Loitering missiles operate from a simple premise: What if a missile could become more accurate by slowing down?
Awkward cousins to armed drones and cruise missiles, loitering munitions were first developed as a specialized weapon to target anti-aircraft systems in the 1980s and now exist as an alternative to everything from airstrikes to mortar rounds or grenade tosses. Loitering munitions can be as small as a model airplane or longer than a surfboard. Typically fixed-wing and powered by pusher propellers, they can resemble everything from matchsticks with wings to Klingon Birds of Prey. Categorically, loitering munitions are autonomous missiles that can stay airborne for some time, identify a target, and then attack. A munition’s loiter—or the amount of time between launch and detonation—is a function of the missile’s sensors and the kinds of targets these weapons are wielded against.
For decades, loitering missiles have been on the forefront of autonomous lethality. Historically, loitering munitions were used to target things like radars but are increasingly being used to attack humans. And as they make this transition in targeting capability, loitering munitions represent a bridge between today’s precision-guided weapons that rely on greater levels of human control and our future of autonomous weapons with increasingly little human intervention.
The origin of loitering munitions
Loitering munitions trace their origin to the dawn of the jet age. As militaries embraced jet aircraft as the method of choice for delivering explosives against enemies, WWII-era anti-aircraft weapons, designed for slower- and lower-flying propeller aircraft, fell out of use. Jets, which could travel at high speeds and higher altitudes, were initially countered by rival jet fighters. But fielding jets to counter jets was an expensive proposition, even accounting for the cheaper and simpler fighters of the early and mid-Cold War. Surface-to-air missiles paired to radar stations, on the other hand, offered an asymmetric answer to jet warfare and proved devastating to jet aircraft. In Vietnam alone, the United States lost more than 200 aircraft to SAMs. The loitering munition was borne out of a desire to counter this new weapon.
The loitering munition was inspired in part by “Wild Weasel” flights, in which a lead aircraft would flash its radar to the SAM battery, letting its compatriots know how to target it, and then maneuver in a way that was hard for the missiles to hit. Loitering munitions take a similar approach, performing both the search and the destroy roles of a squadron of aircraft flying a Wild Weasel mission. By removing the pilot from the anti-missile aircraft, the designers of loitering munitions reduced the risk of bodily harm in seeking out these weapons. By expanding the flight time, the loitering munition could look for both known and unknown missile installations. On their own, loitering munitions would clear an area of anti-air threats. Combined with following aircraft, the loitering munition allowed jet fighters to retain their utility, without needing to be stealth from launch.
What makes a loitering munition?
In his book Army of None, Paul Scharre defines loitering munitions as a type of fully autonomous weapon that can “search for, decide to engage, and engage targets on their own” in such a way that no human can intervene. This is directly contrasted with semi-autonomous homing missiles, which are human-directed and then have a short flight time in which to find a target. Both the HARM homing missile and the Harpy loitering munition are designed to attack and destroy surface-to-air missile installations. Between launch and detonation, the HARM missile has about 4.5 minutes and a travel range of about 55 miles. By contrast, once airborne the Harpy can search for 2.5 hours and travel a distance of more than 300 miles.
One way to understand loitering munitions is as a kind of airborne mine. Like landmines and naval mines, these anti-radar loitering munitions are an explosive placed into an environment and responsive to a set of characteristics it can detect with onboard sensors. Unlike terrestrial or oceanic mines, which can remain in place for long durations of time, a loitering munition is limited by its flight time, though some loitering munitions can land inert and then be refueled for future flights.
Modern loitering munitions are more than just a weapon to attack anti-aircraft systems. While initial models were designed to autonomously attack one kind of defensive installation, developments in communications technology, computing, processing, and miniaturized sensors means that loitering munitions can now serve a range of functions in war once reserved for crewed aircraft or artillery. As these weapons increasingly integrate autonomous capabilities, they are a test-bed for using weapons on a battlefield independent of human control.
When Azerbaijan and Armenia fought a brief war last year over the disputed Nagorno-Karabakh region, loitering munitions were used by both sides, with Azerbaijan enjoying special success against Armenia’s older model anti-air defenses. Loitering munitions were integrated into the greater strategy of the war, their success against anti-air defenses making all other aircraft more effective in combat. For countries that cannot afford stealth aircraft, the ability to take out anti-air missile batteries with expendable radar-seeking drones is powerful and likely more affordable. By using loitering munitions to first remove anti-air defenses, Azerbaijan was able to then attack other heavy equipment, like tanks, with relative impunity, showcasing how a modest technological advantage can turn into a major strategic benefit.
What makes loitering munitions most useful is the ability to find targets without needing direct human supervision. When those targets are weapons platforms, maintained but not directly crewed by people, a drone turning into a missile and crashing into a target becomes a story of military hardware destroying military hardware. When a flying robot instead uses these tools to hunt people, it becomes a profound question of responsibility and the laws of war.
In March of last year, a drone—or what may have been a loitering munition—was used in this way against a human target for what might have been the first time. That month, a convoy of the Libyan National Army of Khalifa Haftar was attacked by drones that may have included a Kargu-2 autonomous quadcopter and/or loitering munitions, causing significant casualties. Whether or not a robot made the call to actually kill people is hard to definitively say, but the report confirmed the worst fears of human-rights campaigners. “The technology that enables weapons systems to operate without meaningful human control and to target people is here now and is being used without regulation,” the Campaign to Stop Killer Robots observed.
