By Colleen Mondor
Aviation technology company Merlin, formerly Merlin Labs, completed a 25-flight trial of its Merlin Pilot autonomous flight technology in Alaska earlier this summer. Operated between June and July, the flights were supported by a widely announced $1 million award from the FAA last April.
Merlin has drawn considerable attention for its efforts to commercialize its Merlin Pilot flight control system, and the Alaska activity came on the heels of announcements for $105 million in Series B funding, a partnership with the U.S. Air Force to retrofit the C-130J Super Hercules with the Merlin flight control system, and large-scale fleet conversion agreements from Ameriflight and Dynamic Aviation.
The FAA award supported Merlin’s Part 91, non-revenue operation of a Grand Caravan from Fairbanks International Airport. The Boston-based company’s CEO, Matt George, who last year described his company as “founded to define what’s possible in the next 100 years of aviation,” has portrayed the Alaska exercise as transformative. He told AIN, “The villages we are going to are off the road network and rely on the air cargo network for everything from milk to mail to medicine.
“We firmly believe,” he continued, “if there is a place to go to with autonomy with a human pilot to do some good, Alaska’s a great spot to do it.”
The rural destinations selected for the test flights—Tanana, Galena, Huslia, Fort Yukon, and Prudhoe Bay—are described by Merlin as “underserved by existing air cargo operations.”
However, not everyone is convinced by Merlin’s assertions of improving air services, with some critics highlighting claims made regarding the Alaska exercise as being overstated. For example, with the Merlin’s Pilot system’s acknowledged reliance on air traffic control infrastructure, some have questioned its effectiveness in a state that has 391 public-use airports but only 13 control towers—five of which are at military bases.
Others have questioned Merlin’s assessment of current air service to Tanana, Galena, Huslia, Fort Yukon, and Prudhoe Bay, an evaluation that relies more on Last Frontier myths than modern air traffic statistics and schedules.
“The language this company is using, talking about ‘lack of service’ or ‘dire need of supplies’ is the same boilerplate language that anyone attempting a start-up operation in Alaska uses,” Dan Knesek, v-p of commercial operations at Grant Aviation, told AIN. “They are relying on the same old thing. They are saying what you have to say to get investors interested. It doesn’t reflect reality.”
Grant Aviation, one of the largest Part 135 operators in the state, flies scheduled passengers and cargo in 50 aircraft, including 25 Caravans, throughout rural Alaska.
None of the five destinations selected for the Alaska trials have control towers and all of them commonly receive VFR-only traffic. Further, absolute radar coverage is not available. “ADS-B and radar have black holes all over the state,” noted Knesek, “and we don’t have ATC communication everywhere. When the FAA shows you a map of full ADS-B coverage for Alaska,” he added, “it doesn’t tell you that you have to be flying at high altitude to access it.”
Merlin’s awareness of these infrastructure deficiencies is unclear, although George did tell AIN that the presence of uncontrolled fields with instrument approaches in the state was “really interesting” and “that’s why we’re doing this in Alaska.”
But he further noted that, “speaking broadly,” Merlin Pilot must interact with the ATC system. “Air traffic controllers need to be able to go talk to it just like any other airplane,” he said, “and it needs to go talk back.”
Fairbanks-based Part 135 carrier Wright Air Service, which has 20 Caravans, is the primary scheduled operator for the five destinations named by Merlin for the trials. (Everts Air Cargo, Warbelows Air Ventures, and 70 North are also active in the region.)
“This is such a dynamic environment,” noted Max Hanft, Wright’s chief pilot, “not only with existing weather but also with the availability of ADS-B, radar, ATC, and weather reporting. Inconsistent reporting of weather and runway conditions is a definite issue,” he told AIN. “Taking the human component out of here is not something I would advise,” he added, concluding, “We interact with VFR-only traffic all the time; it’s common in these routes.”
With its heavy reliance on efficient and effective aviation infrastructure, Merlin appears to be seeking something that Alaska cannot provide.
When he spoke with The Verge in 2021, George described the technology for Merlin Pilot as “pretty simple,” and said, “The reason that autonomy up in the air is so much easier is that you have complete vision, at least in the United States, of everything that’s up in the sky, with ground-based radar.”
But for Grant Aviation’s Knesek, such generalities are a problem. “We have known, documented infrastructure failures statewide,” he told AIN. “If this technology is contingent on that infrastructure, it will fail.”
The collision between Merlin’s vision of success and Alaska’s reality begs the question: Can the state provide what the company requires?
The company frequently uses the words “partnership” and “collaboration.” Merlin mentioned Everts Air Cargo, a combined Part 121/135 operator with bases in Anchorage and Fairbanks, as a key partner in the tests. Everts has issued no public statements on Merlin, and when contacted by AIN, a spokesperson said the company would be providing only ground support.
