The Technology for Fully Autonomous Attack Is Already Here | Alex Liannyi, NORDA Dynamics

0:00What I'm really interested in is how artificial intelligence is being introduced to the drone operations. Do your modules have any artificial intelligence in them? Artificial intelligence is very broad term. It's very speculative. In our case, our last mile targeting is based on classical algorithms. It is mostly math and troll theory and so on. How many drones can you launch at a time right now? I mean, how many are you? One of the systems we use, it allows connection from one pilot to 32 drones simultaneously. And as I'm being deployed in all military organizations, you need to prove that you've hit the target.

0:38You need to monitor what has happened with the drone, whether it engaged in a proper target or you can engage in a false positive scenario. And if you can monitor them, you can control them. Technology is close, but I don't think it is completely necessary to use it. How close do you think the technology is to moving the human earlier in the process to the launch of the drone? Once the drone's up, it searches, acquires and attacks. Technology is very close to that. My name is Alex and I'm CTO of Nordic Dynamics. Before the war, I was a regular founder of a tiny IT business. I was doing software development stuff with a tiny team of 15 developers.

1:26When the war started, we started to search for ways to help. Tried all kinds of different things, from volunteering to developing drones. Not successful. And at some point of time, I focused mostly on specific parts of drones, mostly avionics. I started developing different algorithms and at some point of time met other guys, Bazar, Dima and Roman actually, and we founded Norda. And when you said that your initial efforts at making drones were not successful, what was the problem? Not really a problem because there were other guys who already done a huge amount of work developing fixed-wing drones and developing quadcopter drones.

2:25and it was just more efficient to help them in things they're not done yet. Right. That's it. So there was no reason to compete with them. There was a reason. It was much more efficient just to collaborate with them. Yeah. And when did you start working on drones? When did all of this drone work start? Actually, first efforts I think started like mid-2022. Yeah. Like some research and so on, but work on avionics and pre-Norda stuff started in 2023. And the avionics you're talking about navigation as well as last mile target? Yeah, exactly. It started with last mile targeting, researching on flight controllers and so on.

3:24What is the problem, first of all, with navigation and second of all, with last mile? First of all, the first problem is the absence of GPS navigation. It's the first issue which appears on all drones in the Ukraine war. And the second problem is direct radio control. Distance, earth's curvature, trees, any obstacles, they introduce issues with radio signal. This is why drones don't usually don't have... pilots don't have connection to the drone on last 300 meters, 400 meters, depending on distance. When you fly further, you lose connection earlier before the ground. And that connection is because of jamming, jamming of the signal?

4:22There is such a thing as a radio horizon. Because of Earth's curvature, the further you go, the higher need to be to get line of sight connection. When you lose altitude, you lose your connection with the drone. Right. And the distance before the ground where you lose your connection, it increases with the distance you fly. In the beginning, the drone range is much shorter, and so pilots were piloting all the way to the target. Yeah, it works on a shorter distance, but it still won't work if you have a different ground altitude. Basically, especially on Donbass, you can have like between pilot and drone you can have this uh artificial mountain of uh like a berm yeah yeah i just don't know the proper word so first of all the navigation a lot of people are now using starlink for large drones but for uh these smaller drones either small fixed ring drones or quadcopter drones, there isn't enough.

5:41You can't fit a Starlink panel on it. Sure. Yeah. I mean, what I'm really interested in is how artificial intelligence is being introduced to the drone operations. So do your modules have any artificial intelligence in them? Artificial intelligence is a very broad term.

6:09It's very speculative. In our case, our last maltargeting is based on classical algorithms. It is mostly math, control theory and so on. It doesn't have any fancy AI. It has some neural networks which can be, I don't know, someone can name them AI. Their usage is quite isolated. Drone does not take any, it does not do anything without pilot approval. Basically, pilot selects target, drone does its work afterwards. We have some developments on target detection. A drone can search for the target itself, but still it won't engage without pilot approval. Yeah, this is something that I'm really interested in because, I mean, currently do you have object detection on the drones?

7:11On this, you've seen? No, but overall, yeah. There are drones in the field with object detection that are trained to recognize a tank, for example, as opposed to a passenger car. Yeah. And the drone then searches for that. It helps the operator by scanning for that and recognizing it. It highlights like estimated target. And is that done in the cloud or on device? No, it's on device. How large is the model? Where do you get the models for that? Actually, these modules are not big. And these are mostly, as I said before, it's a hobby level modules. Basically, we don't usually use high-end edge devices because it's hard to make proper supply of them.

8:06It is always in short supply. So what kind of chips are you using right now? Actually, we are big fans of Raspberry Pi stuff. It's a hobby level single board computers. We use them a lot. Yeah, although I've been reading the different units are developing Jetson-level chip modules that have higher accuracy in their targeting. Most of them have higher speed. I see. Not accuracy. There's no direct connection between accuracy and chip type. Right. Well, it's just a number of operations you can perform on it. It's it. How widespread is it? And when you say speed, it's a matter of speed. You're talking about the speed of the compute?

