The new AI self targeting drone is MASSIVELY important sea change in this war. If Ukraine can produce these fast, before Russia catches up, there will be a slaughter of Z AFVs...
But, isn't 'Bulava' the name of the drone operator unit in the Presidential Brigade, not the drone itself ? Or have they self named a drone after their own handle?
And yes, as it turned out (Alex called back to correct his original message): it's a Wild Hornet UAV, with AI-guidance, operated by the Bulava Battalion of the Presidential Guards Brigade. Have corrected correspondingly.
Not really. It makes the drone a much cheaper loitering munition or slightly longer ranged Javelin with a worse seeker head and smaller warhead. It's still cheap. Also, "AI" aka machine vision can be spoofed with silhouette changes.
I agree...there is no artificial intelligence and there cannot be, but a banal image recognition system is used. and it can only recognize those images that are in its database. that is, very quickly there will be a way to discover it. which will require constant updating of the drone database, but this also has technical limitations. Computer image recognition systems have been known since the 70s, but for 50 years they have not found application in military affairs. so there is no need to announce a new miracle weapon.
It's not a miracle, but substantial improvement. Contrary to SCALP etc. missiles target locking, which are carefully preprogrammed in advance, FPV must flexible lock on target with the operator in real-time. I guess some machine learning (AI) is used there.
Machine learning is the same hype and meaningless term as artificial intelligence. a computer cannot do anything beyond what the programmer has programmed into it.
large Western arms corporations with powerful scientific potential have not done anything like this, but Ukrainian amateurs in a garage have done it? Do you believe in miracles?
The Western military industrial complex had been researching and building loitering munitions since the 2000s. Like the switchblade, which costs about 10,000 dollars+, & is dog shit according to those who used it. Or the Lancet which costs $35,000.
FPV were a product of a war being the mother of invention, again.
It took necessity and a lack of high tech and money, to force the ukrainians to come up with the killer FPV.
They hotrod a normal 500g FPV racer by adding massive motors & a huge power bank, allowing it to lift a 2 kg rpg7 warhead, for just five 500 dollars. So cheap they can make thousands of them, enough to go hunting individual soldiers by FPV.
These FPV drones have TOTALLY CHANGED THE BATTLEFIELD. Everything the army does is now about negating FPV. And this had nothing to do with the Western military industrial complex. It came out of poverty and necessity and the ingenuity of smart people.
And I know that a lot of talented Ukrainian computer techies have been tinkering with this AI terminal guidance system since 2022. Ukraine had a huge tech industry before the war, & luckily has lots of very skilled engineers, coders, & developers, with good links to money foundations like Pritula who fund research into things the actual frontline soldiers are asking for.
It's a much quicker & better system than bloated Western military procurement.
Direct feedback from frontlines is used to rapidly modify iterations so that a failure is a learning opportunity & successes are quickly scaled up.
ISIS has also started using cheap drones. it is the weapon of the poor. it is easily purchased and just as easily neutralized by means of jamming. It surprises me that both sides are still using standard consumer drones, with open control channels, instead of setting up the production of ones that are resistant to interference.
A Javelin seeker head and similar anti tank missiles don't need preprogrammed targets. This drone guidance package's BIGGEST advantage is cost. It lets the operator get hands off on certain situations, possibly negating jamming on a terminal dive.
"Wunderwaffen" are mostly a meme, the last real wonder weapon was used in August 1945.
What this does underline is the absolute necessity of soft and hard kill active protection systems on any AFV that can support it going forward.... Something that Russia has shown little interest in at scale.
No, it's not images in DB, that require update. Cuz there's no task for AI to identify the target. It's operator, who's doing that and lock on. What the homing system is doing is only follows it. It's still not a trivial task, cuz target silhuoette is changing permanently. But this is way easier and achievable.
I think You've mis-read the X posting. It's not an AI loitering munition, it's an FPV with automatic final approach.
It says:
"The BULAVA operator detected the tank, locked on, and gave the command, then the drone independently (!) did the job."
The 'AI' is just for flying the drone into the target the human chose , for flying the last 100 meters. All the drone does is stay locked onto the highlighted box the operator provided.
We saw a human drone operator identify the target manually, then highlight & 'boxed' it on his screen . The drone just targeted the boxed item that a human chose, & unless there's a system for constantly re-changing the target's outline in-between the boxing & the hitting, then the drone will hit it.
The current problem for FPV drone operators is that drone jammers are very good at disrupting the final approach, by blocking the visuals being sent to the operator, & causing a miss. But jammers range is only a few 100 meters.
This new system negates jammers, as the human can spot & mark the target from a distance outside jamming range, then the 'AI' drone just fly's itself to the highlighted target, which jamming can't stop.
Yes the enemy can spoof by changing outlines & using decoys, but the human operator will learn & they still choose the target. It's the human you have to fool.
You can't change the outline in the time between the drone is being locked-on & hitting. How can that work?
This new system is for negating jammers, which are being widely used as the sole defence against FPV attacks.
Without jammers the Russians will be back to the situation 10 months ago, where every single attacking vehicle gets hit by an FPV. How will your outline changes stop that? If the tank outline is changed to resemble a shed, the human will just highlight that shed & the drone will self-fly into it. Any moving tank sized object will get hit by a $650 drone, which is a win as even a moving decoy will probably cost more than the drone. It's simpler to churn out & use drones that make & move tank-size decoys.
The Russians have been getting killed in great numbers before this.
This isn't some magical wonder weapon, it's an improved FPV drone. A smokescreen can stop it as can shooting the drone down.
Not every drone out of the thousands Ukraine makes a month have this system.... Maybe someday but not now.
Furthermore, what they didn't isn't magic, the Chinese and their Russian customers will be right behind Ukraine with this feature.
We also don't know this hit rate of this targeting system and a hit doesn't equal a kill. See the "turtle tanks" or even normal tanks ability to soak up multiple stone hits and keep moving and fighting.
As Mr. Cooper has said before, Ukraine needs to kill and wound tens of THOUSANDS of Russians each month to eliminate Russian offensive capabilities and kill and wound even more to seriously degrade their defensive capabilities.
All your points are true...it's not a miracle weapon, it doesn't guarantee 100% hits, but it does negate the main drone defence currently employed: jammers.
So this AI unit, which I saw somewhere is costing just $150, can be added to most FPV without much extra cost & trouble. This should enable a few months of Ukrainian advantage, which could look like the 2023 FPV 'happy time', where the Orcs had no defence against attacks & couldn't move when FPV units were active.
Smoke is another possible counter measure, but would be difficult for individual soldiers to deploy. An afv can use smoke, but it takes a few seconds to form the cloud, which could be long enough for the FPV to strike. An afv continuosly pumping smoke would be tricky to hit from behind, but vulnerable from front.
So this isn't a game changer, just a temporary advantage which should coincide with the summer attack-season & make the Orcs even more vulnerable. They're getting hammered already, so FPV with double the hit rates will make it a grand slaughter ... IMHO
Also I wonder how hard it is to teach AI to follow the smoke source and hit the "root" of it when diving from above. Should be doable but probably depends on many factors. Would also improve ability to finish the targets that are hit once already and are smoking from ... erm.. natural causes.
'Bulava' the name of the drone operator unit in the Presidential Brigade, not the drone itself. ... and Wild Hornets - is a trademark of the FPV drones.
I think that calling Ukrainian long-range air defense systems S-200 would not be entirely correct. I believe that only the constructs are used from them, and the entire control and guidance system is completely new. Perhaps the rocket fuel is more modern, which made it possible to significantly increase the destruction range.
S200 has massive range. In 2001 a Ukrainian S200 accidentally locked onto a Siberia Airlines Tu-154 250km away and shot it down killing 78 people. It's operational range is estimated at 300 km.
Dear Tom, a very important development, that can change a lot! This footage is not clearly showing how this was happening (like capture area, feature recognition, etc.) and from what distance it was self guided, but it would become longer and longer.
I hope and... hope, that all that footage UA has from various drones have been tediously collected and stored in some central database instead of being wiped out every time. Sigh... though it is unlikely.
That footage is a unique advantage and may serve as a largest ever training material for a large vision and recognition model, that could be then used for all kinds of things, that would change the course of war.
Thanks for the update. Given the importance of air support for offensive operations as well as its potenitial for strategic destruction it seems that the air war is extremely important to the overall outcome. This update had some positive News, like the limitation of rockets for Russia. Even if they try to stockpile.
The new AI self targeting drone is MASSIVELY important sea change in this war. If Ukraine can produce these fast, before Russia catches up, there will be a slaughter of Z AFVs...
But, isn't 'Bulava' the name of the drone operator unit in the Presidential Brigade, not the drone itself ? Or have they self named a drone after their own handle?
Yes.
