AIEnabled

Inside Ukraine’s AI-Enabled Drone Campaign Targeting Russian Logistics Deep Behind The Lines

Over the past several weeks, news began emerging about Ukraine’s use of modified, AI-enhanced kamikaze drones to target Russian cargo trucks, fuel tankers, railroad cars and even vessels as far as 150 miles behind the front lines. Dubbed the mid-range strike campaign by Kyiv, this effort is having a devastating effect on Russian logistics, cutting off key highways to Crimea, helping to halt Moscow’s gains and pave the way for Ukrainian advances.

To learn more about this campaign, we reached out to one of the soldiers leading this effort from the unit that created it. In an exclusive interview, an Unmanned Systems Department Officer for the First Corps Azov of the National Guard of Ukraine offered us unique insights into how the program began, how it’s going, where it is headed and how AI is helping to identify and hit targets deep behind the lines. It’s a capability we described in great detail — about how it was rapidly gestating and would arrive on the battlefield around now. The officer spoke on condition of anonymity to discuss operational details.

Some of the questions and answers have been edited for clarity.

Q: Walk me through the genesis of the mid-range strike campaign.

A: I will first of all start with answering the question of why. So the reason why we are currently conducting this campaign is to hunt enemy logistics capabilities, and currently we have a lot of quite cheap assets that can strike enemy targets at quite a deep range. So the overall meaning, the overall sense of this campaign is to strike enemy concentrations of cargo in the places where this concentration is the highest and the protection of those enemy logistical assets is at the lowest level.

So, if we’re talking about enemy logistics very close to the front line, to the line of contact, we are mostly talking about a person with a backpack. Then, if we are moving deeper into the enemy controlled area, further from the front line, then we’re talking about cars. Then even deeper we’re talking about trucks. Even deeper we’re talking about long haulers with trailers, so a higher concentration of cargo. Even deeper into the enemy controlled area, we’re talking about the railroads, the trains that are carrying supplies. So the deeper we go from the line of contact, the higher is the concentration of enemy cargo.

The other part of answering this question is basically about how the enemy protects their logistical assets. So the deeper we go from the line of contact, not only is the distance larger, but also the areas that the enemy needs to protect. So just simple mathematical calculations show us that the distance of 50 kilometers deep from the line of contact is usually the most concentrated area the enemy has to defend against our strike assets. But if we go to the depths of 300 kilometers, then the areas that the enemy should protect are much larger, and it’s basically impossible to saturate that area with all the necessary protection assets that the enemy needs to strike down our drones.

The point I would like to make is that we conduct those strike operations at deeper ranges with basically the same assets that we use for the tactical depths, for example, at the depths of 50 kilometers. That means we do not increase the cost of our assets. We just introduced some technical modifications, and we can use the same assets to strike in the deep area.

How Ukraine Uses the US-made Hornet AI Strike Drone To Disrupt And Cut Russian Logistics Lines thumbnail

How Ukraine Uses the US-made Hornet AI Strike Drone To Disrupt And Cut Russian Logistics Lines




Q: What drones are you using to carry out these attacks and how are you increasing the ranges?

A: Basically, we’re using fixed-wing kamikaze drones with some modifications in terms of communication systems. We’ve installed Starlink systems on those drones, and we have also introduced a number of modifications in terms of the engine of the drones – in terms of the overall propelling unit. And this allowed us to increase the range of the assets that we used previously for the range up to 50 kilometers, with some modifications for longer distances.

Q: What kinds of drones are you using?

A: Just the simple fixed-wing chemical drones, such as Hornet systems, such as the Darts systems, and also some other assets that, for now, we prefer not to mention.

Darts thumbnail

Darts




Q: Can you provide additional details about the modifications you have made to these drones and what is their maximum range?

A: Well, I would say that these are quite technical things, and I wouldn’t want to disclose those, because I would want to avoid the situation where the enemy is going to do the same modifications.

Q: Are you reaching as far as 100 kilometers?

A: From what we can say now, we can cover distances of up to 250 kilometers, but as you might understand, with time we will cover longer and longer distances.

Q: Can you describe how these attacks work? Do you operate these as first-person view drones through a video feed? Do they have a terminal seeker? 

