Skip to content
Technology / Aerial Drones

AI drones in Ukraine: how artificial intelligence is changing drone warfare

According to Russian sources, Ukraine has gained a drone advantage of at least 6:1 with the help of artificial intelligence. An overview of AI drones for reconnaissance and attack on both sides – and why caution is advised with the videos.

AI drones in Ukraine: how artificial intelligence is changing drone warfare
© Archiv

Partly thanks to artificial intelligence (AI) – in the form of AI drones – Ukraine has built up a drone advantage on the battlefield this year in its defensive struggle against Russia. Russian sources put the ratio at at least 6:1; for Russia, they say, it is probably even less favourable.

Frequenzbereich
“SIGINT” (“signals intelligence”), the gathering of information through the collection and analysis of electronic signals, is of great importance in drone warfare. · © Archiv

Recently, however, Ukraine suffered a setback. Along a very broad front, Russia shifted the control signals of its first-person view (FPV) drones to the lower frequency range of 480 to 560 MHz. A considerable share of Ukrainian electronic countermeasures, previously aimed at higher frequencies, has since been ineffective.

Such advantages are, of course, not lasting. Six to eight weeks, sometimes only days – neither side can secure a limited lead in this way for any longer.

Drones that recognise objects, vehicles and people by themselves and also steer themselves, by contrast, no longer necessarily depend on video and control signals being transmitted.

Such systems are undoubtedly already being worked on intensively – in Russia and Ukraine, but also by companies in many other countries. Ukraine benefits from having modern western microelectronics at its disposal.

Russia, on the other hand, remains limited to grey imports of commercially available hardware because of the sanctions regime, so the pressure to find substitutes is correspondingly high. China is considerably further ahead than Russia in microelectronics independence and does help – but whatever Russia obtains from China or from Iran, it has to pay for. Russia does not receive free military aid as Ukraine does.

Reconnaissance: AI drones find targets autonomously

The AI-assisted drone systems seen so far can be roughly divided into two areas. The first comprises reconnaissance systems that locate and identify “military targets” autonomously without human intervention. Ukraine is said to have a system called Saker Scout for this purpose, which passes on reconnaissance information on its own. What the drone receives feeds into “Delta”, the Ukrainian military's computer-based real-time situational awareness system.

Russia has a very similar system: Quadrat, also known as “Area Systems”. It grew out of a university project intended to map and survey ground infrastructure automatically, and was aimed, at least officially, at the civilian market. That has since changed. Footage of a test flight over the Russian test and training ground at Kubinka shows numerous objects being detected and categorised in real time during the flight.

Objekterkennungssoftware im Test
Test of object recognition software at the Kubinka test ground · © Archiv

For the second part of the video compilation (see above), it is stated that the computations for analysing the footage ran on servers at Russian headquarters. This suggests that neither the drones nor the reconnaissance units have the necessary computing power. Real-time display is therefore rather unlikely.

Attack: terminal approach without a pilot

The second area consists of (semi-)autonomous attack drones. Either they detect an object fully autonomously and then fly towards it on their own. Or, in semi-autonomous mode, an operator assigns the target and the drone takes over the terminal approach. The boundaries are fluid, and understandably neither side fully discloses the functions.

It recently became known that Ukraine already has Skynode S at its disposal, a system from the US-Swiss company Auterion. The small all-in-one computer gives drones “machine vision” and makes them independent of GPS and radio frequencies. According to Auterion, the system enables swarm control, fully autonomous flight and jamming resistance, offers unprecedented accuracy and raises the success rate from 20 per cent to more than 90 per cent.

Skynode S – @Auterion
Skynode S: the huge heat sink on the tiny board suggests how much computing power is required. · © Archiv

Whether the videos from Ukrainian drone manufacturer Air 3F are connected with this is not known. In any case, the footage shows a system that acquires a target – here an unmanned ground vehicle (UGV) – and a drone that heads for it.

Ptashka Drones recently released a video that is strikingly similar to this example. Known Russian counterparts are called Ovod-S and Rusak-S.

The Ovod drones (Овод) come from a Russian manufacturer in Tula. In openly accessible Russian tests, these FPV drones performed above average; they use Chinese hardware. The video, however, also shows the large on-board computer, which consumes a correspondingly large amount of energy. The losses in payload and range – a consequence of the sanctions – are obvious.

The Rusak-S is also a comparatively large FPV drone. A whole series of video recordings of its test flights exists.

It should be noted: there is no clear, publicly available evidence that any of these systems is actually in use. Moreover, such videos are easy to fake for propaganda purposes – for whatever audience. The editorial team cannot conclusively judge whether each case is a real development or a fake. This should always be borne in mind when viewing the videos.

Photo credits

© Archiv
Advertisement336 × 280
Perimeter Protection 2027 – Secure ticket now! 19–21 January 2027, Nuremberg (incontent)

Continue exploring

The week in unmanned systems.