Google’s New Autonomous Driving Patent Sounds Terrifying And Great

Google just filed a new patent application (PA) through Waymo that would enable a car’s autonomous driving system to take over, automatically, if it detects a bad driver behind the wheel. That’s right, if you or your aunt Bessie or little-Johnny-learner’s-permit sufficiently demonstrate that you suck at driving, the car could just assume control itself. We’ve all been behind bad drivers and wished someone would take the wheel for them, and this patent could do exactly that, which could be a boon to safety.

Of course, it could also be a complete nightmare.

First spotted by Motor1, Waymo actually filed this application back in May. It’s only now becoming public and the details are interesting. Waymo titles the invention as “using driving assistance to detect and address aberrant driver behavior.” First, it’s worth noting that, according to Google (the search engine, not the patent language itself), aberrant means “departing from an accepted standard.”

So this patent, in theory, could not just detect bad driving behavior but then it could “address” that behavior. Let’s sort out exactly what the language of the patent application tells us, why it could be a good thing, why it might be horrible, and why it might not matter at all.

Straight from the patent application:

“The technology relates to identifying and addressing aberrant driver behavior. Various driving operations may be evaluated over different time scales and driving distances. The system can detect driving errors and suboptimal maneuvering, which are evaluated by an onboard driver assistance system and compared against a model of expected driver behavior. The result of this comparison can be used to alert the driver or take immediate corrective driving action. It may also be used for real-time or offline training or sensor calibration purposes. The behavior model may be driver-specific, or may be a nominal driver model based on aggregated information from many drivers. These approaches can be employed with drivers of passenger vehicles, busses, cargo trucks, and other vehicles.”

To accomplish this very complex task, the system will go through a set of actions. First, it’ll track the actions of the driver and log that data. As it analyzes that data, it’ll look for what it determines is aberrant (though this isn’t defined in clear hard terms in the patent application), and, if applicable, “select an action from among one or more classes of actions in response.”

While the patent application doesn’t define exactly what it’ll consider aberrant, it does provide some hints. It likely won’t be a single act but several that break away from what it considers the standard model of driving behavior. In addition, it’ll judge based on things like how long one has been driving and how far they’ve gone.

Interestingly the application also mentions the controls the system could engage with, including braking, accelerating, steering, signaling, navigation, and communication systems. In fact, the car could in theory turn itself off if it determined there was reason to do so.

Just imagine sitting in traffic, maybe on your way to Kroger or wherever you go, and watching as a person makes a series of terrible driving decisions. Then, just as you’re losing your hope in humanity, the car pulls its own driver over, stops the car, and calls a relative to help them figure their life out. Who doesn’t want to see that? I know I do.

More likely, this system could in its own robotic way actually train drivers to improve their skills. Waymo even says as much in the filing. Speaking about the data collected from a bad driver, it says “This information may be stored … for instance to … provide training.” Of course, there is a potential downside here. Who is to say the system actually works as described and doesn’t present other problems?

[Editor’s Note: This all sounds similar to an approach to semi-automated driving that I’ve always called “reverse Level 2,” though there are other names, often things that suggest guardian angels or ghosts. It’s where the driver is in control, but the automated system is “watching,” essentially the reverse of Level 2 where the car has control but only under the constant supervision of the human in the driver’s seat. I’m not really sure how this is different, other than the process is laid out in detail. I’ve had conversations with Volvo AV people where they’ve told me this is an approach Volvo favors. – JT]

What happens if someone is trying to rush a loved one to the hospital? What if you’re trying to avoid creepy dudes? In theory, the car has no idea what is going on and could impede the journey. How will it manage things like road construction where road signs and road markings are far from perfect or uniform?

Further, consider just how much data we’re talking about here. People were up in arms when they realized that firms like LexisNexis were tracking their every maneuver, brake engagement, and hard acceleration. How will they feel if cars are not just tracking these things, but actively changing the way they drive? Will this data even be secure? Will this data even be secure? Examples of large, seemingly secure holders of personal data suffering breaches are hardly hard to come by.

Finally, it’s worth saying that as of this writing, this is just a patent application. The USPTO hasn’t granted the patent to Waymo. Even if that time does come, there’s no guarantee that Waymo will use it in public-facing vehicles. Some patents are obtained simply so that the owner can stow the tech away from the light of day.

If fully autonomous driving for all becomes a reality, it would seem that the days of complete car control are limited regardless of this new proposed tech. Frankly, I’m all for taking the wheel out of the hands of negligent and disinterested drivers. Let those of us that care to try and suck less at driving do it ourselves. Let everyone else nap on their way to Kroger.

Topshot base image: Andrey Popov/stock.adobe.com