A: The process is really simple. So, for example, we select a section of the road, or the road as a whole and we distribute that road or the sections among our units. And I would like to mention that this is an all Ukrainian operation. It’s not a task for just one single unit, and basically each unit has its own area of responsibility and its own section of the road. 

And then we just send drones there in hunting mode, and I cannot disclose how this works from the standpoint of intelligence, but the overall principle is that the intelligence just gives us the priorities for the targets. For example, they say that you need to target this type of vehicle with these types of cargo.

And in terms of the targeting, I remember that you were also asking the question about how the targets are selected. For this we use both the AI tools and also the operator himself, so we’re using both methods for that.

Q: I want to get back to AI in a bit, but what area of operations do you cover?

A: Our area of interest and area of responsibility is related to all the roads that the enemy is using to bring in supplies to the area of operations of our corps, which is roughly in the area of the town of Dobropillia in eastern Ukraine, and for example, this may be the route from Rostov that goes to Mariupol and then goes to Donetsk, and this could be any other route that the enemy might use to bring military cargo to the area of operations of our corps.

First Corps Azov works in and around the area of Dobropillia in the Donetsk region of Eastern Ukraine. (Google Earth)

Q: What were some of the more successful mid-range drone strike missions you carried out?

A: That’s basically an easy principle. When I said that we are hunting the enemy’s logistics, I mean that this is a daily process. This is a continuous process, and this is not about one standalone mission, when we fly somewhere, when we hit a target and the mission is over.

Q: How are these mid-range strikes on Russian logistics affecting their ability to fight?

A: This is causing shortages of fuel for the enemy. This was already commented on by our media, and also the media resources of the enemy, and now we understand that this is true, not only for Crimea, where there is a severe shortage of fuel now, but also this is true about the other Russian-controlled territories. And basically, when we are striking enemy logistical supplies of fuel, this affects the overall situation, because what is fuel? Fuel is basically the blood of war, which is necessary to fuel the generators that the enemy is using for the FPV operations. The fuel is used for the vehicles that basically bring in those FPVs to the combat zone.

I’d like to add an additional explanation of how this works. If we destroy a car that brings fuel to the line of contact, then we’re talking about destroying like two canisters, so that’s approximately 40 liters of fuel. But if we strike a fuel tank and a tank vehicle, which is bringing fuel to the overall area, so then we can talk about several tons of materials.

Q: Is this setting the stage for a Ukrainian advance in the east or Zaporizhia?

A: I would say that blocking the logistical supplies to the Zaporizhia Oblast and also to Crimea is a byproduct of our attacks aimed at cutting enemy logistics to our area of operations. Because we’re targeting the roads that go through Mariupol, also to Crimea, to supply the Zaporizhia region, and as well our area of operations. But as for the most strategic aspect of this question of whether this might influence some counter-offensive capabilities of the Ukrainian Defense Forces, I think that is not the question within my scope of responsibility, because I’m an officer responsible for the UAS component of the corps, and I think that this is more of a question that should be directed to the general staff of the Armed Forces of Ukraine.

Q: Tell me how you came to use the U.S.-made Hornet drones.

A: I would say that this is also a question that should be directed at one level up from my level. We get those drones, we deploy those drones, but the questions of how we get them and what are the specifics of their supply, this is not a question that I can answer.

A one-way attack Hornet Drone is set up during a demonstration in the 7th Army Training Command’s (7th ATC) Grafenwoehr Training Area, Germany, March 25, 2026. The demonstration provided leaders with insight into how AI-enabled one-way attack systems operate alongside Army fires formations. (U.S. Army photo by Spc. Thomas Dixon)
A U.S.-made Hornet drone. (U.S. Army photo by Spc. Thomas Dixon) Spc. Thomas Dixon

Q: When did you start getting them?

A: We have been receiving these drones for almost a year.

Q: When you got the Hornet drones, were you given any instructions on what missions to accomplish with them or what targets to pursue?

A: I think that the principle thing to explain here is that the Hornet drone is a tactical level UAV system. It can be used at the maximum range of 50 kilometers – the maximum range of the basic configuration of this drone. The drones we are currently using are the modification of the Hornet drone, but the basic configuration of this drone is only a tactical-level UAV. This is why our mid-strike operations, which have started only recently, are using modified drones.

