Drone360 Menu
X

SEARCH SITE

Enter keywords or a search phrase below:

News & Notes

Drone Wars

America's hunter-killer drone program is plagued by human error and a shortage of pilots, but drone swarms might help.

March 8, 2016

Early on the morning of Oct. 3, 2009, U.S. Army soldiers based at Combat Outpost Keating in Afghanistan woke to intense enemy gunfire from hundreds of militants. The outpost sat in a small, rocky valley filled with shrub-like trees and surrounded by mountains. Insurgents were firing down into the base from all sides. Severely outgunned, the troops were forced to surrender the base section by section, as they gathered the wounded, consolidated their forces, and called over the radio for air support. The battle waged throughout the day, with 19 aircraft eventually providing support through intense bombing and machine-gun fire. And Air Force pilots 7,500 miles away at the March Air Reserve Base in California provided surveillance, coordinating the attack with an unmanned Predator drone cruising high above the fray.

As the first Predator pilot’s long shift ended, a second pilot sat down and took the controls. Within 13 minutes, he had slammed the drone into a nearby mountain, turning the multimillion-dollar UAV into a fireball. In the heat of battle, a pilot on a faraway continent had gotten distracted and failed to control his aircraft. An official Air Force investigation said the pilot suffered from what’s known as “channelized attention.”

Out of the loop

Just 20 years after America first flew the Predator drone in combat, UAVs have become the military’s weapon of choice in its war on terror. The once-overlooked technology is now so prevalent, it’s forcing a complete rethink of how modern wars are waged. The Pentagon had a mere 50 UAVs in the year 2000. Current estimates indicate that America now has more than 6,000 military drones in operation around the world.

Most of those are short-range reconnaissance aircraft with hand-held controls and a wingspan that looks more like a model airplane. But the Congressional Budget Office forecasts the U.S. Department of Defense will spend almost $40 billion more by 2020 to add an additional 730 medium- and large-sized UAVs to its arsenal.

This seismic shift hasn’t been without growing pains. Drones like the Predator came into favor because of their long flight and surveillance capabilities, as well as their ability to strike in places American forces aren’t currently operating. But the robotic planes crash far more often than manned aircraft. Crash rates vary by aircraft and complete studies are hard to come by, but even the most conservative estimates show the military’s large UAVs crash three times more often than the rest of the fleet. A sweeping Washington Post investigation in 2014 showed that more than 400 large military UAVs have crashed since 2001. Aside from mechanical problems, boredom, distractions, and other human errors are often to blame.
The military has also quickly learned that the term “unmanned” is a misnomer; most drones actually require more than one human to operate them. The U.S. Air Force now trains more drone operators than fighter and bomber pilots combined. Yet the Pentagon is several hundred operators short of the number it needs to fight the Islamic State and maintain air operations around the globe.

To fix these combined problems, automation experts say the machines should take more control of themselves. Drones have already proven that they can take off, land, refuel in mid-air, adapt to changing flight conditions, and even paint a target — all without the need for human supervision. So why not take humans out of the loop entirely?

Welcome to the Terminator-esque, not-so-distant future of drones, where decisions will be made by machines. Cadets with only a few months training will watch over entire robotic squadrons. And, if the machine loses contact with its coach, it’ll turn around and fly itself home.

An MQ-1B Predator prepares for takeoff behind an MQ-9 Reaper at Creech Air Force Base in Nevada. The aircraft are operated by the 432nd Wing, which trains UAV crew members and runs worldwide combat and surveillance operations. Christian Clausen/U.S. Air Force
Humble origins

Like many disruptive technological advances, the current drone revolution traces its origins to a California garage. Israeli engineer Abe Karem left his homeland in 1980 after meeting resistance to his ambitious plan to build the first ever long-flying, remotely piloted vehicle. He launched Leading Systems Inc. from his Hacienda Heights home and set to work building his dream machine. The Albatross was no one’s idea of a warplane. Its tiny frame and nearly 30-foot wingspan were built simply to soar — its go-cart engine could keep the craft aloft for days at a time.

By 1984, the modest machine had caught the eye of the Defense Advanced Research Projects Agency (DARPA), which saw the value of such a constant presence over conflict areas. The secretive government research group soon helped Karem get the project out of his garage. Albatross became Amber, which eventually led to General Atomics’ Predator. And while the Predator first proved its value in 1995 during the Balkan wars, it wasn’t until after the September 11 terrorist attacks that America’s military realized how much it needed drones.

