X-47B Takes Off from USS Bush in May.

[Updated Thursday with details on third, aborted landing attempt] Two out of three ain’t bad, if you’re trying something no one’s ever done before.

Landing on the narrow, pitching deck of a Navy aircraft carrier is one of the hardest things a human being can do. Today, for the first time in history, a robot did it — twice.

After all the VIPs and reporters had been hustled off the USS George H.W. Bush, however, a third attempt had to be aborted when the experimental X-47B‘s electronic brain “self-detected a navigation computer anomaly,” a Navy spokesman emailed reporters this evening at 7:30. The drone diverted to a landing field ashore without further incident.

[Updated Thursday: In fact, not only did the drone detect the problem itself, it made its own decision to break off the landing, Navy officials told reporters in an after-action review the next day.

“It was on final approach,” said Rear Adm. Mathias Winter, the one-star admiral who oversees all Navy drone programs. “[It was] about four miles aft of the ship, the [tail] hook was down, the [landing] gear was down, and as it’s supposed to do, it continues to check the health and status of all its subsystems.”

Specifically, the X-47B has three navigation computers that constantly cross-check each others’ results: “When you’re flying that close to personnel and other aircraft, you just have to be sure,” said the Navy’s X-47 program manager, Capt. Jaime Engdahl. But, he said, “about two minutes prior to landing,” one of the three computers came up with a result the other two disagreed with.

That discrepancy was enough for the drone to decide — on its own and without human intervention — that it should “wave off,” to use the Navy term for breaking off a landing attempt. So the X-47B flew past the carrier instead of touching down and reported the problem to its human controllers, who told it to execute the better-safe-than-sorry back-up plan and return to shore.

The Navy and contractor Northrop Grumman are still diagnosing the details of the computer anomaly, but the program is getting right back on the horse: There are two X-47 aircraft, and at least one will be ready for the planned second round of tests aboard the carrier this Monday, Winter said.]

So rather than undermining the historic achievement, the abort underlines just how difficult and revolutionary it is to get an unmanned aircraft to land itself on a carrier. It’s also worth noting that it’s way more impressive that the X-47B drone detected the potential problem itself than it would have been if the humans on the ground had to figure it out. And it’s way better to have spotted the problem in flight than discovering it only because the drone crumpled itself into a multi-million-dollar fireball on the crowded carrier deck.

In fact, while it’s easy to be blasé about drones nowadays — with missile-firing Predators acting as the long arm of the war on terrorism abroad and the FAA working to get (unarmed) unmanned air vehicles, aka UAVs, flying in domestic airspace — the X-47B is something different, and not just because it can land on a carrier.

“Traditional” drones like the infamous Predator are best described as “remotely piloted aircraft”: There’s a human being — or two, or three — sitting in a cockpit-like control station on the ground, complete with joystick, where they eyeball video feeds on screens and fly the UAV much like a kid playing a computer game. But an operator on the ground never has the same feel for what the aircraft’s doing as a pilot physically in the plane; screens alone can’t substitute for the fine-tuned balance detection of the inner ear, to start with. There’s also always at least a little lag time between something happening to the aircraft, the data getting transmitted to the operator on the ground, and the operator’s reactions getting transmitted back. That’s part of the reason why drones crash uncomfortably often.

That kind of lag and error is intolerable in the split-second conditions of a carrier landing, when a slight slip can send the aircraft splashing into the water or smashing into the “island” housing the ship’s command bridge and radars. (Small drones such as the Scan Eagle have landed aboard ship before, sort of, but only by being so small they could crash into a net or snag a cable extended from the side of the vessel). Landing on a carrier deck requires the X-47B to be much more autonomous than earlier drones.

“We did not have somebody [aboard the ship] with a stick and throttle and a rudder to drive this thing,” Rear Adm. Winter told reporters after the first two (successful) landings. “We have automated routines and algorithms that were knitted together in an autonomous logic to tell the system to go do different activities, and it does it autonomously.”

As of the first two landings, Winter said, the drone was performing precisely as the simulations and computer models said it would. “We knew we were going to touch down X number of inches past the second wire, the [tail] hook was going to bounce X number of feet, and that hook point was going to engage the [No.] 3 cross-deck pendant at a specific time,” Winter told reporters in the cavernous hanger of the USS Bush, describing which of the carrier’s three “arresting gear” cables the drone was supposed to catch to brake itself to a halt. “We’re going to [analyze] the data, but my visual eyeball showed that exactly happened.”

But even if a test event didn’t proceed precisely as planned, Rear Adm. Winter said prophetically, it would put the X-47B’s failsafes through their paces in valuable ways.  “In every event there’s success,” he said. If, for instance, the drone had come in for a landing and not caught the arresting gear, he explained, “there is ‘bolter’ logic in this system, so that if you do not engage, you have to be able to keep flying and be able to go around.” The Navy actually had the X-47B perform a “touch and go” — coming in for a landing, hitting the deck with its wheels, but then zooming off again — back in May, months before trying for an actual landing today.

Even today, the first event was not a landing but a “wave-off” to test the X-47B’s ability to come in for a landing but then break off when a control officer on the flight deck hit an abort switch. (The term “wave-off” originates from the guy on deck frantically waving his arms to tell the pilot to pull up, but modern jets come in too fast from too far away to see such hand signals). Human pilots will hear hear the order “wave off, wave off!” over the radio, but a robot has to be told electronically. [Editorial note: Our thanks to Navy pilot Benjamin Kohlmann for correcting us on how the wave-off procedure works]

Only after making sure that emergency system actually worked did the X-47B come in for the actual, historic landing. Then the drone taxied to the end of the flight deck, got loaded into the carrier’s launching catapult, and shot back into the air again, at which point it came round and made its second successful landing. Only on the third try did something glitch — and that it’s possible for this evolution to go wrong only underscores how historic today’s successes were.

