The state of drones, 10th July 2024
Aerial warfare in Ukraine busily evolving.
As you can you can see from these two interviews, the participants expect more changes happening soon, and the feeling that it could be a quantum leap in a way:
However what could be the nature of this quantum leap, and how it could change the landscape of the war?
First of all let us extract information on critical pieces of technologies to enable such change.
In this article we can see three critical components:
Mini computer required for implementation on small drones. This called also SOM in modern slang — System on Module, in other words all in one solution for various purposes. And we have to be aware that any AI model requires significant computational resources to function. More on that later.
Machine vision — this technology need for terminal guidance. If you have seen videos of those oil plants hits, then you noted how drone lands on a very specific target, and very precisely, not just somewhere. It is not possible with GPS, it could be either operator driving the drone or? And at 1000 km distance it is surely not an operator. So its machine vision with a preprogrammed feature to target.
Terrain mapping — the drone compares the photos below it with hi-res maps stored in its memory and thus follows the pre-designed route.
Homing on Signal — basically lots of warfare equipment are sources of radio signal, each one is having its own specifics and signature so to speak. As well as visual signals, its is possible to train the model to recognise those and “fly towards them”, where optical sensor would take control for final targeting.
In this short video it is shown and then explained how simple set up allows to find the signal and triangulate the source of it, by use of simple manual technologies. Here we can identify few important things as well:
Air burst technology — this was practiced for quite some time where bomb assigned to FPV would bust not by contact, but by manual action of the operator akin to pulling the trigger. With having good charge of shrapnel it allows to kill enemy drone quite easily.
Every drone is a source of signal — and you have to understand that such simple technologies are not a kind of radar. If drone switches off transmission it would not be possible track it with those means, unless it is in range of optical sensors.
However note, that all kinds of radars are themselves huge source of signal “seen” from far away.
Another important technology is this video:
Speed — in order to reach aerial target a speed is required, since it is very easy to miss the target moving fast, and then not be able to catch up with it.
Of course ruscist are working on their own technology, though with some significant difference, namely radar and laser tagging. Which I don’t think is a way to go, though the device shape overall, launching approach, etc. is good.
AI implications and processing power
First of all a bit a about AI or more precisely deep learning models. First of all such models, whether it is about computer vision, radio signature detection or terrain mapping are quite large. But memory is not that much of constraint, the main constraint is processing power, the larger model the more processing power it requires.
There is generally two ways to deal with that, one is to make processing local, autonomous, on device itself. The other is to make the processing central, like for example all the video footage from reconnaissance drones is transmitted to central station. There are people in some location watching those videos and taking decisions. Making processing in central location is much more simple.
However, any transmission is a signal, and thus such device is a source of signal, and this means homing is easy on such device. This will be a factor, not at once, but more and more.
So developing a local, autonomous processing unit is inevitable. This allows at least part of the flight time not transmit any signal at all. The technical solution for this is called SOM, system on module. It is a kind of Adruino mini computer or similar system, but on steroids, plus built in a more compact and sturdy way.
The cost of such system is between $50 and $250. So it is okay in terms of overall FPV or similar class drone. The trouble is software, and there is no way to know who, when and how will develop a better software. But generally those who had expertise before, in terms of components of the needed technologies, are likely to have faster solutions. For example see this.
The implications of radio signal homing
Homing on radio signal would definitely make all larger, more expensive drones to be a target of anti-air drones.
This means that the way drone communicates with operator must be changed somehow.
One of the ways, is when drone reaches required position, does detection of targets using local computer, while being in the radio silence mode. Then once initial mission is done sends a packages of necessary information in short bursts. And then goes again into radio silence mode, waiting for next mission task.
Implication mesh networks
In Magyar interview, mesh networks are specifically mentioned:
”Eventually, piloting within the framework of separate networks that will replace both Starlink and cellular communications.”
Example of such technology could be this.
Basically it requires you to keep in the air several units that would transmit information from one to another. However theoretically it could allow reaching further, by using chains of drones spaced let say 2 km apart (but even 10-15 km is comprehensible).
A drone like this, which is larger then standard FPV, but still still quite inexpensive at say about $1200, can fly as far as 30 km with a 5 kg of payload, and up to 70 with a smaller payload (enough for air burst).
Organising such drones in packs, chains, swarms would allow for deep strike missions like for example anti radar, anti EW missions.
This type of campaign combined with something like this will allow for full fledged ongoing SE-AD/SE-EW operation, that would give the opportunity for manned air support as well as various ground maneuvers.
Combined with a naval carrier(plus signal extender) platforms, it will allow to extend significantly impact areas, to cut the logistics, reach air bases near the shore, reach command and control targets in mass.