As air travelers suffer through a summer of flight delays and cancellations fueled in part by a pilot shortage, Merlin Labs is one of a slew of companies working on a long-term solution: taking pilots out of the cockpit by developing computer systems that allow planes to fly themselves.
The Boston-based startup has won $105 million in fresh funding, impressing big-name investors including the Scottish firm Baillie Gifford with a project to robotize the U.S. Air Force’s workhorse C-130J Super Hercules transport planes that the company is revealing publicly for the first time, as well as its achievement last fall of a key regulatory milestone: New Zealand’s Civil Aviation Authority approved Merlin’s plan to prove the safety of a single-engine Cessna Caravan cargo plane kitted out with an initial limited autonomy package, making it the first company to win so-called “certification basis” for such a system.
Merlin Labs CEO Matthew George says that its initial product will be the first commercially available flight control system that will enable an airplane to automatically follow a programmed path from the taxiway to takeoff to landing, as well as the first featuring an artificial intelligence element: It will be able to understand radio instructions from fast-talking air traffic controllers and follow them.
What it won’t be capable of: The company isn’t seeking yet to kit out planes with sensors to automatically detect and avoid obstacles on the ground, like animals or airport vehicles, or other aircraft in the air. A pilot will be required to be onboard to do that, as well as to correct any misunderstandings of air traffic control instructions.
That falls a good deal short of the holy grail of full autonomy, but the 32-year-old George says Merlin’s initial system will significantly reduce the workload for pilots, enabling them to act more like safety monitors.
It’s part of a calculated strategy by George to devise a product that Merlin can bring to market soon through a regulatory gantlet of aviation safety agencies that are moving cautiously in the face of a startling array of novel aircraft under development, from cargo drones to urban air taxis, some of which are being designed to be fully autonomous from the start, on top of pushing the boundaries with electric propulsion systems and vertical takeoff and landing capabilities.
“We think that the hardest problem truly is certification,” George told Forbes. Of Merlin’s roughly 70 employees, about 30 are working on safety certification.
That eyebrow-raising share is partly due to the fact that while Merlin is seeking initial air safety approval in New Zealand (the country has an experimentation-friendly regulatory environment that has attracted other cutting-edge aerospace startups that are betting they’ll be able to launch commercial service sooner there, including Boeing-controlled robot air taxi developer Wisk), Merlin is submitting all stages of its work for simultaneous review by the U.S. Federal Aviation Administration to ensure that it will get a green light to operate in the U.S. as well.
George’s game plan is to start with a revenue-producing cargo service in New Zealand. He thinks that’s a goal Merlin could reach well inside of five years.
In the U.S., Merlin, which was founded in 2017, struck a deal last year to automate 55 twin-engine King Airs operated by Dynamic Aviation – George says an initial use case will be to free up the company’s pilots on fire watch and maritime missions to direct more of their attention to surveillance. He declined to disclose pricing. Merlin also inked a partnership with Ameriflight, the largest U.S. operator of small cargo planes for UPS, FedEx and DHL, whose chronic problems with retaining pilots intent on moving up to better-paying jobs with the big passenger airlines have worsened as those carriers have swung from pushing pilots to retire early amid the worst of the pandemic to recruiting feverishly to expand service as travel has rebounded.
A retirement wave of Baby Boomers was already expected to fuel a pilot shortfall – CAE estimated in 2020 that 38% of commercial pilots globally were over 50. Pandemic disruptions have tipped the scales more, including increased volumes of air deliveries of e-commerce orders, further whetting investor interest in Merlin’s B Series fundraising round. George says the company reaped almost double the amount it initially set out to raise.
The $105 million round, which boosts Merlin’s total raised to $130 million, was led by Snowpoint Ventures, a private-equity fund focused on dual-use technology that was founded by two current and former Palantir executives, and Baillie Gifford, which until recently has rung up top-tier returns with long-term bets on tech high-flyers like Amazon, Tesla and Netflix. Prior investors including Alphabet’s GV fund also increased their stakes.
If safety clearance of Merlin’s initial commercial system takes longer than expected, the company still stands to make money from a contract with the Air Force to automate the flight deck of the C-130J, which George says is worth in the tens of millions of dollars. The goals of the project mirror the general aspirations for autonomy for larger commercial cargo aircraft, and eventually passenger airliners: to go from two pilots to one, and eventually from one to zero. Congress is supportive: In a report on the annual defense budget bill the House is currently considering, the House Armed Services Committee urges the Department of Defense to use commercial solutions to give its aircraft autonomous capabilities to mitigate the Air Force’s persistent pilot shortage and requests a draft of an implementation plan.
Merlin isn’t the only startup working to robotize older aircraft. Competitors include Skyryse, which has pivoted from an initial goal of launching an urban air taxi network to focusing in the short run on simplifying flight controls to cut down on pilot training time; Xwing, which has built up its own air cargo delivery fleet; and Reliable Robotics, which has also won Air Force development contracts.
In contrast with Merlin, both Xwing and Reliable initially plan to remove the pilot from the cockpit and have remote safety monitors on the ground oversee their aircraft and handle communications with air traffic control. That promises cost savings – one pilot on the ground could eventually oversee multiple aircraft. But George believes that the FAA is wary of the risk of disruptions to communication links between planes and the ground station, and that Merlin, with its onboard safety pilot, will gain thousands of hours of operating experience before competitors are allowed to get off the ground.
Merlin Labs’ strategy of relying on an onboard pilot to detect and avoid hazards and handle emergency situations “is not bad” to address regulators’ concerns in the short run, says Ella Atkins, director of the University of Michigan’s Autonomous Aerospace Systems Lab, but she says it’s not clear whether the company has a path to full autonomy from the “crawl stage.”
And while developing a natural language processing system to interact with air traffic controllers may make sense now, when the FAA is not yet ready to transition to data link-based methods that will have to replace voice communications in order to coordinate the large numbers of delivery drones and air taxis that are expected to crowd national airspace in the next few decades, it’s still a difficult problem, Atkins warns, with fast-talking controllers speaking over sometimes static-fuzzed radio connections. “You get to a point where you’re 90 plus percent accurate in processing what you’ve heard, but getting to 100 is extremely hard.”
Merlin’s system, which will repeat ATC instructions in a synthesized voice to confirm them, could get stuck in a loop of miscomprehension akin to the old Discover Card commercial in which a customer service agent’s mentions of “fraud protection” are repeatedly misheard by a caller as “frog protection.”
“That’s kind of the nightmare for natural language processing,” Atkins says, which could end up distracting the air traffic controller as well as the onboard safety pilot.
And that’s leaving aside the knotty problem of whether Merlin’s system will reliably translate the instructions into the proper actions, even if it transcribes them word for word.
The new fundraising round is going to enable Merlin to hire a lot more people to work on solving those problems. George plans to add another 50 to 60 employees over the next year, taking headcount to 120 to 130.
“We’re really excited about what the cash is going to be able to do for us as we continue to innovate here,” he says.