MILTON KEYNES, England — Autonomous vehicle startups have raised tens of billions of dollars on promises to develop truly self-driving cars, but industry executives and experts say remote human supervisors may be permanently needed to help troubled robot drivers.
The central premise of autonomous vehicles — that computers and artificial intelligence will dramatically reduce accidents caused by human error — has driven much of the research and investment.
But there’s a catch: Making robot cars that can drive safer than humans is extremely difficult, because self-driving software systems simply don’t have the ability of humans to quickly predict and assess risks, especially in the event of unexpected incidents or ‘edge cases’. .
“Well, my question would be, ‘Why?'” said Kyle Vogt, CEO of Cruise, a division of General Motors, when asked if he could see a point where remote human supervisors should be removed from operations.
“I can provide my clients with peace of mind knowing that there is always someone there to help if needed,” said Vogt. “I don’t know why I would ever want to get rid of that.”
This is Cruise’s first time recognizing the long-term need for remote human operators.
Like air traffic controllers, such human supervisors can sit tens of hundreds of miles away, monitoring video feeds from multiple AVs, sometimes with a steering wheel, ready to board and stall robotic drivers getting back into motion — AVs invariably stop when they’re not there. find out what should we do.
Waymo and Argo of Alphabet Inc, backed by Ford and Volkswagen, declined to comment when asked the same question.
GM recalled and updated software in 80 self-driving Cruise vehicles this month after a June crash in San Francisco that injured two people. US safety authorities said the recalled software could “incorrectly” predict the path of an oncoming vehicle, and Cruise said the unusual scenario would not repeat itself after the update.
For some, the idea that human supervisors could be here raises more doubts about the technology.
Truly autonomous vehicles are lagging far behind the optimistic rollout schedules predicted just a few years ago.
In 2018, GM asked the U.S. government for approval for a fully autonomous car with no steering wheel, brakes, or accelerator pedals to be added to its commercial fleet in 2019. That vehicle, the Cruise Origin, will now go into production in the spring. 2023, Vogt said.
In 2019, Tesla CEO Elon Musk promised a million robot axes “next year for sure” — though his company’s “Full Self Driving” offering has been criticized for its cars not being able to drive themselves without a human behind the wheel. and ready to drive manual control in case of emergency.
In a June interview on YouTube, Musk said developing self-driving cars is “much harder than I initially thought, by far”. But when asked for a timeline, he said Tesla could make it “this year.”
Tesla did not respond to a request for comment for this story.
The unfulfilled promise of true autonomy has increased the stakes for the AV industry.
“If these companies fail in the next two years, they won’t exist,” said Mike Wagner, CEO of Edge Case Research, which helps AV companies assess, manage and underwrite risk. “It’s a matter of hang up or shut up at this point.”
WATCHING PEOPLE FROM DISTANCE
Many AV startups today use humans as remote supervisors, in addition to safety drivers sitting behind the wheel.
Those remote people are an added expense, but self-driving cars help deal with edge cases. This could be something as basic as an unknown series of lane closures during road works, or erratic, unpredictable behavior by pedestrians or human drivers.
When a robot driver encounters an edge case, “he raises his hands and says, ‘I don’t know what’s going on,'” said Koosha Kaveh, CEO of Imperium Drive, which uses humans as remote controls for cars in the English town of Milton. Keynes. In time, those people will act as “air traffic controllers” and oversee a growing number of autonomous cars.
Cruise’s Vogt says the company’s AVs on San Francisco’s roads are currently dependent on people less than 1% of the time. But for hundreds, thousands, or even millions of AVs, that would add up to a significant amount of time on the road waiting for human guidance.
Kaveh of Imperium Drive said that as more self-driving cars — which are more predictable than humans — hit the road, the number of edge cases will drop, “but you’ll never get to zero edge cases.”
“Even decades from now, you won’t get 100% truly autonomous vehicles,” Kaveh added.
Yet competition is increasing. Some Chinese cities are pushing to allow active AV testing faster.
The need to address edge cases and reduce the cost of everything from sensors to the number of people in the loop to get to market has also increased as investor funding for autonomous cars has plummeted.
Doubt has arisen as investors wonder how quickly autonomous companies will become profitable. Simpler or slower AVs like trucks or last-mile delivery services that operate on highways or on low-speed fixed routes are likely to become profitable first, but it will take years to get there.
Overall investment in future mobility startups has slowed, with AV-focused companies being hit particularly hard, accounting for less than 10% of venture capital investment in the second quarter, according to investor website PitchBook.
Investments in AV startups fell to $958 million in the quarter. Just two years ago, AV investment was booming as Alphabet’s Waymo raised $3 billion, Didi’s AV unit raised $500 million and Amazon.com Inc. Acquired AV startup Zoox for $1.3 billion, according to PitchBook.
‘RUSH TO MARKET’
Autonomous systems are not as capable as humans because their “perception and prediction algorithms are not as good as how a human brain processes and decides,” said Chris Borroni-Bird, an independent consultant who previously led advanced vehicle programs at GM and Waymo.
For example, if a human sees a ball rolling on the road — harmless in itself — he’ll assume it can be followed by a child and hit the brakes much faster than an AV, Borroni-Bird said.
“I fear AV companies will rush to market without proving that safety is better than human-powered vehicles,” he added.
The problem is that there are “tens of billions of potential edge cases” that AVs can encounter, said James Routh, CEO of AB Dynamics, which tests and runs simulations on cars, including on the Advanced Driver Assistance Systems (ADAS) that form the basis of autonomous driving.
Auto data startup Wejo Group Ltd receives 18 billion data points from millions of connected cars every day and helps with simulations for AVs, said Sarah Larner, executive vice president for strategy and innovation.
“But there are so many variables, like the weather, you can take an edge case and then have to layer all the different variants,” she said. “It’s really millions of outputs.”
DRIVER-FREE DELIVERY
In its automotive track tests, AB Dynamics is using a robotic arm it plans to retrofit on slow-moving mining and farming trucks to make them largely autonomous.
Routh provides a team of remote people who oversee fleets of, for example, self-driving mining trucks operating in closed environments.
He sees that scenario not working for vehicles in faster, more open environments, as it can be difficult for remote human supervisors to react quickly enough to hazards.
Within the next 12 months, UK online food delivery and technology company Ocado Group Plc will roll out a small fleet of self-driving delivery vehicles with Oxbotica, an autonomous vehicle software startup – supported by remote human supervisors – that will operate on set routes just a few blocks away in a small town in the UK and never drive at speeds exceeding 30 miles per hour.
“If a vehicle panics at 30 miles per hour, it can step on the emergency brake and seek help,” said Alex Harvey, Ocado’s chief of advanced technology. “This feels like a very viable low-speed strategy.”
“But you can’t play that game on a highway,” Harvey added, because hard stops in edge cases would pose a safety risk.
Harvey said it would take about five years for Ocado to develop a profitable driverless delivery system. More than half of Ocado’s UK customers could be reached with AVs traveling no more than 40 mph, he said. Eventually, the service could be rolled out to Ocado customers such as US retailer Kroger Co.
Related video: