Pardon the bad pun. But imagine that you've taken one of the new driverless Waymo taxis from the San Antonio airport right after a big rain. The car exits the freeway, and on the access road ahead you see where a nearby stream has overflowed, and water the color of coffee with cream is rushing across the road. The Waymo vehicle keeps going. You can't be sure, but the water looks deeper than you'd like, and if you were driving, you'd follow the poetic advice to turn around rather than drown. You look for a panic button in the car: there is no such thing. You think about just opening a door and jumping out, but the car's going too fast for you to try that without risking your neck. Just as the front wheels hit the water, you feel the car skidding, and you decide to try jumping out anyway. Only now the water pressure on the door is too great, and it won't open. . . .
This didn't actually happen. But it could have.
On April 20, after heavy rains caused flooding on several roadways in San Antonio, an unoccupied Waymo vehicle was swept away from an interstate access road in floodwaters arising from Salado Creek. The vehicle was later found about half a mile downstream.
Following this and another incident earlier in the month, when a Waymo car became stuck in high water in San Antonio and had to be towed out, Waymo suspended all service in San Antonio and five other cities in three states, including Nashville and Atlanta. Another Waymo car got stranded in Atlanta during heavy rains as well. Waymo is working on updating its software to prevent such incidents in the future, but as of today, service still appears to be suspended in Austin and Atlanta, although it has been restored in San Antonio.
Waymo is to be commended for taking swift action in the face of obvious problems that could kill passengers in a peculiarly horrifying way. Especially with high-tech operations of super-wealthy companies such as Alphabet (the parent firm of both Waymo and Google), profit sometimes takes a back seat to safety, and that is as it should be.
Up-to-date information on high water in roadways is currently available through various means, including municipal emergency services and mapping organizations such as Waze. But especially in Texas, floodwaters can rise within minutes, and few systems can respond that fast. It looks like the challenge of directly detecting the presence of high water on a road through on-board sensors is something that is either too hard to do, or too unreliable. So Waymo and other autonomous-vehicle operators may rely mostly on third-party data about where it's too dangerous to drive because of high water.
Despite the drama surrounding the flooded Waymo cars, this is actually part of a process I would call normal engineering. No matter how many problems a designer anticipates before a new technology is fielded, there will always be some issues that don't show up until the technology has been in use for a while. This is doubly true for a complex multidisciplinary system such as an autonomous vehicle. The world is simply too complicated for non-omniscient engineers to anticipate every single thing that might go wrong before deploying a system in the field.
Many of the locations where Waymo suspended service were new to the company, so Waymo had its first encounter in those cities with this spring's rainy season. While California also has flash floods, I don't think they are as prominent a part of the landscape as they are in central Texas, for example.
If I'd been riding in a Waymo car headed toward a flooded road, I wouldn't have had a lot of options. There is no panic button as such in a Waymo. But on the Waymo website there is a section titled "Pulling over, collisions, and security events." They do list an option to ask the car to pull over out of traffic, but the decision as to how and when to do that is up to the system, not the passenger. There is a way to contact "Support" (presumably a human being on the monitoring staff that every autonomous-vehicle company maintains), but having a conversation with someone possibly thousands of miles away as water surrounds the car is probably not the first thing you'd want to do.
All things considered, it was the wisest course for Waymo simply to shut down all service until their engineers can figure out a way to keep their cars out of water, hot or otherwise.
This is also a good example of how semi-serious incidents can flag a problem that could easily lead to more serious consequences, such as customer fatalities. If you look into the details of many historic engineering disasters, you will commonly find that less serious precursor incidents often happened in the days or years leading up to the big disaster. This isn't always the case. But it happens often enough that system designers should train themselves to take very seriously any safety-related problems that constitute narrow escapes. And they should address the issues allowing such narrow escapes before some unlucky person encounters the combination of circumstances that proves deadly.
In shutting down all service in certain cities until the flood problem can be addressed satisfactorily, Waymo did the right thing. This doesn't guarantee that nobody will ever drown in a Waymo car. But in taking such a public move that obviously cost the company revenue, Waymo has both given itself time to fix the problem, and has sent a strong message to the public that they regard passenger safety as paramount.
Sources: I referred to articles in the Austin American-Statesman at https://www.statesman.com/business/technology/article/waymo-freeway-robotaxi-safety-pause-22272052.php, the BBC News website at https://www.bbc.com/news/articles/ckgplyxxl75o, and Waymo's own safety information at https://support.google.com/waymo/answer/9449023?hl=en.
