How Hot is Ukraine Gonna Get?

[akd]

Yet another view of same failed Russian assault at Avdiivka:

 

[Battlefront.com]

2 hours ago, chrisl said:

Yes, it's a "difference detection" problem, but if you're looking for something that's near your limit of resolution it gets harder.  And all measurement systems have noise. For most human optical applications you're used to signal levels that are *way* above the noise, but someone trying to hide from a satellite is going to work hard to keep the changes that you see down close to the observer's noise level. That makes life much harder for the automated system and its trainer.  And aside from the changes they're looking for, there are lots of other changes going on above threshold because stuff just moves around on earth.  That's noise, too, but of a different sort, and part of the developer's goal is to be able to discriminate benign activity from targetable activity when they both show up in the signal.

You and Kimbosbread need to listen to a couple minutes of this if you haven't already.  It's Perun at his finest (I cued it to the relevant spot):

Steve

[Battlefront.com]

30 minutes ago, akd said:

Yet another view of same failed Russian assault at Avdiivka:

 

Thanks.  This one gives us the best look, so far, about what the dismounts did in connection with the vehicles going boom.  Apparently they mostly decided to bugger off.  Which means Russia made a huge investment in a fairly small attack, got the infantry within striking range of their objective, dismounted most of the infantry intact, and then they ran away as their vehicles were blown to pieces.

As any CM player knows from experience, this is not a high quality formation in these videos

Steve

[chrisl]

9 minutes ago, Battlefront.com said:

You and Kimbosbread need to listen to a couple minutes of this if you haven't already.  It's Perun at his finest (I cued it to the relevant spot):

Steve

Yeah, I'm very aware of that, and I suppose that's a cue for getting a little into why we don't have anti-drone drones.

There are two parts to the anti-drone drone: detection and attack.  The detection is the hard part.  Destruction is easy - we already have no end of systems that can very accurately destroy anything that you give them coordinates of.  We can accurately fire projectiles, exploding projectiles, exploding projectiles full of razor sharp hoops, high energy beams of photons, rings with chains on them, rings with strings on them, giant wads of gooey stuff, or anything you want to take out a drone.  But you have to detect it.

For an anti-drone drone, there are sort of two categories of drone you're targeting: open loop (no comm back to the sender) and closed loop (some comm back to the sender, whether full two-way control, occasional updates, or whatever).  

Detection of the first type (no comm), which includes Shaheds, is tricky - unlike the F-35, these *start* with the radar cross section of a goose* and then you can make that even smaller.  These things are all small on visual and radar cross sections because you can paint them and they don't have a lot of metal.  You're going to track them with frustrating "visual" algorithms, where "visual" can mean different things in the optical vs. radar wavelengths, but you're still trying to pick out changes in the scene to decide where the thing is.  I'm not going to spend much time on it, other than to say that unless you have really high signal to noise and high resolution (both of which the target is trying to reduce), it's a lot harder than you think, and in general you're not going to get there with simple image differencing.  And this problem exists for commless drones whether you're using another drone, a gun, or a death ray to take them down.  Shaheds at least have a very characteristic sound that you can probably use for detection and targeting once they're within audible range.

Detection of the second type (active comm) is easy.  It's transmitting, and transmitting enough to get clear signal back to its operator, who is farther away than you are if it's attacking you.  Triangulation is old technology.  Piece of cake: you lock onto the frequency, have some kind of sensor so you know your own orientation relative to the sensor, and just maneuver in a way to make the signal from the drone stronger until you hit it and destroy it with whatever mechanism you prefer.  Or have a few sensors that are networked to give you the position (helloooo MLAT) and shoot it with your favorite method of action-at-a-distance.

Except for one problem: whose drone did you just destroy?

In the Ukraine environment, IFF is the hard part of doing radio based anti-drone systems.  There are tons of things flying around, as evidenced by the daily releases of yet another view of every bit of ground combat we ever see.  It's not quite Diamond Age concentrations of them, but they're working on it.  And they're all sorts of random drones, including commercial drones, custom drones made with commercial off the shelf parts, custom drones with a mix of commercial and special mil parts, totally custom mil drones, and who knows what else. And they're all using similar frequencies, because the combination of physics and the atmosphere force you to the same frequencies if you want a particular range and data rate at powers that you can reasonably supply to both the ground operator and drone with batteries.  If you don't sort out the IFF thing and you set an autonomous anti-radiation based anti-drone system loose, it's just as likely to attack its allied drones as the enemy drones, because it has no way to tell them apart.  That means you have to have your complete drone ecosystem integrated (ring that cash register over at Lockheed/Northrop Grumman/Raytheon!!) or you're just going to be attacking your own stuff.  

