Inside Ukraine's Azov Drone R&D: The Engineer Building AI Weapons 18 km From the Front Line | Alexander Palamarchuk

Autonomous weapons on the battlefield

0:00The U.S. has said that we'll not develop fully autonomous lethal weapons for ethical reasons, but it seems on the battlefield, every side is trying to get an advantage. In two years, you'll be able to field fully autonomous drones. I mean, that's a frightening world. I would guess that that technology is going to be on the battlefield very soon if it's not there already. The drone with AI is on the battlefield. So now, there is a presence. There is some part of AI that is a good drug of license in targets, but it footage recognizes, for example, right target, wrong.

0:39It's very good drug of license vehicle, but it is not good enough to know is it enemy or not. And that's the problem. The NATO countries are trying to get the right weapon. I mean, right model shoes. They are trying to study the small reactive.

Azov drone development unit

0:54Paradise is the call sign of a man named Alexander, a Ukrainian drone specialist serving with the Azov Corps of the Ukrainian National Guard. Azov began in 2014 as a volunteer battalion. It drew international notoriety for its links to far-right movements and was later incorporated into Ukraine's National Guard, where it has grown in one of the country's most prominent combat formations.

1:27Paradise joined in 2023 and now works in an R&D unit developing drone technology near the front. I spoke with him remotely about jamming, autonomous navigation, AI target recognition, and how drones are rapidly reshaping the battlefield and the future of warfare.

History of drones in Azov

1:49First of all, I wanted to ask you about the history of Azov, when you joined, when Azov started using technology. I joined Azov in 2023 and work only with drones. If you are talking about Azov, as I know, the first usage of drones in Azov was like in 2017.

2:25They were used to reconnaissance firstly and then to strike some bombs. To strike drones as anti-drone warfare, is that what you mean? They were used to hit different kinds of targets. As a fast forward, it was used as reconnaissance drones. So you can just be in some kind of tension and use it when you can even see your enemy with your own eyes.

3:00But it gives you a good angle to view what is going on on the enemy side. Right. But the war didn't start, the invasion didn't happen until 2022. So in 2017, where were you using this? Actually, it started in 2014 for us. Actually, even on YouTube, you can find a video with Russian regular army in Donetsk region.

3:31And so it was used with the same enemy from 2014. I see. Yeah. In Donetsk. That was started with own-made drones. Then when some kind of DJI fandom survived, guys tried to use them. And as I know, they were glad that it gives good possibility for reconnaissance.

4:01Okay. And when you joined, they were still using DJI or home-built or when you joined the effort?

4:13Even for now, we are using some Mavic and Matrice drones for reconnaissance for some kind of bombers with light bombs. Mainly, we use them as bomber. And now, on this here, we use them mostly for reconnaissance. Because for now, we have more possibilities. And to take bigger bombs or fly lambda, stay in the air for longer, actually.

4:50They have good cameras now. And we are on the way to change the Chinese drones for our own-made drones. I mean, Ukraine-made drones, our partners-made drones. We are on the way for this.

Jamming and communication methods

5:07One of the big problems with drones on both sides is jamming. And I've seen all of these different jamming systems, but also systems to evade Russian jamming. I mean, how do you handle that? I know there's the fiber-optic line that unspools up to 40 kilometers or something.

5:38But I've also seen drones with Starlinks mounted on them. So there's satellite communications, which I understand is unjammable. Can you talk about the jamming issue on both sides? Yeah, I can. I will divide it into three parts. Like radio, like fiber-optic, and like internet connection. Let's just start it with radio. This is a classic drone switch I started to use in 2023.

6:13First, we use its common frequencies. It's the same as LTE, for example, or some 2.4 GHz. It was regular. Then we meet some jamming from enemy's side. So we are trying to use the possibilities which we had with ELRS. And we can move from 700 MHz up to 1000 MHz.

6:47And in this frequency range, we are trying to find the possibilities for us to fly over the enemy's side. And actually, until, I guess, the meet of 2024, it was more than enough. Then we moved to our LRS, maybe, you know, kind of ELRS, which are made in Ukraine. This is based on common ELRS, but it's made for our needs to have a stronger communication with the drone and ground station, which was helpful to fight the jamming.

