I wonder how these things work when the target is still and camouflaged. I’m sure adding infrared would be trivial but infrared can be foiled by a tarp, I’m sure we could figure that out, a tarp with camouflage on one side seems pretty easy
I guess it is like most things. There is always a work around but what is the cost of said work around. Even if you can camouflage yourself from things like this, you still take on additional burden of having to address this.
They’re light weight optic models. You’re right that infrared and other wavelengths are trivial to train on but the compute becomes expensive and the combined battery + chips become very heavy for these drones that also need to carry an explosive payload AND remain nimble to function. Flying through a forest means rapidly adjusting course constantly for branches and whatnot.
We’ve had the technology to do this for over a decade. The reason we haven’t seen these in war yet is because they don’t work especially well and there are cheaper + better options.
I suspect that as SOC performance continues to improve, the power demands and AI complexity will eventually over come this. Yes, your phone can play subway surfers at 8K but that same chip could also end up on the receiving end of a drone.
I mean a Snapdragon X2 can give a lot of image processing performance for about 20watts of power draw. That same work load would have been 200-1000 watts a decade back, and power draw is still being reduced with larger ASIC designs.
I wonder how these things work when the target is still and camouflaged. I’m sure adding infrared would be trivial but infrared can be foiled by a tarp, I’m sure we could figure that out, a tarp with camouflage on one side seems pretty easy
I guess it is like most things. There is always a work around but what is the cost of said work around. Even if you can camouflage yourself from things like this, you still take on additional burden of having to address this.
They’re light weight optic models. You’re right that infrared and other wavelengths are trivial to train on but the compute becomes expensive and the combined battery + chips become very heavy for these drones that also need to carry an explosive payload AND remain nimble to function. Flying through a forest means rapidly adjusting course constantly for branches and whatnot.
We’ve had the technology to do this for over a decade. The reason we haven’t seen these in war yet is because they don’t work especially well and there are cheaper + better options.
I suspect that as SOC performance continues to improve, the power demands and AI complexity will eventually over come this. Yes, your phone can play subway surfers at 8K but that same chip could also end up on the receiving end of a drone.
I mean a Snapdragon X2 can give a lot of image processing performance for about 20watts of power draw. That same work load would have been 200-1000 watts a decade back, and power draw is still being reduced with larger ASIC designs.