avs

Q&A: Inside Waymo's 2021 Game Plan for Autonomous Ride-Hailing and Trucking

If autonomous driving had a 2020 valedictorian, it’d be Waymo
article cover

Waymo

· 6 min read

It’s no secret that this year hit the autonomous driving industry hard: layoffs, problems collecting data, timeline setbacks, hardware shipping issues, you name it.

But Waymo still landed an additional $3 billion in external financing—and launched both Waymo One, its fully autonomous ride-hailing service, and Waymo Via, its self-driving truck and freight venture. The company has New Year’s resolutions for them both.

We spoke with Saswat Panigrahi, Waymo’s senior director of product management, about everything from simulation during Covid-19 to the pain points of deploying a robotaxi fleet—and why 2021 will be the “the year of the long tail.”

What was the Waymo team’s single biggest obstacle during Covid-19? How did you address it?

We initially paused all of our driving operations back in March to ensure the health and safety of our entire team and local communities. As on-road testing is critical to our development, this was definitely a challenge for us: We not only had to figure out how to stand up that infrastructure from our work from home setups, but the volume of simulation testing increased significantly.

Simulation is critical for two things: speeding up development and supporting rigorous safety testing. Roughly speaking, it’s been responsible for about 80% to 85% of our progress. While our cars were off the roads, we were still able to “drive,” just in simulation, and that was a way for our team to continue to make advancements from home.

We ended the year surpassing over 20 billion miles in simulation, a major accomplishment for the team, and the work we were able to do in this area during Covid helped enable other key successes for us in 2020—including the opening up of our 100% fully autonomous Waymo One service to the public in October.

With an additional $3 billion in external financing, can you talk about the specific tech advancements that the team will now prioritize? What's possible now that wasn't before?

Definitely a huge and exciting milestone for us here that helped validate all of the great work the team has been doing over the last 11 years. The additional capital has allowed us to deepen our investment in areas we were already focusing on—including our operations across both Waymo One and Waymo Via.

From a technical standpoint, 2021 is probably going to be “the year of the long tail”. We’ve already demonstrated with our Phoenix service that a successful fully autonomous ride-hailing service is possible, and now we’re focused on solving the long tail of problems that will allow us to efficiently scale and one day bring the benefits of that service to even more people. For example, we’ve found from over 11 years of testing in major cities that autonomous vehicles in dense urban environments need to be prepared to encounter many more types of vehicle and road users, far more frequently—and obviously we need to continue working to improve our technology so that we can handle those edge cases safely.

Tell us about Waymo's 2021 roadmap. Can you give us specifics on what's next for Waymo One ride-hailing, and Waymo Via trucking and delivery? And has the comparative focus shifted in the past year?

With Waymo One, we’ve heard extremely positive feedback from riders in Phoenix following our public launch last year, and look forward to continuing to expand and serve more riders. We’re testing in multiple locations around the country, too, and as we put the 5th-generation Waymo Driver on the road, you’ll start seeing more of our all-electric Jaguar I-PACEs integrated into the Waymo One fleet in Phoenix, as well as driving down the busy streets of San Francisco, where we’re ramping up testing. We’re excited about this opportunity to continue advancing our Waymo Driver in our home state of California.

Stay up to date on emerging tech

Drones, automation, AI, and more. The technologies that will shape the future of business, all in one newsletter.

When it comes to Waymo Via, we’re excited to continue our strong local delivery partnership with UPS and AutoNation in Phoenix. WIth trucking, we’ll continue expanding our testing across the Southwest U.S., advancing our partnership with Daimler to build a fully redundant truck for the Waymo Driver (meaning it has the backup systems built in that enable us to safely remove the human backup driver), and continuing to deliver freight on behalf of our fleet customers.

Walk us through the pain points of deploying and scaling a robotaxi fleet. John recently mentioned how hard it is—can you talk about that challenge at a tactical level?

It’s true. On the technical side, you’re looking at ensuring your stack can handle the everyday nuances of driving on city streets safely and reliably, over and over again. That means not only assessing how your technology performs in theory, but how it handles the messy reality of real-world driving.

We assess each layer of our hardware and software technologies carefully: how our steering systems (and their backups), our on-board compute, and sensors all perform, as well as how our software completes trips and obeys local driving rules. But operating that technology as a full robotaxi fleet then requires additional systems—rider support, making software updates over time, protocols for how to handle emergency situations—that you also have to consider.

Then, you need to take this technology and build it into a product that will be both convenient and comfortable for a rider to use. The first ride or two without a human driver may be novel and exciting, but at the end of the day we are developing a service that riders will be able to use in their everyday lives. This is where a lot of our current work is focused, whether it’s around improving pickup and drop-offs, fine-tuning routing, or making maneuvers smoother.

Can you talk about how the team leveraged sensor tech from its rideshare efforts for trucking and delivery? How have the projects overlapped, and what's been especially tough about juggling the two different self-driving applications?

Absolutely—there is tremendous overlap between vehicle platforms on ride hailing and trucking. Because we’re developing one core autonomous driving system, the Waymo Driver, that can be applied to multiple vehicle form factors, we’re able to use the same set of sensors and core software algorithms across both cars and trucks. Our simulation infrastructure is also directly applicable across both platforms.

Of course, there are some unique challenges to both cars and trucks where we have to create custom solutions and where tradeoffs and prioritization come into play. But by pursuing both Waymo One and Waymo via in parallel, what we learn with one area directly helps improve the other.

That relationship goes both ways: we aren’t just taking what we’ve learned from cars and applying it to trucks; we are taking what we learn with our trucks and the business use case and applying that back to ride-hailing and the car platform. As an example, all of our surface street experience on the car side will help us unlock full hub-to-hub delivery for our fleet customers on the truck side, while all of our freeway driving experience on the truck side will advance those driving capabilities on the car platform.

Stay up to date on emerging tech

Drones, automation, AI, and more. The technologies that will shape the future of business, all in one newsletter.