Waymo Recalls 3,791 Robotaxis After Driving Into Flooded Road
Credit: Waymo
Waymo has announced a recall of 3,791 self-driving vehicles after an incident in which one of the vehicles drove into a flooded section of the road and became stuck. This is a troubling development for the autonomous vehicle market, not just because of the incident itself: if Waymo’s robotaxis fail in a basic scenario like standing water on a highway, questions arise not only about the pretty demos, but about engineering discipline, maps, weather restrictions, and the update process.
This week, Waymo suspended its self-driving taxi service in the US on highway routes. The company also suspended its robotaxis in Atlanta, Georgia, to improve their control software. After the software update, the automated vehicles should be better able to navigate flooded roads and areas undergoing construction. Waymo notes that some improvements will be made to the control algorithms, after which robotaxis will be able to return to US highways. Until now, Waymo robotaxis have been available on highways in the San Francisco, Los Angeles, Miami, and Phoenix areas.
Within city limits in the regions where Waymo operates, taxi services continue as usual. Robotaxis continue to experience problems in areas where roads are flooded during heavy rainfall. Last Wednesday, an unoccupied vehicle stopped in the middle of a large puddle in Atlanta. This month, Waymo has already updated its robotaxis’ software to train them to navigate flooded areas safely. Specifically, if a flooded area cannot be avoided, robotaxis will now significantly reduce their speed.
The self-driving car industry is increasingly confronted not by the question of “can a car drive itself?” but by the far less dramatic question of “where exactly should it be able to say no to itself?” The flooded road incident hits the sorest spot of autonomous vehicles: the promise that software will be more careful than humans not in ideal conditions, but in dirty, wet, and unpredictable scenarios. Until Waymo’s robotaxis learn to more confidently recognize such boundaries, the market will evaluate not the beauty of the technology, but the quality of its failures.
