4 ms·
And that’s ok if it’s failing to do the job as intended, learning is acquired, and it looks fun to build, I am in the field and I find it great homemade concept
by tamimio 8mo ago
And that’s ok if it’s failing to do the job as intended, learning is acquired, and it looks fun to build, I am in the field and I find it great homemade concept.
Realistically, I doubt there’s ANY system out there will be able to counter small weaponized drones that are flown manually let alone with AI, you might have some workarounds, but never a real counter.
- tzury 8mo ago"let alone with AI" what's falling into the AI category here perhaps is the key question, since microseconds counts, and LLM are very slow! Even the fastest "real-time" LLM frameworks currently report sub-second latencies around 120ms. This is fine for high-level mission planning (e.g., "fly to the red house") but too slow to prevent a drone from hitting a tree at 50mph (80 KM/h)[1] Whilst the Shahed-136 kamikaze drone typically flies at a maximum speed of around 185 km/h (roughly 115 mph or 100 knots). [1] https://arxiv.org/html/2602.19534v1 https://arxiv.org/html/2602.19534v1 [2] https://en.wikipedia.org/wiki/HESA_Shahed_136 https://en.wikipedia.org/wiki/HESA_Shahed_136
- ninjagoo 8mo ago> "let alone with AI" what's falling into the AI category here perhaps is the key question, since microseconds counts, and LLM are very slow! LLMs (Large Language Models) are far from the only type of AI around. It's a pretty broad field, and there are real-time AI systems, for example, self-driving cars, which have the response times you're thinking of. [1] [1] https://en.wikipedia.org/wiki/Artificial_Intelligence:_A_Modern_Approach https://en.wikipedia.org/wiki/Artificial_Intelligence:_A_Mod...
- tzury 8mo agohence: "what's falling into the AI category here perhaps is the key question" YOLO is a good example for something that can work. https://docs.ultralytics.com/models/yolov8/ https://docs.ultralytics.com/models/yolov8/
- lukan 8mo ago"Realistically, I doubt there’s ANY system out there will be able to counter small weaponized drones that are flown manually " Why would lasers not work? Those cheap drones are made from plastic, if you have a laser powerful enough and a target guidance system (like a camera and a PI) - then you would just need enough of them.
- condensedcrab 8mo agoAt long distances the small cross section of the drone requires tight focusing (expensive optics) or a high power, preferably pulsed laser (expensive laser) or both. Not impossible but many times more expensive than the drone
- tjoff 8mo agoExpensive is fine since it is reusable.
- rdtsc 8mo agoAt some point it itself becomes a target. It has to be able to get almost 100% kills, otherwise the enemy can swarm it with cheap drones, destroy the expensive installation, then continue as before.
- tjoff 8mo agoSure, but it needs many "many times" for that to be a factor. And even in the case it could be useful as an addition to or paired with a tank etc.
- rdtsc 8mo agoMany times more is about what it comes out to. There are some companies selling laser defense systems but they are many times more than cheap FPV drones with grenades attached.
- tamimio 8mo ago
- skybrian 8mo agoI don’t know if it will work, but here’s a startup that seems to be building an AI-controlled shotgun: https://www.ycombinator.com/companies/9-mothers-corporation https://www.ycombinator.com/companies/9-mothers-corporation Given the war in Ukraine, wanting to build such things is certainly understandable. But still, this is the stuff of nightmares.
- wombatpm 8mo agoDavid Suaez in Kill Decision had swarms of small single shot drones with the targeting intelligence of ‘00 camera. Identify a face, fly towards it, fire when close. It was an implementation of quantity has a quality all its own.
- danmaz74 8mo ago> Realistically, I doubt there’s ANY system out there will be able to counter small weaponized drones that are flown manually let alone with AI What kind of systems are you thinking about? Jet airplanes for sure are completely safe from small drones.
- chrisweekly 8mo ago"Jet airplanes for sure are completely safe from small drones." That feels like a bold and unsupported assertion. Ask a pilot how they'd feel about takeoffs or landings through airspace filled with adversarial drones.
- fragmede 8mo agomaybe in the air, but I seem to recall the Ukrainians being successful at attacking Russian planes on the ground.
- rdtsc 8mo ago> Jet airplanes for sure are completely safe from small drones. Until they land then, due to their cost, they become a very juicy target to aim for.
- 15155 8mo agoA flock of unlucky geese can knock out a jet turbine, how is this a "for sure" conclusion?
- ninjagoo 8mo ago> Realistically, I doubt there’s ANY system out there will be able to counter small weaponized drones that are flown manually let alone with AI, you might have some workarounds, but never a real counter. Weaponized drones (say D_A) can be countered by other weaponized drones (say D_B), equally cheap or cheaper than D_A because the D_A is usually targeting something larger (so more payload) and typically has a longer range. D_B only needs to wreck D_A at a shorter defensive range. That's what Ukraine is doing. You can also use drone swarms with coordinated action so that each drone in the swarm is only targeting one other drone, and automatic re-targeting if one node misses. [1] [1] https://en.wikipedia.org/wiki/Swarm_robotics https://en.wikipedia.org/wiki/Swarm_robotics
- maratc 8mo ago> equally cheap or cheaper I doubt it, as D_A's target is stationary (and could be reduced to GPS coords) while D_B's target is moving.
- ninjagoo 8mo ago> I doubt it, as D_A's target is stationary (and could be reduced to GPS coords) while D_B's target is moving. It's a good point, though I should point out that GPS denial is assumed in those sort of contexts as a first countermeasure so D_A likely has alternative targeting, and that smaller drones can move faster with less energy storage, which itself requires less weight, compounding the benefits of being smaller.
- maratc 7mo agoI think you underestimate that. D_A only needs to get to coordinates (X, Y) with a minimal requirements for Z as "some meters above ground." It doesn't matter if it gets there in 15 minutes or in 20 minutes, or in 2 hours. D_B needs to absolutely nail X, Y, Z -- but also t! To be at the time t0 at the point (X0, Y0, Z0), the defence drone needs to be at some other point (X1, Y1, Z1) at the time t1 -- given the speed and the direction of the attack drone. But what the defence drone has is just an estimation of these. The amount of back-propagation calculations for the defence drone is simply immense.
- torginus 8mo agoFrom what I can tell, Ukrainians are having some success with converting guns into automatic turrets that can track and shoot down drones via sensors, and the rifle-equivalent of birdshot.