Will the next U.S. fighter aircraft be manned, contrary to the predictions of technologists and policymakers like Secretary of the Navy Ray Mabus? That is the prediction of Col. Mike “Starbaby” Pietrucha, who argues against the viability of unmanned fighter aircraft on the grounds that they can never match the abilities of human aviators. I have great respect for Pietrucha’s insightful and often unconventional thinking on air warfare issues, but his conclusions are deeply flawed. In envisioning future unmanned fighter aircraft, Pietrucha falls victim to two common pitfalls in thinking about the role of robotics and autonomous systems in military operations: conflating robots with warfighters and expecting “fully autonomous” systems to replace humans.
Throughout the article runs an implicit assumption of human vs. machine. Numerous times, Pietrucha mentions unmanned aircraft “replacing” humans. But the real future of combat — in the air and elsewhere — is human–machine teaming: physical teaming between “manned” and “unmanned” vehicles, and cognitive teaming that blends automation and human decision-making. Robots will not replace humans, but will expand and augment their capabilities, helping warfighters to better accomplish their mission — to fight and win.
Robots are Not Warfighters
Pietrucha argues robots cannot replace humans: “Warfare is a human enterprise and combat even more so.” In this, he is right. To envision a world where this is not the case, where war becomes devoid of human purpose or decision-making, is a terrifying prospect. But conflating the role of humans as warfighters and where they physically reside — inside or outside the aircraft — is mistaken. The U.S. Air Force already has warfighters who sit outside of the aircraft, who make crucial human judgments about the use of force from the other side of the globe.
Wars are fought by people, but technology has continually changed how humans fight. Since the first time a human picked up a rock in anger, warfighters have sought ever-greater standoff. The tools of warfare have changed over time, from sling and stone to arrows, bullets, missiles, and aircraft, but it is still humans doing the fighting. Robots are merely the next step.
In the future, humans will still fight wars. But they may fight them with robots.
Unfortunately, our terminology often encourages misperceiving robots, which are tools, as warfighters themselves. Calling robotic systems “unmanned” is unhelpful, which is why the Air Force avoids the term. “Unmanned” conveys the false impression that there is no person involved, but there is always a person involved. Humans accomplish missions. Robots perform tasks, at the direction of human warfighters.
The current Air Force term of art — “remotely piloted” — is also problematic, however. It implies a continuation of the current command-and-control paradigm, where humans still directly control the aircraft, only from the ground. “Remotely piloted” accurately describes Predators and Reapers today, but as Pietrucha points out, this approach has severe limitations. Robbed of the situational awareness and immediate kinesthetic feedback that comes with being physically in the cockpit, and saddled with a sometimes-significant time delay, remotely piloted aircraft are in many ways the worst of all possible options. It would be as if instead of pursuing an autonomous car, Google built a remotely-driven car that lacked robust sensory feedback and had a costly time delay. The result would be more accidents, not fewer.
A better term for vehicles that lack human occupants is “uninhabited,” which more precisely captures what is new: a vehicle without a human on board. This does not mean, however, that humans are not involved in the vehicle’s operation.
Just as an “autonomous car” may drive itself from point A to point B but a human decides the final destination, command-and-control of uninhabited aircraft depends on a combination of human and machine decision-making. Human–machine teaming means adopting a “centaur” model of command-and-control: Simple tasks that can be automated are automated, and human control is retained for context-dependent decisions that require human judgment.
Uninhabited aircraft have many potential advantages — smaller size, greater endurance, maneuverability, and the ability to take more risks (depending on the mission and aircraft cost). In many ways, the human pilot is the greatest physical limiting factor in aircraft design and operations. At the same time, the pilot is the greatest cognitive advantage. Pietrucha rightly points out, “Aircrew are the ‘fighter’ in ‘fighter aircraft.’” And therein lies the one tremendous shortcoming of uninhabited aircraft — by removing the human warfighter and placing her somewhere else, the aircraft loses the most advanced cognitive processing system on the planet: the human brain.
“Full Autonomy” is a Mirage
Pietrucha assumes that in order to provide relevant combat capability, uninhabited aircraft must replicate all of the functions human aviators currently perform on board the aircraft. He argues, “To replace manned fighters, a fully autonomous — not remotely piloted — aircraft is necessary, and this will remain a showstopper for the foreseeable future.”
But “full autonomy” is a mirage. Human warfighters don’t even have full autonomy. They operate under constraints shaped by training, rules of engagement, and commander’s intent. Operationally relevant autonomy is the bar uninhabited vehicles must surmount to make it into the force — that is, sufficient autonomy to accomplish an operationally relevant task under realistic conditions.
In the centaur model of human–machine teaming, command-and-control for uninhabited platforms depends on the combination of their onboard autonomy and their connectivity to remote human operators, not one or the other. Remotely piloted aircraft occupy one end of this spectrum — with little automation they are almost entirely dependent on their connectivity to human operators. Simple uninhabited platforms such as missiles occupy the other end of the spectrum — with no connection to human operators, they are entirely dependent on onboard automation.
