Tallinn's Defence EXPO Made the Case That the Drone War Is Now a Robot War

Introduction

In a converted industrial hall in Tallinn last week, more than a hundred companies set out to sell Europe the machines that will fight the next war. The drones dominated the floor, as they have dominated every conflict since 2022. But the more telling detail sat on the ground: tracked unmanned ground vehicles, the kind that carry troops' casualties out of fire zones and drag wounded soldiers home, displayed beside systems built to shoot drones out of the sky.

That pairing is the story. For three years the arms industry has sold counter-drone systems as an airborne problem — something to be solved with jammers, guns and radar pointed upward. Tallinn Defence EXPO, one of six events in Estonian Defence Week, made the commercial argument that the ground has joined the argument. The European firms showing there are increasingly selling robots to operate the counter-drone fight, and to hold ground while the sky is contested.

The backdrop is unforgiving arithmetic. Estonia's defence spending rose from 1.9% of GDP in 2013 to an average of 5.4% through 2029, and the country now puts 5% of GDP into core military capabilities, well beyond NATO's 3.5% target. The week drew around 4,500 visitors from 56 countries across 68 sessions. The money is real, and so is the urgency behind it.

Detecting the Threat Is the Hard Part

The first problem in any counter-drone fight is seeing the drone. Estonian firm Marduk Technologies brought a camera-based detection system that uses optics and AI to identify flying objects — birds, drones, helicopters, anything airborne. The system's most interesting property is what it does not do: it emits nothing.

"It's a camera-based drone detection system. It's fully passive, and with this we can see whatever is flying," says Leet Rauno Lember, Marduk's CEO. Passive optical detection is hard to locate because it radiates no signal of its own, unlike a radar. Lember is careful not to oversell it, and that caution is the most credible part of his pitch. Acoustic sensors reach only a few hundred metres and can miss electric drones, which are very quiet. Radar covers long distance but can miss drones made of lightweight foam.

"All of the layers are complementing each other, and end users are able to make better decisions," he says. The industry's emerging answer is layered detection rather than a single magic sensor — a point that matters enormously for procurement, because it means no single vendor's box is the whole solution.

Marduk has been building these systems for more than a decade. Lember recalls the reaction to its early warnings before 2022: "Back in 2016, everybody said that drones will not be a problem, at least in the near future." The war in Ukraine settled that argument.

From Detection to Interception

Estonian company DefSecIntel bundles that layered approach into a single mobile system called Eirshield, combining optical, radar, acoustic and radio-frequency sensors with effectors that range from jammers to interceptor drones to lasers.

"This is a turnkey counter-drone system," says Jaanus Tamm, DefSecIntel's founder. "Eirshield detection sensors will detect the drones, then the operator will select which one to go after, and then they launch the interceptor drone. Then the hostile drones will be destroyed and they will not hit any infrastructure elements or do any harm for the civilians."

Tamm calls Eirshield one of the few TRL 9 counter-drone systems in Europe available to buy in quantity, with TRL 9 the top of the technology readiness scale. The open architecture matters strategically: "If there are new threats coming, we can just add new effectors, or new sensors to detect better."

That last point is now a procurement requirement rather than a feature request. The rate at which counter-drone threats evolve has outpaced the procurement cycle. A Belfer Center analysis published in September put the number plainly: by the middle of 2025, output at Russia's Alabuga facilities alone had reached roughly 170 Shahed-type drones per day, rising to about 190 per day by the end of 2025, with Ukrainian military intelligence estimating total Russian monthly capacity near 2,700 drones. A system that must be re-contracted to add a sensor every time a threat class changes is a system that will always be one revision behind.

The interceptor itself is becoming a product category of its own. Latvian company Origin Robotics displayed Blaze, a radar-guided autonomous interceptor designed to fly at fast-moving aerial threats including drones and loitering munitions, taking radar data directly into its onboard autopilot. France's defence procurement agency selected it at Eurosatory 2026 after a competitive evaluation, according to Defence Industry Europe, with local assembly to be built with French integrator DSV. Origin Robotics describes it as the first NATO-codified autonomous interceptor drone with a STANAG-compliant warhead module, and says it is deployed operationally in Latvia, Belgium and Estonia.

The Ground Robot Argument

The second half of the story is what happened on the floor. Milrem Robotics, majority-owned by the Emirati conglomerate EDGE Group, showed its THeMIS unmanned ground vehicle, and made a claim that goes well beyond logistics.

The company describes the concept as an "unmanned kill zone" — a stretch of ground where robotic vehicles detect and push back an attack while human troops stay out of the line of fire. Milrem is explicit that the vehicles are remotely controlled and partly autonomous, not fully autonomous killers. "Instead of sending troops on the battlefield, we can send in these machines and save lives," says Raul Rikk, Milrem's capability development director, speaking at the company's Tallinn factory.

The platform is modular, and that modularity is the commercial argument. THeMIS can carry a machine gun, an anti-tank system or other weapons, but it can equally handle reconnaissance, mine clearing, logistics and casualty evacuation. One chassis, several missions, and a supply chain that does not require a new vehicle for every new job.

Rikk's production claim is deliberately vague. "We are talking about hundreds of vehicles, and different sizes," he says, adding that production can be set up anywhere and scaled rapidly. He also pointed to the reason the argument lands with buyers: "Nothing is theoretical anymore. Ukrainians are using them on the battlefield."

That claim is checkable. Breaking Defense reported that Milrem secured an order to deliver over 150 THeMIS units to Ukraine, on top of 15 already operational there since 2022, as part of a Netherlands-led initiative. The same reporting puts Milrem's director of industrial partnership, retired British colonel Paul Clayton, on the record explaining why ground robots won: unmanned vehicles became dominant for frontline logistics because manned transport had become too dangerous to accept. What began as supply runs and casualty evacuation has expanded into weaponised systems for direct fire, counter-drone operations and electronic warfare.

The pattern extends well beyond one company. Estonia's defence and aerospace industry association forecast that the country's 138 defence companies could see $842 million in sales and $518 million in export turnover for 2025, growth of nearly 350% since 2021. Threod Systems, founded in 2012 with 50 staff, has since grown to nearly 200 employees and credits a 1,100% increase in sales over five years. Its Cata launchers — designed to jolt hostile drones and loitering munitions in the air — won a $6.6 million UK contract, following trials under the British Army's ASGARD digital targeting programme. Threod's Eos C UAS has been supplied in quantities of 100 to 200 units to Ukraine, and the company now runs a training and service centre inside the country.

Why the Architecture Is the Bottleneck

The most pointed critique of Europe's response came not from a vendor but from an analyst. A Belfer Center paper published in September argues that the European Drone Defense Initiative, the EU's multi-billion-dollar programme to build a "drone wall" along eastern borders, will probably underdeliver — and that the money is large enough to build the wrong thing at scale.

The core objection is structural, not political. In September 2025, nineteen Russian drones crossed into Polish airspace. NATO destroyed four; Dutch F-35As from the 313th Squadron did most of the work. Poland invoked Article 4. In an unrelated accident during the response, a Polish AIM-9X fired at a target and damaged a civilian house in Wyryki-Wola. The paper's read is that Europe watched its most sophisticated fighters and Patriot batteries scramble against roughly $35,000 decoys at a deeply questionable efficiency.

The Belfer authors identify a deeper problem: detection is being neglected while money flows to effectors. Radar networks optimised for jets and ballistic threats perform poorly against low-flying, small-signature, low-thermal drones. The paper's sharpest detail is that Poland's own SkyCTRL anti-drone system — arguably the capability that could have detected the September incursion at source — had been delayed by 18 months for lack of funds and sat inactive that night. Their recommendation is blunt: any funded effector must be able to plug into any funded detection layer, which means open APIs as a condition of funding rather than an afterthought.

This is the same insight the Tallinn exhibitors were selling, arriving from an unexpected direction. A detection architecture that cannot see a Shahed at range is a European version of older debates about radar coverage and autonomous systems applied to a new battlefield, and it explains why the more open the sensor stack, the more valuable the effector sitting on top of it.