And yes, as it turned out (Alex called back to correct his original message): it's a Wild Hornet UAV, with AI-guidance, operated by the Bulava Battalion of the Presidential Guards Brigade. Have corrected correspondingly.
Not really. It makes the drone a much cheaper loitering munition or slightly longer ranged Javelin with a worse seeker head and smaller warhead. It's still cheap. Also, "AI" aka machine vision can be spoofed with silhouette changes.
I agree...there is no artificial intelligence and there cannot be, but a banal image recognition system is used. and it can only recognize those images that are in its database. that is, very quickly there will be a way to discover it. which will require constant updating of the drone database, but this also has technical limitations. Computer image recognition systems have been known since the 70s, but for 50 years they have not found application in military affairs. so there is no need to announce a new miracle weapon.
It's not a miracle, but substantial improvement. Contrary to SCALP etc. missiles target locking, which are carefully preprogrammed in advance, FPV must flexible lock on target with the operator in real-time. I guess some machine learning (AI) is used there.
Machine learning is the same hype and meaningless term as artificial intelligence. a computer cannot do anything beyond what the programmer has programmed into it.
large Western arms corporations with powerful scientific potential have not done anything like this, but Ukrainian amateurs in a garage have done it? Do you believe in miracles?
I disagree 100%.
The Western military industrial complex had been researching and building loitering munitions since the 2000s. Like the switchblade, which costs about 10,000 dollars+, & is dog shit according to those who used it. Or the Lancet which costs $35,000.
FPV were a product of a war being the mother of invention, again.
It took necessity and a lack of high tech and money, to force the ukrainians to come up with the killer FPV.
They hotrod a normal 500g FPV racer by adding massive motors & a huge power bank, allowing it to lift a 2 kg rpg7 warhead, for just five 500 dollars. So cheap they can make thousands of them, enough to go hunting individual soldiers by FPV.
These FPV drones have TOTALLY CHANGED THE BATTLEFIELD. Everything the army does is now about negating FPV. And this had nothing to do with the Western military industrial complex. It came out of poverty and necessity and the ingenuity of smart people.
And I know that a lot of talented Ukrainian computer techies have been tinkering with this AI terminal guidance system since 2022. Ukraine had a huge tech industry before the war, & luckily has lots of very skilled engineers, coders, & developers, with good links to money foundations like Pritula who fund research into things the actual frontline soldiers are asking for.
It's a much quicker & better system than bloated Western military procurement.
Direct feedback from frontlines is used to rapidly modify iterations so that a failure is a learning opportunity & successes are quickly scaled up.
ISIS has also started using cheap drones. it is the weapon of the poor. it is easily purchased and just as easily neutralized by means of jamming. It surprises me that both sides are still using standard consumer drones, with open control channels, instead of setting up the production of ones that are resistant to interference.
A Javelin seeker head and similar anti tank missiles don't need preprogrammed targets. This drone guidance package's BIGGEST advantage is cost. It lets the operator get hands off on certain situations, possibly negating jamming on a terminal dive.
"Wunderwaffen" are mostly a meme, the last real wonder weapon was used in August 1945.
What this does underline is the absolute necessity of soft and hard kill active protection systems on any AFV that can support it going forward.... Something that Russia has shown little interest in at scale.
No, it's not images in DB, that require update. Cuz there's no task for AI to identify the target. It's operator, who's doing that and lock on. What the homing system is doing is only follows it. It's still not a trivial task, cuz target silhuoette is changing permanently. But this is way easier and achievable.
I think You've mis-read the X posting. It's not an AI loitering munition, it's an FPV with automatic final approach.
It says:
"The BULAVA operator detected the tank, locked on, and gave the command, then the drone independently (!) did the job."
The 'AI' is just for flying the drone into the target the human chose , for flying the last 100 meters. All the drone does is stay locked onto the highlighted box the operator provided.
We saw a human drone operator identify the target manually, then highlight & 'boxed' it on his screen . The drone just targeted the boxed item that a human chose, & unless there's a system for constantly re-changing the target's outline in-between the boxing & the hitting, then the drone will hit it.
The current problem for FPV drone operators is that drone jammers are very good at disrupting the final approach, by blocking the visuals being sent to the operator, & causing a miss. But jammers range is only a few 100 meters.
This new system negates jammers, as the human can spot & mark the target from a distance outside jamming range, then the 'AI' drone just fly's itself to the highlighted target, which jamming can't stop.
Yes the enemy can spoof by changing outlines & using decoys, but the human operator will learn & they still choose the target. It's the human you have to fool.