And part of why we aren't seeing even rudimentary versions of it in Ukraine is that it's not a function that people were already spending much effort on for commercial/hobbyist drones. You can't just pop over to Robotshop.com or Alibaba and order tunable RF sensor kits (or a few thousand of them) the way you can other types of sensor, or actuators for operating your 3D printed grenade dropper.  It's possible to get relatively inexpensive software-defined radio modules that are small (that's what feeds ADSBExchange so you can see who's flying around Ukraine), but the environment is so variable, along with the need to confirm what drone you're attacking, that at least for now you're going to need a human in the loop, even if you can semi-automate your remote control drone sensor.  And even with a human in the loop, nobody is painting national flags on their drones, so unless you know "this is one that our side makes" after you get up close to it (assuming you're doing that, rather than sending a death ray at it from 5 km), you really don't know who you're shooting down.  So the basic tech isn't all that hard, but because it's not just point and shoot or point and drop, it's a lot more dependent on integration of the whole system to be usable.

*geese, like all waterfowl, are incredibly mean and probably deserve to die. That's why there's a book entitled "Ducks and how to make them pay".  If we can do an autonomous system for drones, it should probably be immediately applied to geese and ducks.

[Maciej Zwolinski]

10 hours ago, Grossman said:

 

Oleg Tsaryov once said to be the potential leader of a Kremlin controlled puppet government in Ukraine shot twice by unknown assailant.

 

Good. Sic semper traditores.

[Maciej Zwolinski]

On 10/26/2023 at 11:28 PM, Zeleban said:

Be that as it may, I can say that times have changed. compared to last year. I had a development project. But because of the war, I decided that civilian projects would not benefit my country and decided to join the Ukrainian armed forces. Moreover, recently chronic diseases are not considered an obstacle to recruitment

Good luck and happy hunting!

[zinz]

https://nitter.net/GloOouD/status/1718321869089092003#m

Interesting details about the dinipro river crossing 

[fireship4]

1 hour ago, Maciej Zwolinski said:

Good. Sic semper traditores.

I believe he had recently made some anti-war comments.

[Jr Buck Private]

3 hours ago, Battlefront.com said:

Thanks.  This one gives us the best look, so far, about what the dismounts did in connection with the vehicles going boom.  Apparently they mostly decided to bugger off.  Which means Russia made a huge investment in a fairly small attack, got the infantry within striking range of their objective, dismounted most of the infantry intact, and then they ran away as their vehicles were blown to pieces.

As any CM player knows from experience, this is not a high quality formation in these videos

Steve

I was wondering if they retreated a little to easily, not that I blame them.   Not exactly a modern day charge of the light brigade.   

[Battlefront.com]

17 minutes ago, Jr Buck Private said:

I was wondering if they retreated a little to easily, not that I blame them.   Not exactly a modern day charge of the light brigade.   

The thing is, it was closer to engage the Ukrainians than it was to disengage.  At least if you get up close the artillery will have to stop out of concern of hitting friendlies.  On top of that, what are the chances that the Russian commander of this op is going to say "oh well lads, you gave it your best shot.  I understand why you bolted.  Take a rest and we'll try again when we have a better plan"?  Zero.  Instead it is going to be "OK, tomorrow you're going to assault again, but this time without rides".

The militarily sound, logical thing for these guys to have done was to press on because it was their best option.  Good quality NATO standard units would all understand this, right down to the most junior enlisted.  It is part of their training and should have been part of their mission briefing (i.e. covering various contingency plans).  Chances are none of that happened and that, more than anything, sort of ignorance is what feeds panic.

Steve

[Kinophile]

https://mil.in.ua/en/news/germany-hands-over-sonobot-5-surface-drones-to-ukraine/

Interesting design. 

[Kinophile]

And this:

https://www.pravda.com.ua/eng/news/2023/10/27/7425995/

 

 

[Kinophile]

Russian Imperialism, old and new. 

https://www.politico.com/news/magazine/2023/10/27/russia-colonization-alaska-ukraine-00123352

[acrashb]

Another demonstration of the fragility of MBTs.  Down the hatch:

 

[Harmon Rabb]

Gosh I like the M270...From the days I played C&C 95 for the first time.

