Drone Technology in Ukraine - Automation, Lethality & The (Scary) Development Race

Drone Technology in Ukraine - Automation, Lethality & The (Scary) Development Race

Show Video

In early 2023 I covered the evolution of small drones in Ukraine when the technology level and production rates were already scary. But now just over a year later the leaps in quantity and capability have been immense. Production targets have ramped up, in some cases by more than an order of magnitude. And even as the quantities have ramped up, the technology involved has continued to get better.

And with no sign of that breakneck engineering and industrial evolution slowing down, I think it's time to ask some important questions. How are we seeing these small UAS evolve in Ukraine, what impact are they having on the war? And what might it mean for the rest of the world when this technology and the tactics developed for its use inevitably start to go global? To do that, as normal I'm going to break this up into a couple of segments. We'll start with an overview of some of the most common types of drones that are operating in Ukraine and how they may be getting better. Looking at how some of these drones are evolving in terms of their sensors, range, targeting, lethality or counter-measure resistance. Then we'll go to the second scary part of the picture, because drones are not just getting deadlier, they are getting more common.

So I'll cover some elements relating to production estimates and techniques, before closing with a zoomed out look at just how impactful these systems have been, and what their role might be going forward. Some quick caveats then in terms of scope and content. Firstly, while there are a huge array of drones out there, to keep things manageable I'm going to focus just on smaller Unmanned Aerial Systems [UAS] that are used in attack roles. So bomber drones, FPVs, loitering munitions and one-way attack systems. I've talked about dedicated ISR drones like the Russian Orlan before, so they won't be in this episode. Nor will naval or ground drones that very much deserve their own billing.

I also need to flag that drone technology and the tactics for using them in Ukraine are evolving incredibly quickly, and some of the very latest technical and tactical developments are usually fairly sensitive. So I'll generally try and focus on examples that are at least a month or two old, because I want to put together a high level picture here, not a how-to guide for example on how to build a more jam-resistant FPV drone. OK, so let's do some quick categorisation of the sort of drones we are going to be talking about today. Ranging from small ISR platforms, all the way through to long-range one-way attack drones. Your smallest ISR drones are usually going to be quadcopters, sometimes octocopters, think your off-the-shelf DJI Mavic 3 or equivalent for example. And while they're probably intended to be reusable, they are cheap, relatively short ranged, and a lot of times they are being used for things like increasing situational awareness, identifying attacks, calling in and correcting artillery.

And giving even relatively low echelon units the ability to metaphorically see what is on the other side of the next hill. Weaponise one of those quad or octocopters and you might have yourself what are sometimes called "drop drones" or "bomber drones". These keep the same fundamental characteristics as the base system, so intended re-usability for example, but they add a small payload to be used in attacks.

For example knocking out disabled vehicles by dropping grenades through hatches. There is of course a massive variation in scale and cost here, ranging from off-the-shelf hobby drones dropping hand grenades, all the way through to purpose-built monsters capable of carting around artillery projectiles, but the basic concept is broadly similar. FPVs, or First Person View drones, are also generally quadcopters, but with a different piloting method.

These originally developed out of racing drones, so compared to the more reconnaissance-focused quadcopters they tend to be much faster, require a lot more in terms of pilot skill, and are also often a lot cheaper. A lot of these systems tend to be intended as single-use attack options against tactical targets. So you are still operating in a relatively close range bracket, 10 kilometres for example is a pretty long range FPV strike, but you're probably doing it with a lower cost, a bigger boom, single use only. Although as always those points should come with a giant asterisk, because there are FPV designs out there that are designed to operate as reusable bombers. Loitering munitions in Ukraine, like the Russian Lancet, tend to represent a bit of a step change from the FPVs. The designs are usually fixed wing, not quad-rotors, so optimised for speed and range.

And are often going to be used by specialised teams at ranges of tens of kilometres to hit point valuable targets. Lancet and some of its Ukrainian equivalents for example are perhaps best known for their counter-battery role. Hunting enemy artillery systems that aren't exactly going to be parked in a frontline trench.

Finally, you have long-range strike or one-way attack drones. Here you are still mostly looking at fixed-wing designs that are intended to be single use only, but the range, payload and intended target type are all going to be different. We are talking ranges in the hundreds of kilometres, not tens of kilometres. Targets that are operational or strategic in nature, think fuel refineries or air bases for example. And usually a larger payload to reflect both the different target type and the fact you are probably not going to get pinpoint accuracy.

The Iranian-designed Shahed 136 probably fits into this category, along with a number of Ukrainian designs. And you can see them as representing a sort of bridge between the other cheaper drone systems, and traditional long-range munitions like cruise missiles. Compared to a cruise missile, the one-way attack drones we've seen so far tend to be much slower, have smaller warheads, and be missing a lot of the technological bells and whistles.

But they also tend to be available at a fraction of the cost while being much easier to produce. OK, so now we do get to the scary part where we start talking about how these systems, which are increasingly everywhere and very affordable, are also continuously evolving and have become more dangerous even just over the last year. But in order to give an overview of some of the factors that might be making drones in Ukraine more dangerous over time, what I thought I'd do is take a concept that we've looked at before, the survivability onion, and flip it around a little. The survivability onion usually just describes all the things that have to go wrong sequentially for you to get killed in your vehicle on the battlefield. You have to be somewhere the enemy is a) looking and b) can reach.

