EVERETT, Wash. – The Boeing Co. says it’s on a path that could ultimately lead to self-flying commercial passenger airplanes, starting with simulations and ground-based experiments this year and progressing to flight tests by 2019.
“In commercial air transportation, we are not yet in this business,” said Mike Sinnett, vice president of product development at Boeing Commercial Airplanes
However, he said aviation industry trends are pushing the company to look into autonomous options.
Boeing’s market outlook projects that nearly 40,000 airplanes will need to be purchased over the next 20 years to meet the demand for air travel.
“More of those will be growth than replacement,” Sinnett told reporters last week during a briefing at Boeing’s Everett plant. “It begs the question, where are all those experienced pilots going to come from?”
In the past, airlines could turn to pilots with military experience, but Sinnett said that’s “not as true today as it was 10 years ago or 20 years ago, and it will become increasingly less true as time goes on.” It’s not clear whether pilot training schools can fill the gap, he added.
Trends in air travel also point to a shift toward smaller airplanes and air taxi services, which will add to the demand for pilots.
“There’s going to be a transition away from the requirement to have a skilled aviator operating the airplane tactically, to having a system that operates the vehicle autonomously – if we can do that at the same levels of safety, integrity and availability that we have today,” Sinnett said. “And that is a huge ‘if.’ That’s not a small ‘if.’ That’s a really big ‘if.'”
Boeing has long experience with autonomous systems – ranging from underwater robots and robotic surfboards to the military-grade drones built by its Insitu subsidiary. There’s even a robotic co-pilot system called ALIAS that made a simulated Boeing 737 landing all by itself.
Pilots already make use of autopilot systems and fly-by-wire controls for level flight and landings, and Sinnett acknowledged that the software could handle auto-takeoffs as well.
“The airplane is capable of doing it, but not capable of doing it at the same level of integrity as with a pilot in the loop,” he said.
There’s the rub: Most experts acknowledge that autonomous driving systems should reduce the nation’s traffic death toll of 40,000 annually. But when it comes to regularly scheduled commercial airline flights, no one has died in a jet crash in the U.S. since 2009. That’s one reason why the bar is set higher for planes than for cars, Sinnett said.
“They’ve got to be better than 40,000,” he said. “We’ve got to be as good as zero.”
Sinnett said Boeing’s studies would look into how close computer vision, machine learning and other artificial intelligence tools can come to the decision-making processes that pilots use.
The classic example is the “Miracle on the Hudson,” the 2009 incident in which Capt. Chesley “Sully” Sullenberger successfully ditched an Airbus A320 in the Hudson River after a wayward flock of geese took out both engines.
On one hand, Sully’s unconventional landing choice saved the crew and passengers. On the other hand, computer simulations showed that he could have just barely brought the plane safely back to New York’s LaGuardia Airport – if only he were able to weigh all the options as quickly as a computer could.
Boeing’s ecoDemonstrator test program will help push the envelope on autonomous control technologies. In 2019, the research plan calls for using a modified 787 to expand the use of fly-by-wire to control airplanes.
“You could imagine auto-takeoff, auto-taxi, things like that nature on a 787 – not, obviously, in a certified environment, but with an experimental ticket,” Sinnett said.
More from the Boeing tour: How robots and humans will join forces for 777X
Between now and then, Boeing will be conducting other tests, on the ground and with smaller experimental aircraft, aimed at developing what Sinnett calls an “enterprise autonomy architecture.” He said reporters shouldn’t be surprised if they see “a van driving around a field in Moses Lake with a big boom on it that looks like it’s collecting information, and it’s got a Boeing logo on the side of it.”
If Boeing makes enough progress, the company could work with the Federal Aviation Administration and other stakeholders on ways to validate machine systems for flight.
“I’ve had discussions with regulators around the world, and in general, a very interesting discussion goes like this: ‘We don’t know how to certify non-deterministic systems. For the same set of inputs, we always get the same set of outputs,'” Sinnett said. “And I’ll say, ‘That’s not true. You certify non-deterministic systems all the time. They’re called pilots … Same set of inputs, different outputs.'”
But Sinnett emphasized that autonomous flight systems would have to have as much integrity and reliability as human pilots. “The machine has to be capable of making the same set of decisions, and if it can’t, we can’t go there,” he said.
Boeing’s efforts could eventually result in self-flying airplanes, but in the nearer term they could lead to less drastic changes – for example, AI systems that would allow for smaller cockpit crews in specific flight scenarios, or software that would provide a better backstop for human pilots.
“One outcome of the study that I’m doing right now may be that we have an improved safety algorithm,” Sinnett said.
The study may also show that it’s absolutely essential to keep humans in charge when humans are in the air. And that would be just fine with Sinnett.
“I would be perfectly happy if what came out of this whole study is that we need more pilots, and we’re going to commit to training and ensuring that we have the right levels of experience and competence,” he said. “This isn’t a quest to go take pilots out of the cockpit. This is a quest to ensure that the system maintains the same level of safety that we have today.”