[Goose]

My crystal balls says:

 

There will be a class of small, relatively short range FPV reconnaissance drones. They will use fiber optic cables instead of radio transmissions for control uplink and video downlink. They will be therefore "jam proof". I  am hoping they will be company-level assets and operated by 11Cs. Targets located by these drones can be passed directly to the company mortar section or the Fire Direction Center (FDC) as appropriate. If adequate bandwidth via other links is available the FPV video may be forwarded to battalion. Unlike the Russian and Ukrainian battalion level drone management system a quick kinetic response will be immediately available at the company level. I am guessing they will be mostly winged drones for better efficiency, but that is by no means certain. These drones need to be cheap and easy to operate, as well as plentiful.

 
There will be a class of medium size longer range and longer endurance reconnaissance drones. These drones will be in the same class as the drone shown in [WA-1]. FPV will be optionally configurable on these drones, but if fitted it will also be capable of recording the video and downloading video when not being jammed or after landing. These drones will be a battalion asset and operated by 15Ws. They will be capable of operating over a wide frequency range and utilize signals for both uplink and downlink that are as "jamming resistant as possible for the money available". It may be these drones end up utilizing satellite or airborne relay for both uplink and downlink because that is the only practical approach. The "jam-proof" terrestrial radio solutions may be too expensive, too power-hungry, too large or all of the above.These drones will have waypoints loaded into the navigation system before launch so that they can operate in the absence of any communication at all. The MAY also have some AI capabilities that allow them to pursue high priority reconnaissance opportunities.
 

Both of these drone classes will require development. The US has not been real good at that for quite awhile. These drones should use COTS hardware whenever possible, recognizing that sometimes trade-offs must be made between capability, affordability, and availability. The small drones could conceivably be derived from commercial drones, since no radio system is involved. However, it may be that the "signature" of the commercial drone is such that it is extremely vulnerable to interceptor drones. While the small drones are in theory recoverable problems with the fiber-optic cable ensure a steady attrition. They need to be inexpensive.

 

That takes care of the reconnaissance drones. Now the "fun" begins. The existing kinetic commercial FPV drones suffer from several problems.

1. First and foremost, they can be totally defeated by EW (jamming).
   
2. Unlike both mortars and howitzers FPV drones cannot deliver lots of ordinance quickly, i.e they lack "weight of fire". Each drone carries a relatively small warhead (5 kg or so). At present, they are normally operated as independent single units. Even if launched in a group, no more than 5 to 8 can operate within range of any given controller. Beyond that they fatally interfere with each others video.
   
3. FPV drones are slow compared to other "heavy" weapons. They are normally launched from well behind the line of contact. Going full speed a Mavic 3 will take about two minutes after launch to reach the line of contact. Any distance it must travel beyond that is just added on. If a fiber-optic reconnaissance drone  locates a target, that target has at a minimum several minutes to withdraw or take measures to avoid the subsequent inbound drone attack.
   
4. FPV drones are inherently manpower intensive. The "pilot" has to fly the drone for several minutes before he selects the target and attacks it. During that time the pilot is unavailable for other tasks. This restriction limits the "rounds per minute" a drone operator can deliver.
   

 

Of course, FPV drones have several advantages as well. Going forward, the goal of the new drone designs will be to eliminate/mitigate the disadvantages of current FPV drone designs while preserving their advantages.

 

There will be several classes of "killer" drones. Notionally the classes will be anti-personnel, anti-armor (vehicles) and interceptors (anti-drone drones). These drones will have no FPV or downlink. The size, weight and power (SWAP) made available by these features being omitted will be used for target identification sensors, hardware and software. This design eliminates problems 1, 2 and 4 completely. These drones can't be jammed and do not require a "pilot". They also can't interfere with each other, so as many as desired can be launched as fast as physically possible. This allows drones to deliver a very large kinetic effect at almost a single point in time, a very important capability they previously lacked. These drones are not limited by mutual interference or the number of highly-trained specialized operators available. They can be launched at will right on the line of contact. This mitigates problem 3. Interestingly [WA-6]  shows the Russians recognize this problem and are trying to mitigate it using current technology. The FPV requirement however restricts the number of drones flying at any one time to 5-8.