9:01Yeah. Empty speed. Empty power. In what conditions is that speed of compute important? Is it how fast the drone is moving? I would say it's a complexity of environment. Yeah. Yeah. And, but from my experience, good engineering usually overcomes cheap limitations. Yeah. Because with bad engineering, you can overload even high-end Jetson processor. Right. So it still will perform very poorly. Yeah. Actually, when we started developing our system, like a few years, two years ago, or something like that. People were making fun of us using Raspberry Pi 4 because they were saying that it is not possible to run proper computer vision algorithms on that.

9:52And after that, we started using even lower. We started using Raspberry Pi 0.2. It's a smaller thing, which is much less performance than Raspberry Pi 4. So it still runs in real time. On the navigation, so there are two parts, right? navigation and target acquisition. Actually, it's mostly target acquisition and control. Navigation part in classical terms, when we know exactly our location, is not introduced yet. These are not loitering drones that you're using right now. Because my understanding, what's coming are drones that you can geofence an area. Yeah, yeah. And you can say to the drone, you can give the drone instructions.

10:38or the module instructions, look for military vehicles in this geofenced area. If you see a military target, lock on it, and then what? The operator has to give permission for it to attack? Yeah, yeah. Actually, we are working on such kind of scenario, actually very close to what you described. And yeah, it is important that the operator approves the target. Basically, we hard code this condition that operator needs to select with already. That last switch that the operator has to toggle to allow the drone to attack. I mean, one of the things I'm interested in at some point, either Russia or Ukraine or some other state will decide that the human in the loop happens when you launch the drone.

11:44And once the drone is launched, if it's restricted to a geofenced area, and that area you're confident is an enemy line, then you could allow the drone to go ahead and hit the target autonomously. And I know there's a lot of discussion about international human rights law. And right now, everyone's very careful not to take that step. But at some point, someone's going to take that step. How close do you think the technology is to moving the human earlier in the process, to the launch of the drone? Once the drone's up, it searches, acquires, and attacks. I think people are very close, and technology is very close to that.

12:41I'm not sure if it's going to be an efficient way to use the drone, but overall, the technology is very close. And not efficient because the models are not yet accurate enough? Yeah, it's mostly about accuracy, about accuracy of positioning, of how the navigation part will perform in different conditions. Another important thing, you need to have proof of, in all military organizations, you need to prove that you had hit the target. Otherwise, you just used your resources, you used 10 drones for what. and in this case you need to monitor what had happened with drone whether it whether it engaged in a proper target or it or it engaged in a false positive scenario right so in this case you need to monitor them and if you can monitor them you can control them and if you can control it why need a complete antenna yeah so in my opinion technology is close but i don't think it is completely necessary to do to use it right although part of the problem uh i mean certainly there's a supply problem but uh one operator uh can really only keep his eye on one or two drones.

14:09But if you had that autonomy, you could launch a thousand drones in a geofenced area and the operator could watch. Yeah, exactly. That is why the drone can notify you when it sees a target and then you can pull up the picture and approve it. So you still can control lots of drones simultaneously but still be in a loop. Yeah. And so how many drones can you launch at a time right now? It's not part of our system to do like a swarming part. Right. But it mostly depends on a radio connection system used. Right. One of the systems we use, it allows connection from one pilot to 32 drones simultaneously.

15:05to how many? 32. 32, wow. And is that being deployed? As far as I know, some squads using. The targeting module, so you have a Raspberry Pi on device, you have a small model on the Raspberry Pi that has object recognition or what is the targeting? We have different options. The one which is mostly used doesn't have any models at all. It runs regular math algorithms on it, so nothing fancy as a model. We have other and it doesn't have any object recognition. It just locks on a specific target. We have a higher version which has object recognition and it runs a model. And the drones I'm seeing that you've been showing me, do those have object recognition or you're just the pilots doing the locking?

16:16No, the one we've shown there just have a lock on a target. So pilot selects an object, a vehicle, or a place in the field that our pilot chooses, and we just targeting the drone there. Right. And then once the target is acquired, whether it's through object recognition or by the pilot locking on a target, the module takes over control of the drone. and so these drones currently have a range, like the fixed wing you were telling me is 30 to 40 kilometers, is that right? No, 30 to 40 is a with a higher payload. That's right, yeah. I would say it's up to 80 kilometers of these electrical fixed wings, up to 80.

17:07And how long is the target, I mean, how far out can you lose radio contact and have the drone take over? Usually our distance of terminal guidance is like up to one kilometer. It depends on the situation because like when you have fog or lower clouds or just the sun is low or something like that, it may vary. But overall, it's up to kilometers. And is that getting longer or is that a pretty much a fixed parameter? No, we can do better. Yeah. But by using different optics, by using gyro-stabilized cameras, we can increase distance, I think up to two and a half kilometers. And the drones themselves, is Norda supplying the drones with the module built in?

18:13No, we supply modules or we supply software, either this or that. We don't do drones. Basically, we collaborate with other drone manufacturers in Ukraine. It's much more efficient because we can focus on research and develop part of the stuff and people on all manufacturing do their job. And how many manufacturers are there today? Like we are collaborating with? Or first of all in Ukraine? Oh, I know, hundreds. Yeah. Hundreds of manufacturing manufacturers and we are in contact and and collaborating in one way or another is like about 40. 40 or 14? 40. 40, wow. Yeah.