And then you have to be seen, identified, targeted and hit by something that is lethal enough to penetrate your defences and do significant damage. So in order to identify some changing trends in drone lethality, we're going to take that same concept and sort of flip it around. Instead of a survivability onion we have a much tastier (and more traditionally Australian) lethality Lamington. A layered list of the characteristics your system is going to need to maximise the probability of a kill. You need the sensors to see a target, the range to reach it, a guidance and targeting system to hit it, and enough of a payload to make sure they know that they've been hit. And of course you want your system to be as resistant as possible to whatever counter-measures or defences your opponent might be using to maximise the probability that you make it to the end of that process.

We've observed technical changes in Ukraine that impact each layer of the Lamington. So now what I'll do is go through some of them in sequence. OK, so step one for our hypothetical drone system is finding a target. And here the war in Ukraine has already witnessed some pretty significant technical improvements, particularly when it comes to night operations.

While we talk about flocks of drones being an omnipresent fact of life on the Ukrainian battlefield, the reality is there's actually a lot of conditions during which a lot of the systems out there just can't operate effectively. Bad weather can be even more of a problem for a lot of UAS than it is for manned aircraft. And while already in 2022 a lot of the larger and more expensive drones were equipped with things like thermal and night vision optics, for the vast majority of FPVs and quadrotors out there the sensor suite on board, which would often be just a simple visual camera, didn't operate effectively in low light or night-time conditions. Indeed, what you get with a lot of FPV drones is visual spectrum only, just like the Mark 1 eyeball, only with resolution levels reminiscent of the Xbox 360, and a level of grain and interference that wouldn't look out of place in an old analogue TV broadcast. The obvious tactical adaptation that that sort of limitation imposed is that in environments where the opponent has the ability to fly a lot of these drones, we have seen reports of Russian and Ukrainian forces doing for example a lot of their low-level logistics and resupply operations and manoeuvre at night. Now of course there were still small drones and ISR assets that could see you even if you were manoeuvring after sundown.

As well as the quite advanced sensor packages on things like the TB2 drone, smaller tactical level UAS also had some night vision capability. Off-the-shelf DJI drones for example have some thermal options, the 30T and the 3T being some of the most common. And that was fine if your goal was to see enemy units manoeuvring at night and do something like call in artillery on them. Because unguided artillery shells don't tend to care about the ambient light level when they are addressed to a particular set of grid coordinates. But that arrangement brings with it two limitations.

One, if there is a target type that you would rather engage with something like an FPV as opposed to artillery, perhaps for example because you are short of artillery ammunition, that might be difficult to do when the FPVs can't fly at night, and so you either inefficiently task artillery to deal with the problem, or you let the target go. The other limitation was that thermal-camera equipped drones were generally just much more expensive than those that didn't have that level of bling. Which means both a) probably not everyone gets to have one. And secondly, if you want to make cheap FPVs thermal capable you probably can't do it by adding those sort of thermal optics otherwise your cheap FPV isn't going to be so cheap anymore. Make no mistake, it'll probably still be cheap compared to many of the other weapon systems on the battlefield, but is it really the quintessential Ukraine War experience if you are not touching off multi-million dollar armoured vehicles with something that cost a couple of hundred bucks? And yes, I am joking around a lot, but it's mostly to distract from the fact the next bit is kind of terrifying. As we go into 2024, I think there are plenty of signs that more and more drone systems, including relatively cheap FPVs, are going to be capable of operating at times that were previously at least if not safe, maybe safer.

We've seen more and more FPVs, loitering munitions and drop drones that integrate thermal cameras into their designs. And you also see some innovative approaches like the one mentioned on the right there. That's a post by the Ukrainian organisation Wild Hornets that manufacture FPV drones, showing off what they describe as "An affordable solution for night-time FPV drone operations." And when they say that, they don't mean it in the corporate buzzword sense, they actually mean affordable as in 50 dollars US per unit.

At $1,000 for a thermal camera, you might be tripling the cost of your average FPV. At $50 per drone, it's only a 10% premium for giving something the ability to operate at night. And at that point wide-scale adoption might start to make a greater degree of sense. So at the big picture level, what's changed and what matters? Basically if in 2022 thermal or night vision was an expensive, rare capability for small drones, in 2024 the direction of travel has been towards cheaper and more common. All else being equal, you should expect that means that in the future there will be more drones out there capable of identifying more targets during more hours of the day.

The protection that night-time provided (which was already highly imperfect) is likely to be reduced, and the military-industrial complex as a whole will take one more gigantic step towards what I can only assume is the ultimate goal of equipping every drone with a mechanical equivalent of the Eye of Sauron for less than the cost of a Happy Meal. OK, so increasingly drones can solve problem number 1 of needing to be able to see a target, what about actually being able to reach it? After all, just because I can see the top of a mountain doesn't mean that I can climb to it. The war in Ukraine has arguably really highlighted the value of weapons with longer reach. Exploiting the limitations of your opposing systems reach has also been one way commonly used to protect important targets. When the Russians pulled a lot of their ammunition depots out of HIMARS range, that was a range-based adaptation.

And while yes, there are some limitations on the ability of forces to just move everything of note out of opposing range (good luck shifting an oil refinery for example), particularly in the earlier stage of the invasion, a lot of the cheaper drone systems out there had comparatively limited endurance and practical range. Now while there is always a lot of uncertainty in this analysis, here I just want to flag there's more uncertainty than normal. But if you've gone through some of the relevant publicly released interviews and reviewed a lot of the drone footage out there that has been geolocated from 2022 and 2023, one thing you might notice is that as time has gone on the number of attacks being launched by small UAS, like FPVs, at longer and longer ranges has increased. And we have seen that trend towards increased practical range play out across a range of systems with various effects. For an FPV drone it might mean being able to hit relatively short-range artillery systems like the TOS-1A.

