A Skynode S-equipped drone during field testing in Ukraine
The FPV video shows a Ukrainian kamikaze drone closing in on a Russian armored vehicle. Russian radio jamming cuts out the video feed, causing the pilot to lose control. Normally this would end the mission, but video from a second drone shows the AI-enabled FPV guiding it straight to the target and hitting it directly. This is one of the first attacks using a new smart autopilot supplied by US company Auterion, which is able to ignore the jamming and has a higher hit rate than a human pilot.
The Skynode S autopilot was only announced last week, but is already being deployed on the Ukrainian front line. It can lock on from a long distance, making jamming pointless. I’ve seen some combat videos, which for security reasons cannot be made public and come with some caveats. But as far as we can judge, the system is effective. But as Auterion CEO Dr. Lorenz Meyer told me, automatically attacking targets in FPV is just the beginning.
Auterion’s app store allows developers to create their own apps for the Skynode S or download the ones they need. Meier says a terminal guidance app for attack drones is currently the most popular.
“Device Guidance is like MS Office on your laptop,” Meier says. “It’s the bare minimum that everyone expects.”
Open Source Drone Intelligence
Auterion has a long history in the drone world and has been leading the open source movement for over a decade. In 2008, Meier, then at ETH Zurich, began developing autonomous drones. The result was Pixhawk 4 (commonly known as PX4), an open source flight controller that provides drone developers with a flexible set of tools. Today, there are 10,000 PX4 developers worldwide, some of whom are using it on robots and other machines.
Open source software is at the heart of the Blue sUAS initiative for safe American-made drones.
The Defense Department has such trust in Auterion’s open-source software that it has made it central to its Blue sUAS initiative for small U.S.-made military drones, implemented as an alternative to Chinese-made drones that have built-in security risks as standard and are banned by the U.S. military.
Auterion previously developed the Skynode X, a powerful autopilot and mission computer that gives drones autonomy with machine vision, object recognition, visual navigation, a range of third-party apps, etc. The newly released Skynode S offers similar capabilities but at a significantly reduced cost.
“We offer a ‘Ukraine Aid’ price point in Ukraine, where we license our software to the Ukrainian government for free,” Meyat said, “typically at the same price point as an Android smartphone, in the hundreds of dollars.”
This puts the Skynode S in the same class as thermal imaging in that it offers significantly more functionality, albeit at a cost well above the $500 FPV, and the cost is likely to come down as production scales up.
“Skynode S will be produced in the tens of thousands, which will be unprecedented,” Meyer said.
It’s no surprise then that Auterion is also involved in the Department of Defense’s Replicator program, which aims to deploy large numbers of low-cost AI-enabled drones.
100% fatality rate (so far)
The Skynode S is small and affordable enough to be mounted on an FPV drone. … [+]
The guidance system features optical lock-on, meaning the pilot identifies a target and flags it to the autopilot while the drone is well outside jamming range. The drone can then continue through the “jamming bubble” even if the pilot loses contact. Currently, 75 percent of drones are shot down by jamming, according to General Pierre Cyr, chief of staff of the French army.
Meyer said the company’s AI terminal guidance system being used in Ukraine has achieved a 100% hit rate so far, compared with 20% to 40% for human-piloted FPVs. He doesn’t expect to be able to maintain a consistent success rate, but believes they can achieve 90% in the long term.
This means that the biggest cause of error – pilot error and interference – can be eliminated.
This means that with a guidance system like the Auterion S, four times as many drones can pass through and have a higher hit rate due to smart guidance. It’s scary to think that such a simple change could dramatically increase the already high kill rate of an FPV.
While the current version simply guides the drone to a target, future developments will likely feature an aim-point selection feature where the AI would choose where it is most vulnerable — for example, the drone might swing around to attack the rear of a Russian tank’s turret, or an ammunition dump on an SPG, which tends to cause devastating explosions.
Open-source software makes it easier to develop new applications. In June, Auterion’s nonprofit Dronecode hosted a 48-hour hackathon in Krakow for NATO, tasking a team of programmers with developing a system that could visually identify targets and plan a flight path to intercept them. The team used an early version of the Skynode S running AuterionOS, which is similar to the just-released product.
“This was a world first,” Meyer says. “Developers could just focus on building their apps.”
But this technology can do much more.
