Around midnight on June 28, Calvin Hu was driving with his girlfriend near San Francisco’s Golden Gate Park when he pulled up at an intersection behind two white and orange autonomous Chevrolet Bolts operated by Cruise, a subsidiary of General Motors . Another was stopped to his right in the adjacent lane. The light turned green but the cars, which operate in the city without drivers , didn’t move.
When Hu prepared to reverse to make space to go around the frozen vehicles, he says, he noticed that several more Cruise vehicles had stopped in the lanes behind him. Hu, another driver, and a paratransit bus were trapped in a robotaxi sandwich. After a few minutes of bemused waiting, Hu says, he resorted to driving over the curbs of the street’s median to escape.
When he returned on foot a few minutes later to see if the situation had resolved, the Cruise vehicles hadn’t budged. A person who appeared to work for the company had parked in the intersection, Hu says, as if to indicate the street was closed, and was trying to direct traffic away from the immobile self-driving cars. Hu estimates that the robot car blockade, which has not previously been reported, lasted at least 15 minutes.
The Cruise vehicles that trapped Hu weren’t the only autonomous cars holding up traffic in San Francisco that night. Internal messages seen by WIRED show that nearly 60 vehicles were disabled across the city, over a 90 minute period, after they lost touch with a Cruise server. As many as 20 cars, some of them halted in crosswalks, created a jam in the city’s downtown, in an incident first reported by the San Francisco Examiner and detailed in photos posted to Reddit.
In a written statement, the California Department of Motor Vehicles, which oversees autonomous vehicle operations in the state, said it was aware of the incident and would meet with Cruise to “gather additional information. ” The June 28 outage wasn’t Cruise’s first. On the evening of May 18, the company lost touch with its entire fleet for 20 minutes as its cars sat stopped in the street, according to internal documentation viewed by WIRED.
Company staff were unable to see where the vehicles were located, or to communicate with riders inside. Worst of all, the company was unable to access its fallback system, which allows remote operators to safely steer stopped vehicles to the side of the road. A letter sent anonymously by a Cruise employee to the California Public Utilities Commission that month, which was reviewed by WIRED, alleged that the company loses contact with its driverless vehicles “with regularity,” blocking traffic and potentially hindering emergency vehicles .
The vehicles can sometimes only be recovered by tow truck, the letter said. Images and video posted on social media in May and June show Cruise vehicles stopped in San Francisco traffic lanes seemingly inexplicably, as the city’s human pedestrians and motorists navigate around them. By Alex Davies and Aarian Marshall Cruise spokesperson Tiffany Testo provided a written statement that said the company’s vehicles are programmed to pull over and turn on their hazard lights when they encounter a technical problem or meet road conditions they can’t handle.
“We’re working to minimize how often this happens, but it is and will remain one aspect of our overall safety operations,” the statement said. Testo did not respond to specific questions about multiple incidents in which Cruise vehicles stopped in traffic. The outages come at a vital time for Cruise, which is accelerating its autonomous vehicle program on the tricky streets of San Francisco as it competes with well-capitalized rivals like Google’s sister company Waymo , Aurora, and Zoox, which is owned by Amazon.
In the spring, General Motors bought out the SoftBank Vision Fund’s $2. 1 billion stake in Cruise and invested another $1. 35 billion into the self-driving unit.
Just over two weeks after the May outage that froze Cruise’s entire fleet, the CPUC approved Cruise’s permit to charge money for Uber-like ride-hail rides—opening a path to a full commercial robotaxi service that could help the company start to recover the billions it has poured into building its technology. Cruise began testing its autonomous technology in San Francisco in 2015, with safety drivers behind the wheel to intervene if something went wrong. Five years later, the DMV granted approval for the company to test its cars without humans onboard.
Early this year, Cruise invited members of the public to apply to join a select group of testers in the city, who can summon completely driverless rides with an app. The service is available between 10 pm and 6 am, and covers 70 percent of the city, but isn’t permitted to operate in rain and fog. Around midnight on June 21, nearly two weeks after Cruise won permission to charge for rides, San Francisco resident Stephen Merity was walking through the city’s downtown Tenderloin neighborhood when he saw a driverless Cruise stopped in a crosswalk, blocking a right-hand turn lane.
When he returned a few minutes later, he found two more Cruise vehicles stopped behind the first. When another driverless car appeared and started to navigate around its stuck brethren, an apparently inebriated onlooker cheered it on: “You can do it!” Late one night in June, Stephen Merity encountered three autonomous cars stuck in downtown San Francisco. Merity and the robot’s cheerleader told a human driver patiently waiting between two of the motionless Cruise vehicles that she should steer her SUV around the robotaxis.
