[Goose]

If this article is accurate, it certainly steps up the value of FPV drones.  

https://www.forbes.com/sites/davidhambli...autopilot/

It is accurate. It is also not really all that "new". The "new" parts are 1) the addition of "AI" allowing a relative novice to describe targets and to some limited extent attack trajectories and 2) delivering these capabilities in a video sensor and electronics package that is small enough and cheap enough to use in vehicles that have a very very limited payload capacity. The US military has been able to do these things for many years, but our drones were (are) larger and can carry a bigger guidance payload. They are much more expensive. They are also programmed by specialists that know how to do it. Open source software is an advantage in Ukraine partly because the US gives nothing away in its use. It is not necessarily the "best". However, it may be. Consider the trajectory of Linux :-).

The Maverick missile had software that could lock on to a "designated" target. It was primitive software and the sensor resolution wasn't great. It works very well. Subsequent platforms do much, much better, as one would expect. The "FPV" video requirement has always been the Achilles Heel of these small attack drones. The video back link can be jammed as (much more easily) can the the command and control uplink. The technology described in the article allows the drone to fly through jamming once the operator designates the target. The enabling assumption is that the "jamming bubble" is small enough the designation by the operator (via FPV) is possible. There is no reason to believe that has to be true. More and better concentrated jamming power defeats that approach. It is already quite possible to launch "jammer" drones that fly a bit higher and blanket an area with RF power. These drones would be recoverable and therefore cost effective. For these reasons the "FPV" feature is inevitably doomed to failure fairly quickly.

The "big deal" here is the software described is in principal capable of allowing a drone to be launched in a "search and destroy" mode where absolutely no up or down link is required. It also doesn't need GPS, as the INS can keep it inside a fuzzy box in enemy territory. Landmark recognition can optionally allow it to "re-calibrate" and do even better. In principle, it carries neither a transmitter nor receiver. You can't jam it no matter how much power you generate. The people launching these drones don't need to know how to fly them or view targets on video. The drones do all that without human aid. There is no inherent reason multiple spectrum cameras can't be used to even further refine the target recognition in the drones. That is the "gold standard" that the Ukrainian war is pushing the world towards. Ukraine isn't there yet, but is getting much closer. Imagine a drone launching platoon saturating an area with a hundred killer drones to "sanitize" an area just before an attack. Even if not all the defenders and their vehicles are eliminated, the fear and chaos will be debilitating.

The above scenario is one reason the US has not gone "all in" on jamming enemy drones. The assumption is that quite soon that will become an irrelevant capability. Lasers that can rapidly kill drones either by damaging their electronics or blinding them is therefore the approach of choice. While AFAIK the US is not heavily pursuing it, anti-drone drones will certainly become a significant requirement. Along with that will be "IFF" capability to prevent various forms of fratricide. This requirement encompasses air to ground, ground to air, and even air to air. It is in many ways a nightmare, but it is happening, like it or not.

[Goose]

If this article is accurate, it certainly steps up the value of FPV drones.  

https://www.forbes.com/sites/davidhambli...autopilot/

It is accurate. It is also not really all that "new". The "new" parts are 1) the addition of "AI" allowing a relative novice to describe targets and to some limited extent attack trajectories and 2) delivering these capabilities in a video sensor and electronics package that is small enough and cheap enough to use in vehicles that have a very very limited payload capacity. The US military has been able to do these things for many years, but our drones were (are) larger and can carry a bigger guidance payload. They are much more expensive. They are also programmed by specialists that know how to do it. Open source software is an advantage in Ukraine partly because the US gives nothing away in its use. It is not necessarily the "best". However, it may be. Consider the trajectory of Linux :-).

The Maverick missile had software that could lock on to a "designated" target. It was primitive software and the sensor resolution wasn't great. It works very well. Subsequent platforms do much, much better, as one would expect. The "FPV" video requirement has always been the Achilles Heel of these small attack drones. The video back link can be jammed as (much more easily) can the the command and control uplink. The technology described in the article allows the drone to fly through jamming once the operator designates the target. The enabling assumption is that the "jamming bubble" is small enough the designation by the operator (via FPV) is possible. There is no reason to believe that has to be true. More and better concentrated jamming power defeats that approach. It is already quite possible to launch "jammer" drones that fly a bit higher and blanket an area with RF power. These drones would be recoverable and therefore cost effective. For these reasons the "FPV" feature is inevitably doomed to failure fairly quickly.

The "big deal" here is the software described is in principal capable of allowing a drone to be launched in a "search and destroy" mode where absolutely no up or down link is required. It also doesn't need GPS, as the INS can keep it inside a fuzzy box in enemy territory. Landmark recognition can optionally allow it to "re-calibrate" and do even better. In principle, it carries neither a transmitter nor receiver. You can't jam it no matter how much power you generate. The people launching these drones don't need to know how to fly them or view targets on video. The drones do all that without human aid. There is no inherent reason multiple spectrum cameras can't be used to even further refine the target recognition in the drones. That is the "gold standard" that the Ukrainian war is pushing the world towards. Ukraine isn't there yet, but is getting much closer. Imagine a drone launching platoon saturating an area with a hundred killer drones to "sanitize" an area just before an attack. Even if not all the defenders and their vehicles are eliminated, the fear and chaos will be debilitating.

The above scenario is one reason the US has not gone "all in" on jamming enemy drones. The assumption is that quite soon that will become an irrelevant capability. Lasers that can rapidly kill drones either by damaging their electronics or blinding them is therefore the approach of choice. While AFAIK the US is not heavily pursuing it, anti-drone drones will certainly become a significant requirement. Along with that will be "IFF" capability to prevent various forms of fratricide. This requirement encompasses air to ground, ground to air, and even air to air. It is in many ways a nightmare, but it is happening, like it or not.