Whenever I talk to John Krafcik, CEO of Waymo (the former Google Self-Driving Car Project), I’m always left with one overriding feeling: sadness for his competition. I’ve been lucky to ride in many an ostensibly self-driving car. Some were. Others left me giving thanks the moment I stepped out. But none of the dozens of companies out there working to develop autonomy systems has ever given me the impression that they’re anywhere close to what Waymo is doing.
That feeling was reinforced by the announcement today that Waymo’s cars have crossed their 10 millionth driverless mile on public roads. 10 million miles. No other company is as transparent as Waymo when it comes to sharing information about testing, and so it’s impossible to know for sure, but I don’t think any of its rivals come close.
Really, though, numbers don’t matter. It’s more about lessons learned, and Waymo has evolved to the point where its technology is now having to solve the kind of minutiae that wasn’t even on the radar back when this whole thing was still called the Google Self-Driving Car Project.
Those first 10 million miles, Krafcik told me earlier this week, were “really all about safety.” The next 10 million? “It is moving to that next level of comfort and convenience, getting people more easily to where they want to go.” As it turns out, that includes taking the “where” to a new level of specificity.
Take a moment to consider a future ride in a self-driving car. I think most of us envision climbing in and saying something like “take me to the grocery store” before kicking off our shoes, opening a bottle of champagne and forgetting about all our worldly problems. That’s all possible, but what happens when you get to the grocery store?
Does the car drop you off on the closest street? No, that’s no good. The main entrance? That’s straightforward enough — if your car has the necessary mapping data to know where the main entrance is, plus the sensing and processing power to safely traverse the maelstrom of confusion that is your average parking lot.
So you get dropped off, you buy another bottle of bubbly for the return trip plus a few bags of other necessities, and now it’s time to go home. Where do you want to be picked up? Back at the main entrance again? Actually, probably not. As Krafcik explained when walking me through this scenario (minus the bit about the booze), loading all your groceries into a car parked outside of the main entrance causes bottlenecks for other shoppers trying to get in. Many of Waymo’s early users instead indicated they’d rather be picked up by a shopping cart return area.
“They have been teaching us about a lot of things beyond safety,” Krafcik said of Waymo’s early riders. “They’re teaching us things that we really need to know about comfort and convenience, things that in retrospect seem so obvious. When we started this early rider program, they really weren’t obvious at all.”
In summary: Waymo is now teaching its cars how to be polite at the grocery store. Meanwhile, the competition is still running red lights
“We know we need to solve for inclement weather,” Krafcik said. “A lot of our 25 cities have been in places where the weather is nice. … We’ve now spent a lot of time in Tahoe and a lot of time in Michigan to make sure that our services are able to handle all sorts of weather.”
To that end (and perhaps also thanks to some) Waymo is re-engineering its entire sensor package. The current version, as seen on the company’s Pacifica minivans, isn’t entirely cold-weather compatible. The next version, set to debut on the Jaguar I-Pace, will be a major change. “We’re going to have new lidar, radar and cameras, which is pretty much the whole thing,” Krafick said. “One of the things to really focus on is ensuring four-season robustness. And that means that sensor cleaning is a critical, up-front design parameter.”
It’s worth pointing out that this new focus on comfort, convenience and all-weather performance is not coming at the expense of safety. Waymo is far from satisfied that its cars are as safe as they ever will be. Its extensive simulation program provides the verification that none of these tiny steps forward cause a regression elsewhere in the system.
The company is approaching 7 billion miles driven over simulated roads, and the beauty of instrumented testing is that every new tweak can be checked against every previous scenario, ensuring nothing got missed along the way. With Waymo adding more and more nuance to its ultimate driver-less driver, keeping things from slipping through the cracks is only going to get more important.
That still leaves the rather important question of when the heck the Waymo service will actually launch. According to Krafcik, it already has. “I think it’s going to be hard to define any one moment where it’s suddenly launched. I think from our perspective it is launched, and the service is out there now, but we’ll have more and more people riding it to more and more places.”
Waymo’s service is already running 24/7 for roughly 400 users in the greater Phoenix area, and it continues to expand outward slowly from there. The company is currently testing in 25 cities, but Krafcik says the San Francisco Bay Area is likely the next major expansion point. Further locales could prove more difficult. Why? Winter is coming.