7:33For 20-25 year, we moved to not regular ELRS receivers, transmitters. And for now, we can use almost from 100 MHz up to 3000 MHz. And we are trying to find some frequency ranges where can we walk freely. And it mostly depends on where you're trying to fly.

8:07Just for example, we couldn't fly to criminal region on common frequencies. But the guys who stand five kilometers next to us could freely fly on these frequencies on their point of interest and feel good for them. So we are trying to resolve our problems and try to find the solution for this by ourselves in the unit.

8:46For now, we are using not ELRS for radio communication, but some kind of Ukrainian stuff or our partner stuff. Okay. And then what about internet communication via satellite? Actually, maybe, you know, some names of drones that use it even from 2022 or 2023. Yeah. Like math or something like that.

9:16This is a common internet connection, actually, via Starlink or some satellites like this. Actually, there is nothing, there is no, actually, secret from this. If you open, I use Marlink, which can be routed via internet or via some local network or something like this. And you have your control and you have your control and you have your control and you have a video from a drone.

9:51So it gives you possibilities to fly where the something else wouldn't fly. You will have actually no problems to fly over some kind of buildings or something like that even far from the ground stage. Is there a possibility to jam the Starlink connection? I guess that is possible.

10:22So we know actually what happened in Iran, so it is possible. I haven't made it on Battleground for now, but I think that is possible somehow. It is not, this is not the thing where you could find easily, but you need some kind of very expensive stuff, which could easily be hit by something else. Are the Russians also using something like Starlink?

10:57Yeah, yeah, yeah. You can find even videos where they are using Starlink. Right. And so at this point, neither side has an advantage, is that right, with technology? This is the kind of, uh, that we could, uh, uh, find the solution in our own unit and, uh, by, uh, one person. Uh, and, uh, when you are trying to, uh, solve some problem, like, uh, UI jammed or something else you can find actually in, in three parts of the internet.

11:34Very, very, very, very, very many, um, very much information about, uh, what you can do with it. Uh, and, uh, this is good when we have some, uh, kind of, uh, engineers who can, uh, sources. So we are good at, uh, making some, um, very interesting, uh, projects or very interesting solutions, uh, to, uh, some kind of jamming.

12:05If you are talking about jamming, but, uh, I guess the Russians are good in, um, to do it more, I guess, in kinds of, uh, as a part of system. Uh, they are good at producing, uh, many, many, many drones and this, uh, of using this kind of, uh, communication, for example. Yeah. Okay.

Artificial intelligence in operations

12:33Are you, are you guys using any artificial intelligence in your... Yeah, using this as, uh, early programmers to, uh, to make our lives with code is much, much easier. For example, um, actually, I mean, very, uh, using, uh, when we are working with some data, for example, it is very helpful too. Yeah. Um, uh, uh, it's, uh, uh, it's, uh, uh, it's a common, uh, I guess, as a way to use, uh, AI, mainly using it in actually the same way.

13:10Yeah. And that's, are you using cloud code or what, what model are you using for, for, uh, Actually, uh, when they have some, uh, something we should code. We are trying in this model or this model or this model and which would good, uh, which would get us better results. We use that. Yeah. So I couldn't tell that this one model is great for us, but, uh, they use different models, which works well.

13:42Okay. We use it. Yeah. Yeah. And are you using, uh, uh, AI in, uh, in drone number to help the drone navigate or in, to auto-nive? Yes, of course. Yeah. In terms of how, how you're using, uh. There are two kinds of, uh, ultra-con recognition for drones. Like, uh, where you, uh, track some pixels or when you recognize some kind of UAVs or something like that.

14:15Or not UAVs, but, uh, some kind of cars or this. Um, when you're tracking the pixels, uh, you move, uh, some, uh, little, for example, forest and you just lose your target and the drone couldn't recognize this car. Uh, this is armored vehicle or this is civilian cars. It's not the grade, but cause you, um, always, uh, the operator should track what the drone is doing.