I think we've actually seen videos of two FPV strikes on those systems in the last week at time of recording. But also they are used in more of an interdiction role, where you are attacking supply trucks or reinforcements trying to move up to the front line. Meanwhile for a system like Lancet that started with a double digit range, enhancing it considerably might mean putting longer-range artillery systems, or things like forward air strips that were out of range in range. With unfortunate results for example for the occasional Ukrainian MiG-29 airframe that's been caught out of position. In a sense, when you see the range of these systems improving, often there are two different elements that are moving together. First are the improvements to the physical performance of the system, how far can it physically fly? In that sense the move from off-the-shelf drones to more dedicated military models has a lot of potential.

Compared to a civilian version, a military FPV might have a lot of shit you don't strictly need stripped off. It doesn't need to last particularly long or look particularly good, so you can make some economies there. And to an extent you can also always just add more fuel tankage or battery. Problem two then is trying to find a way to control said drone when it does fly further. Here again from a mechanical perspective, there are a couple of related but still distinct problems.

Firstly, you just need an underlying signal strength and quality that lets you reach the requisite distance. Especially in a battlefield context where both sides are likely to have EW guys roaming around the place doing their best to do selective violence to different parts of the electromagnetic spectrum. The second problem is just maintaining signal line of sight.

With all due respect to the flat earthers out there, the earth isn't. So all else being equal, the further a drone flies at low altitude the more likely it is that a bunch of dirt, rock and burnt out tanks get between it and the transmitter. And while you can compensate to an extent by increasing altitude to maintain line of sight longer, that also leads into a phenomena we've seen in lots of videos of FPV and other drone attacks where during the final terminal phase of an attack the drone dives towards the ground (because that's where the target is) and you can physically see the signs of the signal breaking up in real time. To an extent, a good FPV operator knows that disruption is likely to come and can pre-aim the drone so it's going to make the hit regardless. But it's still a factor relating to signal range and quality that can significantly impact just how effective these weapons practically are. OK, so how do you then increase the range of the signal? Well, one approach we've seen used is the employment of repeaters.

Where you might have one or more drones that instead of carrying munitions or a sensor package is carrying a signal repeater. These then can form an aerial relay. Noting there's no reason you have to put them on the drone, you can have ground-based repeaters as well. But putting them on the drones has an obvious advantage from a line of sight perspective. Now, instead of having to send a signal to the drone that's making the attack, you just have to reach the repeater, which then has to reach the next repeater, which can then reach the drone that is making the attack.

If you think of this from a line of sight perspective, a repeater that is relatively close to the site of the attack is less likely, all else being equal, to lose sight of an FPV making an attack in its terminal phase. And so an operator might be more able to make those last-minute flight path adjustments that lead to a hit rather than a swing and a miss. Repeaters also have potential implications for the survivability of the drone operators themselves. It means you might not need the drone operator metaphorically kneeling in a front-line trench in order to get the maximum possible reach out of their drones. Instead you can send the repeaters forward while the operators are sheltered in a position further back.

Depending on the particular set up and opposing capabilities, you might still have a signature problem from a concealment perspective. But you may still end up being more survivable as a drone operator using this sort of system. Another thing to note here is that the repeater drone and the drones it is controlling don't have to be the same design, in fact that might be a disadvantage. I feel like I can talk about this case because it has a casual 732,000 views on Twitter, but recently there was a case of a Russian serviceman describing a situation where Ukrainian drone operators allegedly attacked a position using a larger so-called "Queen" drone with a signal repeater controlling a bunch of smaller, cheaper FPVs. The possibilities of that sort of configuration are obviously massive, because there's a lot of signals equipment you could fit on a larger drone that you can't (and probably don't want to) try and fit onto an FPV.

A large enough quad or octocopter for example, might be able to carry a satellite internet terminal worth thousands of dollars, which would not exactly work from any perspective if you tried it with an FPV. Plus of course in this environment a lot of the most expensive equipment is concentrated on a reusable system, the Queen drone, rather than the disposable systems, the FPVs that are going and slamming into things. If the Queen isn't shot down it can be recalled and then sent out again with a new wing of supporting attack drones. It's worth noting that this idea of a Queen system providing information and controlling other systems in a flock is not at all new.

You could argue that the pairing of ISR drones and loitering munitions already to an extent fits that definition, with the ISR drones generally carrying the more expensive sensor and communication package and the loitering munition carrying the warhead and a Mad Maxian desire to be "witnessed". Back during the Cold War for example, the Soviets introduced the P-700 with the highly creative NATO reporting name "Shipwreck", which had an attack mode where a flight of missiles directed towards an enemy fleet would have a single missile act as the target designator. That Queen missile would climb to a higher altitude so it could better see the potential targets, and communicate via data link with the others that were hiding closer to the deck.

