An autonomous Mayflower aims to prove A.I.’s captain skills by sailing in the Pilgrims’ wake

November 27, 2021, 12:30 PM UTC

Just days before Thanksgiving, Brett Phaneuf, a voluble middle-aged American entrepreneur, sat in a café in Plymouth, England, not far from the spot where the Pilgrims embarked on The Mayflower in September 1620, and talked about their voyage and changing perceptions of risk.

“What I find inspirational, is that these people were willing to take such an incredible risk to sort of jump off into something new, with not just no guarantee of success, but, like, no guarantee of survival,” Phaneuf says.

Now Phaneuf plans to follow them—in a sense.

For the past five years, a non-profit Phaneuf heads, ProMare, has been building an unmanned autonomous ship, powered by a hybrid diesel-electric motor and solar panels, which it hopes will follow in the Pilgrims’ wake, crossing the Atlantic from Plymouth, England, to Plymouth, Massachusetts. If successful, The Mayflower Autonomous Ship, named in honor of its famous nautical forebearer and known as MAS for short, will be the first such trans-Atlantic voyage by an autonomous vessel.

As he plans to recreate the pilgrims’ risky crossing, Phaneuf finds today’s culture illogically risk-adverse. “As a society, we really simply have completely lost the ability to think about risk,” he says. This is particularly true, he says, when it comes to the regulation of new technology, such as artificial intelligence. “There’s just this bias against automation and autonomy,” he said.

Meanwhile, says Phaneuf, the biggest hazards on the water today are inexperienced pleasure boaters.

“But no one talks about banning them,” he said.

The promise of autonomous ships

Phaneuf’s modern Mayflower was envisioned as a testbed for the technology needed to make autonomous ships viable. The market for such vessels, for container transport, research and military applications, is expected to reach more than $160 billion by the end of the decade.

Phaneuf, a Boston-area native who moved to Britain a decade ago for work, has been in discussions with regulators in both the U.S. and the U.K. about what rules should govern ships piloted by artificial intelligence software. One company Phaneuf runs, Marine AI, makes such software. Two other companies he also heads, M Subs Ltd. and The Submergence Group, build manned and unmanned submersibles for use by the U.S. and British Navy. In his view, uncrewed ships captained by A.I. pose far less risk than those helmed by human skippers—but most regulators don’t see it that way, he says.

Dispensing with crews could help ease future supply chain disruptions, not to mention saving the shipping industry millions in labor costs. So it’s not surprising that a number of maritime industry giants, including Maersk and Happag-Lloyd, have been teaming up with software and robotics startups to explore the creation of autonomous ships. Rolls-Royce, a major producer of ships’ engines, is making a big push to develop A.I.-captained vessels too. A Norwegian autonomous container ship, MV Yara Birkeland, this month began testing in the fjord outside Oslo. And Japan’s largest shipping company, Nippon Yusen, plans to send a captainless containership on a 236-mile coastal voyage from Tokyo Bay to the Japanese city of Ise in February.

But so far, none of big league shipping companies has proposed a journey as lengthy and audacious as what Phaneuf has planned for MAS. The 49-foot long, aluminum trimaran was supposed to have made its trans-Atlantic voyage to commemorate the 400th Anniversary of the Pilgrims’ journey, in 2020, but the COVID-19 pandemic delayed the vessel’s completion. In June this year, MAS belatedly embarked from Plymouth bound for America. The journey was expected to take about three weeks. But, after three days at sea, an exhaust pipe broke, preventing the hybrid electric generator from operating. MAS was forced to limp back under reserve power to the Isles of Scilly, which lie just west of the English Channel, and then the ProMare team towed her home to Plymouth.

“Boats always break, so our boat broke—but not the A.I. part,” Phaneuf shrugged.

Phaneuf isn’t one to give up easily. “One setback doesn’t define the whole project,” he said. After all, the original Mayflower also had to turn around—twice—for repairs before it eventually completed its 66-day Atlantic crossing. So Phaneuf redesigned the ship’s engine, Promare upgraded much of the boat’s software for good measure, and the group is planning another attempt to cross The Atlantic this spring. But first, the boat—which has cost Promare about $1.2 million so far, with a variety of corporate partners donating more in equipment and labor—is undergoing extensive testing in the waters near Plymouth, with a longer shake-down cruise to Rotterdam and back planned for February.

It should be easy

In theory, building an autonomous ship ought to be easier than creating a self-driving car. Even in busy shipping lanes, the oceans are less congested than most city roads or highways. “Last time, we travelled 500 miles offshore and we never came within 10 miles of another vessel,” Phaneuf said of MAS’s maiden crossing attempt. Except when coming in and out of port, ships generally have a lot of more freedom to navigate than cars do. Also, in most cases, ships travel much more slowly than cars, meaning there is a lot more time for A.I.-systems to identify potential hazards and maneuver to avoid them.

But training an A.I. system to identify buoys, channel markers, boats and all sorts of other things that float in the sea, from crab pots to logs—and to do so reliably in rain or fog while the ship is also pitching in heavy swells—that is a bit of a trick, says Don Scott, the tall, bearded Canadian, who is the chief technology officer at MarineAI and who oversaw the development of the software systems deployed on MAS. To do so, Scott developed a labelled database of millions of images of marine objects in different light and weather conditions. He installed cameras and sensors at various points around Plymouth Harbor to record activity and create a set of training data for MAS’s A.I.

Mayflower Autonomous Ship at sea
Photo courtesy of IBM

Sitting still alongside the pier in Plymouth, MAS cuts an impressive figure: its main hull and outriggers have reverse bows, designed to cut through the waves more efficiently, giving the ship the profile of some exotic spacecraft. A central platform rises on four struts rises from the middle of the main hull, sporting an impressive array of antennas, radars, sensors and cameras.

IBM took an early interest in the project and became ProMare’s principal technology partner, its corporate logo prominently displayed on the ship’s hull. The company provided some of the software the ship uses to “see” objects around it as well as the software the ship uses to decide how to steer when other vessels obstruct its path. This decision-making system was adapted from software IBM first used to help banks and insurance companies. IBM has also provided MAS with computers that can run A.I. software without needing to send information back and forth to on-shore datacenters. Being able to handle these computing demands “on the edge” without consuming too much of MAS’s limited electric power was an essential requirement. MAS may need to be make quick navigational decisions while in the middle of the ocean, beyond the reach of cellular communications and where satellite connectivity can be slow and bandwidth limited, or sometimes nonexistent.

Other computers onboard the ship will be helping to gather scientific data. One of the big advantages of autonomous ships, as Phaneuf sees it, is their ability to advance human understanding of the oceans. An unmanned vessel can remain at sea longer, at a tiny fraction of the cost, of a human-crewed research ship. With its own propulsion and the ability to navigate independently, it can also cover a wider geographic area than a sensor-laden buoy.

IBM researchers have designed three science projects that MAS will conduct while it is crossing the Atlantic. One is a hydrophone that will be used to listen continuously to sounds underwater, feeding the data to software that has been trained to identify whale songs. Another sensor, called Hypertaste, is an “electronic tongue” that will sample the water around MAS at 15 minute intervals, identifying its chemical composition. The data will be used to study ocean acidification, a process, linked to climate change, in which, as the amount of carbon dioxide in the atmosphere increases, the oceans also absorbs larger and larger amounts of carbon dioxide, becoming more acidic. Finally, researchers will be using video from the ship’s cameras, combined with data from inertial sensors onboard, to train an A.I. system to predict the energy a wave carries based on its appearance. This might one day help scientists make better use of video to understand the potential for storms and erosion, as well helping future autonomous ships steer the safest, most efficient path, through high seas.

Big Blue has contributed in other ways too. IBM owns The Weather Company, which transmits meteorological data to MAS that it will use to help determine the best route across the Atlantic. (In return, the ship is also relaying weather data from its own sensors back to shore-based researchers.) Weather company forecasters are also consulting with the ProMare team to help them find the best weather window for maximizing MAS’s chance of having a successful crossing.

Test drive

It’s a chilly, but calm and sunny early afternoon as Scott and I step aboard a small cabin cruiser in Plymouth Harbor. We watch as Matt Shaw, the 24-year old who serves as chief engineer on the MAS project, opens a laptop inside the pilot house and pulls up a digital dashboard displaying readouts from MAS’s instruments. Using an Xbox controller and a Bluetooth network, Shaw remotely pilots MAS as she motors placidly in front of us, attached to the cabin cruiser by a slack, day-glow orange line. As a precaution, local officials have mandated that MAS be under human control and tethered to a manned boat while navigating in and out of the harbor.

Brett Phaneuf, co-director of the  Mayflower Autonomous Ship project.
Photo courtesy of IBM

Once we reach a designated testing area in Cawsand Bay, just outside the harbor entrance, the tether is released, Shaw taps a key on the keyboard and MAS is off, making her own way between preprogrammed waypoints at a speed of 6.8 knots (about eight miles per hour). “This is her optimal speed for fuel consumption efficiency,” Shaw said. Meanwhile, James Sutton, a software engineer from IBM Research who has worked on the science experiments MAS carries, sits with a laptop perched on his knees, watching the readouts from the scientific instruments onboard.

The cruiser motors along next to MAS and then, pulling slightly ahead, crosses her path. “We’re testing to make sure the sensors can correctly see us and that the collision-avoidance software is working properly,” Scott said. The authorities won’t actually let MAS operate fully autonomously this close to shore, so MAS doesn’t deviate from its set course. But back on shore, in a squat redbrick building right alongside the wharf, other ProMare employees can monitor MAS’s camera feeds and radar data in real-time and also see what decision MAS’s A.I. captain would have taken to avoid a collision if it had been operating autonomously. They radio Shaw to tell him everything appears to be working. “This is the reality of being at sea, nothing much happens,” Scott said.

Phaneuf grates at what he sees at the needless caution of government maritime authorities. “It’s just this sort of hyperbolic sort of safety culture that is being used to justify irrational attempts to regulate things in a way that will retard the growth of the industry,” he said. Of all the many hardships the Pilgrims had to bear, at least excessive maritime regulation wasn’t one of them.

Subscribe to Fortune Daily to get essential business stories straight to your inbox each morning.

Read More

Artificial IntelligenceCryptocurrencyMetaverseCybersecurityTech Forward