Don't see that much footage of M270s being used compared to HIMARS, always a treat to see some.

 

 

[Battlefront.com]

9 hours ago, chrisl said:

Yeah, I'm very aware of that, and I suppose that's a cue for getting a little into why we don't have anti-drone drones.

There are two parts to the anti-drone drone: detection and attack.  The detection is the hard part.  Destruction is easy - we already have no end of systems that can very accurately destroy anything that you give them coordinates of.  We can accurately fire projectiles, exploding projectiles, exploding projectiles full of razor sharp hoops, high energy beams of photons, rings with chains on them, rings with strings on them, giant wads of gooey stuff, or anything you want to take out a drone.  But you have to detect it.

For an anti-drone drone, there are sort of two categories of drone you're targeting: open loop (no comm back to the sender) and closed loop (some comm back to the sender, whether full two-way control, occasional updates, or whatever).  

Detection of the first type (no comm), which includes Shaheds, is tricky - unlike the F-35, these *start* with the radar cross section of a goose* and then you can make that even smaller.  These things are all small on visual and radar cross sections because you can paint them and they don't have a lot of metal.  You're going to track them with frustrating "visual" algorithms, where "visual" can mean different things in the optical vs. radar wavelengths, but you're still trying to pick out changes in the scene to decide where the thing is.  I'm not going to spend much time on it, other than to say that unless you have really high signal to noise and high resolution (both of which the target is trying to reduce), it's a lot harder than you think, and in general you're not going to get there with simple image differencing.  And this problem exists for commless drones whether you're using another drone, a gun, or a death ray to take them down.  Shaheds at least have a very characteristic sound that you can probably use for detection and targeting once they're within audible range.

Detection of the second type (active comm) is easy.  It's transmitting, and transmitting enough to get clear signal back to its operator, who is farther away than you are if it's attacking you.  Triangulation is old technology.  Piece of cake: you lock onto the frequency, have some kind of sensor so you know your own orientation relative to the sensor, and just maneuver in a way to make the signal from the drone stronger until you hit it and destroy it with whatever mechanism you prefer.  Or have a few sensors that are networked to give you the position (helloooo MLAT) and shoot it with your favorite method of action-at-a-distance.

Except for one problem: whose drone did you just destroy?

In the Ukraine environment, IFF is the hard part of doing radio based anti-drone systems.  There are tons of things flying around, as evidenced by the daily releases of yet another view of every bit of ground combat we ever see.  It's not quite Diamond Age concentrations of them, but they're working on it.  And they're all sorts of random drones, including commercial drones, custom drones made with commercial off the shelf parts, custom drones with a mix of commercial and special mil parts, totally custom mil drones, and who knows what else. And they're all using similar frequencies, because the combination of physics and the atmosphere force you to the same frequencies if you want a particular range and data rate at powers that you can reasonably supply to both the ground operator and drone with batteries.  If you don't sort out the IFF thing and you set an autonomous anti-radiation based anti-drone system loose, it's just as likely to attack its allied drones as the enemy drones, because it has no way to tell them apart.  That means you have to have your complete drone ecosystem integrated (ring that cash register over at Lockheed/Northrop Grumman/Raytheon!!) or you're just going to be attacking your own stuff.  

And part of why we aren't seeing even rudimentary versions of it in Ukraine is that it's not a function that people were already spending much effort on for commercial/hobbyist drones. You can't just pop over to Robotshop.com or Alibaba and order tunable RF sensor kits (or a few thousand of them) the way you can other types of sensor, or actuators for operating your 3D printed grenade dropper.  It's possible to get relatively inexpensive software-defined radio modules that are small (that's what feeds ADSBExchange so you can see who's flying around Ukraine), but the environment is so variable, along with the need to confirm what drone you're attacking, that at least for now you're going to need a human in the loop, even if you can semi-automate your remote control drone sensor.  And even with a human in the loop, nobody is painting national flags on their drones, so unless you know "this is one that our side makes" after you get up close to it (assuming you're doing that, rather than sending a death ray at it from 5 km), you really don't know who you're shooting down.  So the basic tech isn't all that hard, but because it's not just point and shoot or point and drop, it's a lot more dependent on integration of the whole system to be usable.

*geese, like all waterfowl, are incredibly mean and probably deserve to die. That's why there's a book entitled "Ducks and how to make them pay".  If we can do an autonomous system for drones, it should probably be immediately applied to geese and ducks.