Strengthening bombers and supporting interceptors
GPS and other satellite navigation have been severely jammed in both Ukraine and Russia, even affecting some military systems, misdirecting U.S.-supplied Excalibur guided artillery shells, JDAM guided bombs, and drones. The usual solution to this problem is expensive inertial navigation modules, which become inaccurate over long ranges and are expensive.
The AI system would also be capable of visual navigation, using the drone’s cameras to find its way around objects on the ground, much like early pilots used imagery from previous reconnaissance flights. Some Ukrainian drones have already done this using AI software called Eagle Eyes.
Long-range attack drones, such as the Ukrainian Bobur, are equipped with sophisticated AI guidance systems, … [+]
Meyer said the Skynode S will have a new twist: terminal guidance. Once the drone reaches the target area, an object recognition app will allow it to visually acquire the target and strike with pinpoint accuracy. In principle, an attacker could locate a specific machine in an oil refinery, hit the center of a radar antenna or fly through a specific window from hundreds of miles away. Or target a tank parked far from the front line that was spotted by satellite.
This setup has already been tested on Ukrainian drones equivalent to Russia’s Lancet and Shahed.
“Testing has demonstrated that it can accurately strike targets,” Meyer said. “The system is expected to be deployed to frontline locations within the next few weeks.”
The Skynode S will also be useful for air-to-air combat, and Meier predicts that an autonomous “dogfight” app will soon be available that will allow an FPV drone to plot an intercept course, outmaneuvering and destroying evasive enemy aircraft without operator input, then exploding if it comes within kill distance.
Ukrainian FPV drones have recently begun intercepting Russian reconnaissance drones at high altitudes, and the Skynode S has the potential to significantly increase the number and success rate of these engagements, while also ensuring that Ukrainian drones can fly faster than Russian interceptors.
AI pilots may also be useful for other tasks – for example, some skilled FPV pilots can perform dive-bombing attacks with kamikaze-like precision without sacrificing the drone – the main limitation seems to be a lack of experienced pilots, but dive-bombing apps may soon be just a download away.
It could also become much easier to use automated drones to lay mines, or to use anti-clogging drones to drop tire-smashing iron balls on specific roads behind enemy lines, or simply to resupply friendly forces.
Overall picture
Looking at the bigger picture, Meyer says Skynode S is already integrated with battlefield management software from the likes of Ukraine’s Delta and Kropyva. These command and control systems collect information from drones, satellites and other sources and compile it into a single picture that commanders on the ground can see. He demonstrates this in a video where an AI-equipped reconnaissance drone finds an object and feeds back data.
The Skynode S hardware is small and affordable, so it can be fitted onto any reconnaissance quadcopter.
While targets are always reviewed by a human operator, AI systems already do much of the work of finding items of interest from video feeds. By eliminating the need for an operator on each drone, it provides more detailed 24/7 coverage of the battlefield in real time. And by only needing to communicate when it spots something, AI-enabled reconnaissance can transmit far more useful information using far less bandwidth.
The AI could also incorporate features such as change detection, automatically comparing a scene to a previous one to find new minefields, trenches, vehicle tracks, or camouflaged locations.
Putting AI at the edge has the potential to dramatically accelerate the process of collecting, processing, and disseminating battlefield information. Thousands of smart, autonomous reconnaissance drones could eventually become more important in combat than terminal guidance.
It’s called Skynode, not Skynet.
Skynode S also supports drone swarm control, allowing one operator to control multiple drones, so one operator can arm multiple scout drones and lots of FPVs, use the scouts to find targets, and then turn and attack with the FPVs.
While this may sound like the entire kill chain could be automated, the key point here is that there is always a human in the loop overseeing the operation, even if they no longer need to be in the loop.
Meyer emphasizes that the Skynode S isn’t about creating an autonomous killing machine. He compares the level of automation to how guided missiles in use today are operated: An FPV operator locks on to a target, just like a Javelin missile operator. Both are “fire-and-forget” weapons that aim at targets pre-designated by a human.
“We have serious concerns about being more autonomous,” Meyer said, “but there’s still a lot that can be done without allowing the drone to autonomously select targets.”
It’ll take some work to fully explore the new capabilities Skynode S has to offer, but with a large team of software developers, low-cost hardware, and a plentiful supply of matching drones — and a war to win — things are likely to happen soon.
