Imagine the Starling: Peak Fighter, the Swarm, and the Future of Air Combat

February 17, 2016

Considering the future of air combat? Consider the black birds, not Blackbirds.

For special access to experts and other members of the national security community, check out the new War on the Rocks membership.

Software is eating the war, and it’s coming for airpower. In the future, air dominance will be determined not by large, non-expendable fifth-generation platforms, but by nimble drone swarms. The only way to stay ahead of the curve and ensure American air dominance is to begin designing the sixth-generation “fighter” as a rapidly reprogrammable adaptive aerial network. Because the next critical performance envelope isn’t hardware — it’s software.

At the opening of the 20th century, the next hundred years of military aviation could be discerned in the ungainly biplanes whose control surfaces and engines form the aeronautical foundations of today’s sleek supersonic fighters. Wings, engines, pilot/bombardier, and the occasional air-to-air weapons system. Man plus machine, optimized to meet target.

From the Wright brothers to Skunk Works, aircraft continued their march towards modernity. Higher. Faster. Sleeker. Deadlier. And while scramjets and laser weapons offer potential improvements towards the asymptote of instant death from above, we are, for all intents and purposes, nearing what we might call “peak fighter.”

In dogfights of old, fighter pilots squared off like gunslingers in engagements that mimicked sporting challenges: one on one, four on three, six on seven. These were human-scale engagements in human-scale platforms to achieve human-scale ends.

With the proliferation of additive (3D) manufacturing and commercially available drones, it is likely that air-to-air engagement will shift from human-scale to digital. Smaller platforms. Smaller payloads. With printable, payload-carrying drones and cheap processing power, how far away are we from a few thousand quadcopters and a cheap machine learning algorithm deploying them as aerial three-dimensional pickets — or some similar heretofore unimagined asymmetric hack? Once that happens, we will need countermeasures, and will suddenly find ourselves entering a swarm-on-swarm scenario with aerial force that was designed for just the opposite type of combat.

What might swarm-on-swarm air combat look like? To tackle one possible answer, consider the starling.

Go to YouTube and search for the word “starling” and you’ll see that swarm engagements won’t look anything like human-driven air combat.

Swarm air combat at scale (say, a 1,000-on-1,000 engagement) may mean that in a single second, a single small air combat vehicle might considered to be “engaged” with 10–20 enemy drones as it banks through a cloud of hostile platforms, firing when the Venn diagram of “firing solutions” and “current scheme of maneuver” overlap by the microsecond.

The only way to prepare for this future is to design platforms that are open and reprogrammable on the fly. Why reprogrammable? Because swarm software will need to evolve in real time to adapt to the enemy swarm. Just like today’s hardware, drone swarm software will have a performance envelope just as likely to determine victory as the actual physical parameters of the platform.

Combat philosopher John Boyd understood that platforms could enable victory by allowing their pilots to rapidly change their speed, angle of attack, and position. Boyd’s efforts produced some of the most nimble, and lethal aircraft in history. But the Boydian approach to air combat assumes a human operator cycling from the Observe-Orient-Decide-Act cycle. In swarm combat, that cycle is being executed by software. And the cycling is happening at nearly the speed of light.

In drone warfare, enemy operating systems will become important targets of intelligence collection. Confronting an unknown schema of maneuver, vulnerability data will begin to roll in within seconds of engagement, at which point human or machine analysis will be able to evaluate potential platform and software weaknesses. At this point, we will need the ability to reprogram our swarms on the fly to capitalize on enemy weaknesses — an OODA loop on the operating system level.

Unless the defense-industrial complex takes seriously the fact that the hardware and software ought to be treated as separate and complementary, we are doomed to be overtaken by those who do. In a world where the ability to perceive, understand, analyze, and counter enemy algorithms makes the difference in combat, Moore’s Law will make the difference in air-to-air combat.

Were Boyd coming of age in the era of semiconductors, he might have better defined Orientation as both biological and digital processing power. To remain dominant in the skies, the United States must build platforms that can take advantage of computational upside through rapid upgrades of both operating systems and processors.


Joshua Steinman is a former member of the Chief of Naval Operations’ Rapid Innovation Cell. He left active duty in 2015, and now runs Special Projects at ThinAir, a digital security startup company in San Francisco. In his spare time he is a Senior Visiting Research Fellow at National Defense University.


Photo credit: Tim Regan (adapted by WOTR)

We have retired our comments section, but if you want to talk to other members of the natsec community about War on the Rocks articles, the War Hall is the place for you. Check out our membership at!

7 thoughts on “Imagine the Starling: Peak Fighter, the Swarm, and the Future of Air Combat

  1. Great article, and prescient. If we look to ground combat, we might see army ants in place of blackbirds, but the hypothesis is the same. Unfortunately, given our track record we’ll likely only figure it out only after the first defeat…

    1. Excellent article, particularly the points made regarding software and the ability to speed the decision cycle beyond human capabilities. I am unconvinced on small drones, and swarming drones. The tyranny of distance is well known to PACOM commanders, and the physics of flight and chemistry of energy density still reign as dominating features of long-range flight. Small weapons dropped from a large “truck” like the B-52 is closer to reality, but the B-52 needs to be 150nm behind the furball to stay alive…so even then, the drones are not tiny.