Good discussion (apart from the duck sentiment).

I read this stuff pretty much every day and I just don't see anybody with a good idea how to reconfigure air defenses to handle hundreds or thousand of simultaneous targets EVEN IF they can be detected consistently.  And of course nobody wants to mention ground hugging drones that can use terrain masking in ways no previous aircraft could ever even conceive of.

This is a big concern for this war because Russia is able to operate sufficient drones to cause Ukraine a lot of problems, both in the rear and at the front.  And because nobody has a good solution for drones, including the ones that are dependent upon GPS and 2 way coms, it's going to get worse as drone numbers continue to increase.

ISW even reported some are saying Russia is now actively using fully autonomous drones with autonomous targeting. 

Quote

Russian forces are reportedly using a new version of the “Lancet” kamikaze drone that can autonomously identify targets. Russian sources claimed that Russian forces began using the new “Izdeliye-53” kamikaze drone as of October 21.[16] The sources claimed that the “Izdeliye-53” drone reportedly has an automatic guidance system that can distinguish types of targets and increase strike success rates.[17] Russian forces are reportedly not using the “Izdeliye-53” drones on a wide scale yet, but Russian sources claimed that Russian forces are currently testing the drones for mass synchronized swarm strikes.[18] ISW previously reported on October 24 that Russian forces also allegedly recently used the new “Italmas” (also known as “Izdeliye-54”) drones during a drone strike on Kyiv Oblast.[19] ISW also previously assessed that the Russian command may believe that a large number of strike drones will allow Russian forces to overwhelm Ukrainian air defenses, although the payload of the “Izdeliye-53” drones, which is reportedly between three and five kilograms, may not be sufficient to significantly damage most critical military targets.[20]

If it's anything like Soviet Dog Mines this should be fun to watch

Steve

[chrisl]

2 hours ago, Battlefront.com said:

Good discussion (apart from the duck sentiment).

I read this stuff pretty much every day and I just don't see anybody with a good idea how to reconfigure air defenses to handle hundreds or thousand of simultaneous targets EVEN IF they can be detected consistently.  And of course nobody wants to mention ground hugging drones that can use terrain masking in ways no previous aircraft could ever even conceive of.

This is a big concern for this war because Russia is able to operate sufficient drones to cause Ukraine a lot of problems, both in the rear and at the front.  And because nobody has a good solution for drones, including the ones that are dependent upon GPS and 2 way coms, it's going to get worse as drone numbers continue to increase.

ISW even reported some are saying Russia is now actively using fully autonomous drones with autonomous targeting. 

If it's anything like Soviet Dog Mines this should be fun to watch

Steve

I'd think the Russians training waterfowl strapped with mines (two birds with one mine?)  would probably be more effective than anything "autonomous" that they can field today with their current resources.  Automatic?  Sure, that's Shahed.  Automatic with a basic IR sensor for picking out warm targets?  Probably, but not widely available.  Truly autonomous and using some kind of AI to pick targets?  Bring in the trained ducks.

Air defense against the Diamond Age swarm isn't going to be any single system - it's going to have to be many layers dispersed over a large volume, and insensitive to any particular node or 20 being knocked out.  Basically attack of the killer bees on both sides.  

Ground hugging drones is a nice approach for anything that can be a lightweight payload.  Lots of background to make them hard to pick out, but without the mobility issues of a UGV. Arguably a dumptruck full of ground hugging drones will be more effective than most UGV concepts.

[dan/california]

11 hours ago, chrisl said:

Yeah, I'm very aware of that, and I suppose that's a cue for getting a little into why we don't have anti-drone drones.

There are two parts to the anti-drone drone: detection and attack.  The detection is the hard part.  Destruction is easy - we already have no end of systems that can very accurately destroy anything that you give them coordinates of.  We can accurately fire projectiles, exploding projectiles, exploding projectiles full of razor sharp hoops, high energy beams of photons, rings with chains on them, rings with strings on them, giant wads of gooey stuff, or anything you want to take out a drone.  But you have to detect it.

For an anti-drone drone, there are sort of two categories of drone you're targeting: open loop (no comm back to the sender) and closed loop (some comm back to the sender, whether full two-way control, occasional updates, or whatever).  