If the Queen, perhaps owing to its higher altitude and more exposed position, were shot down a drone would be promoted to Queen duty, climb to altitude, and resume where the previous missile left off. Here there was no physical difference between the different P-700s in the salvo. But you still have a split between attacking missiles and a coordinating or target designating missile in order to help solve the underlying line of sight issues. So we probably shouldn't be that surprised more than 2 years into the invasion of Ukraine that this old concept is being leveraged to deadly effect. So in summary (and as always this is oversimplifying) if in the early stages of the war the norm was for small drones to attack front-line targets and face significant line of sight issues, in 2024 the meta (if you will) increasingly involves long-range interdicting strikes supported by technologies that mean that range and line of sight are no longer always as much of an issue.

The reason that matters is because it establishes more and more tactical and operational use cases for these systems. At 2 or 3 kilometres, your anti-armour FPVs can supplement the role of something like a Javelin missile. But if you start reaching out to 10 kilometres, you're playing in territory that might normally be serviced by indirect fire options, artillery. And once you have loitering munitions that might be able to go out 70 kilometres, you are playing in territory that was previously dominated by long-range precision rocket artillery and air support. That's obviously a factor on the battlefield, because you are putting more and more of the tactical and operational depths at risk from systems that might be able to economically engage targets that are not worth dumping an entire battery of MRLS on.

It may also raise questions in some armies as to who should control these longer-range fire systems? Because suddenly you might have a cheap system whose primary target is tanks and armoured vehicles, so you think it belongs wherever the ATGMs go, but it has a range bracket that's much, much longer than your traditional platoon-level asset. You have a family of weapon systems here with the potential to, for lack of a better word, democratise access to long range precision. And it may not be fully understood just yet what all the implications of that are going to be. But moving on, what all those changes have basically done at this point has increased the odds that our drones will find a target and then be able to get to it. The next step, and a pretty important one at that, is actually hitting it.

Even though you can argue that drop drones, loitering munitions, FPVs are all precision weapon systems. They all after all have guidance systems and tend to go after point targets, even if the guidance system in question is often just the brain of the pilot at the controls. It's important to understand that in practice the evidence suggests that these things do miss a lot. And it's an area where I think it's pretty logical to assume that the video evidence available to us is probably a bit skewed.

Sure, we do get videos of drones and loitering munitions appearing to miss their targets fairly often. Just like people on social media who might love to post images of the fast car, the fancy vacation or gourmet dinners while conveniently leaving out images of their credit card statements, in drone warfare it's the operator who holds the footage and makes the decision on releasing it. And so you probably get to see the $500 drone slamming into the $20 million air defence system or totalling a T-90M. And what you don't see are the drones that miss, or are jammed out, or miss, or suffer technical failure, or miss. Remembering of course that a lot of these systems are hard to fly, built by the lowest bidder, and often provide the operator with a solid potato-tier video feed.

One trend we've seen which might have an impact on that particular problem is the increased utilisation of machine assistance for target identification and also for guidance. These sort of visual targeting solutions for example might include a system that looks at a video feed and flags things that might be a potential target and/or once the target has been selected and designated by a human operator allow the weapon to guide in without further human assistance. Russian sources for example often claim that the latest versions of the Lancet loitering munition are capable of autonomous or semi-autonomous guidance. Basically the system can look at its surroundings, identify things that look like a tank, an infantry fighting vehicle or a piece of farm machinery, say, "Hey, I think that thing over there might be a target." And then, depending on the exact autonomy settings (the information available is very vague at the moment) either with or without a thumbs up from a human, go in and engage one of those targets.

The exact answer to the question of how autonomous systems like Lancet can be is still very much contested. You have claims that run the full gamut from the system is just highlighting potential targets to give the operator a clue as to what he might want to fly into, all the way up to this is an AI-powered super drone that can recognise a Ukrainian tank that the human eye would never spot and then autonomously guide to it. But while there's a lot of debate about how autonomous exactly systems like Lancet are, we do know we've seen a significant amount of footage from Russian sources of Lancet attacks being launched with some sort of mechanical assistance appearing to be in play. At the moment at least the evidence seems to suggest that the systems that are out there, including the one on Lancet, isn't exactly perfect. Think more 10% Skynet, 50% chatGPT, and 30% that New Zealand supermarket AI that suggested people put mosquito repellent on their potatoes or turpentine in their French toast.

OK, so to make this just a little bit more real, what I want to do is now dissect some stills taken from the footage of a recently documented Russian Lancet attack. This is footage released by a pro-Russian source that appears to show at least some autonomous targeting assistance being used. And I think it demonstrates both the strengths and also potential weaknesses of systems like this, at least right now.

We'll start with this image I have on screen there from relatively early in the video, and while those of you on mobile phone screens might be struggling a little bit, I think most humans would be able to parse this image pretty successfully. That large object at the end of the tread marks that looks awfully like an armoured vehicle is in fact an armoured vehicle. I don't know about you, but for me it was the turret and main gun system that kind of gave it away.

All of the stuff immediately surrounding the vehicle, including the bit I've highlighted for reasons which will become obvious in a moment, lack those same obvious visually identifying features and are probably not the primary target for this sort of system. Next we have a still, seemingly taken from the sequence where the loitering munition is approaching the potential target. At this point you see a green bounding box appear on the right hand side of the screen, which suggests that maybe the automated targeting system is identifying another potential target off to the right.

But the munition, still potentially under operator control, continues towards that original target we were just looking at a top down image of before. And indeed you can see that central crosshair over that target at this point. And as the munition approaches that target, the expectation is probably that the automated system will kick in, identify the target, and assist with the final approach. So of course the munition races towards the target and then proceeds to not identify it. Instead, you can see it throws a bounding box around what appears to be an object that the potential target is hiding behind. We can go into theories about why that misidentification may have been made, but the fact is it happens, with negative consequences for the engagement.

