THE SECRET FUEL POWERING CHINA`S SELF-DRIVING CARS / THE ECONOMIST

Autonomous driving

The secret fuel powering China’s self-driving cars

Responsive regulation—and lots of experiments

Illustration: Rose Wong

IN MARCH THIS year a student driving down a motorway in the Chinese province of Anhui turned on an autopilot system built into her car which is designed to assist—not replace—human drivers. 

Her SU7 electric vehicle (made by Xiaomi, a Chinese tech giant) reminded her to keep her hands on the wheel; eight minutes later, it warned of an obstacle ahead. 

She took control but the car crashed within seconds, killing all three of the people inside. 

Images of the wreckage spread across China. 

They sparked heated discussion about China’s fast-and-furious roll-out of self-driving tech, and the pros and cons of the way it is governed.

China is building a decisive global lead in getting cars with autonomous features on the road. 

Motoring nerds divide cars into six categories. 

A vehicle at level zero has no automation; those at level five do everything by themselves. 

In the consumer market, much of the attention at present is on tech that matches level two or level three. 

An L2 car can steer, accelerate and brake by itself, but requires a human driver’s sustained attention. 

A driver in an L3 car can check e-mails and watch TV, so long as he or she is ready to take over swiftly in an emergency.

These days more than half of new cars sold in China are classed as L2. 

Morgan Stanley, a bank, estimates that by 2030 China alone will account for half of the world’s L2+ cars, an especially clever subset. 

But China also looks ahead of other countries in its roll-out of “robotaxis”: driverless cars classed at L4 that are permitted to ferry around passengers, albeit within limited areas. 

Goldman Sachs, another bank, expects China will have 5,000 of these on the road by the end of this year and 632,000 by 2030 (see chart)—far outpacing America’s 1,800 and 35,000, respectively.

How has China moved so fast? 

In autonomous driving, as in so many spheres of technology, hyper-competition and strong supply chains enable the “China speed” foreign firms covet. 

China’s self-driving dreams have also been brought closer through strong government backing, censorship of news related to accidents and a population optimistic about technology. 

But there is another factor which, though crucial, tends to get less attention: differing approaches to regulation. 

In China, myriad low-risk local pilot schemes have helped engineers and policymakers gain experience. 

Regulators watch carefully to see where the technology is going before they cement national rules.

China’s road-traffic safety law still assumes that all cars have fully active human drivers. 

But leaders in Beijing, the capital, have pulled plenty of other levers to give the industry room to develop. 

They have set gutsy targets for adoption of autonomous features and also pumped out much useful guidance. 

National standards that will apply to L2 and L3 systems are still in draft form (even though lots of L2 cars are already on the road). 

But piecemeal standards on software updates and data recording take effect in January.

If the national government sets the direction of travel, it is in local pilot schemes that the rubber meets the road. 

By 2024 more than 32,000 kilometres of road across China had been approved for testing autonomous driving of some sort. 

In Beijing, by the end of 2023 techies had clocked up 39m kilometres of testing, mostly on the capital’s outskirts. 

Wuhan handed out its first licences to fully driverless robotaxis in 2022; now passengers are able to hail them on one-third of its streets. 

This kind of real-world experience is crucial for ironing out kinks before technology goes national. 

“There’s more room for trial and error. 

If problems arise, you can reverse course quickly,” says Zheng Lin, a law professor at Dalian Maritime University, who studies autonomous driving.

The pilot schemes are not just a test bed for engineers. 

They promote innovation in rule-making too. 

Local officials in China are keen to make their hometowns hubs for new technologies, so as to attract investments and boost their own careers. 

To enable this, more than 50 cities have written their own competing policies on things such as liability in accidents, guidelines for responsible testing, and what subsidies they will make available to firms that test in their neighbourhoods. 

Even boosters admit that this has become a touch confusing. 

But it is nonetheless providing plenty of good and bad examples for lawmakers to learn from.

All this contrasts with the more cautious approach in Europe, where carmakers grumble about the obstacles to testing on public roads. 

In China a “‘test first, regulate later’ model helps regulators and industry learn together and speed up innovation”, says a spokesperson for Volkswagen Group. 

The comparison with America is a little less stark: use of L2 systems, such as Tesla’s “Autopilot” and “Full Self-Driving”, is rising. 

But American automakers face a higher risk of costly lawsuits, public protests and government investigations than do their peers in China.

Will China slow down? 

Since the start of the year, regulation of L2 systems has been tightening. 

In February officials said automakers would have to do more testing and get approval before sending remote updates to systems that are already out in the wild. 

This ruling aims to reduce the risk of bad fixes, and give carmakers more incentive to ensure their technology is top-notch when first sold. 

The well-publicised accident in Anhui has given sceptics ammunition (even if its precise causes do not yet seem to have been clearly ascertained). 

The deaths should “force all car companies to stop their fanatical promotion of intelligent driving”, read a headline from Vista, a Chinese magazine.

Weeks after the crash, China’s industry ministry gathered 60 automakers and sector representatives to warn them against marketing their L2 driver-assistance systems as “autonomous driving”, for fear of misleading drivers into thinking they can stop paying attention. 

Executives scrambled to comply before a big auto show. 

In September regulators published draft national safety standards that all L2 cars will soon have to meet; it may require cars to turn off the driver-assistance system if they detect the driver has become disengaged. 

That same month regulators said they were requiring Xiaomi to fix 117,000 SU7 cars because the driver-assistance system “may not be able to adequately recognise, warn against, or handle extreme scenarios”; the firm said it would correct the problem with an over-the-air update.

The Communist Party is navigating a narrow pass. 

It is adamant that China must dominate the world’s most important new technologies. 

But it also worries that attention-grabbing accidents will threaten faith in the party, which it prizes above all else. 

Its approach has been to give automakers more than the usual space to innovate, but also to have regulators ready to hop into the driver’s seat if things look like they are veering off course.