The vehicles had been stuck for at least 10 minutes before he left to go home, Merity says. The scene initially struck Merity, who works in machine learning, as pretty funny. But after more reflection, he turned anxious about the technology.
When he saw a news report a week about the June 28 outage, he was dismayed. “I had assumed alarm bells were going off at Cruise HQ and they were thinking of pulling the cars off the road and putting drivers back in,” Merity says. Losing connection with its vehicles, and especially its backup safety systems, might violate Cruise’s permits to operate in California, says Bryant Walker Smith, an associate professor at the South Carolina School of Law who studies autonomous vehicles.
The California DMV program that regulates driverless cars requires a vehicle’s operator to certify that it has a link allowing for “two-way communications” between a vehicle—including its passengers—and an employee remotely overseeing the robot’s movements. However, much like autonomous cars themselves, regulations drawn up to apply to the vehicles have not been tested in every possible scenario. Cruise did not respond to specific questions about its permits.
Neither the California DMV nor the CPUC would say how Cruise’s permits might be affected by the outage incidents. The CPUC did not say whether it had responded to the anonymous Cruise employee’s letter, or taken its contents into consideration before approving the company’s permit. Regardless of Cruise’s legal obligations, Walker Smith says that self-driving companies should be open and transparent about what’s happening on public roads.
“From the perspective of the public, when the vehicles do something wrong or weird, it’s on the company—the ‘driver’—to really explain it. ” Screens inside autonomous vehicles stuck near an intersection in downtown San Francisco urged first responders to contact Cruise. People who have seen Cruise’s vehicles on the streets of San Francisco, or ridden in the cars, say the robotaxis generally avoid the busiest streets.
Rodney Brooks, an MIT roboticist and entrepreneur , signed up for Cruise’s taxi service via the company’s online public portal . In May, he and a friend took a series of nighttime rides throughout northern San Francisco. On one trip, Brooks noticed that their driverless taxi went at least 11 blocks further than necessary, roughly half a mile, apparently to avoid the most crowded roads.
The journeys were mostly smooth, Brooks says, though one car he summoned stopped alongside a construction site, forcing him and his friend to walk through traffic to get into the car. For Brooks, robotaxis getting stuck in and blocking traffic is evidence of the challenges faced by Cruise and its competitors as they try to turn promising prototype autonomous vehicles into large-scale commercial services. “A lot of technologists think if you do a demo, then that’s it.
But scaling is what kills you,” he says. “You run into all sorts of things that didn’t happen at a smaller scale. Incidents when Cruise lost touch with its vehicles appear to have caused inconvenience, not injuries.
Some have enjoyed the spectacle. One night in May, Scott Gatz was caught behind four stopped Cruise vehicles, at the same intersection where Hu would be trapped weeks later. A city worker approached the vehicles, looking confused, Gatz says, and later a man with a tablet arrived, who appeared to work for Cruise.
Gatz and other drivers escaped that night by squeaking through a small gap to the side of the robot vehicles. “Cruise really needs to fix their software, and we as a city need to figure out how we can coexist with these cars,” Gatz says—but he’s still glad to see the technology tested in San Francisco. His 12-year-old son, in the car with him that night, thought the whole thing was pretty fun.
Operating large, heavy robots at speed around humans inevitably comes with some danger. On June 3, the day after the company received its permit to charge for rides in California, a Cruise vehicle making a left turn in front of traffic was struck by an oncoming Toyota Prius. A Cruise employee in the car and the Prius driver both had to seek medical treatment, according to a report filed by the company with the DMV, in line with a requirement to report all autonomous vehicle-related crashes.
The report said the Cruise vehicle had stopped in the intersection before completing its turn, and that the Prius had sped straight through its right turn lane, into the motionless robot. In response to that crash, Cruise temporarily reprogrammed its vehicles to make fewer unprotected left turns, according to internal messages seen by WIRED. At an internal meeting, Jeff Bleich, Cruise’s chief legal officer, said the company was investigating the incident, according to a recording reviewed by WIRED.
He also warned employees not working on that investigation to try and tune out crashes or related news reports, saying they were unavoidable and would increase in frequency as the company scaled up its operations. “We just have to understand that at some point this is now going to be a part of the work that we do, and that means staying focused on the work ahead,” he said. On Thursday, the National Highway Traffic Safety Agency said it would open a special investigation into the crash.
In a statement, Testo, the Cruise spokesperson, said the company is “proud” of its safety record, “and it speaks for itself. ”.
From: wired
URL: https://www.wired.com/story/cruises-robot-car-outages/