You can't change the outline in the time between the drone is being locked-on & hitting. How can that work?
This new system is for negating jammers, which are being widely used as the sole defence against FPV attacks.
Without jammers the Russians will be back to the situation 10 months ago, where every single attacking vehicle gets hit by an FPV. How will your outline changes stop that? If the tank outline is changed to resemble a shed, the human will just highlight that shed & the drone will self-fly into it. Any moving tank sized object will get hit by a $650 drone, which is a win as even a moving decoy will probably cost more than the drone. It's simpler to churn out & use drones that make & move tank-size decoys.
The Russians are going to get slaughtered.
Last bit should say: "It's cheaper to churn out and use drones THAN make lots of tank size decoys"
The Russians have been getting killed in great numbers before this.
This isn't some magical wonder weapon, it's an improved FPV drone. A smokescreen can stop it as can shooting the drone down.
Not every drone out of the thousands Ukraine makes a month have this system.... Maybe someday but not now.
Furthermore, what they didn't isn't magic, the Chinese and their Russian customers will be right behind Ukraine with this feature.
We also don't know this hit rate of this targeting system and a hit doesn't equal a kill. See the "turtle tanks" or even normal tanks ability to soak up multiple stone hits and keep moving and fighting.
As Mr. Cooper has said before, Ukraine needs to kill and wound tens of THOUSANDS of Russians each month to eliminate Russian offensive capabilities and kill and wound even more to seriously degrade their defensive capabilities.
Will this guidance package help?
Of course but again, it's not magical.
All your points are true...it's not a miracle weapon, it doesn't guarantee 100% hits, but it does negate the main drone defence currently employed: jammers.
So this AI unit, which I saw somewhere is costing just $150, can be added to most FPV without much extra cost & trouble. This should enable a few months of Ukrainian advantage, which could look like the 2023 FPV 'happy time', where the Orcs had no defence against attacks & couldn't move when FPV units were active.
Smoke is another possible counter measure, but would be difficult for individual soldiers to deploy. An afv can use smoke, but it takes a few seconds to form the cloud, which could be long enough for the FPV to strike. An afv continuosly pumping smoke would be tricky to hit from behind, but vulnerable from front.
So this isn't a game changer, just a temporary advantage which should coincide with the summer attack-season & make the Orcs even more vulnerable. They're getting hammered already, so FPV with double the hit rates will make it a grand slaughter ... IMHO
Also I wonder how hard it is to teach AI to follow the smoke source and hit the "root" of it when diving from above. Should be doable but probably depends on many factors. Would also improve ability to finish the targets that are hit once already and are smoking from ... erm.. natural causes.
'Bulava' the name of the drone operator unit in the Presidential Brigade, not the drone itself. ... and Wild Hornets - is a trademark of the FPV drones.
Also came here to mention this...
Thank you Tom for your support and the comprehensive review!
I think that calling Ukrainian long-range air defense systems S-200 would not be entirely correct. I believe that only the constructs are used from them, and the entire control and guidance system is completely new. Perhaps the rocket fuel is more modern, which made it possible to significantly increase the destruction range.
S200 has massive range. In 2001 a Ukrainian S200 accidentally locked onto a Siberia Airlines Tu-154 250km away and shot it down killing 78 people. It's operational range is estimated at 300 km.
Thank you Tom
Dear Tom, a very important development, that can change a lot! This footage is not clearly showing how this was happening (like capture area, feature recognition, etc.) and from what distance it was self guided, but it would become longer and longer.
I hope and... hope, that all that footage UA has from various drones have been tediously collected and stored in some central database instead of being wiped out every time. Sigh... though it is unlikely.
That footage is a unique advantage and may serve as a largest ever training material for a large vision and recognition model, that could be then used for all kinds of things, that would change the course of war.
Also, according to some reports like this, the Dutch Gepard are entering the fray:
https://www.youtube.com/watch?v=eAZ9tLLht9I
Thanks for the update. Given the importance of air support for offensive operations as well as its potenitial for strategic destruction it seems that the air war is extremely important to the overall outcome. This update had some positive News, like the limitation of rockets for Russia. Even if they try to stockpile.
Thanks for the analysis Adrien and Tom . . .
Do we know what percentage of Ukrainian missiles/long range drones Russians IADs is shooting down?
From a cursory glance, the Russian IADS seems to be incapable of shooting down anything, including weapons attacking the SAM batteries themselves.
Thanks Tom and Adrien for a very informative update. Always glad to read some good news as well.