I think that it is important to add that these modifications are conducted by the units themselves.

Q: When did you begin to make the modifications and was the mid-range strike concept something that Azov developed?

A: Yeah, I would say that basically that was Azov development. So the first tests of these new modifications started in early winter. After approximately a month of testing, we started combat deployment of those drones approximately in January or February of this year. And we use those new modifications to ensure our middle strike operations. In terms of the tactics, I would also like to say that Azov developed new tactics of the deployment of middle strike drones, because the previous tactics they were mostly concentrated on the use of very expensive assets, which are available in limited numbers, so the targeting cycle was different, which is why we needed to develop new tactics and change this targeting cycle.

Q: What systems were you using before that were so expensive?

A: Well I’m not really talking about some specific assets here. I’m mostly talking about the overall tactics and the targeting cycle. So, the tactics of middle strikes were developed for expensive assets that the Corps didn’t have at the moment. So, this is basically why we started developing our own assets that could ensure that we have those middle strike capabilities, and we can also engage targets at the operational level.

Q: When you added Starlink, did you have to seek permission from SpaceX to use their system for these weapons?

A: The first thing that I would like to say is that we do not use only Starlink. It is just one of the possible communication systems and communication solutions that we’re using. And the second thing that I would like to add is that all the international issues, international cooperation questions, are the issues that should be dealt with at the level of our ministries and at the level of our central government, so once again, this is not a question that I can answer.

Q: Can you tell me what other communications systems you use?

A: No, this information is still secret, because none of those assets have been taken by the enemy, so the enemy doesn’t know yet that we’re using those. But I think that in some time we will be able to go public with those assets and tell more about what we use, apart from Starlink.

Q: Are you using Starlink and the other systems on the Darts drones as well, or just Hornet?

A: Yes, we are using Darts for those middle strike attacks. So, basically, the overall tactics of those mid-strike attacks with small UAV systems, it can use any fixed wing kamikaze drones with modifications which is why we are using Hornets, we are using Darts and also the other types of drones.

Q: Can you talk about the other types of drones? 

A: Unfortunately, no. The enemy has not yet got hold of those assets, and if one of them will be shot down, or if one of them will be captured by the enemy, then I think that we will be able to go public with those assets.

Q: Can you talk more about how you are using AI for the Hornets and Darts?

A: The AI is used for the so-called last-mile system. I’m talking about the terminal guidance here, and also the AI can also be used to identify the targets, especially when the drone is flying in the autonomous mode. It can recognize the type of the target and it can automatically engage the target. So all of these processes can happen without the involvement of the operator. This allows us to launch several drones at the same time, and this also allows us to ensure a higher level of coordination for those attacks.

I think that one of the examples of these attacks was captured by one of our videos that you might have seen when one fixed-wing drone is observing the target and the other drone is actually hitting it. 

1-й корпус НГУ «Азов» контролює ворожу логістику під Донецьком. Знищення російської техніки thumbnail

1-й корпус НГУ «Азов» контролює ворожу логістику під Донецьком. Знищення російської техніки




Q: Is there a man in the loop or a man on the loop during the terminal phase of the attack or is it fully autonomous?

A: Since we are trying to control the whole process, usually it is the person who takes the final decision and actually makes this decision to strike the target, because we do not want to over rely on the artificial intelligence. If necessary, this whole process can also happen without human involvement, but as I have already mentioned, our policy is that the decision should be made exclusively by the operator.

Q: So this is man in the loop?

A: Yes, so technically men can be out of this loop, but in our case we involve humans in this system, so that’s man in the loop.

Q: Have the Russians developed countermeasures for this effort?

A: Since the overall intent was to go to the massive deployment of those drones against the enemy logistics at once, as for now, the enemy has not yet had enough time to adapt to these tactics of ours. And now they’re desperately trying to find some countermeasures to find some ways to react to this. 

