The Mercedes-Benz experimental self-driving car, a.k.a. Bertha, has come to California, a lead engineer for the project said yesterday at a San Francisco conference on autonomous vehicles.
"It has driven the roads of Sunnyvale for several months now," said Ralf Herrtwich of Daimler, the parent of Mercedes.
California recently opened up its roads to robot drivers by offering special licenses to their owners. The licenses weren't supposed to go into effect until September, which suggests that Mercedes had already gotten permission earlier, as Google had done for its car.
Unlike Google's car, which has a distinctive roof-mounted laser imaging system, Bertha uses only discreetly stowed radars and cameras. It got its name last summer by plying the 100 kilometers between Mannheim and Pforzheim, following the route blazed 125 years earlier by Bertha Benz, wife of Karl, a founder of the company. She did it unbeknownst to him, making her jaunt the first documented joyride.
Herrtwich said that Mercedes was designing automated cars for two kinds of customers: owners and renters, such as the customers of Cars2Go, a Daimler subsidiary. Though full autonomous driving may be far away for passenger travel, Herrtwich said, a form of it may be a selling point in rental cars. They might drive in “automatic valet” mode to pick up a renter, then revert to a less automated mode once the client is inside.
“The costs that you incur running such a service are mainly driven by the number of vehicles you have to deploy, which is driven by how comfortable you want the cars to be to use,” he said. “If it’s too far to walk to get to the [rented] vehicle, it will be harder to get customers. Automation can reduce the number of vehicles you’ll be needing, and then you’ll have money to automate them.”
He said the main challenge will be to move up to full autonomy from the partial form that’s becoming available right now, notably in the 2014 Class S. So far, driver assistance technologies have reduced the accident rate, but they intervene only occasionally; full automation, he said, would mean taking over the remaining 99.99-plus percent of cases—a tall order.
Perhaps the mode that's just shy of full automation is the most unnerving of all, Herrtwich says. In this case it's the driver who only occasionally intervenes, and that requires more vigilance than a human being is designed to maintain. You might be looking out the window, daydreaming, when your car suddenly screams at you to wake up and take charge.
“So, can we bring back the human, as fallback?” he asks. “With the driver in the loop, arguably partial automation is the safest level of automation. The system is supervising the driver, and the driver is supervising the system.”
This argument for keeping the human being involved directly contradicts the Google position, which decries the slow and deliberate ramping up of automation, favoring instead a direct push to full automation. Herrtwich also rejected the idea that robotic cars would soon be so safe that they wouldn’t need strong metal bodies or sturdy “crush” zones to handle a head-on impact.
“Some say passive safety is going away, but why should it?” he asked. “Somebody else may hit your vehicle.”
Some say? Who could he possibly mean? Why Google, again. It recently unveiled a new, roly-poly version of its car that has a soft body (for the benefit of pedestrians), no crush zone, and no wheel--just an on-off switch.
Philip E. Ross is a senior editor at IEEE Spectrum. His interests include transportation, energy storage, AI, and the economic aspects of technology. He has a master's degree in international affairs from Columbia University and another, in journalism, from the University of Michigan.