Basically the machine was fed a real life version of one of those visual CAPTCHA problems and it appears to have proceeded to fail in spectacular fashion. Other footage I've looked at has shown cases of vehicles being correctly identified as potential targets, but also bounding boxes being thrown around things that are very obviously not military equipment. And other cases where military equipment that is fairly obvious doesn't get the automated target recognition treatment. And some observers have noted that after a rush of videos that appeared to show Lancets being used with this sort of autonomous or semi-autonomous targeting system, it's been comparatively rarer or absent in more recent releases. Potentially suggesting (although this is very much straight guesswork) that potentially the feature got some battlefield beta testing, attracted a little bit of customer feedback so to speak, and may now be in a revision and iteration phase.

But here's the thing, there is every reason to believe that better versions of this technology are coming, and coming quickly. We'll probably talk about this more in the future, but when you are trying to make an AI-driven visual targeting solution work, there's a couple of components that go into it. And some of those components are the literal components, the things that give it the computing power it needs to make the solution work. That compute is already lighter and cheaper than it has been at any other point in human history, and it's only going to get lighter and cheaper. There was some reporting I saw that suggested that some Lancets might be using NVIDEA Jetson TX2 modules to provide some of that compute.

I did ask some people who keep a downed Lancet as a sort of office pet whether or not it contains such a module, but the version they have appears to be a pre-autonomy version, so I can't confirm one way or the other. But as a hypothetical exercise, that sort of embedded AI computing device only goes for a couple of hundred US dollars. And the hardware that's available and the price point it's available at is improving quickly.

Another main input to getting good outcomes out of AI is to have a lot of good data to train it on, and the opportunity to iterate and develop. And so whether the rest of the world likes it or not, the war in Ukraine, where drones are being used on a massive scale in combat conditions, might become a sort of nursery/training ground for the targeting AIs of the future. It's also concentrating a lot of resources, human and financial capital, potentially creating that sort of precinct of capability, that sort of rapid development environment, where you have the full array of engineers, specialists and end users concentrated, communicating and capable of rapid development and iteration.

So in broad summary strokes, if in 2022 and early 2023 drone targeting in Ukraine was mostly a matter of pilot skill, experience, intuition and occasionally just hoping that you didn't lose signal at an inopportune moment, in 2024 more of the options we are seeing deployed include at least some degree of computer assistance. Whether that just be in the area of target identification and recognition, or in the field of semi- or fully-autonomous terminal guidance. And as to why developments like that might matter, I think we're only now starting to unpack the potential implications.

Camouflage and concealment may become even more difficult if you need your scheme to fool both the human operator and the automatic recognition system. The need to have a highly skilled loitering munition or FPV operator might be somewhat reduced, which matters a lot given just how difficult training can be and the premium that is currently being placed on skilled operators. Plus, and perhaps most frighteningly, it might offer a partial answer to jamming. Because even if electronic warfare can successfully disrupt communication between an operator and their drone, that's not going to stop the attack if the drone is capable of autonomously identifying its own targets and guiding into them. OK, so at this point the drones have been able to find a target, reach the target, and then with a little bit of mechanical assistance hit the target, let's now talk about the element that actually makes that hit mean something, the payload the system is carrying and just how lethal it is.

After all, without an effective payload, slamming a drone into something like a tank is probably going to be about as effective as throwing a bike at a battleship. And here's where lethality ties into the purpose that drones can actually fill. With just about any weapon system, there are going to be certain targets it's designed to service and others that it doesn't really provide an effective answer for. A machine gun isn't an anti-tank weapon, and a Javelin missile is going to struggle against infantry in the open.

So an evolution in drone payloads isn't just about making them more effective against certain sorts of targets, it's about broadening the sort of missions they can do and targets they can service. At 1 kilo of payload you are probably talking about infantry or vehicles with the hatches left open. 2.5 kilos plus and you start to have more anti-armour options available to you. And once you start getting into serious double-digit figures, your drone might be able to do its bit for the global emissions reduction movement by pro-actively imposing an output restriction or two on opposing oil refineries. In 2022, while there were some units operating specialised heavier drop drones even at that time, most of the drop drones we saw operating and the first generations of FPVs that started to come through were mostly using very light payloads. There were a lot of systems I saw at that time that were often using effectors in the 1 kilo or under range.

Perhaps the quintessential examples here were either literal hand grenades, or adaptations of the famous VOG. The VOG, often specifically the VOG-17 version, is a lightweight 30mm high explosive projectile primarily intended for the AGS series of automatic grenade launchers. A wooden case of these things from some manufacturers would contain about 108 grenades and weigh 55 kilos, with a lot of that being packaging. The actual rounds themselves from memory are about 350 grams, so about 0.75 of a pound, and particularly when you are only dropping them in singles, you are only really going to do damage to light targets, and even then only with a direct or very close hit. But everyone uses them because with a 3D printed tail kit they could be dropped relatively accurately, and because the Soviet Union was kind enough to leave an absolute shit ton of these things behind.

American 40mm grenades would be another more lethal option, but were only available in much smaller numbers. In 2022 through 2023, the path of drone evolution created a number of FPV designs that were capable of carrying heavier payloads. Some of the most common payloads here, for obvious reasons, were RPG-7 warheads, things like the PG-7VL . So now you have a munition that is, yes, heavier and requires a chunkier drone to carry, but is a) designed to act as an anti-armour weapon, meaning you probably can graduate from killing golf carts to BMPs. And of course, like the VOG 30mm, the Soviet Union built a shit ton of these things and there's plenty of them available. That availability was important not just in a quantity sense, it was also important in terms of making sure that drone operators could actually get access to the munitions.

