The Nova drone uses AI to help ground troops look inside buildings.

The Nova drone uses AI to help ground troops look inside buildings. Artificial Intelligence company Shield AI

2021 Is the Year the Small Drone Arms Race Heats Up

The cat-and-mouse of drone defense and offense is entering a new phase.

As drones become smarter, cheaper, more nimble, easier for rogue adversaries to acquire and more advanced adversaries to evolve, they pose a unique threat for the U.S. military that grows in importance as the objects themselves diminish in size. This year, trends in autonomy will reshape drone capabilities and concepts, making them more offensively useful and even harder to defend against.

“Drones and most likely drone swarms are something you’re going to see on a future battlefield...I think we’re already seeing some of it,” said Army Gen. John Murray, who leads Army Futures Command. “Counter drone, we’re working the same path everybody else is working in terms of soft skills and hard kills via a variety of different weapons systems. It just becomes very hard when you start talking about swarms of small drones. Not impossible but harder.”

The U.S. military plans to spend $83 million this year to buy lasers, electromagnetic devices, and other means to take down small drones. By year’s end, the destroyer Preble will get a 60-kilowatt laser and an optical dazzler, while the Air Force will deploy a Tactical High Power Microwave Operational Responder, or THOR. But the Pentagon will spend $404 million — almost four times as much — to develop new anti-drone defenses, the Congressional Research Service reported Jan. 11.

Future counter drone efforts will be coordinated by the year-old Joint Counter Small Unmanned Aerial System Office, or JCO, which released its first strategy document on Jan. 7. The office was established after individual services had spent “a couple billion dollars” to develop and deploy counter-drone tech, Army Maj. Gen. Sean Gainey, who leads the 60-person JCO, told a CSIS audience recently. 

Such efforts managed to field a few systems, like the Marine Air Defense Integrated System that the Navy used in July 2019 to down an Iranian drone. (The system was mounted on a truck on the deck of the USS Boxer.) But the services’ hurried, disorganized efforts produced “several redundant systems” and “not all of it worked as advertised,” Gainey said. Even the best and promising solutions couldn’t meet their fullest potential in such an environment. “We never followed up” on maturing the technology that worked, he said. He said the JCO’s “enterprise” approach should help to fix that, allowing a much more organized development of counterdrone tech that’s better matched to current intelligence and technology trends.

The real problem will be staying ahead of these trends. The JCO’s new strategy looks ahead to an era when commercial drones will fill the skies over cities, and defenders will have to spot the ones that are acting strangely. The military must aim to “adopt a posture of anomaly detection by seeking ways to highlight abnormal behavior,” the strategy says. In the U.S., at least, this will eventually be aided by the Federal Aviation Administration’s efforts to build a next-gen Aircraft System Traffic Management System. But that is years away, and in any case, doesn’t apply overseas. “Until they are implemented, the burden of tactically detecting and identifying anomalous systems in the vicinity of U.S. forces and facilities remains the responsibility of installation commanders,” it reads. 

Jamming drones, or even blasting them out of the sky, might work fine over the strait of Hormuz or the desert sands of Syria, but it’s a tricker proposition in the cities where the military expects to fight. Just finding them is a big problem. Small drones are often too small for radar, too cool for thermal sensors, and too soft for sound detectors. 

One promising approach combines detection and defense: hijacking radio control signals.

“With our system, we surgically take control of the connection between the remote control operator and the drone and essentially hijack that session. Then we own the drone,” said Josh Montoya, a pre-sales engineer with the Israeli company D-Fend Solutions. “The drone takes commands from our system. What we tell the drone to do is take a safe route from where you’re at and get out of the area, land, in a location we determine you can land safely.”

We were given an exclusive demonstration of the technology at a small farm just outside of Washington, D.C. The system picks up the signal of any drones in the area and assembles a list. Press a button and the drone you were fretting over is now yours to land where you like — and without interfering with other radio-connected devices. 

“That’s kinda what federal law enforcement likes, the border force folks like, the VIP protective detail people like…all of them who have the concern of ‘if the drone crashes, what the collateral impact could be,’” said Montoya. 

But you can’t hijack a radio connection if it doesn’t exist — that is, if a drone can operate autonomously. Smarter drones are the next big challenge for defenders, Gainey said. 

