Robot Cars are Coming to Get You

in by

In addition to hoverboards, unicycles, mopeds, and dog-pulled skateboards – as well as an occasional car or bike – San Franciscans will soon be sharing the roads with driverless robocars, zipping through traffic without the added weight of human passengers. Last October Cruise LLC received a permit to test up to five vehicles at a time within City limits without a human in the driver’s seat. Cruise is the fifth company allowed to conduct such field work in California. So far, it’s the only business pitting driverless technology against San Francisco’s notoriously challenging driving conditions. 

San Franciscans have seen plenty of self-driving cars, but always with human passengers. Usually identifiable by prominent logos and strangely protruding sensors, autonomous vehicles (AV) have been approved for testing on California’s roadways since 2014. The Department of Motor Vehicles (DMV) has issued AV permits to 59 companies, many of which operate in the Bay Area. Previously, however, trail vehicles were required to carry a human backup “safety driver”. 

In 2018, the state issued the first permit allowing autonomous vehicles to operate without the physical presence of a human. Test vehicles are required to have human supervision, conducted remotely over cellular networks with the teleoperator able to take control of the AV at a moment’s notice, such as during a critical traffic event or engagement with a police officer.

Aside from Cruise, companies granted licenses for driverless testing in California are Waymo, Zoox, AutoX and Nuro. Waymo is owned by Google’s parent company, Alphabet, and is piloting a driverless taxi service in Phoenix, Arizona. Zoox is a recently acquired Amazon subsidiary.

According to Marc Hoag, co-founder of the independent consulting firm Hoag+Co and host of the podcast, Autonomous Cars with Marc Hoag, Cruise stands out from other driverless testing projects for its all-electric fleet and willingness to pit its artificial intelligence (AI) capabilities against San Francisco’s narrow, winding roads and chaotic traffic. 

Perhaps the thinking goes if a robot can negotiate a city where cars share the roads with trains and cable cars, traffic cops ignore double- and even triple-parking, an intersection can be mobbed by a horde of pedestrian shoppers or fleet of boombox-carrying cyclists, right-hand turns are allowed some times and forbidden others and topography is more than just a fancy word from elementary geography lessons, then it can drive anywhere. 

While Cruise is the only company to conduct fully driverless tests in San Francisco, others are keenly interested in the Bay Area as a market for autonomous technology. A Zoox representative described San Francisco and Las Vegas as proposed “anchor markets for us to rigorously test, validate, and refine our technology before eventually launching an autonomous ride-hailing service,” though presently “testing of purpose-built driverless vehicles is still confined to private roadways.” According to the DMV, Zoox has secured a permit to conduct driverless testing in Foster City.

“Driving in San Francisco sucks, for human drivers as well as computers,” said Hoag, a California-licensed attorney whose firm specializes in global autonomous vehicle and mobility strategies. “I’ve often said the only other American city more challenging is perhaps Boston. And that, I think, is what makes Cruise so special, so cool. With all due respect to Waymo, driving on the freeway-sized boulevards of suburban Arizona is a vastly less challenging task than driving through the streets of San Francisco.” 

Cruise has been testing with in-car safety operators in the City since 2015. According to a statement, Cruise sees itself as “a hometown company.” Though it was acquired by General Motors (GM) in 2016, it was founded in San Francisco in 2013 and is headquartered South-of-Market. The business deploys its vehicles—with in-car safety drivers — on behalf of the San Francisco-Marin Food Bank and SF New Deal, making more than 100,000 deliveries to Bay Area families since the start of the public health crisis.

Prior to being acquired by GM, Cruise was developing its autonomous technology using the electric Nissan Leaf.  The company’s present AV fleet consists of modified Chevy Bolts. Less than a year ago it introduced the Chevy Origin, which Hoag described as “literally just a pod-car”, designed to ferry up to six passengers in a crowded metropolitan area, with no need to accommodate a driver or fit into a conventional parking space. The Origin has no pedals nor steering wheel, making it a Level 5 autonomous vehicle, according to the Society of Automotive Engineers’ driving automation scale. 

The lowest level, 0, is for vehicles with no driving automation at all, operated completely manually. Level 5 vehicles are fully automated, requiring no human intervention during any part of the driving process. Levels 1 through 4 cover a range of automation, from adaptive cruise control and lane centering technologies featured in many car models to most of the autonomous vehicles being tested on California’s roads, with controls that can be operated by human safety drivers.

Anecdotal research suggests that despite proximity to Silicon Valley and generally tech-friendly views, San Franciscans have considerable concerns about the risks driverless cars may pose to other vehicles and pedestrians. Hoag pointed out that if a company is ready to take the next step of replacing the backup safety driver with a remote teleoperator, then its vehicles have gone through extensive development and testing, with little difference in performance with a human sitting idly in the driver’s seat. 

“There are only five companies in California that have gotten this permit,” said Hoag. “Essentially what it is saying is, the car is basically done. It’s capable of doing its thing. Then, good: let’s put it out there, but as one last layer of safety precaution-to be absolutely certain—we’re going to keep a remote eye on it.”

Advocates of driverless vehicles cite traffic fatalities as the biggest argument in favor of handing over the nation’s keys to an army of AI chauffeurs. Roughly 100 Americans die daily from car accidents and collisions with pedestrians and bicyclists. Robots don’t get tired or drunk, the argument goes, nor do they adjust their speed at yellow lights because they don’t want to be late for a meeting. 

“While we have millions of miles under our belt,” said a Cruise representative over email, “as we have promised, we do not embark on new phases of testing until we are certain we are safe. For this first driverless pilot in the city, we will start small, with just five vehicles in designated areas of the City and grow incrementally from there.” 

DMV guidelines restrict Cruise’s driverless vehicles to roads with a speed limit no greater than 30 miles per hour following a predetermined route. Testing can take place anytime, without duration limitations. Testing isn’t allowed during extreme weather conditions, such as heavy fog or rain. 

Thirty miles per hour or slower means residential roads. Cruise has yet to disclose where in San Francisco testing will take place. 

Hoag predicts that a “truly driverless car” – one without a human teleoperator – is ten to fifteen years away. Even then, he imagines there’ll still be a role for remote human supervision. “Eventually, we may end up seeing an evolution of a sort of air traffic control system of sorts for autonomous vehicles.” 

Hoag believes there’ll be an interim phase where driverless cars are supervised remotely by largely inactive teleoperators who will monitor the vehicle and its passengers and intervene periodically throughout the course of the ride. He predicted that, at least initially, higher level automation will be limited to ride hailing or delivery service vehicles, as opposed to privately owned cars. 

He expected that within five years autonomous taxi services will operate on a restricted basis in San Francisco, while “limited deployment of L4 vehicles should start to really become a thing during the mid-20s, and L5 by the mid-30s. Frankly, if by the end of the 2030s we aren’t seeing hockey stick-growth curves of AV deployment and the beginnings of phasing out human drivers in certain areas then it’s the various government bodies and regulators that have failed us, and it is they who should be held accountable for the 3,000 people that die on American roads every month, and not the engineers.”