In Google’s future, when self-driving cars get confused, they’ll be able to call tech support. And hopefully the human on the other end will be able to make a quick assessment of the situation, because the car might be hurtling down the street when it yells for help.
The plans are revealed in a new patent application that allows for remote assessment of autonomous vehicles. The car can call home and seek advice on what to do next. With the call will come sensor data from the car and even live video, so that a human can assess the circumstances and tell the car what to do.
Here’s an example, taken from the patent application. The car wants to turn left on a traffic signal. The light is green, but there’s no left-pointing arrow. Is it safe to turn across the opposite lane? In this case, “It may be useful in such a scenario to have a remote operator verify (e.g., based on a live video feed taken from within the car) that no traffic is coming from the other direction that might interfere with the turn,” says the application.
If you’ve ever tried to contact tech support for assistance, you’ll know that the typical timescale isn’t measured in anything close to milliseconds. However, Google has thought of this too. The patent also covers the possibility that the car’s computer just isn’t powerful enough to make some decisions. In this case, it could call Google’s own data processors and ask for help. Given the speed of Google’s web services, this scenario seems a lot more plausible. Just cross your fingers that the incident doesn’t occur when your car is disconnected from the Internet.
Other situations are more suited to human intervention. For instance, if the car is picking up passengers, it could ask a human to check that everyone is in the car, posits the blog What A Future. This would be a handy feature for taxi services.
But that’s not all. What about the passengers in the car? Couldn’t the system just ask them for help?
“In further examples, the remote assistor may be a passenger in the vehicle, which may be useful for certain intuitive behaviors that are difficult to automate (e.g., asking a taxi to move forward a few feet before dropping off passengers).”
That’s right. Google just tried to patent taking manual control of a driverless vehicle, aka “driving.” Sarcasm aside, the system seems to be designed to deal with problems in the transition period when autonomous cars are still learning the ropes, and humans still know how to drive. It seems smart for the car to ask an on-board human for tips, or to consult a more powerful computer in sticky situations.
It would certainly assuage the fears that will arise when cars first go fully autonomous, without any human monitor on board to take the wheel in emergencies. But somehow, a fleet of operators at a central location, directing traffic as it calls in for help, seems rather quaint, almost like having human operators connect our phone calls. In fact, I’d argue that lawmakers are more likely to force something like this on autonomous vehicles than Google is to ever implement it.