The Ukrainian armed forces (-> current news from the war in Ukraine) are pushing ahead with the transition to more automated warfare. The aim is to remove soldiers from direct combat situations and replace them with autonomous, unmanned systems.
The aim is not only to protect limited human resources, but also to overcome the growing challenges of modern battlefields: fatigue, mental stress and the overload of ever-increasing amounts of data from a wide variety of sensors and sources are to be overcome through the use of artificial intelligence.
Kateryna Bondar from the Wadhwani AI Center at the Center for Strategic and International Studies (CSIS) in Washington, D.C., has comprehensively analyzed the future vision of an AI-supported military strategy, comprehensively analyzed in a 37-page report.
“It’s a drone war”, the report quotes a Russian soldier as saying. He only saw the first Ukrainian fighter after he had surrendered to an unmanned aerial vehicle (UAV).
What does autonomy mean in warfare?
The US Department of Defense defines an autonomous weapon system as a system that, once activated, is capable of independently selecting and engaging targets without the need for further human intervention.
In contrast, there are non-autonomous or semi-autonomous systems that can carry out individual steps – such as recognizing or pursuing a goal – automatically, but do not take over the entire decision-making and action process independently.
Millions of drones
According to Ukrainian Defense Minister Rustem Umerov, Ukraine produced around two million drones in 2024. More than 1.5 million of these were FPV drones. 96.2 percent of the drones used by the Ukrainian armed forces were locally produced. This means that Ukraine received around 80,000 drones from abroad during the same period.
From January to September 2024, the Ukrainian military approved 140 UAV models and 33 UGV models for operational use. Ukraine plans to produce around five million drones by 2025.
Wide range of possible applications for AI
Artificial intelligence is used to analyze intelligence, surveillance and reconnaissance (ISR) data. AI algorithms help to prioritize images and videos based on the number and intensity of observed events – such as active battles or extensive troop and equipment movements. The aim is to recognize and classify objects. Satellite images, drones and stationary cameras are used as image sources.
Intercepted communication data – such as voice and text messages – is also a field of application for AI. Messenger platforms such as Signal and Telegram deliver thousands of group chats and messages every day, which often contain unstructured and non-standardized battlefield reports. In addition, intercepted audio communication from various sources can also provide valuable information.
Without technical support, it is almost impossible to transcribe and analyze this amount of data in a timely manner in order to gain usable insights.
Griselda: AI in action
To process this unstructured data, Ukraine uses the AI-controlled system Griselda – a platform that is specifically designed to analyze large volumes of voice and text information.
Zvook acoustic airspace monitoring system
Acoustic systems are able to classify and identify different sound patterns. The Zvook acoustic reconnaissance system has significantly improved Ukraine’s air defense capabilities. It uses the sound analysis of high-quality microphones to detect threats from the air at low altitudes – where radar systems are less effective.
Zvook currently monitors around 20,000 square kilometers of Ukrainian territory with its acoustic sensors arranged in a grid pattern.
Artificial intelligence filters out background noises such as cars, birds or human voices and classifies air threats based on their characteristic sound signature – including airplanes, helicopters, cruise missiles or drones with piston engines.
Thanks to Zvook, Ukrainian radar systems can remain in standby mode and are only activated automatically in the event of a specific threat.
Automatic target detection
At the heart of AI-supported automatic target acquisition is the ability to recognize and classify targets faster and more precisely than human operators. Sophisticated machine learning algorithms, which have been trained with extensive data sets from combat scenarios, are able to distinguish between vehicles, human figures and structures in a wide variety of terrain and under varying conditions.
Currently, the average range of automatic target acquisition in combat is around one kilometer; under optimal conditions, it can be up to two kilometers.
The AI must be continuously updated with real combat data in order to adapt to the dynamic conditions on the battlefield.
Autonomous navigation
The disruption of conventional control systems by electromagnetic warfare reduces the success rate of target acquisition to around 10 to 20 percent. This particularly affects measures to disrupt GNSS signals such as GPS, but also the interruption of control connections between the drone and the human operator.
In contrast, the success rate of autonomously navigating drones is 70 to 80 percent.
A distinction must be made between solutions for navigation over the “last mile” and systems for long-range missions designed to penetrate deep into enemy territory.
13 березня дрони СБУ атакували нафтопереробні заводи в РФ. відео з соцмереж pic.twitter.com/8HG8MjS6Ak
– Українська правда ✌️ (@ukrpravda_news) March 13, 2024
AI-controlled navigation in GPS-free environments has established itself as a key element of long-range reconnaissance missions. To avoid enemy air defense systems, flight routes are programmed with over 1,000 waypoints.
Using pre-installed satellite images and terrain data, the drones determine their position autonomously and are able to navigate to targets with high precision even without active satellite communication.