[WA-7] discusses applying "AI" to drone warfare, as does [JAM-3]. Some drones supposedly already use "AI" to steer the drone when it is on "final approach" to the target and therefore too low to receive the uplink. The operator can't receive the downlink anyway at that point. There are many, many more articles about drones using AI to identify targets. IMO lots of the claims need to be taken with several grains of salt, but there is no question that for many years software has been "trained" to identify and kill targets with no "man in the loop" intervention. It is in a sense "AI". However the term AI is being used to create specious associations with the current social media and Internet publicity. It isn't anything like ChatGPT or other large language model (LLM) AI. It is "simple" AI. The sensors don't need to provide megapixels of data. "Old" multi-spectrum technology as used on e.g. the AARGM works just fine.

 

[WA-8] also has a lot to say about AI targeting of drones. Lots of it is pretty arguable, but it does say the one important thing that will drive the future of armed drones.

The solution is to avoid jamming altogether by relying on AI to guide the drone to its target in the final stages of flight.

The "in the final stages of flight" should simply be deleted. This statement is true wherever there is jamming, which in Ukraine today is almost ubiquitous.

 

Helsing and all the other AI drone companies all are careful to say that there is a "man in the loop" who will make the final decisions about attacking. This is of course nonsense and is said for political correctness only. Jamming means the man in the loop can't see what the drone sees and can't command the drone one way or another. The enemy side of the line of contact is a free-fire zone anyway. [WA-9] is VERY hyperbolic and draws some very questionable inferences BUT it does show clearly that fully autonomous  AI is going to be the future. There is no "man in the loop" on most long range cruise missile strikes. There won't be one on anti-tank drones either.

 

IMO closely customizing the various drone classes to their intended targets is going to be necessary to overcome SWAP and cost issues. The flight dynamics requirements, munitions and target detection schemes used for anti-personnel drones, anti-armor drones and reconnaissance drone interceptors are quite distinct.  Anti-personnel drones probably will function best with two CCD arrays sensitive to different wavelengths of infrared light. Anti-armor drones probably want a W band radar and a CCD array. That might work for drone interceptors as well but the CCD array probably will be for a different color. The warhead of the interceptor (if any) will obviously be different than the anti-armor drone.

 

The anti-personnel drones will have a specialized fragmentation warhead optimally detonated by the multi-spectrum thermal imaging system and its target detection software. The warhead will be optimized for kill radius. The Switchblade 600 is protypical for the anti-personnel drone, without the FPV transmitter and electronics.

[Goose]

My crystal balls says:

 

There will be a class of small, relatively short range FPV reconnaissance drones. They will use fiber optic cables instead of radio transmissions for control uplink and video downlink. They will be therefore "jam proof". I  am hoping they will be company-level assets and operated by 11Cs. Targets located by these drones can be passed directly to the company mortar section or the Fire Direction Center (FDC) as appropriate. If adequate bandwidth via other links is available the FPV video may be forwarded to battalion. Unlike the Russian and Ukrainian battalion level drone management system a quick kinetic response will be immediately available at the company level. I am guessing they will be mostly winged drones for better efficiency, but that is by no means certain. These drones need to be cheap and easy to operate, as well as plentiful.

 
There will be a class of medium size longer range and longer endurance reconnaissance drones. These drones will be in the same class as the drone shown in [WA-1]. FPV will be optionally configurable on these drones, but if fitted it will also be capable of recording the video and downloading video when not being jammed or after landing. These drones will be a battalion asset and operated by 15Ws. They will be capable of operating over a wide frequency range and utilize signals for both uplink and downlink that are as "jamming resistant as possible for the money available". It may be these drones end up utilizing satellite or airborne relay for both uplink and downlink because that is the only practical approach. The "jam-proof" terrestrial radio solutions may be too expensive, too power-hungry, too large or all of the above.These drones will have waypoints loaded into the navigation system before launch so that they can operate in the absence of any communication at all. The MAY also have some AI capabilities that allow them to pursue high priority reconnaissance opportunities.
 

Both of these drone classes will require development. The US has not been real good at that for quite awhile. These drones should use COTS hardware whenever possible, recognizing that sometimes trade-offs must be made between capability, affordability, and availability. The small drones could conceivably be derived from commercial drones, since no radio system is involved. However, it may be that the "signature" of the commercial drone is such that it is extremely vulnerable to interceptor drones. While the small drones are in theory recoverable problems with the fiber-optic cable ensure a steady attrition. They need to be inexpensive.

 

That takes care of the reconnaissance drones. Now the "fun" begins. The existing kinetic commercial FPV drones suffer from several problems.

