Google has gotten a patent for an innovation that will help self-driving cars deal with situations where they become stuck and do not have a solution, as Matt McFarland writes for The Washington Post’s Innovations blog. The new technology will be able to identify when an autonomous car is unable to reach its destination without breaking some of its rules, McFarland writes, such as when there are cows on the road. It was invented by Joshua Seth Herbach and Nathaniel Fairfield, according to the U.S. Patent and Trademark Office.
McFarland goes on to say:
Most of us haven’t had a cow wander on to a road in front of us while we’re driving. Still, it’s the type of situation Google is wisely anticipating. A significant challenge for self-driving cars will be handing edge cases, essentially rare situations on the road. As investor Chris Dixon has written, machine learning can quickly solve 80 percent of any problem, but getting the full 100 percent is extremely difficult.
Other examples of problems that self-driving cars will not know how to answer include what to do when a car is stuck in front of it, or when a vehicle is doubled-parked, McFarland writes. To resolve these types of problems, Google’s new patent calls for the establishment of an assistance center staffed by “human experts and/or an expert system,” McFarland writes.
Once a vehicle is in a stuck situation, its sensors will send information about its location, plus other data, to the assistance center, McFarland writes. An expert at the center could then suggest a different route, or request additional information, such as live video or still images from the car’s cameras, in order to assess the situation.
The patent specifies an interface that the assistance center expert would use to communicate with the vehicle, McFarland writes. Such an interface might be controlled by someone in the vehicle, he adds.
Without such an assistance center to solve problems outside of the ability of the autonomous vehicle’s own technology, “that could be a disaster for Google,” McFarland writes. “Imagine hundreds of self-driving cars stuck in heavy traffic as a concert or sport event lets out, all contacting and overwhelming the assistance center,” he adds.
The system would take into account all sorts of parameters in order to solve problems of vehicles being stuck without knowing what to do, McFarland writes. Those include such things as the location, and the time of day or night, he reports. In such situations as being at a sports arena right after a game ends, or being near a school as school lets out, the autonomous vehicles might need to learn to be patient, McFarland writes.
News articles, such as this one by NBC, reported recently that a Delphi self-driving vehicle drove across the United States, from San Francisco to New York City. The trip took nine days, and a team of humans rode in the car; they only needed to take the wheel 1% of the time, such as when police or construction blocked the roadway, NBC writes. In the case of the cross-country trip, the self-driving vehicle “had the benefit of being on highways, an easy situation for self-driving cars,” McFarland points out.
As this blog has written, Google introduced a prototype of its self-driving car back in December. This blog has also reported on a study showing that Americans are willing to pay more for autonomous features in their vehicles.