14:46Uh, you could, you should, you must actually, you couldn't, uh, uh, use drone as like move to this area and hit everything you see. Cause you need, uh, uh, to recognize this civilian car or this, uh, them already destroyed car or some, all of this. And we, uh, if we are talking about, uh, AI models for auto recognition. They are much clever. They are much more clever. They can, uh, recognize by own drone.

15:19Is this, uh, is this correct target for hit or this isn't. But, uh, uh, operator should be present and he has, uh, and don't control just, uh, in the case that, uh, this should be enemy. So you're saying that, uh, they're, they're models for, for recognizing objects.

Autonomous navigation systems

15:48Uh, and what about navigation? I mean, how autonomous are the drones today? Uh, you know, there's FPV, of course, uh, where the operator is watching, uh, uh, through a screen, either goggles or, or, uh, on, on a screen and, and guiding the drone that way. But if you have object recognition in the drone, uh, can you do, uh, can you, can you make it autonomous so that once it locks on an object, you no longer have to guide it?

16:32It's a case where the eyes help. Uh, cause it's put, uh, uh, track the target correctly. If you are talking about navigation, I guess you know that the GNSS is almost everywhere is jammed. And yeah, the ball ground. Uh, so we are, uh, we have two ways to resolve this problem. Uh, the drone couldn't, uh, navigate by itself cause of the GNSS, uh, jamming. Uh, this one is a visual position.

17:04Um, uh, uh, as a drone should have a camera or the bottom of it. Uh, see, uh, it's, uh, check what, uh, is above this. And, uh, uh, uh, compares to, uh, uh, maps, uh, which preloaded to the drone and the result also. I'm here. So it can move, uh, to, uh, uh, position with some kind of, uh, accuracy.

17:38Uh, uh, some problems with accuracy up to 50 meters, for example, um, difference, uh, you know, position for use, uh, different, uh, results. Um, and, uh, other, uh, other way to solve these problems, to use, uh, beacons, uh, like the phone communicates with ground station, with, uh, one, two, three, uh, beacons.

18:10And, uh, it gives you, uh, a point where it is. It depends on the, uh, time with, uh, which it communicates with beacon, with the ground station. It compares, uh, the time it needs to, um, communicate with, uh, ground station. Beacons, it gives a different time. So, you know, uh, you are closer to ground station and more, uh, and, uh, far from beacon.

18:41You know, this time it needs to, um, to communicate. For example, you need to do 10 milliseconds for one kilometer, for example. Uh, if you have 60 milliseconds, you are far as about six kilometers. And, uh, beacons help, uh, it's to angle age where you are. I see. Yeah. Uh, and, uh, have you guys experimented with fully autonomous drones?

19:16Yep. Yep. Yep. Yep. You do. Can you tell me how that works? Uh, actually you just place this one on the ground to, uh, then you, um, make some kind of function for it. Uh, uh, it's, uh, takeoff by itself. So that calibrate, uh, for example, compost or maps or finding the, uh, radio link with beacons,

19:47some kind of kind of this, and then move to the place or area where it should be. And then, uh, that's the case, uh, where is AI needed? Uh, cause the problem is that you should, uh, mark as it is correct target. Um, just for now, it is impossible to, uh, hit only enemy targets. So it would be civilian. So you couldn't hit this kind of targets.

20:19It's civilian car. Even if enemy uses civilian cars as its arm. So you, this, um, we couldn't hit this kind of target. So you need, so be sure that it's enemy 100%. Um, if it is not, uh, uh, coming considered wrong, uh, it is possible to make. So we'll fail to know a solution. So just move to this area, watch this, watch this, watch this, and then it returns back.

20:53Are the Russians using fully autonomous from what you can see? Um, I guess we know about the hits. Uh, it's using GNSS, some kind of CRPA antennas. And, uh, even it's, uh, uh, many jammers to jam CRPA antenna. So up to, I guess, uh, they have some problems with this for now. Uh, cause we saw, uh, CRPA antennas with 16 elements.

21:28So you need, uh, 16 jammers to jam this. They are moved to Starlink. Oh, now they are using it. So if you saw the, the, the one of the last, they hit the train, civilian train, which moved to, uh, or, uh, in fact, I guess, up to, uh, even now you could find some, uh, those fully automated. They just, uh, take over this and move to this area, hit, hit this.