But advances in autonomy and communications are giving warfighters an increased set of options in the middle of this spectrum. Net-enabled weaponswill allow dynamic re-targeting in flight by human controllers, and highly automated aircraft like the Global Hawk and X-47 can already navigate autonomously or perform simple missions like reconnaissance or strike against fixed targets.
Pietrucha is right that machines are unlikely in the near term to be capable of matching human abilities in highly context-dependent decisions, such as targeting and engagement. While aviators already rely on computer processing for beyond-visual-range engagements and targeting radars, visual confirmation is still important in many situations, either to confirm the target itself or to assess the surrounding area and potential collateral damage.
Achieving human-level artificial cognition to make these decisions would be extremely difficult. Even if it could be achieved, it is far from certain that humans would be comfortable delegating such decisions to machines. Many decisions in war simply do not have objectively right answers, but instead hinge on values and moral judgment. But the centaur model of command-and-control means that future aircraft can automate many tasks and lean on humans only for decisions where human judgment is truly needed. Advanced automatic target recognition and image processing will allow computers to compress the targeting process, finding potential targets, tracking them, and cueing them to human operators who make the final decision. This will dramatically reduce the bandwidth required to keep human operators “in the loop” for targeting decisions.
As one example, the below screen grab from a video of an F-15 strike in Iraq is a mere 12 kilobytes in size. While grainy, it clearly shows sufficient resolution to make out individual vehicles, and would allow a trained operator to discriminate military-specific vehicles, such as a tank or mobile missile launcher, from dual-use vehicles such as buses or trucks.
To give a sense of scale, connections on par with a 56K modem from the 1990s could transmit two such frames per second and still have some bandwidth left over for vehicle command-and-control. This would be far from full-motion video, but would allow 0.5-second updates for human operators to respond to changing events on the ground. While this resolution is not sufficient for finding and tracking people in counterinsurgency conflicts, those aren’t the conflicts where the United States is likely to face communications challenges. In high-end anti-access/area denial (A2/AD) environments where communications will be contested, the Air Force will primarily be targeting military equipment.
Even in the example Pietrucha cites of a pilot calling off a strike in Serbia due to the presence of a church, a mere 15-kilobyte image (below) is sufficient to clearly identify the church.
While communications will be challenging in A2/AD environments, they are not an all-or-nothing proposition. Certainly, the kind of high-bandwidth, satellite-dependent operations the United States conducts today, streaming full-motion high-definition video from dozens of feeds, will not be feasible. Satellites will be subject to jamming and any stealthy aircraft emitting detectable communications would give away their position. But that’s why the U.S. military employs low probability of intercept/low probability of detection (LPI/LPD), jam-resistant communications in contested environments. The F-22 and F-35 both use LPI/LPD links for aircraft-to-aircraft communications. (The two aircraft fleets can’t talk to one another, but that’s another story.)
Stealthy line of sight communications will also be essential in A2/AD environments to connect F-22s and F-35s with theater- and national-level assets. They will also allow the aircraft to receive position, navigation, and timing data in GPS-denied environments. Uninhabited aircraft can use these same communications links; in fact, uninhabited aircraft are critical to building the aerial layer network the Air Force needs to support its manned aircraft in satellite-denied environments. Because of their ability to stay aloft far beyond the limits of human pilots — potentially for days at a time — uninhabited aircraft are ideal communications relays.
Uninhabited combat aircraft can also collect and pass targeting data to human-inhabited aircraft forward in the fight. This is precisely the intent of DARPA’s SoSITE project: Small, expendable, and uninhabited air vehicles autonomously find, track, and recommend targets, passing snapshots to manned aircraft pilots for final approval. In this vision of the future, uninhabited aircraft don’t “replace” manned aircraft. Rather, they enhance and augment the capabilities of the aviators who are forward.
The notion of a near-term future without any manned aircraft at all is simply silly. Even if the F-35 is the last manned fighter, manned aircraft aren’t going away for decades. The F-35 will likely remain in the force through the 2040s. (The F-16 and F-15 both came online in the 1970s and are still in service.) The new long-range bomber won’t even enter the fleet until the mid-2020s and will remain in service for decades beyond, likely for the better part of the 21st century. The B-52 has been in service for sixty years.
Decades from now, human aviators will still be forward in the fight, able to exercise real-time command-and-control of uninhabited combat aircraft via protected communications links. Line of sight communications will reduce today’s cumbersome time delays, which are the result of multiple relays through satellites and ground stations sending signals around the globe. Automation will reduce the bandwidth required for that communication. And in the event that there are no manned aircraft present, uninhabited aircraft will build their own line-of-sight communications bridges back to a human controller, an affordable proposition because of their long endurance.