George told AIN that the test site selection was ultimately the result of "a close collaboration with Everts and UAF [University of Alaska Fairbanks].” The Everts spokesperson said the company had no role in the selection, noting that Everts provides no scheduled service to Tanana or Prudhoe.
A more vocal Alaskan presence for Merlin has been the Alaska Center for UAS Integration (ACUASI) at UAF. “We teamed up with UAF to do this,” George said, and the Merlin press release on the FAA award included several quotes from ACUASI director Catherine Cahill. When AIN asked about UCUASI’s role in the destination selection, Olena Ellis, ACUASI program manager, wrote, “We were not involved in the selection process for the destinations for the testing.”
However, Merlin’s v-p of business operations, Ashley Pelzek, noted: “Merlin has a long-standing relationship with UofFA [sic] …these routes are well established and offer diverse terrain and climate. They continued to make the most sense for Merlin’s flight trials.”
Regardless of how the Merlin routes came to be selected, they originated from the ACUASI test site at Fairbanks International. Established in 2012 as a research facility at the UAF Geophysical Institute, ACUASI is one of seven FAA UAS test sites. (A formal working relationship with one of those sites was an agency requirement to be considered for the award.) Its director has been vocal in her support of Merlin and other unmanned technology.
“Alaska’s terrain and inclement weather can challenge the most experienced pilots,” Cahill said in the Merlin press release last April. “And yet remote communities rely on air cargo deliveries for vital supplies such as milk, mail, and medicine. Working with Merlin on these flight trials benefits our residents and provides data with each flight that will create not only a safer airspace in Alaska but technology that is applicable worldwide.
“This program,” she continued, “will help thousands of our state’s remote residents to acquire supplies necessary to sustain life and it’s exciting that the advent of new technologies can drive greater equity and access across our communities.”
Cahill used similar language when testifying before the House of Representatives' aviation subcommittee in March. Speaking on behalf of drone technology, she noted that it would have “a direct impact on the quality of life” in rural Alaska, especially with deliveries of such things as “diapers and milk.”
Both Cahill and George have not discussed current commercial aviation activity in Alaska. For Wright Air Service, that issue hits close to home. “I wonder what level of market inquiry or assessment was conducted,” Hanft said. “We are into Fort Yukon three times a day at least,” he continued. “Whatever perceived lack of service they came up with on these routes is not the reality.”
This point is even more perplexing when considering Prudhoe Bay, which is on the state-maintained road system and accessible by ground transportation year-round. Deadhorse has a 6,500-foot paved runway and serves as a hub for surrounding North Slope oil fields and villages. According to the Department of Transportation, it received 20 million pounds of cargo in 2022. Galena’s airport includes a 7,249-foot paved runway; it is also a hub.
Regarding mail, which is how medicine, milk, diapers, and nearly everything else is shipped to the Bush, Part 135 operators delivered over 71 million pounds in the first 11 months of 2022. Seventeen of the operators making those deliveries provided both scheduled passenger and cargo service, with only two carriers, Ryan Air Service and Alaska Central Express, solely flying cargo. This is by design, due to the heavily regulated nature of flying the mail in Alaska.
“The mail and passengers from the [rural Alaska] ecosystem...you can’t remove one without damaging the other," said Knesek.
In Alaska, whenever possible, Bush mail will always go to passenger carriers first. While both Cahill and George have commented publicly on flying the mail, neither has discussed how it is distributed or its high degree of dependence on passenger air service. This is significant given expectations that freight-only autonomous flights are likely to be approved well ahead of passenger-carrying services.
George formed Merlin in late 2018 after the closure of his first start-up, Bridj, a “pop-up bus service” for Boston that he claimed would “rethink the way mass transportation works.” Despite raising $11 million from investors in three years, Bridj abruptly shut down in 2017 after the unsuccessful pursuit of an exclusive funding agreement with an auto manufacturer.
A licensed pilot, George pivoted to autonomous aircraft and founded Merlin, then called Apollo Flight Research. Coming out of stealth mode in 2021, Merlin announced a Series A investment round of $25 million, led by GV and First Round Capital, followed the next year by a $105 million round co-led by Snowpoint and Baillie Gifford. The company’s current directors include Andy Wheeler of GV, Josh Kopelman of First Round, and Doug Philippone of Snowpoint.
Like some of its rivals in the autonomy field, Merlin is vague about how its flight control system works. Its website describes the product as “configurable software,” adaptable to any aircraft, that can be integrated “into the existing deck to augment current flights.”
How this technology will work when encountering particular situations in Alaska is unclear and the company has not been forthcoming in providing requested details. When asked about the procedure for operating at VFR-only airports, the company replied, “Flight planning is done just like any other human-operated flight. The system leverages waypoints in the national airspace system and can leverage existing approaches, including instrument approaches.”