The Predator and its big brother, the Reaper, quickly found a home in the skies above Afghanistan and Iraq. It was there that military drones evolved from their reconnaissance roles into the hunter-killers that made them both notorious and indispensable.

The drones, now armed with Hellfire missiles, can bomb ISIS in Syria, take out militants in Afghanistan, and incinerate a terrorist training center in Somalia without putting a single soldier in danger. Predators have also proven their worth in combat situations, where they can tell infantry what waits around the corner, on the roof, or behind a wall.

These UAVs weren’t the only robotic aircraft to come out of the fight in Iraq and Afghanistan either. The Department of Defense also poured cash into the Global Hawk, Hunter, Shadow, Gray Eagle, Sentinel, and many more. A flight of drones can now be launched from a backpack or the deck of an aircraft carrier.
“We need to get away from thinking of drones as having pilots operating drones. They’re really coaches.”
The Predator problem

It took 16 years for the Predators to log their first one million hours of flight time — but only two and a half more years to hit the second million. And with them for each of those flight hours is a team of operators using hands-on controls to launch, land, and fly the Predator. One local crew controls takeoff before handing the flight off to drone pilots in the U.S. The Predator’s traditional control station is part fighter jet cockpit and part arcade game. Its array of computer monitors and multiple joysticks scream action, but the overstuffed chairs say “you’re gonna be here a while.” A Predator pilot’s shift can often last 12 hours with little action, as the aircraft circles endlessly above a surveillance target. Such field operations can go on for days, weeks, or even months.

And while the job might sound a lot like playing video games, the long shifts take their toll on the drone pilots’ ability to focus. For comparison, a typical combat pilot will fly 250 hours each year; a drone pilot racks up an average of 900 hours a year, many of them mind numbing.

Layering manual operations over automated systems means that a moment of misplaced human attention can be catastrophic. Crash investigation reports show one Predator pilot crashed in Afghanistan after squeezing the wrong red joystick button. Another didn’t realize she was flying the aircraft upside down. A media study done in 2010 showed that 38 Predator and Reaper drones had crashed in combat in Iraq and Afghanistan, with nine more crashing in training on U.S. soil. Those crashes might not cost lives, but each Predator costs between $3.7 and $5 million. The problem hit new heights in 2015, as pilot error and mechanical failure brought down a record 20 large Air Force UAVs, according to documents obtained by The Washington Post.

Missy Cummings is the director of Duke University’s Humans and Autonomy Laboratory, and she’s overseen much of the cutting edge research on what boredom can do to UAV pilots. Her lab put drone operators at the controls of a Predator and watched what happened over the course of a long shift. She found that bored pilots struggled to react in the rare situations where their input was needed.
Student teams from the 6th Reconnaissance Squadron at Holloman Air Force Base in New Mexico are trained in UAV combat and support tactics during an MQ-1 Predator super sortie simulator mission.
Breeann Sach/U.S. Air Force
“The primary problem of the Predator is a complete lack of what we call human systems engineering,” Cummings says. The cockpit design doesn’t consider human physiology and limitations. “These systems were doomed to fail from the outset.”

The Air Force is already pushing toward better system designs, and says the crash rate has come down. The Pentagon claims the Predator’s crash rate is now similar to that of the F-16 fighter jet at the same stage. But that doesn’t mean the plane’s problems are totally behind it.

As one of the Navy’s first female fighter pilots, Cummings has successfully pushed the military toward change for decades. She served as a target aircraft pilot before the military allowed females in combat positions, and would eventually fly an F/A-18 when that policy was changed in 1993. Her research has prodded the Pentagon once again; ironically, this time it’s by showing the military that intensely trained fighter pilots are unnecessary to command drones. In fact, such pilots often make the problem worse.

“We need to get away from thinking of drones as having pilots operating drones,” Cummings says. “They’re really coaches.”

In recent years, fear of drone crashes — particularly at takeoff and landing — has forced the U.S. military to move foreign drone operations away from civilian centers. For example, operations in the Horn of Africa had to be pushed to the isolated Chabelley Airfield after the tiny country of Djibouti expressed concerns over aircraft accidents at nearby Camp Lemonnier — the only permanent base on the continent and located at Djibouti-Ambouli International Airport.

