A flooded road can look simple from a distance. It may seem like water on the road, but drivers here know better. Water can hide open manholes and deeper sections of the road.
That is why floodwater is one of the hardest tests for any driver. The driver has to know when the road is no longer worth entering.
This came to mind after Waymo recalled 3,791 robotaxis in the United States over a software issue that could allow its driverless vehicles to enter flooded roads.
The recall followed an April incident in San Antonio, Texas, where an unoccupied Waymo vehicle entered floodwater during severe weather.
I do not see this as a funny robotaxi story.
Floods confuse human drivers too. We have all seen vehicles enter water that looked shallow until it was not.
Drivers often follow the vehicle ahead, trust the shallow-looking water and keep moving because the destination is already near.
When the car stops, the decision becomes expensive.
The problem is not only a machine error because floodwater tests judgment.
Sensors and maps can help read the road, but the harder decision is whether the vehicle should enter or turn back.
Human drivers read the road in different ways. They check and watch how smaller vehicles move through the water. Some wait for a larger vehicle to pass first.
Others look at the people on the side, especially when they are warning cars to turn back. Luck may work sometimes, but it is still a poor substitute for judgment.
This is where autonomous driving faces real-world roads. The road is not only lane markings and speed limits. It is also drainage, weather, road work, stalled vehicles, and common sense.
Flooding is not only a rainy-season problem here. It is also tied to public anger over flood-control projects that were supposed to protect communities better.
A familiar street can change after one strong downpour. One lane can be shallow while another section is already too deep. Even experienced drivers get it wrong.
I do not think the lesson here is only for robotaxis.
Many drivers treat floodwater like a challenge. They judge the depth from the car ahead. They trust height, speed or confidence. They hope the engine bay, battery pack, exhaust, wiring and luck are all on their side.
That, to me, seems like guessing, not driving.
Sensors, maps and software will improve, and robotaxis will get better at dealing with weather and flooded streets.
Floodwater will remain difficult because one wrong guess can stall a car, damage expensive parts or put people at risk.
The safer choice is often to wait, turn around or find another route. Arriving late is still better than calling for a tow truck.
That is not exciting. It is just cheaper.
A vehicle that knows when to stop may be more useful than one that only knows how to proceed. Drivers should learn the same lesson.