This was stuff that might already be being pushed down the logistics chain anyway, there are a lot of units with RPG-7s out there, and so you just have to nicely ask the supply officer to let the drone units borrow a few boxes. A heavier option again that we saw for the chunkiest of FPVs and also some loitering munitions were tandem charges. This would be something a bit heavier again like the PG-7VR, which basically gives you two charges, and one of the main useful benefits of that is making the warhead much more effective against explosive reactive armour. Which obviously has utility in Ukraine where 90% of every vehicle upgrade package seems to consist of covering an object in as many ERA blocks as will fit. So that's already a scary enough evolution, but obviously designers don't stand still. And so I'd argue we've seen at least three parallel tracks of continued development when it comes to increasing the lethality of combat drones in Ukraine.

The first is just continued experimentation with larger and larger payloads. Because one of the simplest expedients for increasing the damage something does is either making it make a bigger boom, or having it make more booms. So one of the responses might be to come up with somewhat chunkier drones that can handle the additional payload. And some designs capable of doing so already existed before the full scale invasion. That image you see there on the right for example is a shot that was put up by the Russians of a captured Ukrainian drone allegedly.

There it looks like the payload is actually three PTABs, which is a roughly 2 kilogram anti-tank bomb originally designed during the Second World War. That implies a payload of at least 6 kilograms, and we're pretty confident there are designs out there that can do considerably more. And this movement towards higher payload capacity isn't just something we've seen with drop drones or FPVs, we've seen it in other systems categories as well.

The Russians have increased the warhead size on the average Lancet that you see used in Ukraine by several times over since 2022. And if you look at some of the long-range strike drones Ukraine is using, again, heavier warheads have become increasingly available. And recently it's been reported the Russians may even be following this same evolutionary path with some of their cruise missiles. When the Ukrainians recently broke open a crashed Russian X-101 cruise missile, NATO reporting name AS-15, they found that it looked like the Russians had pulled out some of the fuel tankage to replace it with a second warhead, adding roughly another 400 kilos of payload to the thing while sacrificing range that realistically the system didn't need to hit targets in Ukraine.

The logic in all of these cases is at least somewhat similar. If you are going to invest a lot of money and resources into either smashing an object into something or using it to bomb something, you want it to be effective when it does so, and all else being equal, bigger boom often equals bigger effect. Of course for any given target type there is an optimum amount of boom and it is possible to go too far. Recently Russian TV showed Sergei Shoigu touring an arms plant. During which one of the weapons showed off was the FAB-3000 3 ton aerial bomb. This fragile masculinity special is obviously a very dangerous weapon, but there are serious questions over whether or not it would actually be practical or efficient.

Its range as a glide bomb is questionable, the ability to mount this thing under anything less than a dedicated bomber aircraft likewise questionable. And it's likely to run face first into that basic physics problem wherein doubling the amount of explosive yield does not in fact double the blast radius. So the trend towards increasing payloads probably has a limit. And the dedication of resources to extremely large systems like FAB-3000 is probably questionable. Although to be fair, probably not historically surprising given that after America spent a portion of its very large military budget to develop and field the 9.8 ton GBU-43 MOAB, sometimes referred to as the "Mother of all Bombs",

Russia dedicated a portion of its much smaller military budget to building an even larger one which it dubbed the "Father of all Bombs". A totally efficient and not at all wasteful exercise. The second big trend is not so much to increase the size of the payloads, but rather to change the nature of the payload. This potentially offers the option to take a drone and optimise it better for certain target types without increasing its size or cost. If you swap out the anti-tank RPG warhead on an FPV with a thermobaric payload for example, that might be much more effective against targets in buildings or enclosed spaces without any increase in weight. Air bursting and enhanced fragmentation projectiles might be more effective against infantry targets in the open.

And of course there are a variety of specialised anti-armour munitions available. There's always been a degree of this sort of munition to target matching going on, and there are a bunch of off-the-shelf munitions that you can potentially bolt to a drone to get alternative effects. But as the war has gone on we've seen more focus placed on dedicated munitions for drone operations. This is the process where we see more and more munitions that have either been substantially modified for drone use or designed and built from scratch for drone use. Although in some cases even those "scratch built" munitions will include some recycled or re-purposed components.

In 2023 Ukraine announced that they were creating a new category of ammunition for drone operations. Which sort of puts the organisational scaffolding in place for getting better at designing and procuring drone munitions and allowing drone units to order and be supplied with them. Military supply systems are often like engaging with customer support, which consists of nothing other than the automated options. You can press 1 for VOGs, 2 for RPG 7s, 3 for 152mm shells, or press 4 to hear options 1 through 3 again. Now it might be more possible to streamline that process from an organisational perspective, but we'll see how things actually play out in 2024.

The move over to purpose-built or more significantly modified payloads has a number of potential benefits for drone operations. And it must be said, potentially a lot of drawbacks for those being targeted by them. Purpose-built munitions might be more stable and predictable. If they are being used for a drop drone for example, they might improve safety and handling so you don't have troops having to cut open cluster munitions in order to ad-hoc munitions for their drones.