Robots on Both Sides of the Line

There is a quieter irony in the THeMIS story. Logistical ground robots went to Ukraine first, and the argument for them was casualty avoidance: too dangerous to send people. But as the platforms matured, the same chassis that was bought to evacuate the wounded increasingly carried weapons. The vehicle that exists to keep humans out of the kill zone has become a weapon that shoots inside it.

Counter-drone systems are following the identical curve, one stage earlier. Blaze is a drone whose job is to destroy other drones, supervised by an operator, and Milrem lists counter-drone operations among the roles its ground vehicles now take on. Autonomous effectors are becoming the default, and the only real question left is who supervises them and under what rules.

Estonia's own defence establishment is less interested in the technology than in what it implies. Kaimo Kuusk, Estonia's permanent secretary of defence, put the domestic budget position bluntly when speaking to Breaking Defense: "That's not the place you can economize or optimize your spendings." His framing of the Estonian role in Ukraine's long-range drone operations against targets in Russia was equally direct about what Estonia is willing to say publicly and what it is not.

The last word belongs to Rikk, and it was not a sales line. "Tomorrow's war is already in Ukraine," he said, standing in a factory in Tallinn.

Conclusion

If one exhibition can be said to have clarified something, the Tallinn Defence EXPO clarified that the European drone problem is no longer only a detection-and-interception problem. It is a full-stack problem that runs from passive optics to radar-guided interceptor drones to tracked ground vehicles designed to operate in the seam between them. Every one of those layers is being commercialised by small, well-funded Baltic firms rather than the continent's traditional defence primes, and the customers are arriving as much for political reasons as technical ones.

The unresolved question is whether Europe's procurement model can keep up. The industry's answer at the show was layered sensors and open architectures. The critics' answer was that money is flowing to effectors while detection, the binding constraint, goes unfunded — and that a Polish system sitting inactive for 18 months is what underinvestment actually looks like when the shooting starts. Both claims were made in the same city within weeks of each other.

What is not in dispute is the direction. The two systems that generated the most interest on the floor, an autonomous interceptor drone and an unmanned ground vehicle, share one property: neither asks a human to be in the place where it operates. That is no longer a design philosophy pitch. It is the procurement requirement, and on the evidence in Tallinn it is already the default.

Images

The kill chain, ground to sky: an MRAP fires a guided counter-UAS interceptor during a US Central Command test. Illustrative of the interceptor-drone effector layer, not of the systems shown at Tallinn.

A fifth-generation THeMIS tracked unmanned ground vehicle under field test — the class of platform Milrem Robotics markets for logistics, evacuation, reconnaissance and mine clearing.

The THeMIS with a Polish PIT-RADWAR TUGA 4D radar mast at MSPO 2026 in Kielce — the same tracked base that now carries counter-drone radar and jamming roles alongside its original mission roles.

A large quadcopter UAV airborne during a demonstration at the Portuguese Navy Operational Experimentation Center. Illustrative of the multi-rotor scale now entering NATO experimentation, not a Tallinn exhibit.

References

• Tallinn Defence EXPO: Counter-drone systems and robot soldiers steal the show — Euronews, 1 October 2026
• France orders Origin Robotics BLAZE interceptor drone system after DGA evaluation — Defence Industry Europe, 18 June 2026
• In tiny Estonia, drones lead booming defense expansion — Breaking Defense, 18 March 2026
• What the European Drone Defense Initiative Has to Get Right — Belfer Center, 14 September 2026
• Milrem Robotics — THeMIS and unmanned ground vehicle product pages
• Wikimedia Commons images: MRAP_fires_a_Counter-Unmanned_Aerial_System_(9492518).jpg, THeMIS_5th_generation_UGV.jpg, THeMIS_tracked_unmanned_ground_vehicle_with_a_PIT-RADWAR_TUGA_4D_radar_at_MSPO_2026_in_Kielce.jpg, Unmanned_Vehicles_and_Robots_Work_in_Sync_at_NATO_Exercise_in_Portugal_(9505448).jpg (CC BY-SA / public domain, via Wikimedia Commons).