Detection of the first type (no comm), which includes Shaheds, is tricky - unlike the F-35, these *start* with the radar cross section of a goose* and then you can make that even smaller.  These things are all small on visual and radar cross sections because you can paint them and they don't have a lot of metal.  You're going to track them with frustrating "visual" algorithms, where "visual" can mean different things in the optical vs. radar wavelengths, but you're still trying to pick out changes in the scene to decide where the thing is.  I'm not going to spend much time on it, other than to say that unless you have really high signal to noise and high resolution (both of which the target is trying to reduce), it's a lot harder than you think, and in general you're not going to get there with simple image differencing.  And this problem exists for commless drones whether you're using another drone, a gun, or a death ray to take them down.  Shaheds at least have a very characteristic sound that you can probably use for detection and targeting once they're within audible range.

Detection of the second type (active comm) is easy.  It's transmitting, and transmitting enough to get clear signal back to its operator, who is farther away than you are if it's attacking you.  Triangulation is old technology.  Piece of cake: you lock onto the frequency, have some kind of sensor so you know your own orientation relative to the sensor, and just maneuver in a way to make the signal from the drone stronger until you hit it and destroy it with whatever mechanism you prefer.  Or have a few sensors that are networked to give you the position (helloooo MLAT) and shoot it with your favorite method of action-at-a-distance.

Except for one problem: whose drone did you just destroy?

In the Ukraine environment, IFF is the hard part of doing radio based anti-drone systems.  There are tons of things flying around, as evidenced by the daily releases of yet another view of every bit of ground combat we ever see.  It's not quite Diamond Age concentrations of them, but they're working on it.  And they're all sorts of random drones, including commercial drones, custom drones made with commercial off the shelf parts, custom drones with a mix of commercial and special mil parts, totally custom mil drones, and who knows what else. And they're all using similar frequencies, because the combination of physics and the atmosphere force you to the same frequencies if you want a particular range and data rate at powers that you can reasonably supply to both the ground operator and drone with batteries.  If you don't sort out the IFF thing and you set an autonomous anti-radiation based anti-drone system loose, it's just as likely to attack its allied drones as the enemy drones, because it has no way to tell them apart.  That means you have to have your complete drone ecosystem integrated (ring that cash register over at Lockheed/Northrop Grumman/Raytheon!!) or you're just going to be attacking your own stuff.  

And part of why we aren't seeing even rudimentary versions of it in Ukraine is that it's not a function that people were already spending much effort on for commercial/hobbyist drones. You can't just pop over to Robotshop.com or Alibaba and order tunable RF sensor kits (or a few thousand of them) the way you can other types of sensor, or actuators for operating your 3D printed grenade dropper.  It's possible to get relatively inexpensive software-defined radio modules that are small (that's what feeds ADSBExchange so you can see who's flying around Ukraine), but the environment is so variable, along with the need to confirm what drone you're attacking, that at least for now you're going to need a human in the loop, even if you can semi-automate your remote control drone sensor.  And even with a human in the loop, nobody is painting national flags on their drones, so unless you know "this is one that our side makes" after you get up close to it (assuming you're doing that, rather than sending a death ray at it from 5 km), you really don't know who you're shooting down.  So the basic tech isn't all that hard, but because it's not just point and shoot or point and drop, it's a lot more dependent on integration of the whole system to be usable.

*geese, like all waterfowl, are incredibly mean and probably deserve to die. That's why there's a book entitled "Ducks and how to make them pay".  If we can do an autonomous system for drones, it should probably be immediately applied to geese and ducks.

 

17 minutes ago, chrisl said:

I'd think the Russians training waterfowl strapped with mines (two birds with one mine?)  would probably be more effective than anything "autonomous" that they can field today with their current resources.  Automatic?  Sure, that's Shahed.  Automatic with a basic IR sensor for picking out warm targets?  Probably, but not widely available.  Truly autonomous and using some kind of AI to pick targets?  Bring in the trained ducks.

Air defense against the Diamond Age swarm isn't going to be any single system - it's going to have to be many layers dispersed over a large volume, and insensitive to any particular node or 20 being knocked out.  Basically attack of the killer bees on both sides.  

Ground hugging drones is a nice approach for anything that can be a lightweight payload.  Lots of background to make them hard to pick out, but without the mobility issues of a UGV. Arguably a dumptruck full of ground hugging drones will be more effective than most UGV concepts.

ChrisSL is starting to rival the Captain in terms of useful information. And that is NOT a statement I make lightly.

