Eye of the swarm

“Who can think at this present time without a sickening of the heart of the appalling slaughter, the suffering, the manifold misery brought by war to Spain and to China? Who can think without horror of what another widespread war would mean, waged as it would be with all the new weapons of mass destruction?”

Speaking in 1937 about the aerial bombing of Guernica, Spain, this very prescient speech by Cosmo Gordon Lang, Archbishop of Canterbury, contained the first known use of the term “weapons of mass destruction”. Of course, this was years before the development of nuclear weapons or, indeed, many of the chemical, biological and radiological technologies that the term has since grown to refer to. Ever since this origin, however, the term has proven malleable and indefinite, and the conditions that must be met to warrant the definition has been hotly debated by experts in the field.

This issue had only been amplified by the ever-ongoing development of increasingly lethal and efficient weapons. But it’s a relatively new technology that has taken hold of the ambitions of militaries around the world, resulting in billions of dollars of investment – drone swarms.

What’s in a swarm?

During the past few years, conflict zones around the world have increasingly witnessed the presence of small unmanned aerial vehicles (SUAVs). While the term ‘weaponised drone’ is more likely to bring to mind larger UAVs along the lines of the US’s Predator drones, SUAVs have contributed hugely to what Steve Wright, senior research fellow for the Engineering Design and Mathematics department at University of the West of England, refers to as “the democratisation of aerial warfare, in the same way that the AK-47 democratised landed infantry warfare”.

By that, Wright is referring to the fact that weaponised SUAVs have seen considerable use by militant forces in the Middle East, most notably by Isis fighters during the 2016–17 Battle of Mosul in Iraq. A fleet of small drones armed with bombs attacked Russian bases in western Syria in early January 2018, while in November 2021, an SUAV was used in an assassination attempt on Iraq’s prime minister, Mustafa al-Kadhimi, who was slightly injured.

Drone swarms, however, are a different beast entirely. “At the moment, we’re seeing attacks around the world where these sorts of vehicles, whether it be winged or rotorcraft, are coming in half-dozens,” says Wright. “It doesn’t take much imagination to realise that in a couple of years’ time, they’ll be coming in tens. And in ten years’ time, they’re coming over the horizon in hundreds.”

However, drone swarms aren’t merely defined by numbers. Two drones could technically be a swarm, with the right software in place – ultimately, what separates a group of drones and a drone swarm is the latter’s ability for its various parts to communicate with each other.

“A swarm is where it’s effectively an autonomous formation – where the behaviour of the entire group of drones is an emergent property of simple interactions between the drones themselves,” explains Wright.

Zak Kallenborn, policy fellow at Schlar School of Policy and Government, George Mason University, offers his own definition. “I would define a swarm as simply unmanned systems that are communicating and making collective decisions on their actions, which incorporate some form of artificial intelligence and autonomy,” he says, taking care to point out that this can consist of everything from small numbers of aerial drones to multi-domain drones working across air, surface, subsurface and ground vehicles. Like Wright, for Kallenborn “the real key issue within a swarm […] is that communication”.

Militaries around the world are investing heavily in this still-largely nascent technology. There’s the French Icarus project, the Russian Lightning, the Spanish RAPAZ, the UK’s Blue Bear and the UAE/ South African N-Raven, among others. The US Marine Corps is developing kamikaze drone swarms, while the Army, Air Force, Navy and DARPA are working on their own separate initiatives, with some services working on multiple projects. At the same time, China has also been working on a number of its own swarm programmes.

From past to present

Kallenborn traces the development of drone swarming technology back to the late ’90s, when the work John Arquilla, an American analyst and academic of international relations, was doing for RAND, looking at swarming as a military tactic in which you have a group of small, dispersed units that work together collectively to overwhelm a target from multiple directions. “That’s not unique to unmanned systems,” Kallenborn adds, noting that the Mongol hordes would implement a similar system using horse archers, splitting them into separate units that would then converge on a point from different angles of attack.

Drones are already used to conduct huge light shows, moving together in perfect synchronisation to enthral viewers. The displays are not examples of swarming, however, as every movement is preplanned and controlled from a centralised point. True swarming is much more similar to that carried out by birds, which take place at higher speeds with far greater coordination than drone light shows.

“I live here on the Somerset Levels, and I always see these extraordinary murmurations of birds gathering,” Wright notes, with starlings usually being the preferred example used by drone swarm experts. In a murmuration, tens of thousands of starlings perform complex, intricately coordinated patterns as though part of the same organism, seemingly without communicating with one another.

“A murmuration of starlings is a beautiful and extraordinary emergent property of a handful of very simple rules of interactions between the different birds in the flock,” Wright says, noting that this part of drone swarming is relatively straightforward to program. Moving a swarm of drones from place to place without causing collisions, however, is only the first step towards a militarised drone swarm – the hard part is programming the swarm to carry out tasks independently of a controller.

A swarm can consist of all manner of disaggregated units working together – some might have multiple types of ISR sensors, while others might be armed with explosive devices, firearms or ballistic missiles. The drones with sensors could feed information back to the collective swarm, enabling it to take aggressive action if needed. Kallenborn sees this ability as particularly interesting, and worrisome, when you look at integrating chemical and conventional weapons together into a single swarm. This could involve a few drones spraying a sarin-gas-like nerve agent, paired with other drones with guns or bombs, presenting a complex challenge for armed forces on the ground.

“Do I run away from the guns and bombs?” he asks. “Or do I don my MOPP [Mission Oriented Protective Posture] suit to protect myself from the chemical agent, which is going to make it harder for me to run away from the guns and bombs?” Similarly, there has been some discussion of using undersea drone swarms to create a distributed sensor network to coordinate searches over huge swaths of ocean in order to identify submarines. Such a network could then be used to identify an enemy’s entire nuclear fleet, potentially ending the concept second-strike deterrence – though this remains unlikely. Such sensor networks could also be used to detect underwater mines. Others are looking at drone swarms for chemical, biological, radiological and nuclear weapons defence, using drones to decontaminate an area. There’s also the potential for drone swarms to relegate tank warfare to history, which Wright describes as “a parallel to the end of the battleship era”, where vast numbers of relatively cheap flying weapons were able to be dispatched beyond the range of a battleship, and sweep in and destroy it.

Weapons of mass destruction

First and foremost, the future of military drone swarms lies in their potential as offensive weapons, and as with Archbishop Lang, many experts are reflecting on the potential of the technology to become weapons of mass destruction. For Kallenborn, that moment is here.

“There’s already militaries that are testing what I would consider weapons of mass destruction, and are developing that technology,” he says. As noted, the conditions that a weapon must meet to be considered a weapon of mass destruction are unclear, but for Kallenborn, two commonalities must be present. The first is sheer destructive capacity, such as the ability to destroy an entire city with nuclear weapons. Of course, that aspect isn’t unique to WMDs – you have other weapons, such as the GBU-43/B Massive Ordnance Air Blast, known as the ‘Mother of All Bombs’, which could potentially cause more damage.

The second aspect, therefore, is to do with a lack of controllability. The damage a nuclear weapon provides isn’t just from the initial blast, but also from all the harm that comes from its fallout. Chemical weapons can also be thrown off course by an unexpected change in wind direction. “So, there’s those two criteria – mass destruction and moral concern around the controllability aspect of it. I see drone swarms as potentially meeting both of those criteria, as we get scale with greater levels of autonomy,” Kallenborn says.

In theory, you could make a swarm of a million drones or more – the question is whether you want to. At what size, then, does a swarm have enough destructive capacity to check off that first box in WMD status? Wright brings up the sorites paradox, also known as the ‘paradox of the heap’ – by removing individual grains of sand from a heap, at what point does it become a non-heap? Kallenborn puts the boundary at around 1,000 drones with small explosives as a rough rule of thumb, working off a threshold of 1,000 casualties in a single attack with a single weapon system. Current military systems often struggle to distinguish between non-combatants and military personnel, but the nature of the swarm exacerbates that risk of error and lends to the lack of controllability – specifically, that need for intra-swarm communication.

When an individual drone makes a decision, it feeds that decision back to the entire swarm, which makes decisions based off that information. Therefore, if a drone mistakenly identifies a target, the whole swarm has now made that error. The building that a false target is in could be labelled as one that is enemy-occupied, and that could then increase the likelihood of other people being identified as collaborators.

These problems will need to be ironed out, and soon, as militaries jostle against one another in their race to weaponise drone swarms. In May 2021, Israel claimed to be the first nation to use an actual drone swarm in combat during its conflict with Hamas in Gaza. This involved a simple aerial swarm, consisting of a small number of drones conducting coordinated searches and reportedly feeding information on enemy hiding spots back to mortars and ground-based missiles on the other side of the border. However, there has been some debate over whether this truly constituted the work of a drone swarm.

In January 2021, the Indian army carried out a demonstration of a 75-drone swarm, also claiming that it possessed true swarming capabilities. They also announced their intention to create a swarm of 1,000 drones, which would meet Kallenborn’s threshold for a weapon of mass destruction. At the same time, the US have been developing cluster munitions that would release a number of small drones upon detonation, with the aim of rapidly deploying hundreds, if not thousands, of drones in an area.

The technology is out there, then, even if the many of the kinks still have to be ironed out. The real question will be how quickly militaries can develop the technology to the point that it could have value in peer and near-peer conflicts – and, of course, which nation will get there first. With tensions rising along the Russia/Ukraine border, and around China’s ambitions for Taiwan, drone swarms could prove pivotal in defining the face of 21st-century warfare, just as nuclear weapons were for the 20th.