Several images showing the unusually painted Russian trucks have appeared on social media channels in recent days. So far, examples of Ural and KAMAZ heavy-duty truck designs have appeared. There are at least two distinct patterns so far: a zebra-style application of broadly straight lines, and a more organic leaflike, swirling design. In both cases, they extend over most external surfaces, including the wheels and tires. The white paint is simply applied over the base color of dark green.
Images showing unusually painted Russian trucks have appeared on social media channels in recent days with a paint scheme designed to throw off Ukraine’s AI-assisted mid-range strike drones. (Via X) via X

They are putting on anti-drone nets deep into their controlled territory. They are placing people armed with shotguns every 50 meters of their critical roads, but basically, as for now, there is no Russian tactic that would be effective in protecting their logistics. I do not say that they will not come up with this tactic in a month or so, but as for now, they do not have an effective countermeasure for that.

The only thing that I would like to add is that we approximately know the direction in which the enemy is moving to ensure counter measures to this tactic of ours, and we already have countermeasures for their countermeasures.

Q: Is the AI enhancement helping to mitigate Russian jamming and electronic warfare measures?

A: Well, I would like to say that EW is not the only way to counter UAVs. So, in addition to EW assets, there are other counter drone methods. Nets can be used. Air observation posts can be used. The drones can be shot down with enemy interceptor drones, and also there are hundreds and hundreds of people armed with shotguns who can also try to shoot the drones down. So the EW systems are just a small part of this overall system of drone countermeasures, and it is a general misconception that all anti-drone efforts are from the EW system. I would say that EW accounts for approximately 10% of overall counter drone efforts.

The main idea of the use of the AI systems is not only to help operators to counter enemy drone countermeasures, but the AI also helps the operator with navigation, with orientation, with the identification of the targets. So this is a complex system that helps the operator with many tasks.

Q: What are the lessons that the U.S. can learn from your mid-range drone strike campaign, given the importance of logistics everywhere.

A: Well, I would say that the main lesson that the U.S. can learn from our experience is that drones in their basic configuration, right out of the box, is not something that can work, and this is not something that can bring you the best results at once. This is why every unit should have their own drone laboratory that can reconfigure those drones and modify them, and I’m not only talking about reconfiguring tactical level drones to turn them into middle strike capacities. I’m talking about all warfare in general.

U.S. Soldier Sgt. Kevin Tran, assigned to 173rd Airborne Brigade, sets up a one-way attack Hornet drone during a demonstration in the 7th Army Training Command’s Grafenwoehr Training Area, Germany, March 25, 2026. The demonstration provided leaders with insight into how AI-enabled one-way attack systems operate alongside Army fires formations. (U.S. Army photo by Spc. Thomas Dixon)
U.S. soldier Sgt. Kevin Tran, assigned to 173rd Airborne Brigade, sets up a one-way attack Hornet drone during a demonstration in the 7th Army Training Command’s Grafenwoehr Training Area, Germany, March 25, 2026. The demonstration provided leaders with insight into how AI-enabled one-way attack systems operate alongside Army fires formations. (U.S. Army photo by Spc. Thomas Dixon) Spc. Thomas Dixon

Assets become obsolete very quickly and in three months everything can change, starting from the navigation systems and ending with the control systems. For example, today the GPS signal works, in a month it doesn’t work anymore. Today we use this set of frequency ranges, and in a month all those frequency ranges are jammed by the enemy. So I’m saying this for you to understand that for example, if the U.S. government purchases Hornet drones for their units, then the units are going to have good tactical level drones. But if you want to squeeze all of the possible benefits from those drones, then you will need to modify the structure of your units and include the departments, include the sections that will deal with the modification and reconfiguration of those drones to ensure the most efficient results that can be achieved.

From what I see now, the U.S. is mostly just giving the unit some assets, but they are not thinking about re-configuring or repurposing those drones.

Q: Is there anything I haven’t asked you that you would like to address?

A: I think that we have covered all the topics. The only thing that I would like to add is that I would like your audience to see that Ukraine introduces new innovations that can basically change the battlefield, and this is why trying to forecast the outcome of the war based on the number of tanks or aircraft or bombs that would be incorrect, so the victory will belong to the side capable of adapting faster and not to the side that has more resources.

Contact the author: howard@twz.com

Howard is a Senior Staff Writer for The War Zone, and a former Senior Managing Editor for Military Times. Prior to this, he covered military affairs for the Tampa Bay Times as a Senior Writer. Howard’s work has appeared in various publications including Yahoo News, RealClearDefense, and Air Force Times.