Five Predator drones crashed trying to land at the base between 2011 and 2013. Locals feared the aircraft could collide with incoming passenger planes. According to The Washington Post, officials told the U.S. it was time for the base to move. A similar situation on the island of Seychelles in the Indian Ocean forced the Pentagon to stop launching attacks from a civilian airport after two Reapers crashed on the runway.

“The pilots still want to take off and land these things,” Cummings says. “Well, you know, we can automate that, and we don’t automate that. But they’re slowly coming around to it.”
Northrop Grumman’s X-47B proved it could safely taxi around sailors in tests on the USS George H.W. Bush. In 2013, it became the first UAV to catapult launch from an aircraft carrier. Northrop Grumman
Fun Fact
According to a study by the Humans and Autonomy Lab, pilots of Airbus planes spend about three minutes per flight with their hands on the cockpit controls.
Robots to the rescue

While the Air Force has struggled with its Predators, the Army has seen strong success with its Shadow drones by simply taking the pilot out of the loop during takeoff and landing using what Cummings calls “point and click.” The pilot authorizes a preprogrammed software routine that brings the aircraft to cruising altitude or back to the ground, avoiding human factors in the period where most accidents happen.

Used by both the Army and the Marines, Shadow also had its first successful armed test in 2015. The smaller aircraft can carry two missiles and fly beyond the horizon under satellite-linked controls. Maryland-based Textron has the go-ahead to begin selling these armed drones to America’s allies abroad.

The Air Force is moving in a similar direction. The agency imagines artificial intelligence taking over when time is of the essence, and completing missions even if GPS signals have been jammed; the military wants its drones to be capable of flying back to base all on their own. For the Air Force, the first step toward fuller automation is the more intelligent — and expensive — Global Hawk.

The Global Hawk can stay airborne for as long as 35 hours, has a range of more than 13,000 miles, and can climb to 65,000 feet. Its endurance is unmatched. And while the modest Predator has a maximum takeoff weight of just over 1,000 pounds, the Global Hawk can launch 30 times heavier.

The Global Hawk crash rates are also lower, thanks to its highly automated system. The aircraft sits somewhere in the middle on the spectrum between the Predator and a full autonomy. Takeoff and landing — the most dangerous part of any flight — are fully automated. Its destinations are determined as the mission is being planned, and it is the aircraft itself that sets and adjusts speed, altitude, roll, pitch, and yaw. Instead of a fake cockpit and joystick like those commonly used for Predator operators, Global Hawk operators sit in a desk chair and use a keyboard and mouse to control the drone.

In fact, the drone has been so successful that the Pentagon says it will retire its U-2 spyplane and replace it with the Global Hawk — if Congress will allow it. DARPA and NASA even successfully partnered on a test flight to show that two autonomous Global Hawks could accomplish high-altitude refueling on their own.

And yet, even the Global Hawk doesn’t rise to the level of autonomy the military wants. On a scale of 1 to 10, where 10 is fully autonomous and 1 is completely guided by remote control, the Global Hawk — the most autonomous large UAV in service — only ranks at 2.5, according to the Air Force.

Drones aren’t the only focus of autonomy, either. The F-35 fighter jet has over 8 million lines of code and automation is baked into its missile and threat management systems. And F-16s have ground collision avoidance software that will take control of the plane and make corrections before handing operations back to the human pilot. There’s also a push for pilot-optional aircraft. By the year 2035, the Pentagon predicts its arsenal will include some 5,000 such planes that can either be piloted, or pilot themselves.

Sound terrifyingly futuristic? It shouldn’t. Automation is common in civilian aviation. The Airbus A320 can already take off and land itself. In fact, Airbus has been running many of its airplanes with little pilot input since the early ’90s. And a new study by the Humans and Autonomy Lab shows pilots of those planes only spend about three minutes per flight with their hands on the cockpit controls. Boeing 777 pilots report that they’re in control for less than 10 minutes of each flight.

A Northrop Grumman RQ-4 Global Hawk awaits a nighttime mission. The surveillance aircraft can hit altitudes up to 65,000 feet, stay aloft for more than a day, and has a range of more than 13,000 miles. John Schwab/U.S. Navy
Drone pilots (still) wanted

The Air Force says it needs to train 300 remotely piloted aircraft (RPA) pilots a year to keep up with the pace of additional aircraft and patrols. They’re currently training just under 200 each year.

