Drones have changed war. Miniature, low-cost and deadly robots buzz in the skies high above battlefields around the world, taking photos and dropping explosives. They are complex to counteract. ZeroMark, a US-based defense startup, thinks it has a solution. I want rotate the rifles front-line soldiers in “handy iron domes”.
The idea is basic: make it easier to shoot a drone out of the sky with a bullet. The problem is that drones are brisk and maneuverable, making them complex to hit even for a skilled shooter. The ZeroMark system would add aim assist to existing rifles, supposedly helping soldiers place the bullet in the right place.
“We’re mainly in the software business,” ZeroMark CEO Joel Anderson tells WIRED. He says this works by placing the sensor on a rail on the front of the rifle, the same place you would place a scope. The sensor works with an actuator located in the stock or front grip of the rifle, which regulates the soldier’s aim when he aims the rifle at a target.
A soldier attacked by a drone aimed his rifle at the target, activated the system, and allowed the actuators to stabilize the target before pressing the trigger. “So there is machine perception, a computer vision component. We use lidar and electro-optical sensors to detect drones, classify them and determine what they are doing,” Anderson says. “Some of the ballistics is actually quite trivial… It’s numerical regression, it’s ballistic physics.”
According to Anderson, the ZeroMarks system can do things that humans cannot. “To be able to calculate things like bullet drop, trajectory and wind… For a human it’s very difficult, but for a computer it’s quite easy,” he says. “So we determined in advance where the shot had to land so that when you pressed the trigger, there was a high probability that it would cross the path of the drone.”
ZeroMark presents an enticing idea – so attractive that venture capital firm Andreesen Horowitz invested $7 million in the project. The reasons are obvious to anyone who pays attention to contemporary warfare. Inexpensive and deadly flying robots define the conflict between Russia and Ukraine. Each month, both sides send thousands of compact drones that drop explosives, take photos and spread propaganda.
As militaries around the world look for a way to fight back, anti-drone systems are a growing industry. There are hundreds of solutions out there, and many of them are not worth the PowerPoint slide from which they are presented.
Could a machine learning-based aiming assist system like the one ZeroMark throws out there work? We will see. According to Anderson, ZeroMark is nowhere on the battlefield, but the company has “partners in Ukraine who are conducting assessments. We hope to change this by the end of the summer.”
There is good reason to be skeptical. “I would like to see a demonstration. If it works, show us. Until this happens, many question marks remain around such technology” – Arthur Holland Michel, anti-drone expert AND older guy at the Carnegie Council on Ethics in International Affairs, he tells WIRED. “This raises the issue of the inherent unpredictability and fragility of machine learning-based systems that are trained on data that is, at best, only a small fraction of what the system may encounter in the field.”