Source link

Russian Trucks Get ‘Dazzle’ Paint To Throw Off AI-Enabled Drones

From an early point in the Russian war in Ukraine, we’ve seen many unorthodox efforts to try to improve the chances of survival of fighting vehicles. Now, Russian trucks are receiving ‘dazzle paint,’ borrowing the same kind of tactic Russia has used for some of its most important military aircraft, to try to confuse seekers on standoff weaponry that use image-matching capability.

KAMAZ truck with zebra-style pattern. via X
via X
via X

Several images showing the unusually painted Russian trucks have appeared on social media channels in recent days. So far, examples of Ural and KAMAZ heavy-duty truck designs have surfaced. There are at least two distinct patterns so far: a zebra-style application of broadly straight lines, and a more organic, leaf-like, swirling design. In both cases, they extend over most external surfaces, including the wheels and tires. It is not entirely clear if the white paint is applied over a layer of black, or portions of black, or if the white is simply coated over the standard base color of very dark green. It could be a mix of both application concepts, as well.

Ural truck with leaf-like pattern. via X
via X
via X

At first sight, the truck patterns recall the iconic paint scheme that the U.K. Royal Navy pioneered for its warships back in World War I. ‘Dazzle paint’ or ‘dazzle camouflage’ was devised in 1917 by official War Artist Norman Wilkinson, as a means of reducing losses to attacks by German submarines, or U-boats. The geometric patterns work by using highly contrasting color blocks, often heavily featuring black and white,  as part of a carefully constructed pattern that breaks up the form of the ship and makes it harder to judge range and perspective.

British aircraft carrier HMS Argus wearing dazzle camouflage in 1918. Crown Copyright

As you can read about here, this kind of naval camouflage scheme appeared again during World War II, and on several occasions since then.

The Canadian frigate HMCS Regina, carrying a dazzle camouflage scheme, takes part in an exercise in 2020. Canadian Forces

When first introduced, dazzle paint was intended to trick the human eye, normally looking through a periscope. There was still a benefit to be had in protecting vessels after the introduction of improved rangefinders and radar. For the eye, it made it harder to judge a ship’s course and speed, as well as simply identifying it reliably.

The same basic principle is at work on the dazzle-painted Russian trucks, although now it’s an artificial eye — chiefly using electro-optical and/or infrared cameras — that is supposed to be fooled.

Increasingly, Ukrainian drones are using artificial intelligence (AI) to boost their combat effectiveness. The revolutions that are coming as a result of embedding of AI into lower-end drones is something you can read about in our past feature here. This includes machine vision: a process of the drone learning object recognition, identification, classification, and tracking, as well as providing recommendations for the operator on what to do, provided there is an operator at all and the drone is not running autonomously.

An HX-2 drone in flight. The HX-2 has some capabilities enabled by AI. Helsing

AI-enabled capabilities make lower-end drones more resistant to electronic warfare systems and make it easier for them to be employed in networked swarms. Above all else, they can result in the cutting of the invisible radio frequency tether of constant man-in-the-loop control that in many ways hampers the potential of this class of drones.

The drawback of machine vision that the Russian countermeasure is supposed to exploit is the onboard AI agent’s capacity for learning object recognition. While it may be able to recognize a 6×6 Ural, for example, out of a wide range of potential truck targets, if the appearance of the vehicle is distorted enough, it will not be positively identified, or at least meet the threshold of corroboration that would result in a kinetic act. However, still with many drones that feature AI assistance, a human operator stays in or on the loop for all critical decisions.

This raises the question of how successful the dazzle-painted trucks might be, although the thinking here presumably stresses avoidance of detection during the autonomous target-search phase, rather than the endgame of an engagement. It is also worth noting that these kinds of paint schemes only really have value in areas where they are unlikely to be seen by any Ukrainian human eyes, even remotely via a sensor; after all, they are far more conspicuous than their standard-painted counterparts. It’s also possible that a drone could be taught to specifically hunt these patterns, as nothing else on the battlefield would look like them and they would be confirmed hostile by default.

Overall, paint schemes are another logical, if extemporized response to a growing threat in the Russian rear areas, following the example of the Russian trucks loaded with logs as makeshift armor to protect against kinetic threats in the early phases of the war.