Fuller automation for drones in the future should lower the number of pilots required and ease the understaffing strain. But humans won’t be completely obsolete, and automation isn’t the only answer: Beginning in 2016, the U.S. is paying $15,000 annual bonuses to UAV pilots who sign new contracts of five or nine years. That sum stacks up to $135,000 for drone pilots willing to stick it out long term, with half of that bonus available up front. The Air Force is also putting $100 million into contracting instructors and purchasing the simulators and ground control stations those operators will need.

“In a complex global environment, RPA pilots will always be in demand,” Secretary of the Air Force Deborah Lee James said in a statement announcing the bonus program for drone pilots in July 2015. “Remarkable airmen have ensured the success of the [Predator and Reaper] programs. We now face a situation where if we don’t direct additional resources appropriately, it creates unacceptable risk. We are working hard to put solutions in place to bring needed relief to our airmen and ensure our actions show their value to our mission.”

Future pilots may not just be flying single UAVs. The Air Force’s former chief scientist, Mica Endsley, envisions new forms of warfare enabled by swarms of robots. These drones will not only fly together in formation, they’ll share information and adapt, collaborating on attack or defense missions.

Other military minds see similar uses. The U.S. Navy’s Office of Naval Research carried out test missions on its Low-Cost UAV Swarm Technology (LOCUST) in 2015 using swarms of small, rocket-tube launched Raytheon Coyote UAVs. The agency plans to test launch a massive UAV swarm made up of 30 aircraft from a ship in 2016.

A team of students from the Advanced Robotic Systems Engineering Laboratory (ARSENL) beat that number in August, flying an astounding 50 small UAVs under the control of just one person. The drones relied on computer algorithms and powerful Wi-Fi to communicate with each other. ARSENL’s next step in swarm testing is a planned 50-on-50 dogfight.

And swarms won’t always be confined to small UAVs. DARPA recently called for help developing the technology to allow existing unmanned aircraft to work collaboratively through its Collaborative Operations in Denied Environment (CODE) program. “Just as wolves hunt in coordinated packs with minimal communication, multiple CODE-enabled unmanned aircraft would collaborate to find, track, identify, and engage targets, all under the command of a single human mission supervisor,” said DARPA’s Jean-Charles Ledé in a statement.

In the fall of 2015, the Army’s Autonomous Unmanned Systems Teaming and Collaboration in GPS Denied Environments program (AUSTC) announced it would work to fund and mature technologies that would bring about “revolutionary” levels of drone autonomy. Still, Endsley and others stress the need to keep humans in the loop to counteract mistakes made by future autonomous systems.
Soldiers that don't sleep

Not everyone is excited about the future of autonomy: These thinking robots have sparked debate about whether robots should be able to kill without consulting a human. Human Rights Watch lawyers have pushed the United Nations to ban the practice before it even gets started. The group says action is needed now because “killer robots,” fully autonomous weapons capable of selecting and shooting humans on their own, are only a few decades away.

In response, the Department of Defense issued a directive requiring that humans make the final call about taking lives. And yet, the Air Force chief scientist raises questions about certain limited cases where drones might be able to take over when there’s not enough time for humans to act or loss of life is imminent.
That’s not enough for some opponents. Human Rights Watch lawyers say there simply couldn’t be accountability for autonomously operating killer robots, so they should never be developed in the first place.

“Some proponents of fully autonomous weapons argue that the use of the weapons would be acceptable in limited circumstances, but once they are developed and deployed, it would be difficult to restrict them to such situations,” the group says.

By 2035, the Army expects to have advanced autonomous drones that can steer clear of hazards, making them safe to integrate into America’s airspace. The Office of Naval Research is similarly working on an Autonomous Collision Avoidance System to allow drones to operate over U.S. airspace, keeping them miles away from potential collisions. The Air Force also plans to rely more and more on robotics in the future — but still expects operators will have a strong role.

“Most or all Air Force operations conducted in the foreseeable future will require a combination of both humans and autonomy to get the job done in the face of a broad range of operational conditions and a determined adversary,” says a report from the Air Force’s Office of the Chief Scientist.

The military plans to keep the machines in check — for now.

Note: A version of this appears in the March/April issue of Drone360 magazine.
Featured image: U.S. Air Force