It's worth noting that converting regular munitions into drone-carry munitions can be a dangerous and difficult process, sometimes people are injured or killed. Sometimes munitions fail on impact, and so the extra safety and reliability you might get with a purpose-built option really does mean something. And for a lot of systems you might just see increased yields for no increase in weight. To illustrate that, imagine an early ad-hoc strike drone concept where the way you gave the thing its boom was by integrating an artillery shell. You have lots of shells so you just create an empty space, plop in the shell, fuse it appropriately and off you go.

The issue however, is that most of the weight of an artillery projectile isn't the explosive charge. A lot of the mass in your average 152 or 155mm projectile is just metal, the structure of the projectile that is designed to survive the very violent experience of being fired out of a metal tube at very high velocity. Which as a propulsion method for a warhead is much, much more traumatic than being gently carried aloft and then to target by a drone. So a purpose-built warhead for a long-range strike drone might have a lot less casing and a lot more actual charge. There also just might be a push because of the sheer scale of the number of munitions being demanded. The Soviet Union and other producers did leave behind a truly mind-boggling amount of munitions.

But as hard as burning through the old Soviet stocks is, in Ukraine they are certainly giving it a red-hot go. So you have producers and organisations springing up to try and do dedicated drone munition production at larger scale. Recently for example there was an article covering the Steel Hornets, which is an organisation in Ukraine which is described as functioning like a sort of Amazon for drone munitions. Units can reportedly place orders for all sorts of specialised drone payloads, including some very nasty looking anti-personnel options. And those are then, after some obvious security checks, boxed up and shipped out.

So to summarise: if the old archetype for drone munitions was a small existing general purpose munition, the new overly-generalised archetype would be payloads that are larger, more specialised, and purpose-built. And as for why that matters, hopefully it's obvious for both Ukraine and the wider world. In 2022, 2023 and early 2024 we have already seen drones be terrifyingly effective. Drones and loitering munitions account for a significant proportion of visually confirmed artillery losses, and have inflicted a lot of visually confirmed damage against a lot of target types. But a lot of that observational baseline we are getting was probably established using munitions that were not optimum for purpose. I think it's logical to assume that Russian Lancet teams would have inflicted greater losses on Ukraine for example, if they had started in February 2022 universally equipped with the larger warhead version.

You can filter through huge volumes of Russian and Ukrainian video of drone attacks where with a different payload the result might have been different. And I think it's also safe to assume that when other major militaries start equipping with these sorts of systems in earnest, Uncle Sam is not going to be buying a version where someone has duct taped a $15 grenade to the bottom of a $500 drone. No, your future hypothetical Drone Optimised Payload Effector - Army Future, or DOPE-AF, might add a zero or two to the price of your average Ukrainian option, but we probably have to model in the fact that they might be much more lethal for it.

As drones have evolved they've flown further, found their targets more effectively, and done more damage when they get there. Which is all of course horrifically terrifying, and logically brings us to the question of how to stop them. And in Ukraine, just as the UAS have continued to evolve, so too are the counter-measures being relied on to defeat them.

During the years of fighting in the Donbas pre-full scale invasion and during the early stages of the full-scale invasion, you saw a number of reports of drone operators claiming to suffer because of counter-measure options that were provided by the drone manufacturers themselves. You can think of this essentially as a software-based counter-measure system, where a manufacturer has decided that they don't want their drones to be used for warfare or other such purposes, might for example offer a system like the old DJI Aeroscope. Which was marketed to law enforcement and enabled an operator to see both nearby DJI drones and also the location of their operators. The Ukrainian Vice-Prime Minister would then complain about the Russians using the Aeroscope system to hunt down Ukrainian operators. DJI would then later discontinue Aeroscope.

But that's certainly not the end for manufacturer built-in counter-measures. In the US for example, the FAA has steadily moved towards a world where the manufacturers of many drones are required to build in a remote identification broadcast system. Relying on the software and hardware of the drone itself to provide one of the counter-measures against potential misuse.

However in 2024 you won't really hear anything about any of these systems or options in Ukraine. Based on the evidence we have, they don't really seem to be a problem for drone operations now. Although for hopefully obvious reasons, I won't go into any detail about how that's been done. You could probably argue that this sort of evolution provides a little bit of a warning to governments and security agencies that are relying on this sort of approach to answer the potential security threats drones might pose domestically. And it helps highlight the difficulty the manufacturer of something like a drone might face controlling their product after it's sold, as opposed to a service like Starlink.

With a system like Starlink the customer buys a terminal. But the terminal can't magically provide internet, it can only do that by communicating with satellites, and the satellites are still controlled by Starlink. And what that means is that if Elon Musk and SpaceX decide to shut down Starlink service in a particular area, you probably can't get around that problem by messing with your terminal, you'd have to mess with the satellites. And those are significantly harder to physically reach and tamper with. A drone by contrast, is going to be intrinsically mechanically functional.

Even if you write a line in its code saying it shouldn't be operable in particular areas, say around sensitive military and civilian sites, the drone is physically capable of getting there, it's just a line of software telling it not to. And if hypothetically for some reason that block of code was ever to change or disappear, then the system would likely be physically capable of doing a couple of loop-de-loops over places it really shouldn't be. That's a concern with potentially global implications, but for now let's focus on the counter-measure picture in Ukraine. In the absence of manufacturer-provided options, the most common answer to drones in Ukraine has been electronic warfare.

