Ensayo War on the Rocks, 05.08.2015 Mike Pietrucha, coronel oficial de la fuerza aerea de EEUU e instructor de guerra electrónica
Sometimes a technology is so awe-inspiring that the imagination runs away with it — often far, far away from reality. Robots are like that. A lot of big and ultimately unfulfilled promises were made in robotics early on, based on preliminary successes.
– Daniel H. Wilson
The F-35 should be, and almost certainly will be, the last manned strike fighter aircraft the Department of the Navy will ever buy or fly.
– Ray Mabus, Secretary of the Navy
If ever a technology was awe-inspiring, robotics is it. Robots have a long, storied past in literature, dating back at least to the Iliad — and possibly further back depending on your definition of a robot. Robots have long been widely used in the government and commercial sectors — more so now than ever. But as promising as the technology is in a wide range of areas, air combat is not one of them. The Department of Defense’s interest in unmanned weapons and weapons delivery platforms is understandable, but their actual potential for combat operations is the subject of wild hyperbole. Absent technological breakthroughs in machine sensing, artificial cognition, and machine learning, the unmanned aircraft is going to remain a very limited craft indeed. An unmanned replacement for the manned fighter is often believed to be just over the horizon, but the reality is that it is nowhere close and may not even be possible. Combat aircraft that actually have to operate in contested airspace are just the wrapper — it is the aircrew that really matters. An artificial replacement will have to solve three major aviation challenges now readily and regularly surmounted by the human aircrew: basic aviation (flying the aircraft), tactical execution (rapid adaptation of the plan under combat conditions), and weapons employment (shooting the right weapon, at the right target, at the right time, for the right reasons).
Beyond that, an artificial replacement will have to learn and teach the next generation. There are certainly applications for unmanned and autonomous aircraft; a fighter replacement is not yet one of them, and might never be.
The most fundamental challenge in combat aviation is that the mission must be accomplished in a hostile environment that is both changing rapidly and unforgiving of error. There is a simple mental test to determine if the first major technological step (basic aviation) of three has been achieved. When you personally are willing to hop on an unmanned airliner from New York to Edinburgh in the depth of winter, with all of your loved ones and irreplaceable personal treasures, that airplane’s technology will be almost a third of the way there.
A Look Back
The first attempt to make an aircraft unmanned followed the first flight of the Wright Flyer by 15 years. The Kettering Bug was an unmanned biplane designed to carry 180 pounds of explosive into enemy territory. It was shrouded in secrecy, was costly, and tied up development efforts for years after the war. It had questionable military utility to begin with — a cautionary tale that should resonate today. No warfighting technology should be pursued without a deep understanding of its utility in the real world.
I am an experienced fighter aviator, with a reasonable amount of combat time, doing just about every type of combat mission type the Air Force flies with a fighter (excepting nuclear strike). Like many other fighter aviators, I have a monumental ego, an unwavering faith in my own abilities, and a healthy doubt for the assessments made by those who have never flown fighters in combat. The vast majority of people — even in the community of defense experts — lack the experience to make a reasoned judgment about what fighter aircraft do, or more accurately, what the aircrew makes the airplane do. A fighter aircraft is just a tool, and it’s the tool-user that matters. It’s not about the airplane.
There is an awful lot of writing about the benefits of removing a person from the cockpit. Risk to the aviator is one issue. Endurance requirements, airframe size, and other physical limitations are others. Those are real issues, but they miss the point. The aviator is in the airplane because the aviator is necessary for the airplane to be used in combat to its best potential. Warfare is a human enterprise and combat even more so. The reasons for removing humans from the cockpit are offset by the reasons for keeping them in — humans are superior sensors and decision-makers, are the foundation of the combat aviation enterprise, and are a well of aviation knowledge and experience. Aircrew are not perfect, but in a combat environment they are immeasurably better than any other non-organic option.
Today’s unmanned aircraft come in two flavors: those that are remotely piloted, like the Predator, and those that are autonomous, like the Tomahawk cruise missile. Remote pilotage requires a two-way communications link to operate the aircraft, often with a time lag (up to three seconds). Our remotely piloted aircraft (RPAs) are employed only in areas where the threat from air defenses and enemy aircraft is practically zero. We call this uncontested airspace. Still, RPAs drop like flies because aviation is an inherently hazardous enterprise that is often too complex for a remote operator to manage successfully. Even in peacetime, the aircrew at the far end of a radio link often has a difficult time determining what is wrong and necessitates corrective action. In combat operations, a pilot at the end of a three-second time lag is in the wrong place, and those communications links are subject to attack.