1. First and foremost, they can be totally defeated by EW (jamming).
   
2. Unlike both mortars and howitzers FPV drones cannot deliver lots of ordinance quickly, i.e they lack "weight of fire". Each drone carries a relatively small warhead (5 kg or so). At present, they are normally operated as independent single units. Even if launched in a group, no more than 5 to 8 can operate within range of any given controller. Beyond that they fatally interfere with each others video.
   
3. FPV drones are slow compared to other "heavy" weapons. They are normally launched from well behind the line of contact. Going full speed a Mavic 3 will take about two minutes after launch to reach the line of contact. Any distance it must travel beyond that is just added on. If a fiber-optic reconnaissance drone  locates a target, that target has at a minimum several minutes to withdraw or take measures to avoid the subsequent inbound drone attack.
   
4. FPV drones are inherently manpower intensive. The "pilot" has to fly the drone for several minutes before he selects the target and attacks it. During that time the pilot is unavailable for other tasks. This restriction limits the "rounds per minute" a drone operator can deliver.
   

 

Of course, FPV drones have several advantages as well. Going forward, the goal of the new drone designs will be to eliminate/mitigate the disadvantages of current FPV drone designs while preserving their advantages.

 

There will be several classes of "killer" drones. Notionally the classes will be anti-personnel, anti-armor (vehicles) and interceptors (anti-drone drones). These drones will have no FPV or downlink. The size, weight and power (SWAP) made available by these features being omitted will be used for target identification sensors, hardware and software. This design eliminates problems 1, 2 and 4 completely. These drones can't be jammed and do not require a "pilot". They also can't interfere with each other, so as many as desired can be launched as fast as physically possible. This allows drones to deliver a very large kinetic effect at almost a single point in time, a very important capability they previously lacked. These drones are not limited by mutual interference or the number of highly-trained specialized operators available. They can be launched at will right on the line of contact. This mitigates problem 3. Interestingly [WA-6]  shows the Russians recognize this problem and are trying to mitigate it using current technology. The FPV requirement however restricts the number of drones flying at any one time to 5-8.

[WA-7] discusses applying "AI" to drone warfare, as does [JAM-3]. Some drones supposedly already use "AI" to steer the drone when it is on "final approach" to the target and therefore too low to receive the uplink. The operator can't receive the downlink anyway at that point. There are many, many more articles about drones using AI to identify targets. IMO lots of the claims need to be taken with several grains of salt, but there is no question that for many years software has been "trained" to identify and kill targets with no "man in the loop" intervention. It is in a sense "AI". However the term AI is being used to create specious associations with the current social media and Internet publicity. It isn't anything like ChatGPT or other large language model (LLM) AI. It is "simple" AI. The sensors don't need to provide megapixels of data. "Old" multi-spectrum technology as used on e.g. the AARGM works just fine.

 

[WA-8] also has a lot to say about AI targeting of drones. Lots of it is pretty arguable, but it does say the one important thing that will drive the future of armed drones.

The solution is to avoid jamming altogether by relying on AI to guide the drone to its target in the final stages of flight.

The "in the final stages of flight" should simply be deleted. This statement is true wherever there is jamming, which in Ukraine today is almost ubiquitous.

 

Helsing and all the other AI drone companies all are careful to say that there is a "man in the loop" who will make the final decisions about attacking. This is of course nonsense and is said for political correctness only. Jamming means the man in the loop can't see what the drone sees and can't command the drone one way or another. The enemy side of the line of contact is a free-fire zone anyway. [WA-9] is VERY hyperbolic and draws some very questionable inferences BUT it does show clearly that fully autonomous  AI is going to be the future. There is no "man in the loop" on most long range cruise missile strikes. There won't be one on anti-tank drones either.

 

IMO closely customizing the various drone classes to their intended targets is going to be necessary to overcome SWAP and cost issues. The flight dynamics requirements, munitions and target detection schemes used for anti-personnel drones, anti-armor drones and reconnaissance drone interceptors are quite distinct.  Anti-personnel drones probably will function best with two CCD arrays sensitive to different wavelengths of infrared light. Anti-armor drones probably want a W band radar and a CCD array. That might work for drone interceptors as well but the CCD array probably will be for a different color. The warhead of the interceptor (if any) will obviously be different than the anti-armor drone.

 

The anti-personnel drones will have a specialized fragmentation warhead optimally detonated by the multi-spectrum thermal imaging system and its target detection software. The warhead will be optimized for kill radius. The Switchblade 600 is protypical for the anti-personnel drone, without the FPV transmitter and electronics.