[chrisl]

56 minutes ago, dan/california said:

 

ChrisSL is starting to rival the Captain in terms of useful information. And that is NOT a statement I make lightly.

In a very narrow domain.  Low SNR tracking of large numbers of objects at the resolution limit of my optical system using (deliberately) underpowered computers has been a headache of mine for the past decade. This is for targets entirely unrelated to anything of defense interest, but it's defense-adjacent, and there aren't good general solutions floating around out there.

[Butschi]

2 hours ago, chrisl said:

I'd think the Russians training waterfowl strapped with mines (two birds with one mine?)  would probably be more effective than anything "autonomous" that they can field today with their current resources.  Automatic?  Sure, that's Shahed.  Automatic with a basic IR sensor for picking out warm targets?  Probably, but not widely available.  Truly autonomous and using some kind of AI to pick targets?  Bring in the trained ducks.

It depends on how you define "AI" and "autonomous". I'm involved with the prediction and planning part of autonomous driving, not flying drones and not (directly) the perception part, so take what I say with a pinch of salt.

Anyway, for image classification you don't need to be an AI or expert (which is not to say the Russians don't have any). Most high end research in that field is public, often even including code. What you need to do is train the (usually) neural net to identify stuff you are interested in (not cats and dogs). But that is easy to do if you have enough training data.

Of course it is not only detection and classification but also tracking... but that, too, is really available on the Internet.

Once trained you can either strip the net down to something manageable by less potent hardware or possibly design a dedicated FPGA or similar.

And let's keep in mind that this task is simpler than autonomous driving if you don't mind having a higher misidentification rate and don't care about collateral damage, or even friendly fire as long as sufficiently many of your (cheap?) drones hit something you want them to hit. All of which is probably true for the Russians and less so for Ukraine.

As you said yourself, steering the drone towards that target is the simple part.

Now the level of autonomy is the other variable here. Of course having a drone where you just press the start button and it plans a sensible search pattern, without relying on GPS, using terrain to its advantage etc. is complicated. Steering the drone manually or programming a path to an area of interest and then switching on "hunt mode" with some pre-programmed search algorithm (like a robot vacuum cleaner does) is rather simple.

 

[The_Capt]

3 hours ago, dan/california said:

 

ChrisSL is starting to rival the Captain in terms of useful information. And that is NOT a statement I make lightly.

I have a physics degree and most of this is way above my head.  One thing is becoming apparent though - if modern militaries cannot solve for unmanned air war below 2000ft then we are entering into an different era of warfare.

Denial and Defence will rule conventional warfare until we can crack the unmanned problem.  Military implications for this are enormous, especially considering we have built for Interventions/Offence for the last 30 years at least.  Political ramifications of this are not small considering that entry costs for these technologies are low.

Well if anyone is looking for a career, this sector will be booming for years.

[Butschi]

"Die Zeit" article (in German)

https://www.zeit.de/politik/ausland/2023-10/ukraine-gegenoffensive-ende-unterstuetzung-krieg-russland

The noteworthy aspect is the way the Russian offensive is interpreted. In line with what I had guesstimated the reaction over here would be.

Excerpt:

Quote

At the beginning of the Ukrainian advance, however, hardly anyone expected that instead of liberating large parts of Ukrainian territory, it would result in another large-scale Russian attack. For more than a month, the front in southern Ukraine has been frozen - even though it has hardly rained in the region for weeks and the sun warms the air up to 20 degrees. It has been two months since the Ukrainian army liberated the last village, Robotyne. At the time, it seemed as if the Ukrainians had found a weak point in the occupiers' extensive defence lines. But instead of retreating, Russia is now attacking itself, thus finally ushering in the end of the Ukrainian counter-offensive.

Translated with www.DeepL.com/Translator (free version)

Seen in isolation and from a purely military point of view the offensive looks like a waste of lives and material. However, for the mainstream audience, what sticks is that Russia is still able to mount a large offensive despite everything we've sent to Ukraine. Note also: this is the only article about Ukraine in Zeit today. Israel has completely replaced Ukraine in the news.

[Letter from Prague]

3 hours ago, Butschi said:

It depends on how you define "AI" and "autonomous". I'm involved with the prediction and planning part of autonomous driving, not flying drones and not (directly) the perception part, so take what I say with a pinch of salt.

Anyway, for image classification you don't need to be an AI or expert (which is not to say the Russians don't have any). Most high end research in that field is public, often even including code. What you need to do is train the (usually) neural net to identify stuff you are interested in (not cats and dogs). But that is easy to do if you have enough training data.