Autonomous aircraft are severely limited and often used on one-way missions. The Air Force successfully flew thousands of combat reconnaissance missions over Vietnam, China, and North Korea with fully autonomous, recoverable aircraft called Fireflies in the ‘60s and ‘70s. These aircraft made no decisions and did not fluidly react to conditions — they executed a preprogrammed plan as best they could, given limited navigational aids, and were snatched out of the air by helicopters at the end of the mission. Air vehicle autonomy has advanced little since Vietnam, although navigation certainly has. To replace manned fighters, a fully autonomous — not remotely piloted — aircraft is necessary, and this will remain a showstopper for the foreseeable future.
Of the three challenges, basic aviation is the easiest for automation to overcome, but still comes with major challenges that we have been unable to figure out. Autopilots are fairly old technology, and can fly an aircraft from point A to point B. We can also ask the autopilot to avoid fixed objects, terrain, and cooperative aircraft, and possibly to take off and land. Airbus has embraced a computer-centric cockpit architecture in its design philosophy, which in many respects puts the aircrew in the role of a systems monitor. But no Airbus aircraft has ever been asked to do anything more than transport a load of passengers and cargo on a preplanned schedule and route.
One of the greatest limitations with the flight computer is that it cannot rapidly diagnose systems failures and initiate corrective actions. Not everything that goes wrong is indicated by flashing lights or failure codes. Indeed, some system failures can leave an aircraft perfectly airworthy while disabling the onboard processors, or feed false data to the computer. Air France 447 crashed because the air sensors feeding the flight computers froze over and gave false readings. Similarly, there is no shortage of drone mishaps caused by an inability to diagnose a failure mode from instrument readings alone. This summary of an Air Force Mishap Report for an MQ-1 Predator lays out a common story:
On 26 October 2012 MQ-1B tail number 99-3058 departed Jalalabad AB, Afghanistan. At approximately 2200Z, the crew completed their assigned mission and steered towards Jalalabad. Six minutes later the crew received a “Variable Pitch Propeller (VPP) servo high temperature” caution. This message was the first indication of a problem. While attempting to resolve the problem, the pilot momentarily commanded the propeller pitch to an angle that produced reverse thrust, and the prop froze in this position. The pilot shut down the engine to eliminate the reverse thrust and increase the glide distance, which remained insufficient to make it to a suitable landing location. The pilot deliberately crashed the aircraft into empty terrain to avoid potential injuries on the ground.
The Accident Investigation Board (AIB) found the cause of the mishap was a combination of a mechanical failure of the VPP servo motor and unnecessary movements of the propeller pitch control lever. Furthermore, the AIB President found that incorrect and insufficient checklist guidance, reinforced by incorrect simulator training, were substantially contributing factors to the mishap.
In this case, a mechanical failure combined with incorrect checklist guidance and incorrect simulator training resulted in the loss of this RPA. The contributing causes are critical because the checklist for dealing with the problem was wrong. Had this been an autonomous aircraft, the checklist programmed into the computer would likewise have been wrong, and the flight computer’s attempt to remedy the problem would also have been incorrect, likewise resulting in a crash. As it was, a pilot located on the other side of the world was reliant on limited information and set up for failure when he followed procedures that were faulty from the outset.
Humans have an entire sensory system designed to tell them about threats and conditions — one that is effectively impossible to replicate in a machine, even at great expense. I can recall times where cockpit data was incomplete, contradictory, or outright deceptive. Aviators undergo extensive training to teach them how to deal with contingencies, including when aircraft instrumentation is unreliable. A review of published Air Force mishap data will illustrate how difficult it is for unmanned aircraft to fly with minor mechanical problems.