(One tidbit to bear in mind: the Navy has been very careful to say this is the first time “a tailless, unmanned autonomous aircraft landed on a modern aircraft carrier.” Given the number of classified aircraft fielded over time, and the fact that other unmanned planes with tails may well have landed during classified missions, the Navy may be hinting at something here — or it could just be referring to those earlier mini-drones that crash-landed into nets strung out from the side of the ship).

Landing a drone on a carrier deck isn’t exactly rocket science: It’s actually more complicated than that. “The engineers and our testers will go back and make sure that the signal strength, the different voltages, all the way down to the ones and zeroes, did exactly what we expected them to do,” Winter said in his press conference before the abort. “Where it didn’t, it’ll help [us] to understand.”

“There is more work to do with X-47,” Winter said. The aircraft will probably attempt more landings on and takeoffs from the USS Bush while the carrier’s at sea this week, and while no further flights are planned, only ground tests, that may change if the Navy decides it’s desirable or necessary.

The experimental X-47B itself, in fact, is only the proto-prototype for a fully armed and operational unmanned aircraft called the UCLASS (Unmanned Carrier Launched Airborne Surveillance and Strike System).

“The plans are for the first [UCLASS} squadron to be at sea six years from now, 2019, as a compliment to the air wing,” Navy Secretary Ray Mabus told reporters at a press conference this afternoon aboard the carrier.

Four firms are already working through a “preliminary design review” for potential UCLASS aircraft: Northrop Grumman, which built the X-47B, but also aerospace giants Lockheed Martin and Boeing, as well as UAV pioneer General Atomics. A draft request for proposals (RFP) for the UCLASS will go out next month for industry feedback, followed by a formal RFP in early 2014 and a final decision on one contractor before the end of 2014. Northrop has a leg up with today’s successful test, but the Navy’s made it clear that it’s not committed to any particular design: In fact, the UCLASS need not necessarily be a tailless flying-wing aircraft like the X-47B but could be a more traditional design instead.

The Navy’s objective is to get a long-range reconnaissance aircraft — with what it calls “limited strike capability” to attack targets — whose time in the air is not limited endurance of a human being in the cockpit. The longer the UCLASS can fly, the greater the time and distance it can stay on patrol.

“it gives us persistence,it gives us the ability to do things we can’t do today because of the limitations of the human body in terms of endurance, in terms of distance, in terms of just how long you can stay on station,” Mabus said.”It will give us what the Navy needs most, which is presence — which is being not just at the right place at the right time but being at the right place all the time.”

The objective, in fact, is 24-7 coverage — with one UCLASS replacing another as needed — at a “tactically significant” distance from the carrier, which presumably means at least far enough out to give ample warning of incoming anti-ship missiles. As the UCLASS evolves, it will get new sensors and new weapons so it can hit some targets on its own, instead of just spotting them for manned aircraft and ship-launched cruise missiles.

UCLASS “will bring payload, persistence, and the ability to adapt,” said the Navy’s senior admiral. Chief of Naval Operations Jonathan Greenert, speaking at Mabus’s side.

So how does that future aircraft fit into military’s evolving vision of long-range, high-tech warfare over the vastness of the Pacific, a concept known as Air-Sea Battle? “It fits right in, Sydney,” Greenert said when I asked him at the press conference, “because what you have here is a piece of sovereignty going anywhere it needs to go, the aircraft carrier” — often described as a mobile piece of US territory — “[and] now you have a platform launched from it that has tremendous persistence.”

Such long-range future warfare has been a favorite theme of Rep. Randy Forbes, the influential chairman of the House of Representative’s subcommittee on seapower. “The first successful landing of an unmanned aerial vehicle on an aircraft carrier constitutes a revolution in carrier aviation,” Forbes said in a widely distributed statement this afternoon. “Unmanned carrier aviation is a ‘game-changing’ technology that will facilitate American power projection, particularly in access-denied environments, for decades to come. Fielding this technology must be one of the Navy’s highest priorities.”

It’s tempting to compare the landing to the first aircraft landing on a ship executed by the Navy  in 1917, as Secretary Mabus did: “This’ll rank up there with the day that Eugene Ely strapped himself into a Curtiss biplane and took off from a ship for the first time, starting naval aviation more than a hundred years ago,” the Secretary said. “What you saw here is the first of the next generation.”

Mabus made sure to emphasize that the new drones only enhance the Navy’s old standby, the aircraft carrier, rather than rendering it obsolete. Precisely because the giant ships have room to accommodate all sorts of new technology,  “our carriers are going to be relevant far, far into the future as the mix of aircraft changes, as we integrate unmanned systems into our manned aircraft wings that we have today,” he said. “We will be able to use our carriers… in perhaps dramatically differently ways.”

Some analysts, however, say large drones like the X-47B will compete with manned aircraft for the very limited deck space of a carrier and will duplicate capabilities they offer. Smaller drones may be better suited to carrier duty, they argue.

It’s very difficult to tell at this stage, but large drones may eventually replace many manned platforms. Or they may become a limited force that supplements the manned force. What we’ll see in the near term may look “rudimentary,” Greenert said, but “it will evolve.”

Updated 8:00 pm Wednesday with news that third landing attempt was aborted. Updated 10:00 pm with extended quotations from Adm. Greenert, Rear Adm. Winter, and Sec. Mabus.