event-based transmission—forwarding only operationally significant observations to the operator rather than a continuous stream. The developers claim the architecture supports control over very long distances; independent verification of the range figure is not available.

Sources — 1 source

Telegram channel:Victory Drones friendly22 Oct 2026 Open the post

Topicsdrones ai warfare

Related findings

US LUCAS strike drones planned for Shield AI Hivemind swarm-autonomy integration · 19 Oct 2026
German Sphere Defense LORIN-1 jet interceptor: >600 km/h, 30 km range, AI optical seeker, vertical launch · 18 Oct 2026
GPS-free autonomous drones inherit sensor-trust attack surface via visual nav pipelines, engineers warn · 17 Oct 2026
Russian teams achieve GPS-free drone navigation at 17 m accuracy via onboard map-matching · 13 Oct 2026

← Latest findings · All findings from October 2026