A combat mission involves more than flying to coordinates and releasing a weapon. We can build a system that is a reusable cruise missile, useful only for dropping weapons on fixed targets. Building a system that can handle all of the tasks and coordination involved in a dynamic combat mission will require technological leaps that may not be possible. The successful accomplishment of the mission is the responsibility of every crewmember, led by the flight lead. Command at the strike package level is typically in the hands of the mission commander, a fighter, or bomber aviator flying the mission who deals with delays, changes, and plan shifts on the fly. Subordinate flight leads and individual aircrew will execute based on the mission commander’s intentions, without micromanagement. Sometimes mission accomplishment happens under the most trying conditions. A summary of one such event, drawn from the citations awarded to the airmen involved, dates from Vietnam:
On 8 February 1968, Hornet flight streaked low over North Vietnam. The two F-4D Phantoms were on their way to the airfield at Phuc Yen. Two days earlier, three IL-28 Beagle bombers had attempted to bomb US positions around Khe Sanh — a rare foray into South Vietnam. The commander of 7th Air Force wanted those bombers destroyed ASAP, and the previous day’s strike had been called off for clouds, which extended all the way to 300 feet from the surface. Capt. John Corder and Capt. Tracey Dorsett were in the lead jet of a lonely, unsupported flight, armed with cluster bombs, smoking across the rice paddies at 600 knots and as low as thirty feet. Phuc Yen was heavily defended; yesterday Capt. Corder had assessed the chances of losing at least one of the two Phantoms at 100 percent.
One minute out, the lead Phantom was hit. Part of the left wing was shot off, the left engine seized up, both fire lights illuminated, and Capt. Corder was wounded. Their airspeed dropped from 600 knots to 180, barely above landing speed and too slow for the cluster bombs to be effective. The wingman inadvertently pulled into the clouds and had to abort the attack. It was all down to two 20-something fighter aviators in a crippled aircraft. Capts. Corder and Dorsett pressed the attack, using the jettison button to drop the weapons and fuel tanks from the left wing on top of the first bomber. Circling the airfield, still under fire, they spotted the second bomber, made a second attack run, and jettisoned their remaining weapons, racks and air to air missiles from an altitude of 40 feet, just 19 feet above the Beagle’s tail. Both bombers were disabled. Limping away from the airfield, they flew to Laos before ejecting. Both were rescued, and awarded the Air Force Cross.
Someday, a machine might be able to accomplish that mission. More likely, the mission would be aborted because no conceivable level of programming would enable a robot to accomplish that mission in that fashion.
The need to modify the plan in flight is a common event, although not to the extreme level outlined above. Times change often, targets change occasionally, aircraft “fall out,” systems fail, the weather interferes — these events are routine. There is a pre-briefed “fallout plan,” which can be modified in real time. An acceptable “risk level” is briefed along with current rules of engagement and the basic execution plan. But the execution plan is just that — a common foundation from which to depart. There are no discrete events that a machine might accept as programming parameters. Audibles are expected to be called inflight based on changing conditions — not based on the conditions that might be anticipated before takeoff. Aircrew are not organic computers programmed at launch to execute a preplanned routine. In over 150 combat missions over ten deployments in six named operations, I have never actually seen an American aircrew abort a mission because the risk level exceeded the briefed limits, but I have seen them pull off high risk tasks and make it look easy, although it most assuredly was not.
Aircrew are the “fighter” in “fighter aircraft.” They are also a fully integrated sensor system, a marvelous biological processor, a communications node, and — most importantly — a living, learning being that can make good decisions on incomplete information and predict likely outcomes in real time. They tend to not be personally risk-averse and are experienced in working as a team. For many missions, including all counterair and some missions such as armed reconnaissance and Close Air Support (CAS), the aircrew have to detect, identify, and engage moving targets which emerge during the mission, often in proximity to friendlies or neutrals and at some substantial risk to themselves. Assessing and mitigating risk is a uniquely human trait. Unlike machines, aircrew can estimate likely consequences beyond the next decision tree, including “strategic consequences” borne from tactical actions. Combat video from ALLIED FORCE illustrates one such case:
On 17 April 1999, two F-15E Strike Eagles, Callsign CUDA 91 and 92, were tasked to attack an AN/TPS-63 mobile early warning radar located in Serbia. The aircraft carried AGM-130, a standoff weapon that is actually remotely flown by the weapons system officer (WSO) in the F-15E, who uses the infra-red sensor in the nose of the weapon to detect the target. CUDA 91, flown by two captains (Phoenix and Spidey) from the 494th Fighter Squadron, launched on coordinates provided by the Air Operations Center. As the weapon approached the suspected target location, the crew had not yet acquired the TPS-63. At 12 seconds from impact, the picture became clearer. “Looks like a tower. That’s a tower, dude, that’s a tower.” “Tower?” “Nah, that’s a church, dude.” “What’s that right there?” Three seconds out, the WSO makes the call: “I’m ditching in this field” and steers the weapon into an empty field several hundred meters away. In a mere nine seconds, the aircrew identified an unexpected object, scanned the surroundings, and made the decision to ditch the weapon in as safe a manner as possible. Postflight review of the tape revealed no object that could be positively identified as a radar, but the profile of a Serbian Orthodox church was unmistakable.