Uninhabited Combat Aircraft are the Key to Affordable Power Projection
The United States will have manned aircraft for decades, but they can’t win the fight alone. Uninhabited aircraft have unique advantages that make them not only useful, but essential to affordable power projection in A2/AD environments.
Multi-decade trends in rising aircraft costs continue to drive down aircraft quantities. Combined with the development of long-range missiles that can target U.S. air bases, the Air Force faces a power projection crisis. A 2009 RAND study of a hypothetical U.S.–China air war over Taiwan found that Chinese missiles would so thoroughly hammer U.S. “first island chain” bases that the United States would lose the war for air superiority over Taiwan:
The danger to both [Taiwanese] and USAF operations in the Taiwan Strait is sufficiently grave that a credible case can be made that the air war for Taiwan could essentially be over before much of the Blue [friendly] air forces have even fired a shot.
Absent a revolution in missile defense, U.S. air power projection in A2/AD environments will depend on longer-range aircraft. The new long-range bomberis an essential component of a counter-A2/AD force, but it is not a fighter aircraft.
While short-range fighters can extend their reach via aerial refueling, the key limiting factor in projecting air superiority at long range isn’t the aircraft, but the pilot. U.S. aircraft flying from airbases in Guam, Tinian, or Saipan to Taiwan would spend over three and a half hours in transit each way, exhausting 90 percent of a pilot’s 8-hour sortie just getting to the fight and back. The result is that sustaining combat air patrols at these ranges is entirely infeasible. The entire F-22 inventory located at Guam could sustain only four aircraft on station 24/7 over Taiwan.
Uninhabited aircraft, on the other can, can remain aloft far beyond the limits of human pilots. Uninhabited aircraft have flown for as long as 80 hours in a single sortie. Fighter aircraft, which aren’t designed for long endurance on a single tank of gas, will need aerial refueling, but refueling unlocks even longer endurance. The aircraft refueled endurance record is 64 days, set in 1958 in a Cessna 172.
The result is that while 500+ manned aircraft are required to keep 16 fighters from Guam on station over Taiwan, the same combat power forward could be generated by less than 70 uninhabited fighters. And that’s assuming only a 16-hour refueled endurance. That equates to 85 percent less in procurement costs, or over $50 billion in savings at the cost of an F-35. As a bonus, tanking requirements would fall by over 75 percent.
Human–Machine Teaming is the Future
Manned F-22s, F-35s, and bombers will remain in the Air Force inventory for decades, but there won’t be enough of them. Even if the United States buys all of the planned 2,443 F-35s, the threat from long-range missiles means there will be insufficient numbers of aircraft forward in the fight.
Uninhabited combat aircraft can change that, giving the Air Force affordable power projection. Uninhabited aircraft will operate in tandem with manned aircraft, either as “loyal wingmen” in close proximity or potentially as detached wingmen, moving forward to recon, jam, or strike targets. These aircraft won’t be “fully autonomous.” Instead, they will employ operationally relevant autonomy — enough autonomy to reduce the bandwidth requirements such that they are manageable with stealthy, jam-resistant A2/AD communications. Human aviators will remain in command, “quarterbacking” the fight and making key decisions, such as target selection and engagement.
Pietrucha frets over a world devoid of human aviators, but robots will not eliminate aviators any more than the invention of firearms eliminated soldiers. Firearms did, however, change how humans fight. Robots will do so as well. In some cases, robotics and autonomous systems may eliminate certain jobs entirely or alter them beyond recognition. We no longer train warfighters to fire crossbows or ride horses, after all. To harness the advantages of the robotics revolution, we must avoid becoming overly attached to the current ways of fighting. Sustaining U.S. military dominance hinges on a willingness to consider novel and creative approaches to warfighting, including those that challenge cherished assumptions. And that task — discovering new ways of fighting, including fighting with robots — is a job for humans.
More from CNAS
PodcastCNAS Tech: How (Not) to Talk About AI & Lethality
The U.S. Army recently announced its new Advanced Targeting & Lethality Automated System, or ATLAS program. The announcement generated concern and media headlines about the le...
By Paul Scharre, Kara Frederick & Megan Lamberth
PodcastThe future of war: Autonomous weapons, AI, and cyberwarfare
How will emerging technologies shape the conduct and consequences of war? And how will they impact civilian security? Today on Displaced, Loren DeJonge Schulman and Erin Simps...
By Loren DeJonge Schulman & Erin Simpson
VideoWill WWIII Be Fought By Robots?
What will autonomous weapons mean for how future wars are waged and the loss of human lives in armed conflicts? That's the topic of a new book, Army of None: Autonomous Weapon...
By Paul Scharre
CommentaryA Million Mistakes a Second
Militaries around the globe are racing to build ever more autonomous drones, missiles, and cyberweapons. Greater autonomy allows for faster reactions on the battlefield, an ad...
By Paul Scharre