  2. A great article and future looking concept.

    Additive Manufacturing, commonly called3D Printing, of course has not reached the level where it can cost effectively produce more than (or even) components of drones or other items such as engines. At the current rate of development it will be some time before it can enable lower cost and timely
    production of the needed number of drones to implement the above tactical approach. DOD would do well to invest in contractual efforts to speed up the learning and development curve using that and utilizing the 3D Printing based manufacturing methods and to increase the number of materials which can be used in that efforts. The materials used costs for 3D printing is also much higher currently versus similar costs in the standard CAD CAM CNC Subtractive Manufacturing world. A DOD investment in that area could also help to reduce those costs.

  3. Interesting article, though I disagree with some of your assertions.
    Networked combat drones are certainly the way forward, but I doubt air combat will be decided through drones that size of a quad-copter. An aircraft that size would have an incredibly limited range, and would have to be ground-based, largely defeating the purpose of an air force. They could be air launched, but that would make them expendable.If they were expendable, then they wouldn’t be particularly different than micro cruise missiles.
    In addition, thousands of drones without electronic hardening would produce a perfect target for a localized EMP blast.
    Lastly, I’m not exactly sure a swarm is the most survivable tactic for drones. You know what else traveled in swarms? Passenger pigeons. Which can fly faster and farther than current drones, which didn’t stop them from being knocked out the air by hunters.

    I’m not sure what your angle is. If you advocate drones working in such huge numbers, that would require drones of such low performance that they could hardly carry a hand grenade. If you are advocating swarm tactics backed by software, that is more reasonable, but then why are you advocating we base warfare on the migratory patterns of a prey animal?

    Interesting piece, in summation, but I’m not sure I buy your hypothesis

  4. Consider that miniaturization will probably change the form of combat, too. Fighters evolved not to fight each other, but to clear the skies of hostile observation aircraft (initially) and bombers (later). The only reason for fighter vs fighter combat (and suppression of ground-based air defenses) is to secure airspace for other functions. As platforms shrink to the minimum practical size to house sensors, propulsion, and warheads, and flight performance reaches towards parity, swarm air combat may become more like missile/anti-missile duels — the offensive swarm simply seeking to avoid, evade, and cut its way through defenders with enough survivors to accomplish a given mission.

    In that light, I find the assertion that software and hardware need to be treated separately as rather puzzling. The two very much depend on each other, particularly as hardware gets smaller. The idea of an additional cyber component to a dogfight is interesting, but even now, first-shot (or first contact) lethality tends to determine the winner, and I’d expect this to become more so in a swarm where combat is a last resort for one side. So smart money says allow a minimum of on-board “learning” based on sensor inputs, and seal the code up tight against outside interference.

  5. Along with Mr. Gallagher, I’m struggling with this idea a bit. It is certainly interesting and worth some investigation, but at least two things pop into my mind that make me question its validity.

    First is simple performance of small systems. Quadcopters? Quadcopters?? A swarm of quadcopters, even for point defense and intended to Kamikaze themselves into an attacking aircraft, would have a range of a few miles and would be able to fly up to only a low altitude. Any modern combat aircraft would simply fly over and/or around them, and/or release ordnance well outside their range. If you try to scale up the concept to something with acceptable performance to catch an aircraft, armed at the lightest with something like Stinger, if not still relying on the Kamikaze concept for minimal weight/size/cost, you’re already moving towards a drone aircraft with a fixed wing with a small jet engine, and at what point does the “swarm” become cost effective for the performance? As Sun Tzu points out, to be present everywhere is to be weak everywhere, so even as a point defense system, again, you now have to have all your territory covered with short range swarm aircraft, vice some high performance interceptors at a few bases that can quickly reach out and engage ingressing hostiles. There’s potential in the idea, but not at quadcopter level, unless you’re trying to take out attack helicopters maybe.

    Second, and perhaps a more critical vulnerability, is that the concept relies on a network ability and constantly updated software to even work. The key word in this author’s post is “hack” but not in the cool, hipster, life “hack” sense, but in the old school, cyber sense. A modern concept to control swarms of pests like mosquitoes is to genetically engineer a virus or a mortality/impotence gene or some such equivalent and infect the swarm, vice area spraying with chemicals. Any competent peer or near-peer will be able to introduce a computer virus into this swarm as well and brick them, IMHO. Admittedly, this applies to any highly networked weapons system regardless of scale, and is a serious concern to currently existing and envisioned tactical systems, but this idea seems even more vulnerable to this attack vector.

    I’m still not seeing the end of the manned, conventional fighter guys like Mabus tout, but maybe I’m just too old and still believe in quaint notions like warfare being about pitting human political desires, wills to persist, and innovative abilities against each other. In “Ghost Fleet” P.W. Singer and August Cole had the Chinese brick the F-35, so aerial warfare actually reverted to teen series fighters in dogfights with enemy aircraft. With BVR missiles only somewhat effective in a heavy ECM environment, this led to something more out of Top Gun than Michael Crichton’s “Prey.” I don’t have the knowledge or authority to predict the future of air combat, but on what little I do know, I find the “Ghost Fleet” extreme scenario slightly more believable than the swarm extreme scenario envisioned here.

  6. but what if your enemy doesn’t have any ‘swarms’ and is just some tribesmen armed with BB guns and pea shooters?
    A bit like the taliban in fact? I mean, they have managed to kick America’s backside seven ways from sunday in Afghanistan these past 16 years and they dont need any 3d printed technology to do it.

    weapons dont win wars guys, just a willingness to die for what you believe in, wins them every time.