You could say that over the last 2 years the frequency of electronic warfare in Ukraine has constantly changed, but I'm not sure we could handle a dad joke that bad. Jammers have proliferated to the point where we've seen them strung along supply lines, mounted on armoured vehicles, or even kept in trench systems. At the extreme end, recently we even got images of this monster here.

Where reportedly what the Russians did was put a pallet on top of a tank, and strapped to that pallet a variety of jammers and electronic warfare equipment, batteries and a diesel generator. While the reporting is purely anecdotal, it's suggested the thing did manage to bring down more than a few FPVs. But in the end, as the image suggests, the boom box jammer approach ultimately wasn't enough to save the vehicle. And while the balance, especially at the local level, can swing continuously between the jammer and the jamee, we have seen a variety of different approaches used by drone and loitering munition manufacturers to try and get around at least some of the EW problem.

One very strange solution we've seen is actually a very low-tech one. This was a Russian drone which was reportedly captured which seems to have followed the basic line of thought that if EW systems inhibit your wireless signals how about you just don't use wireless signals? Instead this drone had a reel of fibre optic cable connected to it and seems to have been intended to just literally spool out a line as it flew in order to maintain a connection to the operator. Well that may sound promising, because no one's found a way to jam a wire yet short of, you know, cutting it. And wire-guided weapons very much are a reality with some other systems like ATGMs and torpedoes.

There are some pretty obvious drawbacks to try to operate FPVs this way. We've only seen one example so far, and it failed. So we may not see other producers race to emulate a system that shares characteristics with a kite and has a 0% success rate. Instead what we've seen more of is a constant shift in the frequency that drone operators use, coupled with methods of guidance and targeting that might be more jam resistant. Frequency selection, coordination and jamming can be major factors in drone operations.

If you jam a frequency your own guys are using, you are going to bring down your own birds. Which reportedly has often been a major problem in Ukraine, but is also one that I think other militaries, if given the time and resources, might find some answers to. You can also have cases where systems might be redesigned to operate on frequencies that are less frequently jammed. Ukrainian media for example claims that the country has produced a small counter-drone system that can operate in the 850 to 940 MHz range.

Which is reportedly a pretty standard range for a lot of FPVs, but that same article points out that there are Russian kamikaze drones that can operate on either lower or higher frequencies. Those aren't really going to care about this sort of system and only going to be vulnerable to those that can hit the wider range. Meanwhile, if you want to look at longer-range one-way attack systems that don't require constant user control but instead use GPS, or a Global Navigation System of some kind, we've still seen a number of approaches being used to protect those systems from just being neutralised by jamming.

The Iranian Shahed 136 for example, presumably to decrease its vulnerability to things like GPS spoofing, doesn't have one GNS receiver, it has multiples. And so if two receivers say that it's approaching its target while one is convinced it's 100 miles off course, the majority is going to shout down the problem child and try and continue to target. If GNS is denied entirely, most long-range attack systems include inertial navigation backup.

Which is usually going to be much less accurate than GNS, but will function even if the system is being jammed. We are also seeing some reporting now that things like long-range one-way attack drones might be using other classical methods of navigation. These include terrain recognition-based approaches where you are not reliant on receiving a navigation signal, instead you are using your on-board sensors to look at the terrain that you're flying over, comparing that to the database and mission plan that's been pre-loaded into the system, and using that to find out where you currently are.

Tomahawk missiles from the Cold War era had this sort of technology. And obviously even then you didn't need a GPS connection to recognise a river, a crossroad, or a mountain range. While there are a lot of old approaches to navigating in a spectrum-denied environment however, the thing we're seeing change now is the degree to which new technology makes these approaches cheaper and more accessible. The availability of commercial satellite imagery for example, makes it much more practical to get a picture of the terrain all the way between your launch site and the target. And memory, computing power and good cameras are all much cheaper, more miniaturised and more accessible than they were in the 1980s.

This and other alternative navigation options obviously aren't perfect, but they do point to the risk of systems like one-way attack drones becoming more and more resistance to attempts to counter them using electronic warfare. That of course does still leave the hard kill option of just shooting down the incoming drones, but then you might run into issues like the difficulty of getting sufficient geographical coverage, or shot exchange problems if you do it using longer-range missiles. So in terms of the big picture for the current war in Ukraine, trying to assess where the status quo is in the race between counter-measures and counter-measure resistance is harder than in some of the other areas we've looked at. We know that approaches and technologies used are constantly shifting, and we have some examples of new jammers succeeding or failing. But getting good, at least semi-complete, information on success rates is very difficult, and it would be quickly outdated even if it was available. But one trend that does stand out because it might have wider implications, is the reportedly greater application, even in the case of relatively cheap one-way attack drones, of navigation methods that might not have an easy electronic warfare-based answer.

A defender isn't going to be able to move mountains, crossroads and river systems just to fool a navigation system. And so the trend in 2024 and beyond might be towards more one-way attack drones with greater resistance to existing EW counter-measures. And greater pressures for countries to reconsider how their air defence systems might be configured to deal with exhaustion or saturation-based threats from affordable long-range attack options. And touching on the possibility of saturation or exhaustion-based tactics brings us to another section, a key element to what is making drones so dangerous in Ukraine, which has nothing to do with the capability of the drones themselves.

And while this may be my bias talking, one of those key factors is production. Ultimately sketches, CAD files and project plans don't destroy tanks, physical weapons do. And so if you have a system or technology that's well suited for quick

2024-04-12 00:29

Show Video

Other news