This example illustrates another reality of weapons employment — sometimes the planned target isn’t there. Hitting the wrong target can have significant effects on the conduct and outcome of a conflict. The implications of hitting a Serbian church with a 2000-lb. general-purpose warhead were worked through in real time. The aircrew not only determined that they could not find the assigned target, but identified what they could find and assessed the consequences. It is likely that had they found the radar system in close proximity to the church, they would have ditched the weapon anyway, because the consequences of hitting the church overrode the positive consequences of hitting the radar. This is not a reasonable expectation for a machine, but a routine demand we place upon fighter aircrew.
The final issue that needs to be addressed is the fighter aviation enterprise. The truly irreplaceable role that the fighter aviator plays is in the accumulation of experience and knowledge, and the transfer of lessons learned to the next generation. No unmanned airplane is going to land, debrief its mistakes, and tell stories at the bar afterward. I know about the Phuc Yen strike because Maj. Gen. John Corder told me and a bunch of other Phantom Phlyers about it over dinner one night. John Corder knew that he could fly a burning aircraft as long as the flight controls held together because by then it was well known that Phantoms, unlike Thunderchiefs, did not explode without warning after taking battle damage. They just burned. That piece of knowledge, not found in any technical or tactical manual, entered our bags of tricks through word of mouth.
Similarly, I know about Phoenix and Spidey’s weapon ditch because I was one of the many aircrew in the weapons shop who reviewed the tape, hoping that we could tell if there was actually a radar system there or if this was simply a monumental foul-up by our own command and control. That debrief paid off on a mission I led two weeks later when our entire flight of four aircraft found a church in the crosshairs and again ditched the weapons. Forewarned, a second 4-ship in the same strike package found “a POV (personally-owned vehicle) and a busload of nuns” at their target location and brought their weapons home. Every generation of fighter aviators has built on the often-painful lessons learned by the previous generation, and every generation has passed down its own lessons.
If you have no fighter aviators, you have no fighter aviation. In order to have an effective unmanned fighter, aircrew functions have to be replicated by a machine. In order to have a fighter aviation enterprise, all of the staff, design, developmental, and training tasks that underpin that enterprise need experienced aviators. It is scary to think that DoD could be one bad program mistake and half a generation away from throwing away a century of accumulated fighter aviation experience.
The future of fighter aviation might someday include autonomous or semi-autonomous combat aircraft. The use of unmanned “consort” aircraft that act as very literal wingmen for manned aircraft is nearly within our technological grasp. But those wingmen will not be an adequate replacement for a real wingman; the aviator is too important to both fighter operations and the fighter enterprise. Any air arm which actually contemplates replacing its manned fighters with unmanned ones is surrendering to a technological fantasy and abandoning the ranks of countries which can generate effective combat airpower. Today, no credible aviation expert will advocate the widespread introduction of unmanned airliners, which do little more than fly gingerly from one point to another. To attempt to make the leap to unmanned fighters would be a triumph of misplaced faith in technology over experience, disregarding the fact that combat aviation is substantially more complex than commercial aviation. Meeting the challenges of tactical execution and weapons employment, and maintaining the ability to learn and improve the fighter aviation enterprise are essential ingredients and remain entirely human endeavors. The leap to unmanned fighter aviation will be a long and challenging effort that is some distance in the future. When an unmanned airliner is a safe and reliable transport vehicle, we’ll be almost a third of the way there.