How Drone Swarms Work—From Iran’s Shahed Attack to Ukraine’s Operation Spiderweb

How Drone Swarms Work—From Iran’s Shahed Attack to Ukraine’s Operation Spiderweb

By Deni Ellis Béchard edited by Dean Visser

Six hours after Israel’s air strikes in Iran last Friday, farmers in Iraq could have looked up and seen Iranian drones traveling west: more than 100 of them flew on a 1,700-kilometer journey to Israel, with their propellers buzzing like Weedwackers. Among them was the Shahed-136. Composed mostly of foam and plywood, each Shahed-136 drone is 3.5 meters long and has a 2.5-meter wingspan and a 40- to 50-kilogram warhead at its nose. The drone’s “brain,” a sensor the size of a cough drop, measures every movement while a credit-card-sized GPS onboard listens for microwave chirps from navigation satellites. The Shahed’s route (its waypoints in latitude, longitude and altitude) is uploaded before a booster rocket fires it into the sky. And it is loud: its 50-horsepower motor is slightly more potent than that of a 1960s Volkswagen Beetle and would be as noisy as a lawn mower or a moped at full throttle—now multiplied by 100 in what military strategists sometimes refer to as a rudimentary swarm.

Drone swarms can take different forms. In attacks such as Iran’s recent launch of drones at Israel—or Russia’s use of them against Ukraine, where Shahed drones are nicknamed “flying mopeds”—the swarm’s power is in its numbers. One missile with a similar range can cost upward of $1 million, but a Shahed can be knocked together for $20,000 to $50,000. Iran’s Islamic Revolutionary Guard Corps (IRGC) fires them from portable rails or from racks on trucks, and the small pulse rocket on the bottom of each drone slams it to cruise speed before falling off. The Center for Strategic and International Studies describes such drone salvos as tools “used as much to saturate air defenses as they are to attack targets, cluttering radar screens and forcing command centers to make decisions about where to fire their more capable surface-to-air missiles,” exactly the situation Israel faced.

Last Friday, as the more than 100 Iranian drones flocked toward Tel Aviv and were shot down by fighter jets, Israel’s Iron Dome air defense system and a U.S. Navy destroyer in the Mediterranean, they couldn’t adjust their course based on what was happening on the battlefield. The Shahed, which means “Witness” in Persian, is generally a “fire and forget” drone: it cannot transmit information back or receive updated trajectories (though it is often modeled in different ways, and some Shahed drones used by Russia have reportedly had communication equipment). Rather the swarmlike power of such attacks is based in their cost: in the one late last week, the IRGC could afford to fire drones in a wave so dense that fighter pilots, radar operators and Iron Dome crews had to sort through a moving cloud of identical radar blips.

On supporting science journalism

If you’re enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.

More Complex Swarms: Ukraine’s Operation Spiderweb

At the heart of most experimental swarms is the boids algorithm, a concept introduced by computer graphics researcher Craig Reynolds in a 1987 paper. A boid is a “bird-oid object” or “birdlike object.” In a boids model of a flock, “each simulated bird is implemented as an independent actor that navigates according to its local perception of the dynamic environment, the laws of simulated physics that rule its motion, and a set of behaviors programmed into it,” Reynolds wrote in his paper. The boids concept follows three basic rules: each boid should stay close to the others (flock centering), shouldn’t bump them (collision avoidance) and should fly at roughly the same speed (velocity matching). When 1,000 bird simulations run on a computer obey those three laws, the screen fills with what resembles a flock. This is the skeleton of swarm logic and the goal of using drones in war. Yet even if the drones can’t communicate with one another, they can be made substantially more lethal just by giving each machine GPS, autonomy and a preprogrammed target, as was the case in the Ukraine’s recent Operation Spiderweb drone attack.

On June 1, less than two weeks before the exchanges between Israel and Iran, flatbed trucks carrying wooden sheds were driven thousands of kilometers by unsuspecting drivers that Ukrainian agents had hired. The trucks parked near Russian air bases; the shed roofs lifted, and out rose 117 quadcopters drones. Each was the size of a medium pizza box, had four rotors and a vision-based autonomy system and carried a payload weighing just more than 3.2 kg. Piloted remotely by Ukrainian operators, the drones rushed toward long-range bombers at the Russian airbases. If the signal to the drones lagged or was lost or jammed, their autonomous systems switched on. These systems had been trained on images of long-range bombers to recognize strategic points at which to strike them. When each drone’s live camera feed matched its preprogrammed target, the machine throttled to full power and struck. An absence of continuous human steering and an ability to autonomously identify targets represent the threshold where drone swarms are more than just a mass launch. The Security Service of Ukraine claims 41 aircraft were hit; even conservative counts admit at least a dozen bombers were destroyed.

The State of the Art of U.S. Swarms

Though being able to identify and pursue targets can make even a rudimentary swarm more dangerous, the ability to soak up the defender’s data, share that information with other drones and adjust based on what’s happening on the battlefield is far more lethal. This is precisely the technology that has been tested at New Mexico’s White Sands Missile Range. In 2021 the U.S. Air Force ran a series of trials in its Golden Horde Vanguard Program involving four Collaborative Small Diameter Bombs: when dropped together, they were able to communicate to decide which bomb would hit which target. The tactic was rehearsed inside a cloud simulator called Colosseum, where every weapon had a “digital twin” to develop strategies for use in real time. The initiative continues to simulate battles using collaborative and autonomous weapons systems.

But the Defense Advanced Research Projects Agency’s OFFSET (OFFensive Swarm-Enabled Tactics) program pushes the idea further, running swarming-drone tactics inside a real-time, game-based virtual environment with the goal of eventually having a single pilot steer 250 drones—in an aircraft or ground system—through a mock city. The swarm would map alleys and ping back a three-dimensional model—a Google Street View with teeth. Whereas the recent attacks from Iran and Ukraine respectively bet on mass and audacity, OFFSET and Colosseum have sought to give swarms the advantage of adaptive autonomy. China is sprinting to close the gap by developing Jiutian, an 10-metric-ton “mothership” drone meant to release 100 subdrones at high altitudes.

All that leaves us humans under a sky that may soon host thousands of autonomous flying things, each no smarter than a sparrow yet smarter than us in one narrow way: an ability to immediately share everything they learn. Defenders may someday fire a “spoofer”—an object that will send counterfeit satellite-navigation signals so convincingly that drones will lock on false positions or even bump into one another like confused bees. Israel is developing lasers to cut through the wings of Shahed drones, replacing the expensive rockets that shoot them down with a power surge as cheap as a subway ticket. But as the defense improves, so will the offense—and the iteration cycle will probably spin even faster.