Drone warfare in Ukraine: Martian drones, swarms and mesh technology
Three key stories: April 24 - 30, 2026
Ukraine has hinted at potential breakthroughs in AI-driven defence technologies—but what might those actually look like? This week, I spoke with a leading Ukrainian developer of collaborative autonomy software, Swarmer, and reviewed several interviews to assess how close we are to seeing it on the battlefield. There is clear progress—but no one is willing to predict when this shift will fully materialize.
In this edition:
“Our Goal Is to Replace All UAV Operators”: Ukraine’s Swarmer on Responsibility in Swarm Tech;
Russians Alarmed by Ukraine’s Mysterious “Martian” Drone;
Ukraine Seeks Local Developers of Mesh Modems;
and 5+ additional developments in drone warfare in Ukraine and Russia.
On Saturday, I’ll release the second edition of my 2-page weekly brief on drones in Ukraine and Russia. I plan to continue publishing it even during weeks of reduced availability, when I’m unable to release the full Drone Warfare in Ukraine edition.
In a couple of weeks, the Ukraine Drone Brief will be available only to VIP-tier subscribers.
Thank you for reading and supporting my work!
“Our Goal Is to Replace All UAV Operators”: Ukraine’s Swarmer on Responsibility in Swarm Tech
Recently, the Head of the Presidential Office, Kyrylo Budanov, noted that Ukraine is placing a strong emphasis on AI in defence technology, which could become a game-changer.
A Ukrainian company, Swarmer, is already making progress in this direction and last year secured the largest publicly announced investment in Ukraine’s defence sector.
Swarmer develops software for coordinating drone swarms. It focuses on models in which the operator sets the objective, and the system distributes tasks among the drones and ensures mission execution, even in environments without GPS or with limited communication.
The technology developed by Swarmer is not tied to any single platform. “It operates at the intelligence layer – enabling large numbers of low-cost unmanned systems to function together as a coordinated, resilient force,” Swarmer Global CEO Serhii Kupriienko told Ukraine’s Arms Monitor
Reportedly, since April 2024, Swarmer’s tech has been deployed in Ukraine with more than 100,000 real-world missions in active combat environments, informing the software and machine-learning models that feed into it.
While the term “drone swarm” is often used to describe large-scale group attacks conducted by Russia against Ukraine, Serhii Kupriienko explained that his company uses a narrow definition of swarming: every member of the group knows everything about all the other members of the group, and can make decisions independently, without a leader and without a single point of failure.
In his interview with the Eastern Circles newsletter, he shared the following:
A true swarm relies on decentralized coordination, where each drone, depending on its application (reconnaissance, strike drone, demining, logistics…) and environment (land, sea, air), continuously shares data, adapts to changes, and contributes to a collective objective without relying on constant human control.
One “swarm” can include up to 690 drones (today), split into groups of robotic systems based on their application.
In a swarm, drones navigate through mutual awareness, constantly recalculating their positions and roles. If one drone is destroyed or fails to complete its task, others immediately adjust—approaching the target from different angles or redistributing responsibilities in real time.
Unlike humans, who can typically coordinate their actions only with a few individuals they can reach, computers can simulate and manage interactions with hundreds simultaneously.
Technically, this requires a specific architecture. Swarmer’s approach is based on AI software that connects multiple drones into a unified network. A single operator can supervise the overall mission while the drones themselves handle navigation, targeting, and adaptation.
With a single drone, a human operator can intervene if something goes wrong. With a swarm of drones, intervention becomes impossible.
That is why the system should be designed to handle a wide range of failure scenarios: loss of communication, navigation errors, malfunctioning payloads, or rapidly changing battlefield conditions. Each of these variables depends on context—terrain, enemy position, and electronic warfare interference—making the engineering challenge harder.
Cooperative autonomy has not yet become widespread. Building truly autonomous, coordinated systems requires not only advanced AI, but also new training models, supercomputers, and hardware-software integration.
In an interview with NV, Serhii noted that the underlying technology is not new: early startups working on group control and drone autonomy systems began emerging in the early 2000s. Currently, around one to two dozen companies in Ukraine are working on swarm solutions.
When asked whether a drone could make mistakes and trigger a chain of incorrect decisions within a swarm, Serhii agreed that mistakes are possible. He emphasized that—unlike image generation or systems like ChatGPT—this is about real life, where failures can cost lives.
“In testing, we follow an approach used in the automotive industry—humanity has already partially solved similar challenges in self-driving cars.
Our goal during testing is twofold: first, to reduce the error rate; and second, to ensure that every error can be traced, its root cause understood, and that it is prevented from recurring on the battlefield,” Serhii told NV Business.
The goal of this technology is to remove micromanagement from the human operator without removing responsibility.
“Imagine a scenario where ten drones are deployed with a task: somewhere in a designated area, there are tanks or electronic warfare systems, and the task is to find and strike them. This is a standard concept—in NATO doctrine, it is referred to as a “kill box”, according to which the person assigning the task remains responsible for ensuring that there are no civilians in the area and that all risks are considered, just as in operations involving systems like the Grad rocket launcher or other artillery. The drones do not independently decide what to do; they operate within predefined priorities”.
This is not a scenario where the system “decides on its own.” The level of decision-making autonomy assigned to the system is deliberately limited and does not approach that of a human pilot.
Over the past six to seven months, Swarmer has focused on bringing the solution to operational use on small drones—starting with groups of ten or more FPVs.
Two years ago, Swarmer integrated its system into UGVs, but navigation remains a major challenge for achieving real autonomy. The team is in constant cooperation with UGV manufacturers and partners and is searching for a solution that does not yet exist in the required form.
Background info: Swarmer went public on the Nasdaq on March 17 under the ticker SWMR, with US-based Lucid Capital Markets managing the IPO. The company raised around $15 million in its IPO. The raised funds will be used partly for operational needs, but primarily to continue product development according to the roadmap and to drive adoption. In practice, the bulk of the investment will go toward accelerating integration with different clients, various drone platforms, and information systems.
On April 29, Swarmer announced a strategic partnership with a Ukrainian producer of tactical radios, HIMERA. The collaboration will integrate HIMERA’s jam-resistant radios directly into Swarmer’s next-generation autonomy stack. Read my article about HIMERA: Good Radios Save Lives: HIMERA Tactical Communications in Ukraine.
Russians Alarmed by Ukraine’s Mysterious “Martian” Drone