San Francisco protesters are pranking driverless cars with traffic cones

The driverless cars that are roaming the streets of San Francisco as both testbeds and transportation service are loaded with sensors. They're adorned with lidar bulbs, camera hoods, ultrasonics, and more, feeding into powerful computers that image the world around them in 3D in real-time to predict what will happen next and plan a route to the next destination. And a simple traffic cone on the hood is all it takes to render the entire system inoperative, and it's become the favored prank weapon of a group of irritate public transit advocates in the city.

It's been dubbed "coning" and is a tactic of the group Safe Street Rebel, who argue that it's wrong to use the public streets of a dense city like San Francisco as a testing ground for these technologies, and that more focus and funding should be put towards public transit instead of making it easier to use individual cars.

As profiled by NPR, Safe Street Rebel views the practice as largely harmless and yet still hope that they can draw attention to their cause with the antics. They have rules they follow, like not paralyzing driverless cars that are carrying passengers and not shutting down cars in the middle of a bus route. But anywhere else is fair game, it seems:

"We thought that putting cones on these [driverless cars] was a funny image that could captivate people," says one organizer. "One of these self-driving cars with billions of dollars of venture capital investment money and R&D, just being disabled by a common traffic cone."

Self-driving cars are programmed to be cautious by default. As long as the inputs are correctly interpreted, the car will chose the safest option to accomplish its goal in that moment. And when the inputs are confusing, the car will chose to fail safe and not do anything, rather than risk doing the wrong thing because it didn't understand the situation. Slapping a big traffic cone on the windshield is immediately and unrecoverably confusing the the sensors, so it stays put instead of going forward when the light turns green. To be fair, a human driver would also be confounded and likely to stop if they end up with a traffic cone on their windshield — the big difference is that the human driver could easily get out and remove the cone to carry on their way.

But not everybody is tickled by the conings. Obviously, Google's Waymo and GM's Cruise aren't thrilled about having to send human staffers out to remove cones from windshields. And while it may be amusing, it's also a nuisance to other drivers on the road and especially to emergency responders. They're already struggling enough with driverless cars that aren't programmed to handle emergencies on the road. As San Francisco's police and fire departments noted, "They've tallied 55 incidents where self-driving cars have gotten in the way of rescue operations in just the past six months."

Both of Safe Street Rebel's points are valid, even if the cone paralysis is annoying for other drivers and a hinderance to police and fire trucks. The driverless cars are a potential safety hazard on the city's streets, though their general accident rate appears to be significantly lower than that of human drivers. After all, the car has a real-time 360º three-dimensional awareness of the space around it, can make decisions and react in the blink of an eye, and is designed to fail safe (thus the success of coning) instead of taking a more human approach of "I am reasonably certain that I can execute this maneuver."

Source: NPR

It's always the edge cases that make a system like that difficult, and the data for those edge cases can really only effectively be collected in the real world. There's definitely and argument to be made that these driverless cars aren't ready to not have a human in the driver's seat to take control in situations where the car isn't behaving as expected. It might only be a 0.01% of the time edge case where a human needs to intervene, but those can be literally life-and-death scenarios. Plus the human could also easily remove the cones.

The other argument about needing to invest more in public transit instead of the billions that corporations have pumped into driverless cars is valid, but not as strong. Public transit is of course a great investment and something that a lot of American cities have struggled to do effectively since the late 20th century.

The personal car is king on American roads and has been for decades. More cities should invest more in expanding bus service and building rail transit systems. These systems can greatly enhance the quality of life and mobility across the cities, while reducing congestion on the roads and pollution in the air. A well-designed and functioning mass transit system is a thing of beauty.

But there will always be a need for individual passenger vehicles and making them fully autonomous will also help to reduce the number of cars on the road and the huge amounts of space most American cities dedicate to private vehicle parking. When fully capable driverless cars become a thing, the need for a personally owned vehicle will decline. The car can drive all day long, refuel when needed, and be hooked into a network to efficiently handle routing to pickups and drop-offs.

Most privately owned cars spend most of the day parked doing nothing. Public transit does not and cannot reach every door, and there will always be people that need to get to and from the places it the bus, subway, and tram do not go near. A driverless car that's able to be hailed to pick them up and take them where they need to be, even if that's to the closest transit stop, will be a game changer, especially for the mobility impaired.

The coning protests are amusing, but perhaps shortsighted and sometimes problematic. But maybe they'll have the effect of bringing about the kind of change that will improve these systems for all.