A delay in General Motors Co’s deployment of its self-driving vehicles is fuelling increased scepticism that self-driving cars are not ready to ply America’s roads.
GM Cruise LLC, the autonomous-vehicle unit of General Motors Co, confirmed Wednesday it will delay indefinitely its plan to deploy driverless taxis in the streets of San Francisco – explicit acknowledgement that hype generated by tech enthusiasts and investors is outpacing both engineering reality and safety concerns.
“The technology is simply not ready yet,” said Sam Abuelsamid, principal analyst at Navigant Research who focuses on mobility. “It is not sufficiently safe, reliable and robust to be able to handle all these situations that people are going to encounter in the real world.”
The Detroit automaker set a self-imposed deadline of year-end to launch a self-driving taxi service from the GM Cruise unit. In a Wednesday blog post, Cruise CEO Dan Ammann wrote that the company would need to increase its testing and validation miles to reach “the level of performance and safety validation required to deploy a fully driverless service”.
That means the company won’t begin the taxi service this year. Ammann did not set a new timeline. The news comes as industry leaders and experts become increasingly cautious about the speed at which autonomous vehicles will hit roadways around the world.
“It is not surprising that they have had some delay,” Autotrader Executive Analyst Michelle Krebs said. “I think most people are in agreement that a few years ago everybody got too exuberant on how quickly self-driving cars would get on the roads.”
The delay also is a setback to smart-money investors looking to own the crucial technologies needed to deliver next-generation Auto 2.0.
GM’s timeline for autonomous-vehicle deployment had been one of the most ambitious of any major automaker. When it first announced the 2019 launch, GM was poised to beat crosstown rival Ford Motor Co to market by two years. The Detroit automaker and its Cruise team attracted major investment and garnered favour on Wall Street for its approach to developing autonomous vehicles.
“If we do this right in our first deployment it’s going to allow us to move faster everywhere else,” Cruise spokesman Ray Wert said July 24. “We’ve got to win the tech race and the trust race.”
To be able to launch successfully, he added, Cruise must start ramping up more miles, building EV infrastructure to support the vehicles and heavily engaging the community to win trust.
Winning the trust race will be more difficult. The public’s trust in self-driving vehicles dipped after a pedestrian was hit and killed by one of Uber’s self-driving cars in Arizona in March 2018. Highly publicised accidents involving Tesla electric cars driven in “AutoPilot” mode don’t help.
Research has shown that the more experience consumers have with self-driving vehicles, they more likely will be won over by them, Krebs said: “After that Uber accident there was a lot of concern about safety. I think GM has recognised that, as have the others.”
In an August 2018 Cox Automotive study, safety perceptions of fully autonomous vehicles dropped 19% in 2018 from 2016. For partially autonomous vehicles, perceptions dropped 14%.
What Cruise and other autonomous companies have to do is create a car that can do everything a human can do behind the wheel, but better, experts say. That can be difficult when there are so many elements from pedestrians to potholes that a driver can experience even just driving down the road.
“These systems are still not as good as humans are, despite our flaws,” Navigant’s Abuelsamid said. “Until these systems can become better than humans, people aren’t going to trust this technology. We can’t put this stuff out there and have it crashing into people and crashing into cars and expect people to accept the technology.”
Within at least the next five years, Abuelsamid believes there will be some limited areas that have the right infrastructure and conditions for autonomous vehicles. He expects it will take another 15 years before vehicles can drive themselves everywhere at any time and under any conditions.
Cruise has 180 vehicles currently deployed for testing with AV trainers in them. Last year, they drove the equivalent miles of every street in San Francisco at least once every day, Wert said.
Still, Cruise needs more time to test in San Francisco before it can deploy there, Ammann wrote. He also said GM wants to build trust in Silicon Valley.
“Having our cars running many more miles on the road will further accelerate our rate of learning and safety validation,” he wrote. “It will also give us crucial operational learnings from running a larger-scale fleet and a larger-scale ride service, which we currently operate for our employees.”
Automakers still must contend with the federal government to get self-driving vehicles on the road legally. Federal motor vehicle standards currently require cars to have human operators, but each automaker is allowed to request 2,500 exemptions to those rules that last for just two years. This has cleared a path to be able to test the vehicles, but not for them to be mass-marketed.
Legislation that stalled last year in the Senate would have pushed that number to at least 80,000 exemptions, while a House version favoured 100,000 exemptions per automaker.
The National Highway Traffic Safety Administration took 14 months to respond to GM’s petition to deploy vehicles without steering wheels or pedals on public roads.
Delays are not unprecedented. Uber suspended all testing of self-driving cars last March after the fatal crash. What followed was an industry-wide reckoning with autonomous-vehicle testing practices. Many companies took their driverless test vehicles off the roads while they revamped testing practices. Uber wouldn't resume autonomous vehicle testing for another nine months.
Waymo LLC, the self-driving unit of Google parent Alphabet Inc, walked back promises to take human safety drivers out of its autonomous Chrysler Pacifica minivans. And GM quietly abandoned plans to begin testing autonomous vehicles in New York City.
Abuelsamid said he would rather see companies take it slower rather than rushing to some arbitrary date when it isn’t ready.
“Putting it out there when it’s half-done, when it’s half-baked, and relying on people in the real world to beta-test it is a reckless approach to the problem.” – The Detroit News/Tribune News Service