22:04Uh, and if it would be jumped, it hits. Anyway, it would. I don't know. Uh, if the drone things is there, uh, to hit this, or even if it jam. Yeah. I mean, that would be, I mean, because the Russians don't seem worried about civilian casualties on the Ukrainian side. Uh, so I, I'm wondering if they're gonna move to fully autonomous that will target any,

22:38uh, any vehicle or any, uh, uh, person in the, in its range. Actually, it's, uh, it's much easier for them. They don't care about any civilians or something like that. If you could, uh, hit any targets that can, that move faster, for example, than 20 kilometers and a car with your car, uh, this is a target. It's much easier than if you can just, uh, move to this area, hit everything that is moving.

23:15Uh, so, uh, it's much easier than hitting, but, uh, say, too much, too many civilians was heated like this. Uh, actually they even have telegram channels where they show that, oh, this was, uh, a woman with, uh, for example, bicycle here, even in this Mavic manual control.

23:47They saw what they did. It's okay for them. So it's much easier for them to make some absolutely, uh, and autonomous missions. Cause they, uh, shouldn't care about the type of term. Let's hear that.

Russian tactics and equipment

24:05I mean, do you analyze the remains to see what kind of, I don't know if there would be anything left, but to see how autonomous the Russian drones have become? I mean, do you periodically, uh, capture one before it's, uh, detonated and, and look at how it's being guided? Yes, of course. Of course. We are trying to, um, and get something, uh, interesting from their engineers.

24:39I absolutely, uh, confident they, uh, they, uh, might, uh, do something like we do. Uh, they are trying to, uh, reproduce our good, uh, ideas. We are trying to reproduce their good ideas. And the device that should work. Um, cause if you don't want to use good idea of your enemy, you will use some kind. Yeah.

25:09Of course. Yeah. It wasn't some good ideas. Uh, uh, they have it. Uh, this, uh, shacket is very, uh, efficient, actually. They can produce a lot, uh, and, uh, hits. This is, uh, uh, this kind of drones, uh, isn't so accurate for now, but, uh, they are using some mesh systems, some star links, uh, to make it, uh, uh, to have a control of it,

25:43to get a video from them, to get some kind of telemetry from them. Um, actually, everyone wanted to have a super autonomous drone. Just press six to win. Just press six, 10,000 drones fly, hit every enemy. I mean, uh, but for now I see that both sides of this conflict of our war trying to get, uh, as much control drone as it, as it was.

26:16That's where we, uh, everyone's trying to use in some, uh, network systems, uh, like mesh or star links or any kind of, uh, internet connection to the drone. Yeah. I mean, what you said about launching, you know, 10,000 drones at once, uh, do you think if the war continues, it'll get to that point, uh, that, that Russia could launch 10,000 autonomous drones?

26:46Uh, so, so the drones, uh, kill everything or, or hit everything that's moving. Uh, I mean, that's pretty frightening. Uh, do you think that Russia is closer to this than we are actually? Yeah, except that, except that, as you said, you're, you're fighting over Ukrainian territory and you don't want to hit Ukrainian civilians.

27:17Actually, you know, we have, uh, the Ukrainian army, we have an operation in, uh, uh, Russia, uh, and the same, the same thing. The civilian, you couldn't hit it. But we saw this, uh, battalion commander here, Russian army. No, we couldn't. We couldn't. Just civilians. And, uh, the partners, uh, looks what we are doing. They saw that try not to, uh, to hit, uh, civilians, uh, as possible.

27:52So that's, uh, one of the problems to use, uh, so the autonomous, uh, chemical, the drone force. Uh, what is this possible to make, uh, some kind of, uh, autonomous, uh, like, uh, or, like this missions, uh, when you, uh, couldn't hit someone? Uh, is it, uh, do we, uh, close to, uh, the life 10,000 bonds, for example, 10,000

28:24drones, but surely, yes, I guess, uh, two, three years, I mean, uh, have the situation where we would launch, uh, many, many, many drones. The one problem is, uh, you said correctly. The navigation of the drone. And if you have some kind of systems that, uh, could, uh, and recognize the enemy correctly, uh, without hitting any wrong targets, it's possible to use, uh, many,

28:59many autonomous drones. The one problem with 10,000 was you could, you could have, uh, so much place to launch. Actually, it, uh, takes, uh, much time to, um, prepare for launch. Two, three years. It's, it will be possible. I know. Uh, I am really, I was launching, uh, five with one ground station drones, uh, at the one time.

29:29But I think it is possible to make, if it is fully autonomous drone, if you don't need to control drones, uh, that it's a bit easier. And then if you need to control any, uh, drones, for example, this 1,100, you know, you need to control this in this time. This, uh, much harder. Then it's, uh, maybe it's fully autonomous.

29:59Yeah. Yeah. So you say in two years, it's possible that, uh, you would have, you would be able to launch 10,000 autonomous drones. I mean, you mean the tech will be there. Uh. Uh, yes. So actually it could be even in this, uh, year, uh, not about 10,000, but even if you could, uh, launch the 100 drones, it's, uh, very hard to deal with it.

30:31I mean, do you worry though that Russia is going that direction? Hmm. Everyone, everyone is going in this direction. Russia, Ukraine, and, uh, anyone who wanted to have, uh, uh, an up-to-date weapon. Yeah. About, uh, fully autonomous weapons that could, uh, recognize it. Correct target. Uh, they are, that couldn't be jammed or something like that.

31:03Everyone wants this kind of weapon. Yeah. Yeah. And the development because the, I mean, there's the hardware production, but a lot of that is software is the AI. Do you have. Of course. Absolutely. This is AI. Yeah. And, and do you have, uh, AI developers that are working on models because that kind of model exists today. Uh. Yep. Yep.

31:33Yep. Yep. One of many of them, uh, uh, uh, the NATO, uh, countries, uh, trying to get, uh, the right weapon. I mean, the, uh, right model for this. They are trying to, um, learn, uh, study this model correctly. USA. Actually, anyone who, who wanted this kind of drone are on the way to have this.

32:05If they're trying to use, uh, AI, they're trying to place it on the drone and test it in, in our battleground. Yeah. I mean, the reason I ask is, uh, so much about this is the U S, um, has said that it will, will not develop, uh, fully autonomous, uh, lethal weapons, uh, for ethical reasons. Uh, but it seems on the battlefield, uh, every side is trying to get an advantage and that

32:44would be an obvious advantage. Uh, do you, do you, and you said everyone's working in that direction. And you said in two years, uh, you'll be able to field fully autonomous, uh, drones. Uh, I mean, that's a frightening world. If, uh, two armies have huge numbers of fully autonomous drones that would kill every living

33:14thing in a certain zone, uh, is that where warfare is, is headed? Do you think? Even in this century, we will see the conflict with, uh, if it would be enough conflict, uh, with, uh, uh, some kind of big, uh, countries which have, uh, uh, wanted to have, uh, very expensive, uh, weapons.

33:44It would. Yes. So this is just my opinion. Um, what, uh, you know, about, uh, everyone knows it. Canis are trying to get this kind of drones. They are testing their, their, uh, software here. They send people, them, their people here, uh, to get some kind of, uh, experience, uh,

34:15how to use the rules of the war just to, uh, and just, uh, just, uh, for example, the, uh, heavy armored vehicle like tanks. Russians, for example, in their assaults, uh, don't use the tanks as attacks. This is a armored, uh, vehicle, just armored vehicle and nothing. And they take, um, uh, some infantry on them.

34:45And this is the one reason is to have a big, big, big jammer on it and move as far as it can. That's all. Tank even isn't a tank. No. Actually, the artillery is suffering from drones. Infantry is suffering from drones. Everything is suffering from drones. Everything is suffering from drones. And, uh, as a war, we would have the rules. This, uh, this will provide new rules of what we are doing, what we are afraid of.

