clock menu more-arrow no yes

Filed under:

Google self-driving car patent reveals how you’ll let AI take the wheel

Google

A new patent from Google published today has some fascinating insight into how the company thinks that production of self-driving cars of the future might take control from the drivers they're shuttling around — and how they could give it back, too.

The description of "Google Chauffeur," spotted by Stanford's Reilly Brennan, is a pretty straightforward concept. An arm on the steering column (not much different from a windshield wiper arm) could be pulled to engage a car's self-driving mode; at that point, the system would do a check to see whether it's ready and able to actually take control from the driver. If it isn't — the car can't get a GPS lock, for instance — the driver might see a "Not Available" light on the dash. Otherwise, you'd see a "Ready" light, at which point you can start taking your appendages off the wheel and pedals. (The focus of the patent's coverage is on the operation of these lights, but the remainder of the patent's text gives interesting insight into how Google sees a car actually working.)

google chauffeur patent
A diagram from the patent depicts a driver chilling out and using an iPhone while in a car that looks suspiciously like a late-model Toyota Prius.

Under normal circumstances, a driver might take back control by reversing the procedure — pushing the arm on the steering column in the other direction — but the more interesting scenario is in an emergency:

If the passenger identifies an emergency situation, the passenger may take control of the vehicle immediately. For example, passenger may see an obstacle which computer has not identified, such as a bicyclist or road construction. Without first disarming computer, passenger may grip the steering wheel to return computer to "ready mode" as shown in FIG. 9. The impact of passenger's hand or hands on steering wheel may be received by the various touch sensitive input apparatuses of steering wheel. Computer may receive this information, determine that the passenger would like to take control, and return to ready mode. This allows the user to feel confident that he or she may take control of vehicle instantaneously.

Obviously, Google would like to believe that the computer can drive far more safely than a human ever could — it has been very aggressive in clarifying that all of its accidents have been the fault of other drivers — but this patent allows for the possibility that a driver could see something that the car's self-driving systems don't.

The very existence of the patent is interesting, considering that Google had indicated at the launch of its cars last year that they'd like to push toward eliminating steering wheels and pedals entirely. That said, the fleet of test cars that Google currently has on roadways in Austin, Texas and the Valley is equipped with them, just in case the car screws up and someone has to take over. If we ever get to the point where autonomous vehicles are actually steering wheel-free, it's not clear how this kind of fail-over system would work.

How does emergency takeover work without a steering wheel?

The contingency plan works in both directions, according to the patent: just as the human could take over for the car in a hurry, the car could detect and quickly take over for a drowsy or intoxicated driver, while using "audible alerts" to get their attention.

Of course, none of this is to say that this'll be how production systems actually work — different automakers and suppliers undoubtedly have different ideas, and regulatory concerns could strongly influence the behavior of real cars. A recent NHTSA study says that drivers take an average of 17 seconds to take control back from an automated car after it warns them to do so, which isn't very fast at all, particularly for a car traveling at highway speeds. Can technology solve that issue? And if so, are lights and audible alerts enough to do so?