By Max Hunder
KIEV (Reuters) – Several Ukrainian start-ups are developing artificial intelligence systems to help fly fleets of drones, pushing the war into uncharted territory as combatants race to gain a technological advantage in combat.
Ukraine hopes that deploying AI-enabled drones across its frontline will help it overcome increasing Russian signal jamming and allow it to operate unmanned aerial vehicles (UAVs) in larger groups.
Ukraine's AI drone development can be broadly divided into vision systems that help identify targets and fly drones, terrain mapping for navigation, and more complex programs that allow UAVs to operate in interconnected “swarms.”
One company working on this is Swarmer, which is developing software to network drones so decisions can be made instantly across the group, with no humans required to intervene until an automated attack is given the go-ahead.
“If you try to scale it up (using human pilots), it doesn't work,” Swarmer CEO Serhiy Kuprienko told Reuters at the company's offices in Kiev. “It's virtually impossible for a human to control a swarm of 10 or 20 drones or robots.”
Swarmer is one of more than 200 technology companies that Ukraine has boomed since Russia began a full-scale invasion in 2022, with civilians in the IT sector developing drones and other devices to help Ukraine counter a much larger foe.
Kuprienko said that while a human pilot would have difficulty conducting an operation involving more than five drones, the AI could handle hundreds.
The system, called Styx, would coordinate a network of large and small surveillance and attack drones in the air and on the ground, with each drone able to plan its own movements and predict the actions of others in the swarm, he said.
In addition to expanding the scale of operations, Kuprienko said automation would also help protect drone pilots, who operate closer to the frontline and are therefore priority targets for enemy attacks.
Swarmer technology is still in development and has only been tested experimentally on the battlefield, he added.
Samuel Bendet, a senior fellow at the Center for a New American Security, said AI drone control systems would probably require human intervention to ensure the systems didn't make errors in target selection.
There are widespread concerns about the ethics of weapons that remove human judgement: a 2020 European Parliament working paper warned that such systems could violate international humanitarian law and lower the threshold for starting war.
AI is already being used in some of Ukraine's long-range drone attacks targeting military installations and oil refineries hundreds of kilometers into Russia.
An anonymous Ukrainian official told Reuters that attacks could involve swarms of up to 20 drones.
The main drone will fly to the target, but it's up to the others to disrupt or distract air defenses along the way, using a type of AI, under human supervision, to help spot targets and threats and plan possible routes, the sources added.
Signal Jamming
The need for AI-enabled drones is growing as both sides deploy electronic warfare (EW) systems to jam signals between pilots and drones.
In particular, small, inexpensive FPV (first-person view) drones, which became the primary means by which both sides attacked enemy vehicles in 2023, are becoming less accurate as jamming increases.
“We are already working on a vision of a future where there will be no frontline connection between pilot and drone,” said Max Makarchuk, AI lead at Brave1, a defense tech accelerator founded by the Ukrainian government.
According to Makarchuk, the percentage of FPVs hitting their targets is continually declining. Currently, most FPV units have a hit rate of 30-50 percent, but for novice pilots, it can drop to as low as 10 percent.
He predicted that the hit rate of an AI-operated FPV drone could reach around 80%.
To combat the EW threat, manufacturers like Swarmer have begun developing features that allow drones to lock onto targets via cameras.
The EW system forms an invisible signal jamming dome over the equipment and soldiers it protects.
If the pilot loses contact with the drone, the pilot will no longer be able to control it, and the drone will either crash to the ground or continue flying in a straight line.
Automating the drone's final flight to the target location eliminates the need for a pilot and negates the jamming effects of EW.
AI-enabled drones have been in development for years, but until now have been considered expensive and experimental.
Bendet argues that Russia had been developing AI-enabled aerial and ground drones before the 2022 invasion and had achieved some success.
In Ukraine, a key challenge for manufacturers is to produce AI targeting systems for drones cheaply so they can be deployed in large numbers across the 1,000-kilometer (621-mile) frontline, where thousands of FPV drones are used every week.
Costs can be reduced by running AI programs on the Raspberry Pi, a small, inexpensive computer that is gaining popularity worldwide even outside of educational purposes.
Makarczuk said he estimates the cost of implementing a simple targeting system, which captures features visible to a drone's camera, would be around $150 per drone.
(Editing by Mike Collett-White and Gareth Jones)
