Tesla’s plan to build fleets of self-driving cars suffered a setback on Friday when the main federal auto safety regulator said it was investigating whether the technology was to blame for four collisions, including one that killed a pedestrian.
The regulator, the National Highway Traffic Safety Administration, said it was examining whether the software, which Tesla calls supervised full self-driving, had safeguards in place to require drivers to retake control of their cars in situations the autonomous technology could not handle on its own.
As sales of Tesla’s electric cars have slowed, Elon Musk, the company’s chief executive, has staked the company’s future on software that allows cars to navigate, steer and brake without human supervision. Last week, the company held an event at the Warner Bros. studios near Los Angeles to unveil what it called a cybercab, which Mr. Musk promised would be able to ferry passengers without a human driver.
But such software has faced persistent criticism from regulators and safety experts who say it does not do enough to make sure drivers remain alert and ready to take over if the system makes a mistake. Tesla faces numerous lawsuits from people who blame the software for injuries or deaths of loved ones.
Mr. Musk said last week that the self-driving taxi, which has no steering wheel or brake pedal, would be available by 2027 and cost less than $30,000. But the investigation by the safety agency is an indication that, even if Tesla succeeds in perfecting the technology, it will still face significant regulatory hurdles.
Tesla did not respond to a request for comment. Mr. Musk has frequently said that Tesla cars operating in self-driving mode are safer than human drivers.
The crashes highlighted by the safety agency on Friday took place when road visibility may have been limited by glare from the sun, fog or dust, the federal safety agency said. Tesla’s self-driving software depends on cameras to operate, unlike other manufacturers who also use radar or laser technology that are often better at detecting objects and people when the view is obscured by poor weather or bright sunshine.
The agency said it would “examine the system’s potential failure to detect and disengage in specific situations where it cannot adequately operate.”
In one of the collisions, a pedestrian died. In another, a person was injured, the agency said.
At the cybercab event last week, Mr. Musk said that the self-driving technology would allow riders to “fall asleep and wake up at your destination.” Such promises could mislead Tesla owners to thinking that the technology available today is more capable than it is, said Michael Lenox, a professor at the Darden School of Business at the University of Virginia.
“His rhetoric seems to outstrip the technology,” Mr. Lenox said.
The investigation by the federal safety agency covers 2.4 million Tesla vehicles, including cars manufactured as far back as 2016. All of Tesla’s passenger models are involved, the agency said, including the Model 3 and Models S sedans, the Model X and Model Y sport utility vehicles, and the Cybertruck pickup.
Federal officials have also been investigating a less capable Tesla system known as Autopilot for several years. These investigations may not survive if former President Donald J. Trump is elected next month. Mr. Trump has said he will appoint Mr. Musk, one of his most prominent supporters in the business world, to lead a “government efficiency commission.”
Mr. Musk has often complained about government regulation, although Tesla has also benefited from federal and state programs, including billions of dollars in clean air credits. In the first six months of this year alone, the credits added $1.3 billion to Tesla’s revenue.
Even if a Trump administration eased regulatory scrutiny of self-driving systems, Mr. Lenox of the University of Virginia said, there would still be public pressure to ensure the technology was safe.
“There are some natural dynamics that will take place regardless of who’s in the White House,” Mr. Lenox said.
Tesla is one of several companies competing to create autonomous taxi services that analysts say could one day be worth trillions of dollars globally. But it’s not yet clear if the technology can work in all places and in all weather and traffic conditions.
In China, the tech giant Baidu is testing autonomous taxis on public streets. Waymo, a subsidiary of Google’s parent company, has been operating driverless taxi services in San Francisco, Phoenix and Los Angeles. Zoox, owned by Amazon, is testing a driverless van that has no steering wheel with plans to start a commercial taxi service.
Others have also run into problems with regulators. Cruise, the autonomous driving unit of General Motors, agreed last month to pay a $1.5 million penalty the federal government for failing to properly report an accident last year in which one of its self-driving taxis severely injured a pedestrian.
<