Shulou(Shulou.com)11/24 Report--

Beijing, September 13 (Xinhua) self-driving car startups have made a beautiful promise to develop real self-driving cars and raise tens of billions of dollars as a result. But carmakers may always need remote human supervisors to help troubled robot drivers, industry executives and experts say.

Remote operator of ▲ self-driving car

For self-driving car development, its core premise is that computers and artificial intelligence can significantly reduce accidents caused by human error, a vision that has driven the industry to do a lot of research and investment.

Fatal deficiency, however, there is also a problem: it is very difficult to build automated cars that are safer than human driving, because self-driving software systems simply do not have the ability of humans to quickly predict and assess risks. especially in the case of accidents or "marginal cases" (extreme cases).

Asked if he could one day consider abolishing the remote human supervisor, Kyle Vogt, Cruise CEO of GM's self-driving subsidiary, said: "well, I would ask, 'Why would I do this?'. I can reassure my clients that there is always someone there to help if needed. I don't know why I want to get rid of him."

▲ GM's Cruise self-driving test car

This is the first time Cruise has acknowledged that there is a long-term need for remote human operators in self-driving cars. In June, two people were injured when Cruise's self-driving car crashed in San Francisco. This month, GM recalled and updated the software of 80 Cruise self-driving cars. Us safety regulators say the recalled software may "mispredict" the path of oncoming vehicles. Cruise said that this unusual situation will not happen again after the update.

For some, the idea of keeping human supervisors raises more questions about autopilot technology. Real self-driving cars lag far behind the optimistic launch of the industry a few years ago.

In 2018, GM sought approval from the US government for a fully autonomous car without steering wheels, brakes or accelerator pedals and promised to include it in its commercial shared travel fleet by 2019. But now, Vogt says, the car, called the Cruise Origin, won't start production until the spring of 2023.

When Musk drew the pie in 2019, Elon Musk, CEO of Tesla, promised that 1 million self-driving taxis would be "sure next year", although Tesla's "fully self-driving" (FSD) system had been criticized because Tesla's FSD could not drive without the supervision of safe drivers and required manual human control in emergency situations.

In an interview on YouTube in June, Mr Musk said developing self-driving cars was "much more difficult than I initially thought". But when asked about the launch timetable, he said Tesla could achieve it "this year".

However, Tesla's FSD is still being tested. Musk also tweeted today to remind FSD beta users to be safe. "FSD Beta 10.69.2 is now available. Please note that safety is a top priority, so be careful, especially around pedestrians." He said.

▲ Musk reminds FSD testers to be safe

For self-driving startups, the promise of true autopilot has not been fulfilled, which increases their risk. " If these companies do not succeed in the next two years, they will cease to exist, "says Mike Wagner, Edge Case Research CEO, a company that helps self-driving car companies assess, manage and underwrite risks." on this issue, either take action or shut up. "

Extreme situations still require humans. Today, many self-driving start-ups use humans as remote supervisors while keeping human safety drivers behind the steering wheel.

These remote supervisors incur additional expenses, but can help self-driving cars deal with marginal cases, which may include some basic things, such as unfamiliar lane closures during road construction, or unstable and unpredictable behavior of pedestrians or human drivers.

Koosha Kaveh, CEO of Imperium Drive, points out that when a robot driver encounters a marginal case, he is overwhelmed and says, "I don't know what happened." Imperium Drive uses humans as remote operators of cars in Milton Keynes, UK. Over time, these people will play the role of "air traffic controllers", supervising more and more self-driving cars.

Cruise CEO Vogt says his self-driving car currently relies on humans for less than 1 per cent of its time on San Francisco roads. But for hundreds, thousands or even millions of self-driving cars, that adds up to a lot of time on the road waiting for guidance from human supervisors.

As more and more self-driving cars hit the road that are more predictable than humans, the number of marginal cases will decline, "but you will never drop to zero," says Carvey of Imperium Drive, a real self-driving company.

"even in a few decades, you won't be able to achieve 100% real self-driving cars." Kavi added.

But competition is intensifying. Some Chinese cities are pushing to allow active testing of self-driving cars faster. As investor investment in self-driving cars has fallen sharply, start-ups are under pressure to resolve marginal cases faster and cut costs for everything from sensors to the number of people to enter the market.

Skepticism is emerging as investors wonder how long it will take for self-driving companies to turn a profit. Simpler or slower self-driving cars may be profitable first, such as trucks operating on roads or set low-speed routes or last-mile distribution vehicles, but it will take years to achieve.

▲ 's investment in autopilot has fallen sharply this year.

Overall investment in future mobile start-ups has slowed, with companies focused on autopilot particularly hard hit, accounting for less than 10 per cent of venture capital in the second quarter of this year, according to PitchBook, an investment website.

Investor investment in self-driving startups fell to $958 million in the second quarter, according to PitchBook. Just two years ago, investment in self-driving cars was in the ascendant. At the time, Waymo, owned by Google's parent company Alphabet, raised $3 billion, Didi's self-driving car division raised $500m, and Amazon bought self-driving car startup Zoox for $1.3 billion.

Eager risk Chris Borroni-Byrd (Chris Borroni-Bird) is an independent consultant who has led advanced car projects at GM and Waymo. Autopilot systems are not as capable as humans, he says, because their "perception and prediction algorithms are not as good as the processing and decision mechanisms of the human brain".

Byrd said, for example, that when humans see a ball rolling onto the road, they think that children may follow it and brake faster than self-driving cars.

"I am worried that self-driving car companies will rush to market them without proving that they are safer than human drivers." Byrd said.

The problem, says James Ross (James Routh), CEO of AB Dynamics, is that self-driving cars could encounter "tens of billions of potential marginal cases". The company tests and simulates road conditions on cars, including the Advanced driving Assistance system (ADAS), which is the basis of self-driving functions.

Sarah Larner, executive vice president of strategy and innovation at Wejo Group, an automotive data start-up, says they receive 18 billion data points a day from millions of connected cars and help self-driving cars simulate road conditions.

"but there are too many variables, such as the weather, and you can come up with a marginal case, and then you have to consider all the different variables," she said. "the output is really millions."

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.