Regardless of whether the attack in Libya was in fact the first instance of a drone autonomously engaging and killing a human being, the technical development and proliferation of loitering munitions makes questions about when humans are in the loop regarding targeting decisions far more immediate. Loitering munitions, together with armed and autonomous drones, pose this question in a more direct way than any weapon outside of land- or naval-mines. Land mines that trigger when they detect a person nearby, are banned by the Anti-Personnel Landmines Convention, in part because they cannot distinguish between civilians and legal military combatants, and in part because they can persist for decades after use. While loitering munitions can persist in the sky only as long as their limited flight time, the matter of target discrimination is crucial, especially as these machines are designed for targets other than missile batteries and radar. Autonomy in targeting systems, especially in targeting-and-firing systems, comes with significant risk. In much the same way that a pressure-sensitive landmine cannot distinguish between an armed soldier and a farmer, an algorithm that converts sensor data into attack triggers is at risk of harming the wrong target.
Loitering toward disaster
Ensuring the accuracy of autonomous target selection is a challenge broader than just loitering munitions, but few other weapons combine such a high reliance on sensor data with a wide window of time in which to find a target. Autonomous systems combine inputs from sensors with internal processing to direct actions. Error is possible at any step of the process. A camera that looks out on the physical world can falter in unusual weather. The algorithm that processes an image recorded by the camera can mislabel an object in it, perhaps describing a tree as a person, or worse, the reverse. Programs that direct action, such as flying to within a certain distance of an acquired target and then detonating, can fail, misjudging the distance or misunderstanding the sensor information fed to it. Some of these problems can be worked out in testing, in safe laboratory simulations, or controlled training environments. But the nature of war is for abrupt, irregular, and violent action to change information in real time, and it is impossible for an autonomous system to anticipate all possible errors.
As loitering weapons evolve, their reliance on computer vision poses particular challenges In research and domestic settings, as computer-vision systems responsible for identifying their surroundings have failed in replicable and major ways. In 2020, researchers at McAfee published a demonstration showing that with a little tape applied to a 35 mph sign, a Tesla car in self-driving mode could be fooled into speeding up to 85 mph. Another example of typographic attack, demonstrated by the OpenAI institute, fooled a computer vision system into seeing a fruit as an iPod by attaching a handwritten note with the word “iPod” on it to the fruit. (These are just a few of a large number of adversarial techniques used to fool computer vision systems.)
The certainty of error in computer systems has been the subject of extensive work in the arms control community. One recent report from the United Nations Institute for Disarmament Research found that in addition to error from data or coding, spoofing or other adversarial measures could misdirect an autonomous system. By feeding misleading data to a known loitering munition, an adversary could direct it to crash away from its target, even possibly returning deadly to where it was launched.
Among the solutions put forth in the report are two worth highlighting. The first is “full or partial moratoriums or limits on use”—essentially a ban on the technology. Given the existing and growing availability of loitering munitions in existing arsenals of multiple nations, it is likely a steep task, though there may be a path in constraining loitering munitions to their earlier, anti-anti-air work. The second proposal is for direct human control, placing autonomous weapons under human supervision that allows a supervising soldier to affirmatively approve a target or deny it in action. This is as much a technical as a tactical hurdle, and it undermines the military utility of these weapons flying through GPS- or communication-denied environments, the exact type of battlefield where such a weapon would be most useful.
More than any other weapon, with the possible exception of swarming drones, autonomy is vital to how loitering munitions work. Any international or regulatory regime that governs autonomy will profoundly change what types of loitering weapons can be designed in the future.
Who has loitering munitions?
Nations with advanced drone programs also tend to have loitering munitions in their arsenals. China, Israel, Iran, Russia, Taiwan, Turkey, and the United States all have domestic loitering munition production. Other countries with loitering munitions have bought them from the major manufacturers, including Azerbaijan, Germany, India, and South Korea. These weapons have also made their way into other markets, and even nonstate actors have built improvised loitering munitions out of existing commercial drones.
Self-guided missiles expand the initial promise of loitering munitions and make air-delivered precision explosions more accessible to a range of military formations. Some loitering munitions, like the modern Harop, can fly for up to six hours, while others, like the Switchblade used by U.S. soldiers and Marines, can fly for about 15 minutes. Mortars, the previous gold standard for hurling an explosive further than a grenade, can be complemented by Switchblades, allowing video guidance on the attack before detonation. Grenades can be slung beneath quadcopters with cameras, turning difficult throws into easy flights that can attack forces behind cover. Whereas guided missiles were once the domain of fighter pilots or ship-borne weaponry, advances in loitering munitions mean trusting that kind of power to lower ranking officers and units, and dispersing the ability to hurl self-guided explosions across future battlefields.
Considerations for policymakers
For policymakers, loitering munitions present urgent questions about how to govern the use of autonomous weaponry. Originally created to meet a particular military need, advancement in sensors, miniaturization, computer processing, and communication networks means that loitering munitions are at the cutting edge of autonomous weapons. It is tempting to assume that loitering munitions are a specialized weapon with a finite set of targets, but policymakers should avoid this simple narrative and instead ask hard questions about degrees of autonomy, how machines select targets, and in what way this weapon design can be coded to be compliant with the laws of war.
Loitering munitions are a way for advanced militaries to maintain an edge against modern anti-air defenses and to win fights against otherwise entrenched forces. Because autonomous target selection is crucial to how some loitering munitions work, the weapons are susceptible to error in ways distinct from human-operated weapons, which should be planned for in the future. Because a wide range of nations use and export loitering munitions, preparing for wars with them on both sides is an increasingly common part of future war planning.
Kelsey Atherton is a military technology journalist based in Albuquerque, New Mexico. His reporting has appeared in Popular Science, Breaking Defense, and The New York Times.