Of course it is not only detection and classification but also tracking... but that, too, is really available on the Internet.

Once trained you can either strip the net down to something manageable by less potent hardware or possibly design a dedicated FPGA or similar.

And let's keep in mind that this task is simpler than autonomous driving if you don't mind having a higher misidentification rate and don't care about collateral damage, or even friendly fire as long as sufficiently many of your (cheap?) drones hit something you want them to hit. All of which is probably true for the Russians and less so for Ukraine.

As you said yourself, steering the drone towards that target is the simple part.

Now the level of autonomy is the other variable here. Of course having a drone where you just press the start button and it plans a sensible search pattern, without relying on GPS, using terrain to its advantage etc. is complicated. Steering the drone manually or programming a path to an area of interest and then switching on "hunt mode" with some pre-programmed search algorithm (like a robot vacuum cleaner does) is rather simple.

 

Discussion whether it's AI is mostly discussion of "what do we call AI" so that is not useful. But yeah, I was assigned to do (software version of) Sidewinder-style missile guidance as a homework in college over a decade ago. The articles about the methods we used were from the 80s.

It is not hard to do, but it is hard to do well - you need good models to be able to say what is a tank and what is car, even if both are easy to see blobs in infrared.

The Russians (and the Chinese I assume) will of course have advantage in this area over the "nice"  / "constrained by lawyers" nations, since their objective is terror - so hitting a random civilian car instead of a tank is not a failure for them.

Edited October 29, 2023 by Letter from Prague

[The_Capt]

20 minutes ago, Letter from Prague said:

Discussion whether it's AI is mostly discussion of "what do we call AI" so that is not useful. But yeah, I was assigned to do (software version of) Sidewinder-style missile guidance as a homework in college over a decade ago. The articles about the methods we used were from the 80s.

It is not hard to do, but it is hard to do well - you need good models to be able to say what is a tank and what is car, even if both are easy to see blobs in infrared.

The Russians (and the Chinese I assume) will of course have advantage in this area over the "nice"  / "constrained by lawyers" nations, since their objective is terror - so hitting a random civilian car instead of a tank is not a failure for them.

Don't we kinda already have this?  Modern ATGM guidance systems look pretty sophisticated in their ability to stay looked onto one large hunk of metal vs a burning garbage truck.

"The tracker is key to guidance/control for an eventual hit. The signals from each of the 4,096 detector elements (64×64 pixel array) in the seeker are passed to the FPA readout integrated circuits which reads then creates a video frame that is sent to the tracker system for processing. By comparing the individual frames, the tracker determines the need to correct so as to keep the missile on target. The tracker must be able to determine which portion of the image represents the target.

The target is initially defined by the gunner, who places a configurable frame around it. The tracker then uses algorithms to compare that region of the frame based on image, geometric, and movement data to the new image frames being sent from the seeker, similar to pattern recognition algorithms. At the end of each frame, the reference is updated. The tracker is able to keep track of the target even though the seeker's point of view can change radically in the course of flight.

The missile is equipped with four movable tail fins and eight fixed wings at mid-body. To guide the missile, the tracker locates the target in the current frame and compares this position with the aim point. If this position is off center, the tracker computes a correction and passes it to the guidance system, which makes the appropriate adjustments to the four movable tail fins. This is an autopilot. To guide the missile, the system has sensors that check that the fins are positioned as requested. If not, the deviation is sent back to the controller for further adjustment. This is a closed-loop controller."

[Tux]

18 hours ago, chrisl said:

Yeah, I'm very aware of that, and I suppose that's a cue for getting a little into why we don't have anti-drone drones.

There are two parts to the anti-drone drone: detection and attack.  The detection is the hard part.  Destruction is easy - we already have no end of systems that can very accurately destroy anything that you give them coordinates of.  We can accurately fire projectiles, exploding projectiles, exploding projectiles full of razor sharp hoops, high energy beams of photons, rings with chains on them, rings with strings on them, giant wads of gooey stuff, or anything you want to take out a drone.  But you have to detect it.

For an anti-drone drone, there are sort of two categories of drone you're targeting: open loop (no comm back to the sender) and closed loop (some comm back to the sender, whether full two-way control, occasional updates, or whatever).  

