Robot Soldiers, Neural Networks: How Machine Vision Is Changing Warfare

By John Oncea, Editor

Modern warfare is constantly changing with numerous technologies, including machine vision, both driving and being used to keep up with it.

The defense industry is increasingly integrating machine vision into various applications to enhance surveillance, improve efficiency, and ensure manufacturing quality. The technology is also playing a role in robotics, defect detection, and even training simulations. 

The results have been disruptive with machine vision-enabled systems able to identify intruders, track them across multiple cameras, and reduce false alarms. Drones equipped with machine vision are deployed to scan terrain for snipers or enemy equipment, allowing for remote monitoring and control.

Beyond that, robots with machine vision are being used for tasks like sorting and inventory management in military logistics, as well as to detect defects in ordnance manufacturing, ensuring the quality of weapons and munitions.

According to Cogent Infotech, machine vision is “an offshoot of the relentless advancements in the realm of artificial intelligence, possessing the potent ability to draw profound insights from an intricate array of visual data, thereby marking a new epoch in the defense industry.”

Cogent goes on to share 10 defense industry use cases of machine vision, from enhancing situational awareness on the battlefield to revolutionizing training programs. Here, we focus on two ways Ukraine is using machine vision in its war with Russia: to overcome the jamming and spoofing of its drones and the creation of a robot army.

How To Beat Jammers At Their Own Game

Jamming and spoofing are, according to IEEE Spectrum, the only effective way to defend against drone attacks. This is proven repeatedly, most recently by the Russian army’s use of the two defenses against a myriad of drones launched by Ukrainian forces.

This cat-and-mouse game is taking place at a breakneck pace, with news drone technology made obsolete in as few as three months. Case in point? KrattWorks, an Estonian manufacturer of jamming-resistant high-performance drones, dispatched its Ghost Dragon ISR quadcopters in mid-2022, thinking they’d function for up to six months only to see Russian forces jamming and spoofing make them obsolete in half that time.

Recently, Ukraine took the upper hand over Russia’s tens of thousands of jammers by deploying the third generation of Ghost Dragon, one enabled with “a neural-network-driven optical navigation system, which allows the drone to continue its mission even when all radio and satellite-navigation links are jammed.”

According to Spectrum IEEE, Ghost Dragon’s “original command-and-control-band radio was quickly replaced with a smart frequency-hopping system that constantly scans the available spectrum, looking for bands that aren’t jammed. It allows operators to switch among six radio-frequency bands to maintain control and also send back video even in the face of hostile jamming.”

The drone is equipped with a dual-band satellite navigation receiver that can switch between four main satellite positioning systems: GPS, Galileo, China’s BeiDou, and Russia’s GLONASS. To enhance its security, it features a spoof-proof algorithm that compares satellite navigation data with information from onboard sensors. This system protects against sophisticated spoofing attacks, which attempt to mislead drones into thinking they are flying at a much higher altitude than they are, potentially leading to self-destruction.

At the core of the drone is a machine-vision-enabled computer powered by a 1-gigahertz Arm processor. This technology gives Ghost Dragon drones the ability to navigate autonomously, even without access to any global navigation satellite system (GNSS).

To achieve this, the computer runs a neural network that, much like an experienced traveler, compares views of landmarks with their locations on a map to determine their position. More specifically, the drone uses real-time images from a downward-facing optical camera and compares them against stored satellite images to accurately assess its location.

“Even if it gets lost, it can recognize some patterns, like crossroads, and update its position,” says KrattWorks cofounder and CEO Martin Karmin. “It can make its own decisions, somewhat, either to return home or to fly through the jamming bubble until it can reestablish the GNSS link again.”

An Army Of None?

While drone warfare dominates the headlines, machine-learning technology is becoming more commonplace on the ground as well. Just last month, Ukraine announced it was planning to field 15,000 Uncrewed Ground Vehicles (UGVs) in an attempt to make up for its shortage of personnel.

According to Forbes, “There is certainly plenty of optimism about military robotics, and no shortage of Ukrainian UGV designs, with developers unveiling a new model every week or so. These are wheeled and tracked machines of various shapes and sizes, with roles from minelaying and mine clearing to logistics, casualty evacuation, and direct combat either with explosive payloads or machine guns.”

While 50 types of UGVs have been approved, only 10 to 15 are in regular service due to the challenges that come with them. “Imagine bringing a 1,000-kilogram machine to the front line,” says Kateryna Bondar, Fellow at Wadhwani AI Center, at the think tank Center for Strategic and International Studies (CSIS). “It’s a logistics issue. If it runs on fuel, you have to transport that as well; if it runs on batteries, they will be big and heavy and require generators to recharge.”

These factors, coupled with the cost, limit the use of UGVs to tasks that drones can’t do, such as operating on the ground near the front lines. “Logistics is an interesting use case,” says Bondar. “UGVs are now conducting way more missions and replacing people. Using a UGV removes a person from physical risk.”

In addition to logistics, UGVs are being considered as a way to evacuate the wounded, though most leaders don’t yet trust them to transport soldiers. Still, “Ukrainian news sources described a UGV evacuation last month, in which three wounded soldiers were transported more than 10 miles through an area covered by Russian mortar and artillery fire,” Forbes writes.

Officials remain optimistic as to the value UGVs can provide but acknowledge improvements need to be made. For instance, while the rescue operation referred to earlier was a success, it required more than 50 people to conduct.

“The future is all about autonomy and AI,” says Bondar.

For starters, using smart software would allow the UGV to navigate for itself, using imagery from an autonomous drone flying overhead. In this scenario the operator acts as a mission commander, ordering the UGV where to go and selecting and approving targets for the weapons – an achievement that will be harder to attain than autonomous flight.

“Even in civilian contexts, autonomous ground navigation remains unsolved; self-driving cars still struggle with edge cases on paved roads despite years of investment and defined traffic rules,” says Bondar. “The system must be able to perceive its environment in real time, make context-aware decisions, avoid obstacles, and control the vehicle’s complex mechanical systems – all under conditions of GPS denial, degraded comms, and electronic warfare.

“While some promising prototypes exist in research labs, widespread battlefield deployment – especially where a single human serves as a mission commander for multiple UGVs – will likely take several more years to mature, even in high-urgency environments like Ukraine.”