A query on the procedure for operating in uncontrolled airspace and airports garnered this response: “Merlin has a robust drop-in autonomy kit that is built in partnerships with vendors around the world. The drop-in autonomy kit enables the Merlin Pilot to control a wide variety of existing aircraft.”
Descriptions of the component parts of the system have varied and the company has not revealed its cost to operators. Questions remain about whether it possesses the flexibility required for piloting an aircraft in a demanding environment.
In The Verge, George described the use of future remote human pilots acting as supervisors, “monitoring dozens of aircraft in the sky at once, but leaving the vast majority of the tasks, from communication with air traffic control to navigation, to the autonomous software.”
AIN’s FutureFlight explained that the system consisted of “a combination of software, computers, servos, actuators and sensors that the company claims collectively fulfill the functions of a human pilot.” While the FAA permitted optionally piloted tests at this point, with the company controlling from the ground for some operations, all fight testing, according to George, has a “monitoring” pilot onboard but the system was “fully in charge.”
By 2022, the ground-based pilot was no longer mentioned in interviews, with an onboard human pilot now flying “alongside Merlin’s system.” Popular Science wrote that Merlin Pilot was software “plugged into the plane” that would get sensor information as data rather than via instrument display. Merlin will be “skipping the physical interface and going directly to the electrical controls.” On older aircraft, this would mean “adding servos and actuators so that the system can work with the plane.”
George told the magazine the company was still discerning the best way for the human pilot to check the Merlin Pilot, but said human override will be available. Since then, he has referenced the use of AI technology to enable the system to perform human pilot activities.
Merlin Pilot’s most unique feature is that it communicates with ATC, recognizing and interpreting instructions, although whether it will repeat directives back to controllers for verification is unclear. The human safety pilot will intercede when necessary, which partly explains why the company has pulled back from plans to offer full autonomy and now takes an approach it refers to as “crawl, walk, run.” This means first semi-autonomous flight, then autonomous cargo flight, culminating with autonomous passenger flight. There is no timeline for when this will happen, and in Alaska, the human pilot served as pilot-in-command.
Following the completion of the Alaska test flights, the University of Alaska issued a press release for the August Global Autonomous Systems Conference celebrating Merlin’s “fully autonomous aircraft.” To clarify any confusion over the requirement for a human pilot to legally serve as pilot-in-command, AIN reached out to the FAA. In response, the agency said it does not define “fully autonomous” and added that there are ongoing discussions in the aviation industry on standard levels of automation.
When Merlin announced its second round of funding, George declared in an interview that the company “was founded to define what’s possible in the next 100 years of aviation.” This optimism is a regular part of the company’s presentation and echoes earlier comments with Fast Company, where George said Merlin’s C130J autonomy contract with the U.S. military “far exceeds the traditional million, million and a half dollars. Ours is much larger than that.” A year later he told Forbes it was worth “tens of millions of dollars.” But when AIN sought to ascertain specifics of the deal, Merlin stated that it was unable to provide them as the company was “actively working on news regarding Merlin’s USAF relationships.”
Attempts to obtain information from prospective customers Dynamic Aviation and Ameriflight with details on their years-old agreements with Merlin were similarly unsuccessful. Early this year, Ameriflight did announce agreements to purchase 35 VTOL air cargo drones from Sabrewing Aircraft Company and 20 autonomous aircraft from Natilus. These purchases would apparently supplement its 2021 deal to convert its entire fleet to the Merlin system.
A standard approach for many in the AI industry is to speak loudly about deals and dollars while glossing over the research and development details. And, Merlin has yet to address a myriad of issues, from the lack of certified weather and incomplete radar coverage that Alaskan pilots and operators have struggled with for decades to the refined intricacies of the state’s air cargo network and routine nature of its selected routes.
When asked how the company would address VFR into IMC, the primary cause of fatal Part 135 crashes in Alaska, George demurred, saying, “There’s a wide variety of accidents, and VFR into IMC is a really common one, but we’re focused first on reducing workload in the cockpit.”
Meanwhile, Max Hanft, with decades of experience in interior Alaska, remains perplexed by Merlin’s choices and stressed to AIN that he is all for improving technology for aircraft but, to him, Merlin Pilot is just an advanced autopilot. “They are creating a solution for the wrong set of problems,” he said.
For Dan Knesek, the bottom line to this whole conversation is obvious: “Before advanced air mobility can come to the state,” he said, “we need advanced infrastructure, including weather reporting and ATC, to support it.”
“Alaska does not have the infrastructure to support either semiautonomous or autonomous flight,” Knesek concluded. “The FAA needs to update our current technologies if they want to get serious about improving Alaska’s air safety.”
Read more here.