Detection of the first type (no comm), which includes Shaheds, is tricky - unlike the F-35, these *start* with the radar cross section of a goose* and then you can make that even smaller.  These things are all small on visual and radar cross sections because you can paint them and they don't have a lot of metal.  You're going to track them with frustrating "visual" algorithms, where "visual" can mean different things in the optical vs. radar wavelengths, but you're still trying to pick out changes in the scene to decide where the thing is.  I'm not going to spend much time on it, other than to say that unless you have really high signal to noise and high resolution (both of which the target is trying to reduce), it's a lot harder than you think, and in general you're not going to get there with simple image differencing.  And this problem exists for commless drones whether you're using another drone, a gun, or a death ray to take them down.  Shaheds at least have a very characteristic sound that you can probably use for detection and targeting once they're within audible range.

Detection of the second type (active comm) is easy.  It's transmitting, and transmitting enough to get clear signal back to its operator, who is farther away than you are if it's attacking you.  Triangulation is old technology.  Piece of cake: you lock onto the frequency, have some kind of sensor so you know your own orientation relative to the sensor, and just maneuver in a way to make the signal from the drone stronger until you hit it and destroy it with whatever mechanism you prefer.  Or have a few sensors that are networked to give you the position (helloooo MLAT) and shoot it with your favorite method of action-at-a-distance.

Except for one problem: whose drone did you just destroy?

In the Ukraine environment, IFF is the hard part of doing radio based anti-drone systems.  There are tons of things flying around, as evidenced by the daily releases of yet another view of every bit of ground combat we ever see.  It's not quite Diamond Age concentrations of them, but they're working on it.  And they're all sorts of random drones, including commercial drones, custom drones made with commercial off the shelf parts, custom drones with a mix of commercial and special mil parts, totally custom mil drones, and who knows what else. And they're all using similar frequencies, because the combination of physics and the atmosphere force you to the same frequencies if you want a particular range and data rate at powers that you can reasonably supply to both the ground operator and drone with batteries.  If you don't sort out the IFF thing and you set an autonomous anti-radiation based anti-drone system loose, it's just as likely to attack its allied drones as the enemy drones, because it has no way to tell them apart.  That means you have to have your complete drone ecosystem integrated (ring that cash register over at Lockheed/Northrop Grumman/Raytheon!!) or you're just going to be attacking your own stuff.  

And part of why we aren't seeing even rudimentary versions of it in Ukraine is that it's not a function that people were already spending much effort on for commercial/hobbyist drones. You can't just pop over to Robotshop.com or Alibaba and order tunable RF sensor kits (or a few thousand of them) the way you can other types of sensor, or actuators for operating your 3D printed grenade dropper.  It's possible to get relatively inexpensive software-defined radio modules that are small (that's what feeds ADSBExchange so you can see who's flying around Ukraine), but the environment is so variable, along with the need to confirm what drone you're attacking, that at least for now you're going to need a human in the loop, even if you can semi-automate your remote control drone sensor.  And even with a human in the loop, nobody is painting national flags on their drones, so unless you know "this is one that our side makes" after you get up close to it (assuming you're doing that, rather than sending a death ray at it from 5 km), you really don't know who you're shooting down.  So the basic tech isn't all that hard, but because it's not just point and shoot or point and drop, it's a lot more dependent on integration of the whole system to be usable.

*geese, like all waterfowl, are incredibly mean and probably deserve to die. That's why there's a book entitled "Ducks and how to make them pay".  If we can do an autonomous system for drones, it should probably be immediately applied to geese and ducks.

First of all this was really interesting and clearly-written. Thank you.

Secondly, while solving the IFF issue will clearly confer an advantage, I’m not sure it’s necessary for early-generation drone fleets.  Instead I imagine a world where a fleet of Anti-Drone Drones (ADDs) is released to ‘purge’ the sky over a battlefield at a set time on a set date, designed to catch as many enemy drones in action as possible.  Orders to ground all friendly drones will ensure that the vast majority of ‘kills’ are of enemy drones.  All the ADDs need to be able to do is to tell ‘an ADD’ (so one specific image) from ‘not an ADD’ (nADDs?). 
This could obviously lead to all sorts of efforts to spoof the enemy into deploying their observation/FPV drones en mass so that you get a solid opportunity to “hit them in the nADDs” and hopefully achieve a short period of drone superiority for your own forces.

In that world IFF lends a significant advantage, enabling you to go for full drone supremacy over the battlefield, but it’s not absolutely necessary.