A Russian truck with improvised armor made of logs, in 2023. via X

This has been followed by successive counter-drone measures, best exemplified by the increasingly complicated ‘cope cages,’ ‘turtle tanks,’ nets, and arrays of spikes that have appeared on a range of vehicles on both sides of the war.

Russian ‘turtle tank’ seen operating with additional cage armor and an attached mine roller. via X

Perhaps most apposite, however, is the example of Russian bombers and strike aircraft being covered with disused tires, something that first appeared in around August 2023. TWZ was first to postulate that these were most likely intended to confuse the seekers on Ukrainian cruise missiles and drones that use image-matching capability. This was subsequently confirmed by a senior U.S. military technologist.

A close-up of a Russian Tu-95MS bomber with tires on the wings and top of the center fuselage at Engels-2 Air Base, taken August 28, 2023. Satellite image ©2023 Maxar Technologies

A “sort of classic unclassified example that exists is like a picture of a plane from the top, and you’re looking for a plane, and then if you put tires on top of the wings, all of a sudden, a lot of computer vision models have difficulty identifying that that’s a plane,” Schuyler Moore, U.S. Central Command’s first-ever Chief Technology Officer, explained in September 2024.

Moore said this as part of a larger discussion about AI models and data sets.

It’s also worth noting that Russian combat ships based in Crimea also received unique shading to break up their silhouettes for the same purpose during this period.

As TWZ has explored in detail in the past, AI is now pushing drones toward a major new evolution, if not a revolution in capabilities.

As well as the possibility of operating in large groups or fully networked swarms, it means long-range one-way attack drones can conduct dynamic targeting deep in contested territory. Trucks, for example, can be hunted and struck far behind the front lines, where once they were safe and where air defenses are sparse.

This is a scenario we have set out in the past, too:

“Waves of similar drones could be sent to their own individual geographical ‘kill boxes,’ or defined areas of engagement. Collectively, they could put enemy targets at risk over a huge area persistently without ‘doubling up’ and attacking the same target twice. Using machine learning/AI and associated hardware, they could not just identify targets of interest, but also differentiate moving from still targets, to ensure they are indeed active (not destroyed or already damaged) vehicles. Meanwhile, they can be set to engage other target types, such as surface-to-air missile systems or other high-priority targets, regardless of whether they are static or not. Even troop movements on the ground could potentially be recognized and attacked. All the parameters as to what the drone can engage, and where it can do so, can be defined and tailored to each mission before launch.”

ODESSA, UKRAINE - 2022/08/27: A man with a bicycle passes by a burned-out KAMAZ truck. Units of the operational command "South" of the Armed Forces of Ukraine brought Russian military equipment burned in the battles for the independence of Ukraine for display. The T-90 tank, two transporters MT-LB, Kamaz, and armored personnel carrier are exhibited. (Photo by Viacheslav Onyshchenko/SOPA Images/LightRocket via Getty Images)
A man with a bicycle passes by a burned-out KAMAZ truck, part of a display of Russian military equipment destroyed in the fighting in Ukraine. Photo by Viacheslav Onyshchenko/SOPA Images/LightRocket via Getty Images SOPA Images

It is also worth noting that different types of sensors will be affected to different degrees by passive countermeasures like complex paint jobs. While electro-optical sensors may have issues with the patterns, infrared may not, especially at longer wavelengths.

TWZ has previously highlighted how AI algorithms can be rapidly trained in a digital environment, as well as incorporate data collected from actual real-world employment, to improve their ability to spot, categorize, and engage targets. It is, however, unclear how hard it would be to overcome infinite dazzle patterns. It could, as Schuyler Moore observed, lead to software programmers spending inordinate amounts of time on computer vision with very little to gain, once a new pattern arrived.

While it remains to be seen how effective the dazzle-painted trucks might be, they are another sign of drones, especially AI-enabled ones, being one of the key drivers of innovation on the modern battlefield.

Contact the author: thomas@thewarzone.com

Thomas is a defense writer and editor with over 20 years of experience covering military aerospace topics and conflicts. He’s written a number of books, edited many more, and has contributed to many of the world’s leading aviation publications. Before joining The War Zone in 2020, he was the editor of AirForces Monthly.


Source link