An SFPD policeman stopped a self-driving car. What does the viral event mean for the future of SF?

An autonomous Chevy Bolt headlight from a cruise in San Francisco was stopped for not lighting. It was stopped for a while by police, then departed, crossed an intersection and stopped in front of a Chinese restaurant, after which it “went to a safe place near the traffic stop,” Cruz explained.

The officer and two colleagues surrounded the vehicle, tried to open the doors, peeked out the windows, and the lights were on inside. Eventually, according to Cruise and the SFPD, officers called Cruise, who took over the car’s remote control. No quote provided.

Experts say the incident shows that autonomous-car companies still have a way of detecting human-robot interactions – some of the shortcomings may have been corrected by basic common sense.

Such situations are now more common as General Motors’ spin-off cruise and Google Thai Alphabet’s self-driving division Vemo both drive autonomous cars without anyone behind the wheel on California public roads.

Both companies are testing robot-taxi rides for staff and select members of the public in San Francisco, offering its winding streets, steep hills and numerous two-wheelers and pedestrians the popular “dark” traffic training ground.

Robot car companies “try to lead regimes full of uncertainty and flaws in the small dances between a law enforcement officer and a vehicle,” said Bryan Walker Smith, a law teacher at the University of South Carolina. Internet and community. “We must engage in the structures of the present, as well as the iniquities of the past, as we think fit for the future.”

Cruise has a comprehensive manual and video guide for first responders who should call 1-888-662-7103 when problems arise. Waymo has similar documents and call number. Both said they would conduct exercises for first responders.

But for officials who have not read Cruise’s 24-page document or watched its 19-minute video, there is no clear way for its cars – perhaps a card printed on the dashboard? – Instructing those outside of them on how to reach a man.

Wendy Zhou, a professor of information science at the Jacobs Technian-Cornell Institute at Cornell Tech in New York and an expert in robot-vehicle communications, said the car and / or how it interacts with its human back-up team should be very clear. . He has worked as a consultant for Cruz.

“Police don’t have to punch around to find it,” he said.

Waymo, who has been based in Phoenix for many years, said it has a strong driverless robot-taxi fleet and has developed its contacts with first responders over time.

The Waymo spokesman Nicholas Smith said in an email that the remote monitoring team “will be notified if a vehicle is stopped by police and will immediately roll over the windows and communicate with the officer through the car’s audio system.”

Cruz described the incident in more detail in response to questions from The Chronicle.

All of its cars can recognize emergency vehicles by their lights and sirens. At this point, its car detected the car’s headlights and stopped in its tracks. Remote cruise operators were told to pull it off, “at that point, the AV had identified a safe pullover spot nearby across the intersection,” the company said.

An electronic tablet in the front console of the car showed a message saying, “First responders should contact Cruise at 888-662-7103.” It did not explain why the authorities at first seemed embarrassed to see the news.

After SFPD officials called Cruise, its remote operator rolled up the car’s windows for officers and said it “solved the headlight problem.” Cruise said the headlight snap was caused by human error.

Cruise’s cars are equipped with two orange buttons, which can be pressed to allow passengers to talk to a human. But the company says law enforcement should call its 1-888 number rather than using buttons.

David Zurkin, managing director of communication design at the Center for Design Research in the Department of Mechanical Engineering at Stanford, presents a teaching moment about the cruise incident – and explains the type of communication that can occur most frequently and is expected by a robot. Car companies.

“I wonder to what extent (Cruise) considered the scenes where the vehicle had to interact with people outside the vehicle,” he said.

As cars become more automated, other road users, other vehicles, pedestrians or cyclists may need external interfaces, ”he said. “If a car is towed, or someone starts screaming at it, or throwing a stone – how does the car deal with it?”

During the Stanford Center’s tests on human-robotic communications, people noticed the car being bullied – shouting at it, standing in front of it to prevent it from moving forward, and hitting the brakes as they drove ahead.

“If someone starts abusing a robot car, you may want it to have some sort of escape mechanism,” Surkin said. “You do not want a car to yell back at those who shout at it; autonomous cars will need ways to assess and exacerbate situations like this.”

Smith had similar thoughts.

“One thing is clear, there has to be a less temporary approach to all these interactions: with police officers, first responders, tow truck drivers, even with other road users,” Smith said. “The number of 1-800 numbers that police officers are expected to call cannot be total. There should be a better way to find out who they are contacting and contact the agency.

Overall, Smith said, “This is one of the relatively good news. No one was hurt. Cruise didn’t even get a ticket – how lucky was that? In general, things worked out. But I say ‘sort of’.”

Carolyn Side is a San Francisco Chronicle staff writer. Email: csaid@sfchronicle.com Twitter: csaid