35:20Uh, is some kind of, uh, autonomous drone is present here. Of course, uh, it's very good at recognizing vehicle, but, uh, it is, uh, not good enough, uh, to know, is it enemy or not. And that's the problem. And, uh, this is, uh, what, uh, is, uh, much easier for Russia. If you don't care about, is it enemy or just vehicle, you can just hit. And, uh, this is the point. They have, uh, um, drones with AI, uh, just like U2V or something like that.

35:57But they don't have, uh, many of them. They have, uh, many, uh, stupid, stupid drones. Uh, that you should, uh, have a rate or does it move? That recognize the target, hit the target. But, uh, they don't have many of smart drones. The Bumblebee is smart drone. Uh, but, uh, Russia, uh, don't have something like that.

36:27I would guess that that technology is going to be on the battlefield very soon, if it's not there already. Is that, is, is that fair to say? Uh, yeah. Uh, yeah. As I told later, uh, uh, the drone with AI is on a battleground, uh, for now. They are present here, but, uh, they couldn't recognize, uh, with, uh, 100% accuracy that

36:58it is target what you should hit. Uh, and this is a problem cause, uh, they, they could use just civilian cards, just sit in civilian cards and use it as your own. So would you hit it? Or wouldn't you hit it? Uh, is, uh, the drone, uh, just have no information. It is a source of car. They recognize this as a car, but, uh, the operator should understand, is it enemy car?

37:28For example, if you are, uh, if you have a rear with, uh, there is no civilians, absolutely no civilians. You can use autonomous mission, any car, just hit it. But, uh, if you are, know that, uh, there is some civilians could be there or you couldn't, uh, use this kind of issues. Does Azov have just one R and D group or many R and D groups around the country? Um, actually, uh, for now, um, many units have R and D groups, uh, with, uh, different, uh,

38:09with different results, I would say this way with different results. Some of that are good, some of that are, uh, not so good. They are trying to find, uh, their ways. I think, for example, uh, in Azov, as I know, I am on a call and some other R and D group is present. Um, really, uh, I know some people from other R and D groups.

38:39Um, they are doing their stuff. We are doing our stuff. Uh, sometimes we are doing, um, in communication, uh, and, um, as usually we are different in deep parts. Yeah. And can I ask in, in your group or, uh, is the primary focus navigation, jamming, or, uh,

39:09uh, object recognition? I mean, which is most important? I mean, target recognition. Uh, I understand. Um, actually we are, for now we are focused on navigation.

Current focus on navigation

39:23Uh, we have some good results in it. I, I, I wouldn't, uh, you know, tell about this. We are focused on it. We are, uh, make some results. Um, about, uh, some kind of recognitions. This is an extra. When you can move to this repair, then you can, uh, use some kind of recognition. For now we are focused on navigation. Uh, if we are talking about, uh, autonomy smith.

39:57Yeah. First of all, navigation, then, uh, of course, this navigation is a national reconnaissance drones. Uh, and, uh, this, if it could, uh, uh, do some missions, uh, even without, uh, without recognition and salary, uh, just to work with this thing. Yeah. And again, on the models, the AI models, uh, you don't want to say where you're getting

40:29them. Is that right? And, and on the chips, what, uh, where are you, are you getting the, the semiconductors to run the models? Uh, actually, uh, this is more complicated. Uh, this is not a secret. Uh, everyone uses some kind of, uh, Chinese semiconductors, but now we are trying to move on Ukrainian and we actually have a good result in it. We have actually, even, uh, many kinds of absolutely Ukrainian drones with, with AI in

41:08it or without. Uh, and we have a good drones from our partners. Actually, it's absolutely true. Uh, they're actually NATO countries are very, very good at, uh, software part. But, uh, uh, as you understand, the problem of, uh, auto missions is mostly software. So here's an edge part, but it's, it's much, much closer to have an absolutely autonomous weapon than any other.

41:41T struggling. All right. Sorry. Go ahead. Use the same as our technology. Winless. Hang in. For at the end, handsome monetarily as anything, we are actually most right here. Have a great monday and compliant gun. And we're good doing this last month. Right. 하는 kind of view. And we're at the end, it's going to prakt sand gyms and to really, let's go ahead though. To do this training, whatever training equipment is, that the, um,