“Where we see the threat going in the future is autonomous,” the JCO commander said.  “Massing swarming capability and integrating AI and potentially leveraging 5G out in the future.” These, he said, “are the areas we’re looking to address.”

Yet autonomous drones are also the next big opportunity for the Army, Murray said.

The Army Futures Command leader highlighted an experiment involving a small drone swarm last September, part of the Army’s fledgling Project Convergence experiment in the Arizona desert.  

“I think we got up to about eight to ten out at Yuma and what we primarily use them for was to extend our mesh network,” he said. “So we were replicating a division headquarters which, by today’s doctrine, has the ability to cover about 25 to 30 kilometers. We extended it out to almost 70 kilometers through the use of aerial mesh networks with our drones.”

Murray said the Army is experimenting with a variety of payloads, but declined to be more specific.

“You can think of it from a non-lethal and lethal perspective. We’re able to swarm right now and we’ll continue to try and expand the number,” he said.

Greater autonomy is going to force militaries — and their civilian masters — to rethink the idea of meaningful human control over weapons like drones, Murray predicted. An incoming drone swarm may be too much for any human to deal with; effective defenses might require firing decisions made by artificial intelligence, with no human in the loop after the initial decision to fight. That, in turn, could have international and policy implications. “I was talking about artificial intelligence, where there might not be a C2 [communications] node in the net. The policy of a human on-the-loop, when you’re defending against a drone swarm, a human may be required to make that first decision but I’m just not sure any human can keep up with a drone swarm, so that’s an area where I think, in the U.S., we can have some conversations going forward in terms of how much human involvement do you actually need when you’re talking about non-lethal decisions from a human standpoint,” he said.

Since 2016, drone maker Shield AI, working with the Defense Innovation Unit, has been providing small drones to special operations with the ability to detect their location and maneuver without GPS signalling. Shield AI co-founder Brandon Tseng compared imbuing drones with autonomy to making a self-driving car, teaching software to measure and make decisions about objects in physical space. “GPS is not reliable in dense urban environments, so the cars have to build their own maps of the world,” Tseng said in a phone interview.

In about two months, Shield AI aims to release an upgraded version of its signature Nova drone with “vision-based autonomy,” a system designed to perform better at night than the current LIDAR sensors.

But the company’s most significant work is less about selling specific drones and more about developing autonomic systems that can work on a wide assortment of devices and weapons. “We’ve actually been doing a lot of the work in the DOD on training fixed-wing aerial vehicles to breach integrated air defense systems,” Tseng  said.

He said the company would demonstrate autonomous behaviors and maneuvers on a drone, perhaps from a different drone maker, sometime this year. 

“Once you have a highly intelligent system, you can start to swarm,” he said.

From there, stopping the drones is someone else’s problem. 

Importantly, the same technology that is enabling more autonomy in small drones has big implications for larger drones and the way the two work together in future battlefields. In October, Shield AI entered into a partnership with large UAV maker Textron. The two are making a “proof-of-concept work to integrate Shield AI technology into Textron Systems’ proven air, land and sea unmanned systems,” according to a release from Textron. 

The slow merging of small and large drones in exercises and in autonomy software is particularly relevant for a military facing an adversary like China. China is a market leader in small consumer drones and they’re rapidly moving into larger types. Earlier this month, the China Aerospace Science and Industry Corporation claimed that the country’s first jet-powered long-endurance UAV had completed its maiden voyage. 

Russia is applying lessons from the decisive use of drones in Azerbaijan to new drones and operating concepts, said Sam Bendett, a research analyst with the Center for Naval Analyses.

“Going forward, the Russian military will obtain multifunctional long-range drones that can carry different types of munitions. The [Ministry of Defense] is developing UAV swarm and loyal wingman tactics; and is working on testing and procuring loitering munitions,” as well as imbuing drones with greater autonomy. (The U.S. military has its own loyal wingman program. In December, the Air Force’s experimental Kratos XQ-58 Valkyrie, took its first flight in formation with other jets.)

Russia already trains military units to counter small drones. They will soon be moving larger drones into the mix. Around September, Russia will work cruise missile and drone defenses into its largest annual military exercise, Zapad. The Russian military is “also starting to train in countering larger, heavier drones - its domestic industry has fielded several targeting models whose flight characteristics approximate larger Western UAVs,” said Bendett. “Previously, Russian armed forces mostly trained in countering smaller UAS, with their own smaller UAVs acted as adversarial assets during training.”